Coalescence (film rupture) in evolving foams is measured from radiographic projection image sequences by detecting temporal discontinuities; attributing such discontinuities to coalescence is valid only under a temporal-separation condition: the frame rate must resolve every continuous structural change (bubble motion, growth, film stretching, T1 topological rearrangements) as smooth image evolution, while film ruptures (about 600 microseconds in metal foams) appear as discontinuities. Three coupled gaps remain open: detection when temporal sampling is marginal, because other abrupt structural changes then also appear as discontinuities and no procedure isolates the rupture ones; attribution when coalescence shares its timescale with other fast dynamical processes, in which case no recording rate separates rupture from non-rupture; and quantitative validation, since sufficiently precise mathematical models of realistic foaming processes do not exist and current benchmarks use synthetic, non-physical event signatures. The 2012 optical-flow successor by the same group explicitly leaves discrimination between coalescence and comparably fast rearrangements unresolved; later literature (1250 Hz radioscopy, 3D X-ray tomoscopy up to 1000 tomograms per second) escalates hardware and changes modality rather than answering the 2D-projection attribution question. The problem remains open.

Background

A foam is a dispersion of gas bubbles in a liquid (or solidifying) matrix; in evolving (liquid) foams the structure changes continuously through liquid drainage, bubble growth by gas diffusion, and topological rearrangement. Coalescence — rupture of the thin liquid film separating two bubbles, merging them into one — is the fastest of these processes and a main route to foam collapse, so quantitative, time-resolved coalescence measurements are central to foam-stability studies of aqueous surfactant foams, polymeric foams, and metal foams produced for lightweight structural parts. Because such systems are opaque or hot, the standard in situ probe is X-ray projection radiography: a sequence of two-dimensional images f(x,y,t)f(x,y,t) in which the three-dimensional foam structure is integrated along the beam direction onto a detector plane, so depth information is lost and structures superpose.

Two classes of topological change appear in such sequences. A T1 event is a neighbour exchange in which bubbles rearrange their contacts without any film breaking; in aqueous foams it is reported to take of order 40 ms. Film rupture (coalescence, sometimes called a T2 process when a bubble disappears) is far faster: in metal foams a retracting film is reported to vanish in roughly 600 microseconds. Early radiographic foam studies counted ruptures manually, frame by frame, which is laborious and cannot measure event size; automatic thresholded frame differencing between successive images raises false alarms whenever foam constituents move, so it works only on nearly static foams.

Myagotin, Helfen and Baumbach (2009) proposed the first automatic discontinuity detector aimed at evolving (moving, growing) foams. By the Fourier shift theorem, a pure translation of image brightness between two frames leaves the amplitude spectrum unchanged; therefore a significant value of the normalized amplitude-spectrum difference ε\varepsilon between successive frames — or between recursively refined spatial partitions of them, a scheme called spatio-temporal partitioning — marks a structural discontinuity, which is attributed to coalescence. Validation used synthetic benchmarks with known ground truth: a simulated system of spherical bubbles (correctly localizing 60–80% of the projected coalescence area with 15–20% false-positive area, against more than 80% false positives for difference-based detection) and, more realistically, a binarized micro-tomogram of a real metal foam subjected to simulated motion and expansion with random synthetic material removal (60–75%). The method was applied to synchrotron radiography of an evolving AlSi9Mg0.5/SiC aluminium foam recorded at 1–20 Hz. A 2012 successor by the same group replaced the spectral test with variational optical flow (dense frame-to-frame motion estimation) plus a forward–backward consistency check, which suppresses discontinuities explainable by motion; it reached a receiver-operating-characteristic area under the curve of 0.97–0.98 on a synthetic irregular-foam benchmark and was applied at 2–10 Hz to aqueous foams, in-mold metal foams, and polymer foams filled with multi-walled carbon nanotubes (referred to in the paper as admixed solid nanoparticles).

The whole detection principle rests on a temporal-separation assumption: the recording (frame) rate must be high enough that every continuous structural change — motion, growth, film stretching, T1 rearrangements — appears as smooth image evolution, while rupture, being much faster, appears as a temporal discontinuity. The 2009 paper states as unresolved limitations exactly the regimes in which this assumption fails: at low imaging rates other abrupt structural changes also produce discontinuities; when the rupture duration is comparable to that of other dynamical effects, no recording rate separates them; and the synthetic benchmarks are only approximate because sufficiently precise mathematical (forward) models of realistic foaming processes, including all coupled phenomena, do not exist. The projection model additionally assumes that scattering of the probing radiation is negligible relative to transmission, an assumption that must be verified for each imaging modality, probe energy, and sample system.

Problem Statement

Discontinuity-based measurement of coalescence in evolving foams from time-sampled projection (radiographic) image sequences is currently valid only when a temporal-separation condition holds: the recording rate resolves every continuous structural change — bubble motion, growth, film stretching, T1 topological rearrangements — as smooth image evolution, while film ruptures appear as temporal discontinuities. The open problem, stated by Myagotin, Helfen and Baumbach (2009) and left unresolved by the 2012 optical-flow successor, is to extend this quantitative coalescence measurement to the regimes in which that condition fails:

  1. Marginal temporal sampling. When the imaging rate is too low, other abrupt structural changes (for example T1 rearrangements) also appear as discontinuities; a procedure that isolates, by post-processing or otherwise, those discontinuities that correspond to coalescence is missing, and the 2012 successor likewise left that isolation unresolved. The presence and relative frequency of the competing fast events must be determined beforehand in order to exclude biased detection results.

  2. Overlapping timescales. When the duration of coalescence events is comparable to the characteristic duration of other dynamical effects (rapid film stretching, T1 transitions, bubble rearrangement), no recording rate cleanly separates rupture from non-rupture processes, and the attribution of detected discontinuities to coalescence loses its unambiguous basis.

  3. Validation without realistic forward models. Unambiguous ground truth (for example film rupture) cannot be determined in real radiographs, and sufficiently precise mathematical models for realistic simulation of foaming processes — incorporating the relevant coupled physical phenomena and the projection image-formation chain — are not available; the correspondence of theoretical drainage and coalescence models with real foaming processes is still far from being completely established. Quantitative validation therefore rests on approximate synthetic benchmarks with non-physical event signatures (rupture modelled as spherical material-density removal rather than film retraction) that do not properly simulate the real physical aspects of foaming.

These three gaps are coupled components of the open problem rather than alternatives: a complete resolution addresses the attribution core — isolating coalescence discontinuities under marginal sampling (gap 1) and separating rupture from non-rupture when timescales overlap (gap 2) — and the quantitative-validation demand (gap 3), which is met by acquiring or constructing a sufficiently precise forward model, or an equivalent unambiguous ground-truth source, rather than by the detection procedure alone.

References

  1. A. Myagotin, L. Helfen, T. Baumbach, 'Coalescence measurements for evolving foams monitored by real-time projection imaging', Measurement Science and Technology 20, 055703 (2009). DOI: 10.1088/0957-0233/20/5/055703, https://doi.org/10.1088/0957-0233/20/5/055703
  2. A. Myagotin, A. Ershov, L. Helfen, R. Verdejo, A. Belyaev, T. Baumbach, 'Coalescence analysis for evolving foams via optical flow computation on projection image sequences', Journal of Synchrotron Radiation 19, 483–491 (2012). DOI: 10.1107/S0909049512015099, https://doi.org/10.1107/S0909049512015099
  3. A.-L. Biance, A. Delbos, O. Pitois, 'How Topological Rearrangements and Liquid Fraction Control Liquid Foam Stability', Physical Review Letters 106, 068301 (2011). DOI: 10.1103/physrevlett.106.068301, https://doi.org/10.1103/PhysRevLett.106.068301
  4. M. Paeplow, F. García-Moreno, A. Meagher, A. Rack, J. Banhart, 'Coalescence Avalanches in Liquid Aluminum Foams', Metals 7, 298 (2017). DOI: 10.3390/met7080298, https://doi.org/10.3390/met7080298
  5. F. García-Moreno, P. H. Kamm, T. R. Neu, F. Bülk, R. Mokso, C. M. Schlepütz, M. Stampanoni, J. Banhart, 'Using X-ray tomoscopy to explore the dynamics of foaming metal', Nature Communications 10, 3762 (2019). DOI: 10.1038/s41467-019-11521-1, https://doi.org/10.1038/s41467-019-11521-1
  6. F. García-Moreno, P. H. Kamm, T. R. Neu, F. Bülk, M. A. Noack, M. Wegener, N. von der Eltz, C. M. Schlepütz, M. Stampanoni, J. Banhart, 'Tomoscopy: Time-Resolved Tomography for Dynamic Processes in Materials', Advanced Materials 33, 2104659 (2021). DOI: 10.1002/adma.202104659, https://doi.org/10.1002/adma.202104659
  7. J. Graetz, J. Dombrowski, A. Eggert, A. Rack, U. Kulozik, J. Hinrichs, R. Hanke, S. Zabler, 'Synchrotron micro-CT for studying coarsening in milk protein-stabilized foams in situ', Colloids and Surfaces A: Physicochemical and Engineering Aspects 601, 124832 (2020). DOI: 10.1016/j.colsurfa.2020.124832, https://doi.org/10.1016/j.colsurfa.2020.124832