Pushing the capabilities of high-resolution X-ray 3D imaging

Researchers at the Paul Scherrer Institute have developed and built the a “microscope” probe that produces 3D images of natural or engineered objects, zooming seamlessly from the macro- to the nano-scale. The probe was now used to monitor the quality of computer chips.

Go to the profile of Mirko Holler
Oct 08, 2019
0
0

Fig. 1 Artistic image showing in grayscale a 3D virtual slices representation of the chip. The lowest layer is visible as well as the stacking of the layers to larger structures. In color a 3D volume rendering is shown at various magnification. The resolution of the dataset is 18.9 nm.

Electron microscopy can reach unmatched resolution, but is either mostly surface sensitive in scanning electron microscopy (EM), or can only penetrate thin samples, on the order of a few hundred nanometers, in transmission EM. In contrast, X-rays of photon energy of a few keV can penetrate much thicker samples. The wavelength of such radiation is small enough that nanometric resolution is in principle possible. However, this high penetration, which comes from weaker interaction with matter, also makes it challenging to fabricate high numerical aperture lenses. In X-ray microscopes the imaging optics often sets the limit of resolution. The ability to probe deeper into volumetric structures, and measure larger and more representative volumes, provides an important complementarity to electron microscopy.

At the Paul Scherrer Institut we pioneered Ptychographic X-ray computed tomography (PXCT), an extension to 3D of Ptychography. Ptychography is a lens-less imaging technique that combines scanning microscopy and coherent diffractive imaging: A sample is illuminated by a confined and coherent X-ray beam and far-field diffraction patterns are recorded at many regions of the sample with overlapped illumination [1]. The imaging system is replaced by an iterative reconstruction algorithm, which solves the phase problem and reconstructs a real-space image of the sample. The achievable resolution is therewith neither limited by the beam size, nor by the step size, but determined by the effective numerical aperture of the intensity detector array, the signal-to-noise ratio of the measured intensities, and the positioning accuracy of the scanning instrument.

After the first demonstration of PXCT [2] we started developing dedicated instrumentation for the technique. A special laser interferometry ensures sample scanning with nanometric accuracy. Continuous improvement of such instrumentation allowed demonstrating high-resolution imaging of integrated circuits in 2017 [3], allowing us to identify circuit connections in all layers of the chip down to the transistor layer in Intel’s 22 nm FinFET technology. A caveat of PXCT is that a small sample has to be extracted from the chip and mounted isolated, such that it is accessible to the radiation from all sides. This means that, although the X-ray measurement is non-destructive, the sample preparation is destructive, removes the sample region from the surroundings, and the measurement position must be selected prior the X-ray measurement. That is a clear limitation for many samples whose native state is within a flat extended surface.

A solution to this problem is to migrate to a laminography geometry, for which the axis of rotation is no longer perpendicular to the direction of X-ray propagation. First proposed in 1974 [4] and later implemented in conventional X-ray projection imaging [5], laminography would  significantly simplify sample preparation and allow mounting full chips, if we manage to significantly increase imaging resolution. Ideally, various magnifications should be available to allow selecting a region of interest for a high-resolution measurement based on low-resolution data. The challenge was thus to combine laminography with X-ray ptychography, a technique we now call PyXL (Ptychographic X-ray computed Laminography). As mentioned, nanometric sample positioning accuracy is required, and we thus developed a dedicated instrument. Again, we decided to measure and control the scanning positions using laser interferometry, but the basic concepts used for PXCT were completely unsuited to the task, due to the large field of view specification of 10 mm x 10 mm. New ideas were developed and implemented; engineering and building of the Laminography Nano Imaging instrument (LamNI) took about 15 months. We profited immensely from the expertise we built when designing and testing our tomography instruments. For our very first PyXL X-ray imaging test we prepared and mounted a full-sized chip for general-purpose logic manufactured in 16 nm FinFET technology, a sample we had already characterized with electron microscopy and PXCT. To ensure we would measure the same region in LamNI, the region of interest was marked by depositing a cross shaped structure on the top of the chip using the gas deposition system, with platinum precursor, in a focused ion beam machine. In the 2D projections we obtained during the X-ray experiment we were at first disappointed, because we did not see this alignment structure at all. But we were confident that the individual projections were of high quality and probably the contrast of this structure was just not strong enough. We kept going and recorded low- and high-resolution laminograms.

Novel codes for projection alignment and reconstruction, for generating a three-dimensional dataset from the collected projections were ready but only tested on simulated data. It took some time to get them running properly on the first real PyXL data, but already a handful of days after the measurement we had a first 3D reconstruction, and we were very excited to see the previously mentioned cross marking in a virtual slice in the 3D dataset – we indeed did the measurement at the right spot, and moreover: LamNI works!

Further improvements in the data reconstruction allowed improving image quality to resolve all chip layers and interconnects down to the lowest layer, results now published in Nature Electronics  https://www.nature.com/articles/s41928-019-0309-z [6]. The resolution achieved in this first measurement was 18.9 nm and we now have a new, unique tool at hand, which allows measuring flat samples at high resolution in 3D without the need to undergo a complex sample preparation.

Video 1 3D rendering of the chip dataset demonstrating the zooming capabilities and resolution achieved.

Moreover, various magnification levels are accessible within the same instrument. The resolution obtained is about an order of magnitude better compared to what was demonstrated in laminography before [7]. The method is not limited to integrated circuits but useful whenever a high-resolution image of a flat sample is needed, which has many useful applications for materials that don’t lend themselves well to be milled in pillars due to loss of structural integrity, and it simplifies significantly in situ or operando measurements. We are now starting to explore other areas of science with LamNI. The Swiss Light Source (SLS) at the Paul Scherrer institut is a user facility, and our developments are accessible to scientists and industry worldwide. If you believe that your science project can benefit from high-resolution three-dimensional X-ray imaging, get in touch with us.

Written by Mirko Holler and Manuel Guizar-Sicairos. 

Links to the original work and related stories:

https://www.nature.com/articles/s41928-019-0309-z

https://spectrum.ieee.org/nanoclast/semiconductors/design/xray-tech-lays-chip-secrets-bare

https://www.psi.ch/en/coherent-x-ray-scattering/scientific-highlights/3d-imaging-for-planar-samples-with-zooming

https://sciglow.com/tech/3d-imaging-of-computer-chips-for-security-quality-control-and-reverse-engineering/

References

[1] F. Pfeiffer, “X-ray ptychography,“ Nat. Photonics 12, 9–17 (2018)

[2] M. Dierolf, A. Menzel, P. Thibault, P. Schneider, C. M. Kewish, R. Wepf, O. Bunk, F. Pfeiffer, “Ptychographic X-ray computed tomography at the nanoscale,” Nature 467, 436–439 (2010)

[3] M. Holler, M. Guizar-Sicairos, E. H. R. Tsai, R. Dinapoli, E. Müller, O. Bunk, J. Raabe, G. Aeppli, “High-resolution non-destructive three-dimensional imaging of integrated circuits,” Nature 543, 402–406 (2017)

[4] F. A. Hasenkamp, “Radiographic Laminography”, Mater Eval 32, 169 (1974).

[5] L. Helfen, T. Baumbach, P. Mikulík, D. Kiel, P. Pernot, P. Cloetens, J. Baruchel, “High-resolution three-dimensional imaging of flat objects by synchrotron-radiation computed laminography,” Appl. Phys. Lett. 86 071915 (2005).

[6] M. Holler, M. Odstrcil, M. Guizar-Sicairos, M. Lebugle, E. Müller, S. Finizio, G. Tinti, C. David, J. Zusman, W. Unglaub, O. Bunk, J. Raabe, A. F. J. Levi, G. Aeppli, “Three-dimensional imaging of integrated circuits with macro to nanoscale zoom,” Nat. Electron. 2, 2018, https://www.nature.com/articles/s41928-019-0309-z

[7] F. Xu, L. Helfen, H. Suhonen, D. Elgrabli, S. Bayat, P. Reischig, T. Baumbach, P. Cloetens, “Correlative Nanoscale 3D Imaging of Structure and Composition in Extended Objects,” PLoS One 7 e50124 (2012).

Go to the profile of Mirko Holler

Mirko Holler

Dr., Paul Scherrer Institut

No comments yet.