Using X-ray tomoscopy to explore the dynamics of foaming metal

The complex flow of liquid metal in evolving metallic foams is still poorly understood due to difficulties in studying hot and opaque systems. We apply X-ray tomoscopy –the continuous acquisition of tomographic (3D) images– to clarify key dynamic phenomena in liquid aluminium foam such as nucleation and growth, bubble rearrangements, liquid retraction, coalescence and the rupture of films. Each phenomenon takes place on a typical timescale which we cover by obtaining 208 full tomograms per second over a period of up to one minute. An additional data processing algorithm provides information on the 1 ms scale. Here we show that bubble coalescence is not only caused by gravity-induced drainage, as experiments under weightlessness show, and by stresses caused by foam growth, but also by local pressure peaks caused by the blowing agent. Moreover, details of foam expansion and phenomena such as rupture cascades and film thinning before rupture are quantified. These findings allow us to propose a way to obtain foams with smaller and more equally sized bubbles.

information and unprecedented time resolution provided by the method. Perhaps this can be done in follow-up studies/analysis on these datasets to provide a more in-depth modeling, beyond morphological description. Some of the potential future work on this may be discussed. 3. There have been many high-speed tomoscopy experiments conducted at TOMCAT in the past. Has the time resolution been significantly improved in this work? Or has the same time resolution been demonstrated before so this work is more on the application to the foaming metal? The reviewer is trying understand this point because the manuscript put a great emphasis on the development of the tomoscopy.
Minor comments: 1. In the 2nd paragraph, the authors refer metal foam as a still emerging material -the authors shall reconsider this. Metal foams have well established literatures in a wide variety of topics, and dedicated publications and conferences; it would be fair to say that it remains as an active research field with evolving topics and new categories of materials/processing methods.
2. On page 2 of the manuscript, the author cited 4.9 micron voxel size as 'high spatial resolution'. It should be noted that voxel size is different from the spatial resolution. The authors are encouraged to clarify this.
Reviewer #3 (Remarks to the Author): The authors presented x-ray imaging experimental results on metallic foam coalescence behaviors. They proposed a new tomography reconstruction technique to improve the measurements' temporal resolution. However, the reviewer found the current work is quite superficial. The new results are not major breakthrough upon what have been reported in literatures. The reviewer recommends publishing the work on a more specialized journal with more focused audience. Below are reviewer's concerns.

Dear reviewers,
Thank you for your reviews. We have carried out a number of changes in the paper and the supplementary material. We have assigned a colour to each of the reviewers to make it easier to track the changes in the text, which are written in the colour of the respective reviewer.
We hope that we can convince you that our paper is suitable for publication in Nature Communications now.

Summary
The authors have demonstrated capability to obtain 3-d (tomographic) X-ray images at an order of magnitude higher rate that reported previously. They have applied this technique to observe the structure changes in a metallic foam produced by heating an alloy conatining outgassing agents. The technique being performed at a syncrotron (where the X-ray beam in necessarily fixed) requires that the sample be rotated at 100rpm. The influence of the inertial forces pertaining on the sample have been compared by contrasting with radiography results obtained in microgravity. They have shown dramatic new capability in X-ray imaging, and applied it to a problem of both technological importance and fundamental scientific interest. The paper should be published and I think it is suitable for Nature Communications.
One thing that was not clear to me, is what is the main reason or key technology allowing the greater speed , is there greater flux at Tomcat than previously ? or simply that the rotation stage is superior to previous ones? The article only stated "The work is enabled by our recent experimental improvements that speed up tomography so much, that 3D images can be obtained in a movie-like mode with≥25 tomograms per second" and "The new, self-developed high-speed rotation stage", which is a bit vague.
The flux at the synchrotron has not changed. The rotation stage was an important element. Another key point was the ability to stream the data directly from the camera chip to the computer system. Finally, automated image analysis had to be developed. With all this together, the tomoscopy acquisition rate could be increased by a factor of 10 and the previous restriction of only short experiments overcome. Now we can acquire 40,000 tomograms and more in a series. → New passage in the main text, p. 2, and description of rotation stage in supplementary material, p.

Abstract
In the first sentence there is an ambiguity in the English grammar, the term 'emerging metallic foams' gives the impression that the foams are being extruded or expelled from an opening. I believe the authors mean to stress that metallic foams are 'emerging' as a potentially useful form of material. The opening sentence would be improved by just removing the word 'emerging'. The intention was to emphasize that metallic foams are evolving in the liquid state, not that foams are an emerging material. → We have replaced the ambiguous word "emerging" by "evolving".
I'm afraid I don't agree with creating a new terminology only to create some arbitrary division of time-scales. For exampple, if 12500 tomograms are obtained in 12500 seconds (3 1/2 hours at 1hz) the type of data and analysis woudl be similar (for a process having that time scale) Other authors use '4-d imaging' and I must say I don't like that either, I prefer 'Time-resolved tomogrpahy', but if any new term is used I beleive it should refer to the extension in time regardless of the absolute speed, so I beleive, at least, the notion that one name shoudl be used for greater than 25tps and another for less than 25tps should be eliminated.
Our intention was to move away from terms like "fast tomography", "ultrafast tomography" etc. and to create a term that will be also valid in 20 years after acquisition rates have increased again and what is fast today is then slow.
→ We understand the concern that 25 tps is an arbitrary limit and have decided to drop this part of the definition. We now define in the main text, p. 2 (and shortly in the abstract), that tomoscopy is: • time-resolved 3D tomography, • that is applied to an evolving system to clarify its dynamics, • that is "continuous", i.e. not just the state before and after a change are captured but we are monitoring various stages during evolution of the system (without specifying how many) Now there is an analogy between 2D and 3D: (2D) radiography = static → radioscopy = dynamic (3D) tomography = static → tomoscopy = dynamic

Reviewer #2 (Remarks to the Author):
The work presents a novel operando 3D characterization of foaming metal by X-ray tomoscopy. The direct visualization of the process reveals great information that were not accessible before. The comparison of the structure foamed under normal gravity and micro-gravity is also scientifically interesting. The work, if published, will likely lead to great interests from the readers in this and other relevant fields. Some questions/comments for the authors to consider: 1. Can the authors comment on, would the rapid rotation in tomoscopy/tomography affect the foaming process?
This is an important point since radial accelerations can be high. We are dealing with early stages of foaming in the 208 tps studies, which implies that the foam is still rather incompressible. Moreover, the alloy under consideration is known to be quite stable against drainage effects. This is why we were not surprised (and relieved) that drainage does not lead to visual segregation of bubbles and liquid. In order to support this qualitative argument we have calculated a radial density profile of a sample, which was rotated slowly (1 tps), at medium speed (50 tps) and fast (208 tps).
→ the supplementary information now contains a description of these density profiles on p. 2 and in Fig. S6.

2.
Overall, the reviewer would expect that more quantitative discussions/analysis regarding the kinetics and fundamental mechanisms on the foaming could be conducted, given the rich information and unprecedented time resolution provided by the method. Perhaps this can be done in follow-up studies/analysis on these datasets to provide a more in-depth modeling, beyond morphological description. Some of the potential future work on this may be discussed.
We are discussing fundamental phenomena such as nucleation already but for sure we are planning to extract more data from the tomoscopy series in the future. The limitations at the moment are the algorithms needed to process 40,000 tomograms automatically and in a reliable way. State-of-the art are bubble size, shape and number density calculations on a bulk of tomograms. Moreover, analysis of individual local features such as in Figs. 3, 4 and 6. We still have to develop analyses in cases of weak contrast and correlation analyses (i.e. between blowing agent particles and bubbles) that include spatial and temporal correlations. This will be part of future work, as well as more work on a greater variety of alloys, temperature profiles etc. that could be included in follow-up studies as suggested.
3. There have been many high-speed tomoscopy experiments conducted at TOMCAT in the past. Has the time resolution been significantly improved in this work? Or has the same time resolution been demonstrated before so this work is more on the application to the foaming metal? The reviewer is trying understand this point because the manuscript put a great emphasis on the development of the tomoscopy.
The highest repetition rate at PSI so far has been 20 tps (see literature overview in supplement). Therefore we demonstrate a 10 times higher time resolution now (first time showing this). The development of a suitable rotation stage was an important element. Moreover, we do this continuously over a long time (60 s), which previously was not possible due to the limited memory of the camera chip and the lack of an on-the-fly processing of data. → This is now stated explicitly on p. 2.
Minor comments: 1. In the 2nd paragraph, the authors refer metal foam as a still emerging material -the authors shall reconsider this. Metal foams have well established literatures in a wide variety of topics, and dedicated publications and conferences; it would be fair to say that it remains as an active research field with evolving topics and new categories of materials/processing methods.
Correct, metal foams have been known for many years but still their use is quite restricted. This is what we wish to express and after a slight change we hope will be understood in this way. → An extra reference (Ref. 2) on p. 1 helps to understand how far the history of metal foams dates back.
2. On page 2 of the manuscript, the author cited 4.9 micron voxel size as 'high spatial resolution'. It should be noted that voxel size is different from the spatial resolution. The authors are encouraged to clarify this. This is correct and as we admit was a bit sloppy. The true resolution has unfortunately not been measured this time, but was determined previously with a similar configuration to be approximately 12 µm.

Reviewer #3 (Remarks to the Author):
The authors presented x-ray imaging experimental results on metallic foam coalescence behaviors. They proposed a new tomography reconstruction technique to improve the measurements' temporal resolution. However, the reviewer found the current work is quite superficial. The new results are not major breakthrough upon what have been reported in literatures. The reviewer recommends publishing the work on a more specialized journal with more focused audience. Below are reviewer's concerns.
We deliberately chose a journal for a general audience (Nat Comm) because the new possibilities of tomoscopy at unprecedented rates (now 208 tps), and already aiming at even faster rates, will stimulate studies in other areas of materials science, mechanical engineering, and applied physics.
In the following we will try to convince the referee that also the specific application of tomoscopy to metal foam yields some completely new insights.
1. Aqueous foam systems are mostly studied without tomographic imaging techniques. For instance, avalanche coalescence rupture behaviors in aqueous foam systems were studied with sound detection and high-speed optical imaging. In metallic foam system, x-ray radiology was also applied to study rupture behaviors. If cascade rupture behavior is normal in metallic foam system, why has it never been found in radiographic imaging experiments?
Aqueous foams have also been investigated by tomography in the past years by various groups [1][2][3][4][5][6][7][8][9][10] including our own group [11][12][13][14][15]. Although water is transparent it is impossible to unravel the structure of deeper layers due to strong light scattering. Using X-ray tomography (slower as in this paper however) we showed e.g. that ordered packings occur after ageing initially disordered foams for a few days [14]. Rupture cascades in metal foams actually have been observed before in radioscopic imaging studies [16], added now as new Ref (31) to the manuscript. However, showing that various rupture events are correlated is not always straight-forward as along the viewing direction features are superimposed. Therefore, tomoscopy as applied in this paper provides better evidence. → We have added Ref. [16] as new Ref (31) to the manuscript and a sentence on p. 4 to explain this fact.
Is the 'cascade rupture' behavior found with 'tomoscopy' representative in metallic foam system?
We have found many cascades by tomoscopy in the samples investigated (two already shown in Figs. We think it would overload the paper if we included too many such examples and that 2 are enough ( Fig. 4 and 6).
Radioscopy on larger samples [16] showed many more events in an AlSiCu alloy (with the caveat that some events might be accidentally aligned features and not real neighbours in 3D). 3. The conclusion of 'the deleterious growth of very large bubbles is mainly caused by the action of the blowing agent and to less extent only by gravity-driven drainage' is also repeating the conclusion in ref. 12.
In former Ref. (12) of the manuscript we concluded that coalescence cannot be the consequence of the removal of liquid from a foam driven by gravity. This was based on the observation of foams produced under gravity and without that showed roughly the same level of coalescence. However, the conclusion was based on an indirect argument (no difference between different gravity levels) not on a direct observation of the actual mechanism. This is what we provide now by visualizing the expanding and strongly inflated bubbles around the blowing agent particles in 3D. The results in Fig.  5 are new and have not been shown in former Ref. (12) that -being a review article on the application of X-ray radioscopy -was very brief.
Moreover, we would like to emphasise that this is just one result of the paper that we have emphasised in the abstract in the usual "Here we show …" statement. We put the emphasis on this because it is a finding that can be translated into technological improvements directly.
→ we have squeezed in some words into the abstract to make clear that this is just one of the results of this paper.
Other important results include: • Volume and bubble number evolve in two stages, which we now know are connected to different gas sources. (Fig. 2) • Melting of the AlMg constituent is the trigger for first expansion (Fig. 2) • Liquid actually retracts from films prior to rupture and it was possible to quantify the kinetics of retraction (Fig. 3) • Rupture cascades have been directly shown in 3D images now (Fig. 4) • Coalescence is caused by 3 mechanisms (drainage coalescence, global growth coalescence and local coalescence), and we can say something about the relative importance of the mechanisms (Figs. 2e, 5). This is a completely new finding • The influence of the oxide layer on bubble shrinkage has been shown by following bubble shrinkage directly (Fig. S5) → the conclusions have been slightly modified to emphasise the link between the coalescence induced by the blowing agent and technology.
In summary, various open questions on the behaviour of liquid metal foams have been answered and we would not call this "superficial".