X-ray imaging and 3D reconstruction of in-flight exploding foil initiator flyers

Exploding foil initiators (EFIs) or chip slapper detonators,^1,2 compared to other means of initiating high explosives, offer superior timing and safety performance.³ EFIs consist of a thin conductive foil that is heated and vaporized by a high-voltage, high-amperage electric current. This vaporizing metal accelerates a thin plastic flyer to several km/s across a small (∼100 μm) gap. This flyer then strikes and shock initiates⁴ an explosive such as HNS or PETN. EFIs offer enhanced safety for two main reasons: first, one can directly initiate less-sensitive high explosives^4,5 eliminating the need for often highly sensitive and thus dangerous primary explosives, and second, the initiator and associated electrical hazards are not in direct contact with the explosive prior to detonation.⁶ Slappers also offer precise timing relative to conventional fusing options—the shots in this paper had a standard deviation timing difference at 0.92 mm above the surface of 15 ns, and further much of this is likely due to the hand-soldered and hand-aligned bridges in the field-of-view.

To date, experimental EFI studies have generally relied on variants of velocimetry^2,7 and/or the ability to initiate particular explosives in “go, no-go” style tests.^8,9 Small ∼1 mm length scales coupled with ∼km/s velocities have restricted high fidelity direct experimental measurement of fundamental properties such as the actual shape of the plastic flyer, nature of the metal plasma, and electrical contact performance. In these specific cases, modeling provides higher-fidelity insight,^10–12 but models lack direct comparison to experimentally measured flyer morphologies. The shape of the foil itself affects performance,¹³ but how the shape of the in-motion flyer affects initiation is unknown because techniques to experimentally measure flyer morphology during operation have not had sufficient fidelity. This work combines state-of-the art imaging capabilities with recent implementations of computed tomographic reconstruction algorithms to acquire x-ray images and to generate three-dimensional snapshots of EFI morphology during operation.

Images were acquired at 32ID at the Advanced Photon Source, Argonne National Laboratory, using the 4-camera system from the LANL IMPULSE endstation.^14–16 Images were acquired during 24-bunch mode, which provides sub-100 ps x-ray pulses every 153.4 ns, defining the temporal resolution of the imaging. Imaging used the 1st harmonic undulator white beam, tuned to ∼11.6 keV. The camera frames consisted of 1 K × 1 K images with a pixel resolution of 1.42 μm. Images were normalized to those acquired on the same cameras right before each shot.

EFIs were comprised of square (635 μm)² Cu exploding bridge foils, with electrical contacts at two opposite ends. The bridge foils were covered by Parylene C flyers.¹⁷ The EFI was initiated with 2.5 kV applied to a 0.24 μF fireset, with about 99 nH inductance and.18 Ω resistance in the system. These parameters were chosen to match imaging capabilities such that 3 frames of the operating EFI could be acquired. Although flyer velocity was tuned to 2.5 km/s, images are essentially snapshots: over conservative temporal resolution of the x-ray pulses (<100 ps) the flyer will only travel no more than 0.25 μm, which is nearly an order of magnitude smaller than the pixel resolution. Further, at 11.6 keV, 635 μm of Parylene C has a computed transmission of about 43%, which is optimal for imaging and computed tomography.

Figure 1 presents background-normalized x-ray images of the EFI. The top panes present images acquired perpendicular to the current direction, or 0°, while the lower frames present frames acquired parallel to the ignition current, here referred to as 90°. The electrical contacts are at the top and bottom of the bridge foil in the top panes, while contacts are in- and out-of-plane in the lower frames. The first, leftmost frames were acquired right before initiation, and we define this as t = 0. The second frames image the flyer as the top surface reaches 0.16 mm above the surface, the third, about 0.53 mm, and the rightmost at 0.92 mm. Several features are immediately apparent. In the 2nd frames, the flyer is still attached to the surface by the contacts in the 0° view, and the EFI bursts more rapidly at the interface between contacts and foil. This pre-burst evidently cuts the flyer; in all three in-flight 0° views, the lateral extent is about 0.61 mm, consistent with the 0.635 mm foil size. Conversely, the parallel-to-the-current views show no distinct boundary; rather, the plastic flyer stretches away from the surface. The exploding/vaporizing copper foil is apparent behind the Parylene C flyer, particularly in the 2nd views. This Cu plasma has structure: in the parallel version appearing as lines in the same direction as the flyer velocity, most visible in the 3rd frame. Normal to the current, these straight lines are not as apparent but more rounded structures appear near the contact points.

Flyer velocities can be directly determined from the images and compared to photonic Doppler velocimetry (PDV). Figure 2 presents measured PDV of an EFI initiated under identical conditions as the imaging experiments, the calculated flyer distance above the surface, and the temporal position of the x-ray pulses that imaged the in-flight slapper, and the estimated flyer to surface distances based on the images. As synchrotron x-ray pulses reliably arrive in short, sub-100 ps pulses every 153.4 ns, we can deduce flyer velocity from the images and compare this to the PDV. Images give an average velocity of 2.41 ± 0.05 km/s between frames 2 and 3, and 2.52 ± 0.05 km/s between frames 3 and 4. The PDV of that particular shot gives 2.44 km/s and 2.60 km/s, slightly faster, but consistent with velocity derived from the images.

In addition to 0° and 90°, images on shots were acquired at 15°, 30°, 45°, 60°, and 75° to gain further information about the flyer and to enable 3D reconstruction.¹⁸ CT reconstruction requires a high degree of reproducibility between shots at different incident angles; these shots were sufficiently reproducible but required alignment prior to reconstruction.¹⁸ Alignment of the x-ray images was achieved using a custom image-processing algorithm that implements a 1-D search-based, edge-detection scheme where the magnitude and sign of the smoothed first derivative is employed to identify the front edge of the slapper. The orthogonal, side-to-side, alignment of the slapper additionally required evaluation of the center symmetry axis by fitting of the slapper front edge, the position of side edges (where contrast was sufficient) and center-of-mass. To assist the reconstruction and using slapper symmetry, views at 105°, 120°, 135°, 150°, and 165° were simulated using mirror images of views at 75°, 60°, 45°, 30°, and 15°, respectively. These 11 views were then used as input for the reconstruction. An iterative reconstruction algorithm that excels reconstruction of objects from few-view data was selected to generate three-dimensional images of the in-flight slapper. To this end, Adaptive Steepest Descent Projection onto Convex Sets (ASDPOCS)¹⁹ was implemented in the new software package Livermore Tomography Tools (LTT) and used to reconstruct the slapper flyer. The LTT-reconstructed volumes were down-sampled to a resolution of about 11.3 μm, to further reduce noise, and rendered using Avizo.²⁰ Our results represent the first application of these algorithms to reconstruct 3D images of a dynamically operating ∼1 mm slapper detonator systems using ∼100 ps x-ray pulses.

Figure 3 shows the rendered 3D reconstruction, particularly a region of interest (ROI) of the flyer itself, for the frames at 0.16 mm and 0.53 mm above the surface, corresponding to the middle two panes of Figure 1.¹⁸ In the left, acquired with the slapper about 0.16 mm above the surface, the square shape of the flyer is evident in the top-down view. Differences in images along and perpendicular to the current direction are evident with contacts clearly visible in the orthogonal (bottom left) acquisition. The evolving shape at 0.53 mm appears to elongate perpendicular to the current direction. Cu plasma under the flyer contains some artifacts due to the lack of precise shot-to-shot reproducibility in structure seen in the radiographs. In this view, the contacts continue to expand circularly outward, and the flyer has a slight curvature that has increased compared to the 0.16 mm case. This case represents an extreme test of the new algorithms: reconstructing an object from few-view, noisy radiographs that depend on shot-to-shot uniformity. Given these challenges, standard filtered backprojection fails while iterative reconstruction produces a reconstruction that shows the evolving 3D shape of the flyer at 0.16 mm and 0.53 mm above the surface.

Combining the radiograph and reconstructed images provides detailed flyer morphology with time. At early times, the Cu plasma is accelerating the flyer, but it is still attached to the surface by the pre-bursting contacts in one direction, and by the plastic itself in the other. The pre-burst severs the Parylene C near the foil contact points: in the 0° views the lateral extent of the flyer does not change appreciably and is about 0.61 mm in each of the 0.16, 0.53, and 0.92 mm views. The curvature in this direction does evolve; in the 0.16 mm view, the flyer has a flat region that is nearly as large as the foil. At 0.53 mm in the 0° view, the flyer shape has a radius of about 0.69 mm; this radius shrinks further to 0.54 mm when the flyer is 0.92 mm above the surface. The flyer clearly elongates in the other direction perpendicular to the current as seen in the 90° views and in the top-down reconstructed view: the flyer is over 1 mm long at 0.53 mm above the surface. The flyer material is accelerating both forward and laterally; in late-time views, the edges of the expanding flyer often overrun the 1.4 mm field-of-view, for example, in the final 90° view. The flyer also develops fissures at late times as seen in the rightmost 0.92 mm views in Figure 1. The rich imaging data on EFI and flyer microstructure with time represents a new opportunity to refine understanding of flyer morphology during operation of slapper initiators. Flyer shape, curvature, Cu plasma, flyer break-up, and elongation have all been measured with ultra-fast x-ray radiographs and the corresponding 3-D reconstruction; these parameters can now be verifiably tuned to achieve optimal performance in EFI initiators.

In this paper, the EFI metal plasma and plastic flyer traveling at 2.5 km/s were imaged with short ∼80 ps pulses spaced 153.4 ns apart. Slapper flyer velocities derived from photonic Doppler velocimetry are consistent with velocities extracted from the images. Seven series of images were used to reconstruct the flyer in 3D from images where the front of the flyer was 0.16 and 0.53 mm above the surface. Both the x-ray images and the 3D reconstruction show a strong anisotropy in the shape of the flyer and underlying foil parallel vs. perpendicular to the initiating current and electrical contacts. These results provide the first detailed flyer morphology during the operation of the EFI and such information can now be used to verifiably tune flyer morphology for optimal performance.

See supplementary material for a complete set of the views used for CT reconstruction, views demonstrating shot-to-shot reproducibility, and a multimedia rendering of the flyer.

We acknowledge Kamel Fezza and Alex Deriy of XSD, Sector 32ID, APS, and Charles T. Owen and Mike Martinez of LANL for experimental support, and Paul Wilkins of LLNL. This work was performed under the auspices of the U.S. Department of Energy by Lawrence Livermore National Laboratory under Contract No. DE-AC52-07NA27344. Imaging was performed on the LANL-developed Impulse endstation. EFI imaging was funded by LLNL LDRD-14-ERD-018. Use of the Advanced Photon Source, an Office of Science User Facility operated for the DOE Office of Science by Argonne National Laboratory, was supported by the U.S. DOE under Contract No. DE-AC02-06CH11357. LLNL-JRNL-670442.