REVIEW 4 major objections 4 minor 7 references
Multiscale analysis of the textural atomization process of a rocket engine assisted coaxial jet
T0 review · 4 major / 4 minor · reviewed 2026-08-12 · deepseek-v4-flash
Pith's one-line read Improved multiscale image analysis lets a rocket-engine coaxial jet's textural atomization be quantified from ligament shapes, yielding bimodal blob/drop diameter distributions whose sizes and counts evolve regularly with injector distance.
desk verdict The measurement improvement is real and the application is new, but the reported blob/drop size distributions rest on a non-unique fit that needs uncertainty quantification before the quantitative claims should be trusted. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The central object is the scale distribution $e_2(d)$ of the liquid–gas interface, obtained by eroding the segmented liquid system with disks of diameter $d$ and recording the surface area lost at each scale. Its derivative $-e_2(d)_{,d}$ is proportional to the diameter distribution of the equivalent set of cylinders that has the same scale distribution as the real, arbitrarily deformed ligament population. The paper represents that cylinder distribution by a three-parameter generalized gamma (3pGG) function, and in the fully atomized limit ($\alpha=1$) the blob diameter distribution inherits the same 3pGG form, $f_{0s}(D)\propto D^{q-1}e^{-(D/D_s)^q}$. A two-component version of this model is fitted to the measured derivative, with each component corresponding to one blob family; the fitted parameters directly give the mean diameter, width, and relative number of each family through $D_s = q^{1/q}D_c$, $N_s \propto -e_2(0)_{,d}\,D_s^q/\Gamma(q)$. On the measurement side, the enabling step is an exact-distance computation with fractional-distance binning combined with $4\times4$ bilinear subpixel interpolation, which removes the oscillations and pixelization bias that otherwise corrupt the small-scale part of the distribution.
What would settle it
Measure the actual droplet size distribution just downstream of the ligament zone, in a cold-flow twin of this injector with matched Weber number and momentum-flux ratio, and compare it with the blob diameter distribution predicted by the two-component 3pGG fit; if the predicted bimodality and its spatial evolution do not appear in the measured spray, the blob-equivalence step is wrong. A simpler check is to apply the extraction to synthetic ligament images with known blob sizes and verify that the fitted parameters recover the injected blob diameters.
Extended reading notes
Core claim
The paper's central claim is that the multiscale scale distribution of textural ligaments, measured with subpixel fidelity, carries enough information to reconstruct the diameter distribution of the blobs that structure ligament deformation, and that treating these blobs as drops in formation yields quantitative spray information where conventional droplet diagnostics cannot operate. The argument runs through the identity $e_2(d)=L(d)/2$, which relates the scale distribution to the interface length of the eroded liquid system; differentiating once gives a quantity proportional to the diameter distribution of an equivalent set of cylinders, and a second differentiation connects the fully atomized limit to a spherical-drop distribution. Fitting the measured $-e_2(d)_{,d}$ with two 3pGG components separates blob families by scale. In the present reactive coaxial jet the fit identifies two such families, giving a bimodal diameter distribution: small blobs formed by textural deformation of the ligament surface and larger blobs formed by structural deformation of the ligament body. The mean diameters of both families increase linearly with distance from the injector, while the total blob number peaks between roughly 14 mm and 21 mm and the small-blob family disappears at the farthest positions, marking where textural atomization stops.
Load-bearing premise
The entire drop-size estimate rests on treating the swelling bumps of a deformed ligament as spherical drops and on reading the selected straight segments of a double-log plot as distinct blob families; the paper does not check that equivalence against independent droplet measurements.
Editorial extensions
If this is right
- Textural atomization can be followed quantitatively along the injector: blob mean diameters grow linearly with distance, and the blob number per unit width rises then falls, locating the intense atomization zone near 14–21 mm.
- The bimodal blob diameter distribution points to two coexisting production mechanisms in the textural process, one attached to the ligament surface and one to the ligament body.
- The regular spatial evolution of the fitted 3pGG parameters ($q_1,q_2,D_{s1},D_{s2}$) means a closed-form expression for the blob diameter distribution as a function of injector distance is within reach.
- In reactive flows where laser-diffraction or phase-Doppler instruments fail, this image-based route provides an alternative estimate of the droplet population being formed and of where its production is strongest.
Reading between the lines
- A natural next test would be to run the same analysis on a cold-flow twin with matched Weber number and momentum-flux ratio, where conventional droplet sizing is possible, and compare the predicted blob distributions with measured spray diameters; agreement would validate the blob-as-drop assumption, disagreement would localize the error in the equivalence step.
- Because the method needs only images of the liquid interface, it should transfer directly to numerical simulation data: applying the extraction to simulated ligament fields with known blob populations would provide a ground-truth check of the linear-region selection and the two-component fit.
- If the linear growth of blob mean diameter with injector distance is confirmed as a general trend, the multiscale parameters could serve as a surrogate for the local turbulent scales that initiate textural ligaments, linking image-derived morphology to turbulence-driven breakup models.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper presents a multiscale image-analysis method for characterizing the textural atomization ligaments of a cryogenic coaxial rocket injector operating under combustion. It improves the measurement of the scale distribution e2(d) by replacing the integer distance map with an exact distance transform and a 4x4 subpixel interpolation, and it validates this improvement on a synthetic sinusoidally perturbed cylinder. The paper then models the right-hand tail of the second derivative -e2(d),d as a sum of two three-parameter Generalized Gamma (3pGG) components and converts each component into a number-based blob/drop diameter distribution using previously derived equivalent-cylinder/sphere relations. Applying the model as a function of injector distance, it reports bimodal blob diameter distributions whose peak diameters, widths, and numbers evolve regularly with z, and it interprets these as drops in formation.
Significance. The improved scale-distribution measurement is a genuine methodological contribution: the synthetic test in Fig. 5 shows that the subpixel and exact-distance corrections remove pixelization oscillations and sharpen the measured peaks, and the availability of converged averages from 100 images is useful for future work in harsh combusting environments. If the blob-extraction step were independently validated, the paper would offer a route to quantitative estimates of ligament swelling scales and potential drop sizes where conventional droplet diagnostics cannot be used. However, the paper's central quantitative claim, namely the bimodal blob/drop diameter distributions and their spatial evolutions, is not yet supported, because the 3pGG decomposition is non-unique and the conversion has no ground-truth validation. The qualitative result that ligament scales increase with distance and then the textural population ceases is plausible, but the specific peak diameters and parameter trends should be treated as conditional.
major comments (4)
- [§4, Fig. 13, Eqs. (15)-(16)] The two-component 3pGG extraction is non-unique. At z=12.1 mm the paper fits the 'region 2' blob population twice: as the second component of the [52,124] um fit and as the first component of the [74,180] um fit (Figs. 13a-13d). The two recovered region-2 distributions differ in width, and the first extends to the 190 um region-3 peak. No selection criterion, likelihood comparison, or uncertainty estimate is given for choosing the regions-1-and-2 analysis over the regions-2-and-3 analysis. Because each component is a stretched exponential, many two-component sums can approximate the smooth measured curve over finite intervals, so the reported bimodality and the parameter trends in Fig. 15 are not uniquely determined by the data.
- [§4, Eqs. (6)-(12) and (15)-(16)] The conversion of the scale-distribution tail into blob/drop diameter distributions relies on assumptions that are not tested. Section 4 assumes that the swelling scales of the ligaments correspond to circular blobs, that alpha=1 holds for the equivalent sphere set, and that the linear regions identified in Fig. 12 correspond to distinct physical blob families. The synthetic validation in Fig. 5 covers only the measurement of -e2(d),d; it does not test the blob-extraction model. The paper therefore needs either a synthetic test with known blob populations or independent droplet-size measurements to support the claim that the fitted components represent drops in formation.
- [§4, Fig. 12 and Fig. 14 (z=23.5, 25.6 mm)] The manual identification of the linear regions and the handling of absent regions are not quantified. Fig. 12 identifies linear regions by inspection; Region 0 is excluded with only a qualitative justification; and at z=23.5 and 25.6 mm, where region 1 is not found, the scale intervals [d1,d2] are extrapolated from linear trends of d1(z) and d2(z). These choices determine which data enter each fit, but no sensitivity analysis is reported. Consequently, the disappearance of region 1 and the switch to a single-component model at the farthest positions are not robustly established.
- [§4, Fig. 15] No uncertainty quantification accompanies the parameter evolutions in Fig. 15. The quantities q1, q2, Ds1, Ds2, Ns1, and Ns2 are reported as functions of z without confidence intervals, bootstrap estimates, or multi-start checks. In view of the non-uniqueness documented in Fig. 13, the 'clear spatial evolutions' may be within the ambiguity of the fitting procedure rather than physical trends. The paper should quantify the fit uncertainty and demonstrate that the trends are robust to initial parameter choices and interval definitions.
minor comments (4)
- [§4] The sentence 'An example is shown in Fig. 10 for the result obtained at z = 12.1 mm' should refer to Fig. 12, not Fig. 10.
- [§3.2 and §4] Equation (14) is used twice: once for the distance-bin repartition in Section 3.2 and again for the log-linearity relation in Section 4; the equations should be renumbered to avoid ambiguity.
- [Fig. 5 caption] The caption for Fig. 5 should make explicit that panel b is the corrected measurement without subpixel interpolation and that panels c and d use 4x4 and 8x8 interpolation, respectively; the current wording is easy to misread.
- [§3.2, Eq. (16)] The notation m(r_I) is introduced as the total distance count in bin r_I, but the preceding text describes splitting each pixel's contribution between bins r_I and r_I+1; rewording this passage would improve clarity.
Circularity Check
No significant circularity: the blob/drop size information is an explicitly fitted transformation of the measured scale distribution, not a claim of out-of-sample prediction, and the cited multiscale identities are not re-derived from the present data.
full rationale
The paper's central product is an estimation procedure, not a prediction. The measured object is the scale distribution e2(d) (Section 3), and the blob/drop diameter distributions are obtained by fitting a two-component 3pGG model to -e2(d),d (Eqs. 15-16) and converting the fitted components through previously established cylinder/sphere relations (Eqs. 3-12). No quantity is fit on one subset and then 'predicted' on a closely related subset; the spatial evolutions reported in Figs. 14-15 are trends in the fitted parameters themselves and are presented as such. The identities e2(d)=L(d)/2, the equivalent-cylinder representation, and the d_pc=0 sphere limit are cited from prior work by the same group, but they are parameter-free mathematical/geometric relations with stated assumptions and are not fitted in this paper, so those citations constitute legitimate evidence rather than a circular reduction. The measurement chain is independently benchmarked on a synthetic sinusoidally perturbed cylinder (Fig. 5), and the subpixel improvements are validated against that theoretical distribution. The non-uniqueness of the two-component decomposition at z=12.1 mm (Figs. 13b vs 13d), where region 2 is fitted twice with different widths, is a real identifiability and model-selection concern; however, the paper explicitly discloses the difference and gives a physical reason for preferring the regions-1-and-2 analysis. That is a robustness weakness, not a case in which the output is equivalent to the input by construction.
Assumptions & free parameters
free parameters (4)
- 3pGG parameters q1, Dc1, Nc1 for blob population 1 =
Varies with z; see Fig 15
- 3pGG parameters q2, Dc2, Nc2 for blob population 2 =
Varies with z; see Fig 15
- Scale interval [d1,d2] for each fit =
e.g., [52,124] µm and [74,180] µm at z=12.1 mm
- Equivalent cylinder length Lc =
1 µm (arbitrary)
assumptions (4)
- domain assumption The scale distribution e2(d) satisfies e2(d)=L(d)/2 and the equivalent cylinder/sphere relations (Eqs. 3-12).
- ad hoc to paper The right-hand side of -e2(d),d (scales greater than the modal scale) is dominated by swelling scales and can be modeled by a 3pGG function with alpha=1.
- domain assumption The segmented images faithfully represent the liquid interface down to the minimum resolved scale of about 16 µm after 4x4 subpixel interpolation.
- ad hoc to paper The fitted blob populations correspond to drops in formation.
invented entities (1)
-
Blob populations (Regions 1, 2, 3 families)
Cite this review
Pith. "Pith review of Multiscale analysis of the textural atomization process of a rocket engine assisted coaxial jet." pith.science (2026). https://pith.science/paper/YYMNCBPE
@misc{pith2026241117427,
author = {Pith},
title = {Pith review of: Multiscale analysis of the textural atomization process of a rocket engine assisted coaxial jet},
year = {2026},
howpublished = {\url{https://pith.science/paper/YYMNCBPE}},
note = {Machine review of arXiv:2411.17427}
}
read the original abstract
A method for analyzing liquid ligaments of a textural atomization process is presented in this article for the case of a rocket engine type assisted atomization process under combustion. The operating point positions the atomization process in the fiber type regime carrying an intense textural atomization process. Multiscale in nature, the method based on image analysis associates a scale distribution with a family of ligaments, this distribution being sensitive to the number, size and shape of these ligaments. The quality of scale distributions measured by image analysis depends on the spatial resolution and the precision of area measurements of surfaces with curved boundaries but described by square pixels. Part of the work consisted of improving the method for measuring scale distributions by using a subpixel image analysis technique and refining the surface area measurement method. Another part directed the multiscale analysis towards the estimation of the diameter distributions of the blobs that characterize the large-scale deformation of the ligaments. The analysis describes the atomization process at a level of detail never reached. For instance, assuming that the blobs are drops in formation, the estimated diameter distribution (bimodal in the case examined here) as well as the number of these drops are evaluated as a function of the distance from the injector. This information indicates where the process is most intense and where it stops. Furthermore, these diameter distributions receive a mathematical expression whose parameters report clear evolutions with the distance from the injector. This shows the possibility of elaborating mathematical models appropriate for textural atomization mechanisms.
Figures
Figures from the paper (12 more)
Reference graph
Works this paper leans on
-
[1]
Introduction The atomization of a liquid flow in a gaseous environment is a two-step mechanism referred in the literature as primary and secondary atomization processes. The primary atomization process refers to first events of liquid fragment detachmen t from the bulk flow and the secondary process concerns the atomization of these fragments. The primary...
work page 2009
-
[2]
Experimental setup and optical diagnostic The ONERA's MASCOTTE test-bench is used for the pre sent experiments. This bench is a subscale cryogenic rocket combustor capable of reproducing operating conditions similar to those encountered inside liquid rocket engine combustion chambers, i.e, high pressures and high mixture ratios. It is equipped with multip...
work page 2000
-
[3]
Introduced and applied in several previous works (Thiesset et al
Multiscale analysis – Description and Measurement 3.1 Multiscale description The description of the textural deformation of the liquid gas interface in the injector near field region uses the multiscale analysis of the visualized liquid system. Introduced and applied in several previous works (Thiesset et al. 2019, Dumou chel et al. 2017, 2019, 2022, 2023...
work page 2019
-
[4]
Since by definition f0c(0) = 0 then –e2(0) ,d = 0. In a real atomizing liquid system, the textural str uctures are ligamentous but not purely cylindrical: they can be seen as cylinders showing successive contracted and dilated sections. The distribution – e2(d),d, and thus the equivalent-system diameter distribut ion f0c(D), reflect the distribution of th...
work page 2019
-
[5]
8 as a function of the position z
Results and Analysis A series of segmented images of the oxygen jet uppe r interface in the near injector nozzle region are shown in Fig. 8 as a function of the position z. These images, in which the liquid appears in black, are not correlated in time. The liquid-gas interface shows large-scale and smal l-scale deformations. The large-scale deformation ca...
work page 1992
-
[6]
Conclusion The method presented in this article for the descri ption of textural atomization liquid ligaments is convincing and brings back little-known information on this type of process. An important part of this success is linked to the association of high-quality experimental images with improved image processing and image analysis measurement techni...
-
[7]
References Abramowitz M, Stegun, IA (1964) Handbook of Mathema tical Functions. National Bureau of Standards, Applied Mathematics Series 55, June 1964 Boulal S, Fdida N, Matuszewski L, Vingert L, Martin -Benito M (2022). Flame dynamics of a subscale rocket combustor operating with gaseous me thane and gaseous, subcritical or transcritical oxygen. Combusti...
work page 1964
Reviewed August 12, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.