Behind the paper: A Tabletop X-Ray Tomography Instrument for Nanometer-Scale Imaging: Reconstructions

Published in Physics
Behind the paper:  A Tabletop X-Ray Tomography Instrument for Nanometer-Scale Imaging: Reconstructions
Like

In 2018, shortly after I had agreed with management to concentrate on quantum optics and stop working on tomography, I got a call from a colleague who had moved to our Boulder labs about an IARPA project called “RAVEN” (https://www.iarpa.gov/research-programs/raven).  NIST/Boulder was hosting one of three or four efforts “to develop a prototype analysis tool for acquiring images from all layers in a 1 cm2 area of a 14 nm integrated circuit, within 25 days.”  They had found me because late in the 20th century, I’d led a team which had performed the first tomographic reconstruction of an integrated circuit interconnect (Levine et al., 1999). IARPA’s main effort was a synchrotron-based approach based on ptychography as developed by the Paul Scherrer Institute in Switzerland (Holler et al., 2017) and ported to the Advanced Photon Source at Argonne National Labs (Jiang et al., 2021).   Nevertheless, they were interested in pursuing tabletop instruments (assuming you have a 200 kg table with four pneumatic isolation legs).

The idea behind our approach was to achieve great magnification by applying a platinum x-ray target directly on the sample and directing an electron beam there.  The rig featured an array of superconducting transition edge sensors to distinguish Pt Lα  x-rays generated in the target from x-rays generated elsewhere, as envisioned in the original design (Weichman and Lavely, 2020).

There were a few little, um, challenges.  In this geometry, it isn’t possible to measure the source without the sample.  Fortunately, there was an energy dispersive spectrometer available which could be repurposed to measure the x-ray flux from the backside of the sample.  We worked out a way to find the absolute transmission of the sample through energy-dependent measurements averaged over the whole sample.  Getting this right improved our spatial resolution by nearly a factor of two with greatly improved contrast.  A similar improvement came when we used PENELOPE, a widely-used radiation transport code, to model the x-ray source given knowledge of our electron beam and the target design.  Intuitively, one might think that a smaller source size would always be better, but what you really want to know is how well can you characterize the spot.  In our case, the beam current is such a strong function of the spot size that we do better to model the finite spot than to reduce it and lose our precious photons.

I was on the subteam tasked with deciding the scan protocol.  I was able to reach back to our early experiments (Kalukin et al., 1998) to suggest that we mount the sample with the principal wiring directions 45 degrees from our tilt axis.  In this way, the x and y wiring directions are treated equally.  It also helps with what I at least call the “intruder problem”.   In region of interest tomography (we scan only a portion of the total sample), some of the sample is observed only from a fraction of the scan angles.  A solution is to divide the domain into an inner region which gets most of the photons and an outer region which gets fewer.  We accept a transition zone in which we will have relatively high errors so preserve an inner core.  The length scale for the transition zone is comparable to the sample thickness.

Another surprise came when we set up background subtraction to isolate the Pt Lα fluorescence line.  We found that the photons from the nearby x-ray energies were about as useful in making the reconstruction as the narrow band.  Hence, we could use more of the bandwidth than we expected.  We did find some hints of chemical sensitivity, particularly in the first via layer (as discussed in the paper), but a full two-material basis was not of sufficient quality to publish.

In reconstructing images, there is always the question:  how to you know you’re right?  Here, we compare to design files, and the comparison is very good.

What is the next step in the development of the instrumentation?  High-resolution commercial x-ray microscopes are key contenders, and a “Round Robin” comparison would be in order.  The observation that we do not need such fine energy resolution as provided by a TES detector suggests that other photon-counting detectors with an energy resolution around 100 eV, vs. around 10 eV for the TES, are candidates.  The TES arrays themselves are likely to be much larger in the near future and may provide a route to elemental sensitivity. Higher brightness electron sources are also available.  Of course, any new technology has its competitors:  electron microscopy, high-harmonic generation x-ray sources, and compact synchrotrons are all candidates.  Hopefully, our proof-of-concept experiment of x-ray tomography of an integrated circuit interconnect will kindle interest across the board.

  1. Y. Jiang, J. Deng, Y. Yao, J. A. Klug, S. Mashrafi, C. Roehrig, C. Preissner, F. S. Marin, Z. Cai, B. Lai, and S. Vogt, “Achieving high spatial resolution in a large field-of-view using lensless x-ray imaging,” Appl. Phys. Lett. 119, 124101 (2021).
  2. M. Holler, M. Guizar-Sicairos, E. H. R. Tsai, R. Dinapoli, E. Muller, O. Bunk, J. Raab, and G. Aeppli, “High-resolution non-destructive three-dimensional imaging of integrated circuits,” Nature 543, 402 (2017).
  3. Z. H. Levine, B. K. Alpert, A. L. Dagel, J. W. Fowler, E. S. Jimenez, N.~Nakamura, D. S. Swetz, P. Szypryt, K. R. Thompson, J. N. Ullom, “A Tabletop X-Ray Tomography Instrument for Nanometer-Scale Imaging: Reconstructions,” Microsystems & Nanoengineering, in press.
  4. A. R. Kalukin, Z. H. Levine, S. P. Frigo, I. McNulty, and M. Kuhn, “Effects of feature orientation in tomographic reconstructions,” in X-Ray Microfocusing: Applications and Techniques, Ian McNulty, Editor, Proc. SPIE 3449, 36 (1998).
  5. Z. H. Levine, A. R. Kalukin, S. P. Frigo, I. McNulty, and M. Kuhn, “Tomographic Reconstruction of an Integrated Circuit Interconnect,” Appl. Phys. Lett. 74, 150 (1999).
  6. P. B. Weichman and E. M. Lavely, “Fluorescent x-ray scan image quality prediction,” J. Hardw. Syst. Secur. 4, 13 (2020).

 

Please sign in or register for FREE

If you are a registered user on Research Communities by Springer Nature, please sign in