WorksPhysics and Astronomy
Experimental and numerical simulations of near resonant 2D soliton interaction in shallow water
Abstract
Experiments were carried out in {a} large wave basin (,m) to study the interactions of non-linear waves in shallow water (0.35,m deep). A stereoscopic technique was used to measure the waves accurately over a vast area (,m). The numerical code {Tolosa-lct}, designed for simulating wave propagation in shallow water, reproduces the observations and extends the analysis for larger domains. The diffraction of oblique solitons is discussed. Mach interaction of two solitons occurs for some given incidence and amplitudes, producing a third soliton. A focus is made on the near resonant interaction of a soliton with a vertical wall. An example of a regular Mach interaction is shown, where a steady solution is achieved experimentally, with a maximum amplitude amplification of 3.2 at the wall.
Introduction
The soliton is a mathematical object, localized in time and space, which preserves its identity after multiple interactions. The term ‘soliton’ was coined by Zabusky and Kruskal [31] to define this remarkably stable object. It can be interpreted as a localized hump on the free surface in shallow water (such an observation reported a long time ago by Russell [27]) or as a modulation of the carrier wave of wave groups in intermediate water (Chabchoub et al. [2], e.g.) to deep water (Fache et al. [5], e.g.) but also in other dispersive media such as in optic fibers (Copie et al. [4], e.g.). The soliton is a solution of non-linear equations such as the Korteweg-de Vries (KdV) equation (Korteweg and deVries [12]) or the nonlinear Schrödinger equation. Each soliton interaction causes a phase shift, but each soliton returns to its initial shape after interaction. This is rigorously true with KdV, when the two solitons propagate in the same direction (1D), at the surface of a fluid in shallow water (where the characteristic width of the soliton - its mass divided by its amplitude - is large in front of the water depth) for example. This makes KdV an integrable equation that conserves mass, momentum, energy and an infinity of other higher-order moments, so that the description of the system can be reduced to a few quantities (soliton amplitudes and phases) whereas it would require an infinite number of parameters in Fourier space ([20, 23], e.g.).
From an experimental point of view, in shallow water, it has been shown that the Rayleigh [22] solution is well suited to produce solitons with little dispersive trailing waves with a piston-type wavemaker (Guizien and Barthélémy [7]). The free surface displacement of a soliton is
where is the amplitude of the soliton, the phase speed, the shape factor, the space coordinate and time. Here the soliton is a localized symmetric hump traveling in one direction at constant speed. The parameters of the KdV solution fulfill:
and
while the parameters of the Rayleigh solution fulfill:
and
where is the quiescent water depth, the reduced amplitude and the gravity acceleration. The parameter is a measure of the exponential decay of the trailing edges of the soliton. The Rayleigh soliton becomes wider than the KdV soliton with increasing amplitude. Rayleigh profiles were shown to be close to experimental profiles measured in a wave flume by Redor et al. [24], and corresponding celerities match well. Leduque [13] also noted that the experiments of Li et al. (2011) better fit Rayleigh than KdV soliton profiles.
When solitons propagate in different directions (2D), interactions become much richer [19, 10, 30, 3 e.g.]. The Kadomtsev-Petviashvili (KP) system of equations is designed to describe such situation for small angular spreading [8, 11]. The case of the interaction of an oblique soliton with a vertical wall can be seen as the interaction of two oblique identical solitons. Miles [18] defined this interaction as ‘Mach interaction’, where a parameter depending on incidence angle and amplitude determines if the interaction is regular or irregular. In the first case, the hump or ‘stem’ at the wall is steady. In the irregular interaction case however, the stem width increases continuously. At the transition between the two cases a resonance is achieved where the stem amplitude can reach four times the amplitude of the incident soliton, a result confirmed numerically by Knowles and Yeh [9].
The aim of this article is to discuss ways of reproducing relatively simple interactions between two or three solitary waves in two horizontal dimensions (2D). Experiments in a wave tank have been carried out, pointing out the inherent limitations linked to the constrained space domain and the dissipation (§ 2.1). Numerical simulations (§ 2.2) reproduce the experiments very well (§ 2.3). The Mach interaction of a soliton against a wall is analyzed in more detail in § 3.
Methodology
Experiments
Experiments were carried out in the large wave tank at the Laboratoire Hydraulique de France (LHF, Artelia company, located in Pont-de-Claix, France). The m basin is equipped with a wave-maker consisting of 60 movable vertical paddles 50 cm wide which can be controlled independently. Their maximum stroke is 60 cm. The other three sides of the basin are rigid, wave-reflecting vertical walls (see Figure 1). The bottom of the basin is horizontal on average, but irregularities of up to mm can occur in places. The water depth at rest is m for all experiments.
Figure 1. Schematic top view of the wave tank: wavemakers on the left (they can move cm with respect to their mean position ), areas covered by the stereoscopic systems (pair of PCO cameras in red and pair of JAI cameras in green), positions of wave gauges (blue circles).
The wavemaker can generate solitary waves very close to the theoretically predicted Rayleigh solitons. This was verified using a stereo-video technique by Leduque et al. [14]. Capacitance-type wave gauges were also used to measure free surface displacements at various points in the tank.
The area covered by the cameras, about 180 m, is shown in Figure 1. By carrying out two symmetrical experiments with identical forcing, by changing prescribed motion of paddle () to paddle , and combining their results, we claim to cover around 320 m of the basin. Two sets of two cameras were deployed, rated PCO (5.5 Mpixels, 16 bits) and JAI (12 Mpixels, 8 bits) in Figure 1. They record up to 20 images per second. Stereoscopic reconstruction provides free-surface elevations with a resolution of approximately 5 cm on the horizontal axes, and , and a few millimeters on the vertical axis ([14]). The cameras cannot measure within 0.5 m of the wall . Wave gauges were fixed at 0.01 m from this wall (at , 11, 13.5, 16, 19.4 m) to capture wave amplification against it. A detailed description of the experimental set-up is presented in Leduque [13].
The free surface is covered with floating particles that form patterns required to enable stereoscopic reconstruction ([14, 13]). In Figure 2 are shown video frames of the passage of a wave at the wall recorded with another video-camera. This was meant to validate the calibration of the wave gauges. The maximum wave amplitude (red dots in Figure 2a and c) is estimated within a few millimeters for this 30 cm wave height. This also leads to further observations. The wave profile on the wall is rather close to Miles solution for strong symmetric interactions ([17 eq. 5.11]), plotted in dashed. This solution will be written explicitly in § 3 along with details on Mach interaction theory.

Figure 2. Video frames (from a camera located at , ) of the passage of a wave against the wall , resulting from the generation of an oblique soliton (, ). The wave propagates from right to left in this representation. Time interval between two successive images: . Measurements of the maximum free surface elevation with wave gauges fixed to the wall (red dots). The dashed line is Miles strong interaction [17 eq. 5.11].
Of note, the theoretical profile has been shifted of -4 cm for matching the measured stem amplitude, which is nearly constant over this 2.5 m travel. Miles’ strong interaction solution matches very well the rear face of the wave (Figure 2c). Instabilities in the front of the wave can be seen in Figure 2a. These can be attributed to the presence of the wave gauge. Miles’ solution does not match very well the front face of the wave (Figure 2a), nor the wave profile 0.43 s later (Figure 2b). This illustrates that this stem is not fully steady with this type of soliton generation. This aspect will be discussed further in § 3.
Numerical simulations
Along with these experiments, numerical simulations of soliton interactions were carried out using the Tolosa-lct code part of the Tolosa suite (tolosa-project.com). The Tolosa code suite was originally developed to predict coastal submersion over complex topographies with low computational cost using unstructured meshes. Its new version, Tolosa-lct, extends this framework to fully nonlinear and dispersive wave propagation.
The underlying model is not the Serre–Green–Naghdi system itself, but the hyperbolized Green–Naghdi model derived by Richard [26] from the Euler equations integrated over the vertical under a weak-compressibility assumption. Three variables are added to the classical Saint-Venant system: the depth-averaged vertical velocity , the non-hydrostatic pressure , and a strain-like vector which, following the approach of Bonneton et al. [1], improves the dispersive properties through a coefficient . Over the flat and fixed bottom of the basin, and in the absence of breaking, the system reads
where is the total water depth, the depth-averaged horizontal velocity, the bottom friction force (§ 2.3) and a finite artificial sound speed. The dispersive effects are carried by the acoustic sub-system, whose waves travel at the speed , being the quiescent water depth. A splitting strategy decouples these “fast” acoustic waves from the “slow” gravity waves, the acoustic sub-system being sub-iterated within each Saint-Venant time step. The artificial Mach number is set to 0.2 here, i.e. . In the limit , the linear dispersion relation over a flat bottom reads
For it reduces to the Serre–Green–Naghdi relation, which matches the Airy relation at first order in only, whereas the value used here gives the Padé (2,2) approximant of the Airy relation, accurate up to .
The interest of this formulation is that the Serre–Green–Naghdi equations require, at each time step, the inversion of an elliptic operator over the whole domain, which is costly and scales poorly in parallel, whereas the present system is purely hyperbolic, hence explicit and local. It is discretized by a finite-volume scheme on unstructured meshes with message-passing interface (MPI) parallelism, the extra cost being limited to a factor of 4 to 6 with respect to a standard Saint-Venant solver. The model is fully nonlinear, with no restriction on the relative amplitude, but it remains depth-averaged: it assumes a single-valued free surface and an irrotational flow of an inviscid fluid, and therefore describes neither overturning waves nor the wave boundary layer. Wet–dry fronts are handled by the code, and the unresolved dissipative processes are modeled by the bottom friction law of § 2.3.
For the present study, the code has further been equipped with an Arbitrary Lagrangian-Eulerian (ALE) formulation that reproduces the prescribed motion of the wavemaker, allowing a direct comparison with the experiments in terms of soliton generation, soliton–soliton interactions, and reflection against the basin walls. Following a mesh convergence study, we set the resolution to 2 cm, which corresponds to a mesh of 1350 1500 cells. The ALE formulation requires the use of a deformable mesh, which we chose to apply to the first 10 meters. We verified that this did not affect the accuracy of the simulation.
Validation: interaction of two solitons
These measurement and calculation tools need to be compared. An example is shown in Figure 3 for the generation of two solitary waves. A first wave (A) is generated at with all wavemaker paddles in phase (incidence ) and reflects against the wall at then against the wavemaker at , while a second wave (B) is generated with a set incidence of at . With angles, the aperture of the wavemaker restricts the length of the line soliton. This induces diffraction, the soliton bends and its amplitude decays along the line crest. This effect is more pronounced at large angles. Wave B interacts with the first soliton A (Figure 3 at ) and reflects against the wall at . A second interaction is shown at : the main wave resulting from A is not uniform along , showing a maximum near the wall while the wave resulting from B is strongly curved. The main wave of incidence reflects at the wavemaker, then interacts with two solitary waves of incidence (B′) and at (B″). These were created by the interaction of the B wave with the walls at and . At , the main resultants of these two waves B′ and B″ can be seen interacting again with the wave from A.

Figure 3. Stereoscopic measurements (top row) and simulation Tolosa-lct (bottom row) of the free-surface elevation at different stages following the generation of two solitary waves. A: amplitude relative to water depth , incidence generated at and B: , at ( is the start of the backward motion of the paddles, indicated times are the start of the forward push). B’ and B” are the main resultants of B interactions. Each frame covers 27 m in and 30 m in . The arrows indicate the direction of propagation of the main wave resulting from the soliton A.
Agreement between experiment and simulation is remarkably good after several interactions of these solitary waves. The simulation over the whole basin (bottom row Figure 3) clearly illustrates the richness of the waves’ interactions with each other and with the walls.
A soliton interacting frontally with a wall dissipates its energy by producing dispersive waves ([25 e.g.]). This type of interaction is similar to the 1-D head-on collision of two solitons which is not fully elastic, also known as weak interaction in contrast to the strong interaction of two solitons propagating in the same direction (see [24 e.g.]). These dispersive waves are the vertical stripes seen on the left of Figure 3 at . In contrast, the interaction between the waves within the basin at later times look more elastic. At , wave crests are clearly visible, which could be related to solitons, with a tendency to regularize along , , or in the diagonal of the basin.
The solitons, once generated at the wavemaker, interact with the walls and dissipate through bottom friction. One of the difficulties in simulating experiments is to model this bottom friction correctly. Like the Serre–Green–Naghdi equations it approximates, the model solved by ToSoLa-xt assumes an irrotational flow of an inviscid fluid. The dissipation by bottom friction is therefore parametrized. The resulting force on the flow consists of two terms that simulate the dissipation in both laminar and turbulent boundary layers, as:
The model was evaluated against the measured decay of a frontal soliton traveling back and forth in the basin by Leduque [13]. In the present computation, we take and as a good compromise to model the dissipation of both small- and large-amplitude waves (see details in Leduque [13], sections 2.6.2 and 3.1.1).
The symmetric case was performed, so that following the first frontal soliton, an oblique soliton of incidence was generated 19.8 s later. This procedure was used to compare the water elevation measurements of the stereo video system with a capacitance gauge located in the part of the basin out of the field of view of the cameras.
In Figure 4a are compared the time series measured with the wave gauge and the symmetrical point obtained from the stereo video, for the two symmetrical experiments. This comparison verifies that the stereoscopic measurements are very close to those of the gauge, and that our experiments are reproducible by symmetry. The differences are plotted in Figure 4b, confirming that the agreement between video and gauge is very good. Despite refraction effects induced by bottom irregularities and other generation defaults, the agreement remains within a few millimeters. Discrepancies result from slight wave phase differences.

Figure 4. a Free surface elevations following the generation of two solitons for two symmetrical experiments (A: , , ; B: , , ), measured by a capacitance gauge located at (, ), by video at the symmetrical point (, ) and numerical simulation Tolosa-lct. b Differences between gauge and video measurements (in red) and between video measurement and simulation (in grey).
The signal extracted from the numerical simulation Tolosa-lct is also plotted. Tolosa-lct reproduces the global dynamics of the interactions well, the differences with the gauge at the same point are mainly induced by a slight phase delay at each soliton passage. This small discrepancy might result from a combination of several sources of error, such as bottom irregularities and limited accuracy of the mean level estimation, or the flawed dissipation model in the numerical simulation. Another source of discrepancy is that there is no watertight seal between the paddles of the wavemaker: there is a weak flow in between especially during the receding phase of the motion of the paddles. In contrast, the wavemaker in the simulation is a deformable rigid wall.
To characterize the dynamics of highly non-linear wave interactions, with each other and with vertical walls, and in particular to reproduce the dispersive waves generated, the Tolosa-lct code seems nevertheless well suited.
Mach interaction
Theory
The maximum amplitude reached in the resonant interaction of a soliton with a wall has long been a subject of controversy [21, 18, 11, 9] (Li et al. 2011). Miles [18] was the first to theorize this interaction, which predicts a maximum amplitude at the wall of 4 times the amplitude of the incident soliton, provided that reduced amplitude () and incident angle () of the soliton are both small and satisfy
The parameter
is introduced. The steady solution of Miles [18], called ‘Mach interaction’, predicts the resonance at . A kinematic equilibrium is reached, where incident soliton, reflected soliton and stem all propagate at the same speed along the direction propagation of the interaction point. In the irregular interaction case (), also known as ‘Mach reflexion’, the third soliton against the wall (or ‘stem’) infinitely grows in width, along the steady solution propagating with a given angle to the wall. In a regular interaction (for ), the stem, which is not a soliton, has a constant width throughout propagation. The amplification () of the stem amplitude () relative to that of the incident soliton () is:
Kodama and Yeh [11] have revisited the analysis of Miles [18] and found that this maximum amplitude must be interpreted within the framework of a second-order KP theory. KP is a weakly nonlinear, weakly dispersive equation with little directionality ( small), established by Kadomtsev and Petviashvili [8]. This second-order analysis, allowing larger angles, gives a creditable account of experiments by Li et al. (2011) and numerical simulations by Funakoshi [6] and Tanaka [29], extended to simulations by Knowles and Yeh [9]. In this analysis, the amplitude amplification of the soliton incident on the wall follows Miles’ law as long as a virtual amplification predicted by KP at second order is applied. In this approach proposed by Kodama and Yeh [11] (see Knowles and Yeh [9], (1.3)):
and
For and , and . The resonance theoretically occurs when where can reach 4. The physical amplification of the wave may remain lower than for irregular interactions (). Note that the correction proposed by Kodama and Yeh [11] (13) presents a discontinuity for . The experimental observation doesn’t seem to show any drastic discontinuity between regular reflection and the production of a third wave, as will be shown in the next section.
In this framework, Miles solution of the free surface elevation for strong symmetric interactions [17 eq. 5.11] reads:
with [17 eq. 5.6]:
with our notations and taking from (12).
Experiments
Single oblique soliton
Experiments involving the interaction of a soliton with a vertical wall are generally carried out in a tank, in which a wall is installed at the desired angle to the wavemaker, thus defining a triangular domain. The generated soliton propagates in the direction perpendicular to the wave-maker and interacts with the wall. Our configuration is different, since the soliton is generated in a rectangular domain with an incidence that can be different from 0 by shifting the phase of each paddle appropriately [13, 14]. Due to the wavemaker segmentation (50 cm-wide paddles moving in the -direction) and diffraction (see § 2.3), there is some loss in the process so that, for large angles, the amplitude of the wave incident onto the wall is smaller than the amplitude of the target soliton. Here the reference amplitude is measured by a wave gauge at a location ( m, m) which is not too far from the wall and not too close to the wave-makers (also at the boundary of the area covered by the cameras).
Three sets of runs were conducted, by varying mainly the incidence angle of the soliton. In two cases (triangles and squares in Figure 5), the target amplitude was fixed ( and , respectively). Due to the segmented wave-maker, there is some loss in the generation process and the measured reduced amplitude of the soliton entering the video-surveyed zone decreases with increasing . The target amplitude is increased with increasing incidence in another series (circles), in order to better cover the parameter space (, ). Figure 5 depicts the different incidence amplitudes (Figure 5a), related Mach interaction parameters (Figure 5b), maximum wave amplitude at the wall (Figure 5c). As a result, with this set of parameters and generation procedures, the amplitude amplification at the wall depends mainly on the soliton incidence angle (Figure 5d).
Figure 5. Measurements in Mach interaction for various incidence angles and three different series of runs (triangles, squares and circles): a relative amplitude of incident soliton measured by the gauge located at ( m, m); b Miles parameter (10) (: small symbols) and Kodama & Yeh parameter (12) (: large symbols); c relative amplitude at the wall measured by four gauges located at m and , 11.00, 13.50, 16.01 m; d amplitude amplification at the wall . In c and d the symbols indicate the mean value, vertical bars indicate the minimum to maximum values of the four gauges.
Figure 6 plots the amplification at the wall for these gauges measurements and compared to [18] and [11] theories. The main difference in the two approaches is the computation of the parameter that characterizes the type of interaction: irregular (with amplification of the stem width throughout the wall interaction) or regular. The threshold is (10) for [18] and modified to (12) by [11]. The second difference made by Kodama and Yeh (2016) was to consider that the amplitude amplification for irregular interaction () has to be considered in a second order KP framework, so that the amplification factor has to be corrected (13).

Figure 6. Experimental results of the amplification factor versus , the Miles parameter (10) and b the KP amplification factor versus Kodama and Yeh [11]’s higher-order KP parameter (12). Generation of a single oblique soliton with various incidence angles and relative amplitudes measured at , same symbols as in Figure 5. The symbols indicate the mean value, vertical bars indicate the minimum to maximum values measured with the four gauges located close to the wall for each run. In solid line is Miles’ solution (11) in a, the same formulation as in (11) replacing by and by in b.
Our measurements (Figure 6) clearly show a decay of the amplitude amplification at the wall around . The correction of the amplification factor might even improve the trend (Figure 6b). The measurements nevertheless do not match the theory, lying below the predictions. This might be attributed mainly to the distance needed to establish a steady solution.
Despite these limitations, our experimental configuration can produce solitons in quasi-steady interaction, as shown in the next section.
Three-solitons solution (3S)
In an attempt to produce near resonant interaction established directly at the wavemaker, three solitary waves were generated in succession. A first soliton of incidence is immediately followed by the production of a frontal soliton () on the first four wavemaker paddles ( m). A third soliton is generated () to account for the wave reflected by the wall. This test is referred as 3S in this paper. The numerical simulation of this test is shown in Figure 7. The third generated wave limits the diffraction of the reflected wave produced by the wall interaction, but is not in equilibrium. The reflected wave tends to take an incidence , so that the interaction at the wall resembles a regular interaction.

Figure 7. Simulation with Tolosa-Ict of three-wave interaction 3S: free surface elevation from stem generation, then one frame every 2 s. Each frame covers 27 m in and 30 m in .
The stem produced in this interaction is remarkably stable, as shown by the time series extracted from the video measurements plotted in Figure 8. The profiles of the incident wave, the stem and the reflected wave are very close to Rayleigh soliton profiles. This means that these three waves all travel together as individual solitons along their own direction of propagation. At 6 m from the wall, the incident wave tends to slightly decay in amplitude, while the reflected wave amplitude slightly increases (Figure 8c). This observation is therefore qualified as quasi-steady.
Figure 8. Stereoscopic measurements of the free surface elevation in the three-wave wall interaction 3S (thin lines). a Time series every 0.5 m along m at m, b every 0.5 m along m at m and c along m at m. The thick dashed lines are Rayleigh soliton profiles.
The transverse profile (along ) of the stem varies very little throughout the area covered by the cameras, see Figure 9. As expected, the stem is much longer than a characteristic soliton width, somehow matching the strong interaction profile deduced from Miles [17] for which is replaced by (14-15).
Figure 9. Profiles of the stem along , experiment 3S. Stereoscopic measurements: JAI profile at m (thick solid line), PCO profiles every 0.5 m in (thin solid lines). The symbol () corresponds to wave gauge measurements at ( m, m). Tolosa simulation (thick dotted). The dashed grey line is a Rayleigh soliton profile that would propagate along (for reference). The thick dashed line is Miles solution (14-15) for strong interaction for ( and ).
The parameters and used in the model are different to that of the prescribed oblique soliton and . They are deduced from the stereoscopic reconstruction and appear to reflect the proper balance between the actual diffracting oblique solitons (incident and reflected) and the stem at the wall (see Figure 10). With these values, and the amplification factor .

Figure 10. Snapshot of the free surface displacement (m) in the three-wave interaction 3S measured by stereoscopy. Prescribed angle of the first generated oblique soliton (white dashed, ) and best fit angles (black dashed, and ).
This example shows that it is possible to produce a 2D soliton (a hump localized in both and directions) with very stable length and width. By adjusting the parameters of the experiment, it has been possible to compensate for diffraction effects in a zone covering the soliton’s propagation over ten times its characteristic width. Such an observation gives credit to the second-order KP theory. The Tolosa-Ict simulation reproduces it rather well, see Figure 11. The simulation also provides details of the dispersive waves produced during soliton generation.

Figure 11. Snapshot of the free surface in the three-wave interaction 3S measured by stereoscopy (a) and corresponding simulation (b).
Simulations and asymptotic extrapolations in time
Only numerical simulation is capable of producing solitons in a domain large enough to avoid diffraction effects. Moreover, wave dissipation can be controlled through the bottom friction parametrization. Simulations with Tolosa-Ict in a m numerical tank, canceling bottom friction, were carried out to produce Mach interactions.
The results of these simulations show that the stem requires a long establishment distance when , at the transition between irregular and regular interactions, in agreement with Li et al. (2011) and Knowles and Yeh [9]. In the irregular case, the stem rapidly acquires an equilibrium velocity that is that of the soliton obtained asymptotically in the steady state (Leduque [13]). This makes it possible to extrapolate the results of the simulations and thus verify the results obtained by Knowles and Yeh [9] for amplification at the wall, see Figure 12. The agreement is rather good with the second-order KP theory proposed by Kodama and Yeh [11], albeit indecisive close to resonance , with notable discrepancies for small variations of the incidence angle. In this context, the stability conditions for the 3S solution shown in Figure 7–Figure 11 remain to be fully specified. The measured amplification at the wall (red triangle in Figure 12b) and the simulation with bottom friction (blue star) are reasonably close to the inviscid second order KP theory (solid line). Of note, the simulations without bottom friction at (crosses) also lie slightly below the curve. As a consequence, the bottom friction is likely not the dominant parameter that prevent to catch up the fourfold amplification at . Our experimental and computing constraints limited the exploration of . We nevertheless confirm that near-resonant solutions are very sensitive to the main parameters (, ) through the parameter (12).
Figure 12. Stem amplification at the wall ((12)–(13)) and comparison with Kodama and Yeh [11]’s theory (solid line). a Extrapolated numerical results with Tolosa-Ict in a 80 × 80 m tank for various incidence angles and amplitudes and without bottom friction. b Close-up around maximum resonance and Knowles and Yeh [9]’s numerical results (blank symbols), for different angles: (□), (◇), (○). The quasi-steady 3-wave solution obtained with Tolosa-Ict with bottom friction (blue star) and measured 3S (red triangle) are also plotted. For the latter, the error bars depict the variability with ( cm from Figure 8c).
Conclusions
Experiments have been carried out in the large LHF (ARTELIA) wave tank to study the interactions of strongly non-linear waves. Our experiments are the first to detail soliton diffraction, studied theoretically recently by Ryskamp et al. [28]. Stereoscopic reconstruction of free-surface displacements is achieved with good accuracy (of a few mm) over a large area of the tank (, or using symmetrical experiments).
Diffraction plays a major role in our experiments. This is the cause of the main uncertainty in computing the parameter, where would be the maximum resonance with a fourfold amplification of the soliton amplitude in a Mach interaction at the tank wall.
Validation of the Tolosa-lct code on a few cases enables us to extend the study to configurations that cannot be realized experimentally. This joint approach opens up the discussion of the relative stability of interacting solitons. In particular, in the vicinity of , we have shown on an example that a relatively stable 2D three-solitons solution can be generated, somehow confirming the predictions of Kodama and Yeh [11], among others.
The properties of the solutions experimentally confirmed are also helpful to understand the so-called “solition-gas” state in 2D, in which a large number of solitons constantly interact generating stems that outlive the incident solitons. Predicted theoretically by Zakharov [32], such a state has only recently been obtained experimentally in 2D shallow water by Leduque et al. [14, 15] and presented in detail by Leduque [13].
Acknowledgements
This study was funded by the French National Research Agency (project SOGOOD, ANR-21-CE30-0061) and the Simons Foundation (Simons Collaboration on Wave Turbulence). The Tolosa collection of software suite is developed by the Institut de Mathématiques de Toulouse as part of the national plan for wave-submersion vigilance, with funding from SHOM. The technical support of ARTELIA and LEGI was essential to the success of this project, in particular thanks to H. Michot and M. Lagauzère on wave generation and F. Bonnel (Grenoble-INP) for capturing images. HM is deeply indebted to Lili Kimmoun who made some very significant contributions to a French project (PATOM) on wave breaking in the late 90’s and 2000’s.
References
- [1]Bonneton, P., Chazel, F., Lannes, D., Marche, F., and Tissier, M. (2011). A splitting approach for the fully nonlinear and weakly dispersive Green–Naghdi model. J. Comput. Phys., 230(4):1479–1498.DOI
- [2]Chabchoub, A., Kimmoun, O., Branger, H., Hoffmann, N., Proment, D., Onorato, M., and Akhmediev, N. (2013). Experimental observation of dark solitons on the surface of water. Phys. Rev. Lett., 110(12):124101.pubmed.ncbi.nlm.nih.gov/25166807
- [3]Choi, J., Bokhove, O., Kalogirou, A., and Kelmanson, M. (2022). Numerical experiments on extreme waves through oblique–soliton interactions. Water Waves, 4(2):139–179.DOI
- [4]Copie, F., Suret, P., and Randoux, S. (2025). Controlled manipulation of solitons in a recirculating fiber loop using external potentials. Phys. Rev. Res., 7(4):043232.DOI
- [5]Fache, L., Bonnefoy, F., Ducrozet, G., Copie, F., Novkoski, F., Ricard, G., Roberti, G., Falcon, E., Suret, P., El, G., et al. (2024). Interaction of soliton gases in deep-water surface gravity waves. Phys. Rev. E, 109(3):034207.DOI
- [6]Funakoshi, M. (1980). Reflection of obliquely incident solitary waves. J. Phys. Soc. J., 49(6):2371–2379.DOI
- [7]Guizien, K. and Barthélémy, E. (2002). Accuracy of solitary wave generation by a piston wave maker. J. Hydraul. Res., 40(3):321–331.DOI
- [8]Kadomtsev, B. B. and Petviashvili, V. I. (1970). On the stability of solitary waves in weakly dispersing media. In Doklady Akademii Nauk, volume 192, pages 753–756. Russian Academy of Sciences.
- [9]Knowles, J. and Yeh, H. (2019). Fourfold amplification of solitary-wave Mach reflection at a vertical wall. J. Fluid Mech., 861:517–523.DOI
- [10]Kodama, Y. (2010). KP solitons in shallow water. J. Phys. A: Math. Theor., 43(43):434004.
- [11]Kodama, Y. and Yeh, H. (2016). The KP theory and Mach reflection. J. Fluid Mech., 800:766–786.DOI
- [12]Korteweg, D. J. and deVries, G. (1895). On the change of form of long waves advancing in a rectangular channel, and on a new type of long stationary waves. Phil. Mag., 39(5):422–443.DOI
- [13]Leduque, T. (2024). Étude expérimentale et numérique d’ondes non-linéaires multi-directionnelles en eau peu profonde: application aux gaz de solitons. PhD thesis, Univ. Grenoble Alpes.
- [14]Leduque, T., Barthélémy, E., Michallet, H., Sommeria, J., and Mordant, N. (2024). Space-time statistics of 2D soliton gas in shallow water studied by stereoscopic surface mapping. Exp. Fluids, 65(84).DOI
- [15]Leduque, T., Kaczmarek, M., Michallet, H., Barthélémy, E., and Mordant, N. (2025). From deep to shallow water 2D wave turbulence: Emergence of soliton gas. Phys. Rev. Fluids, 10(11):114801.arxiv.org/abs/2510.09637
- [17]Miles, J. W. (1977a). Obliquely interacting solitary waves. J. Fluid Mech., 79(1):157–169.DOI
- [18]Miles, J. W. (1977b). Resonantly interacting solitary waves. J. Fluid Mech., 79(1):171–179.DOI
- [19]Nicolas, P. (1993). Utilisation de variables distordues dans la théorie des ondes de gravité: application à la réfraction et à la génération. PhD thesis, Univ. Joseph Fourier, Grenoble.
- [20]Osborne, A. R. (2010). 30 - nonlinearity in duck pier data. In Osborne, A. R., editor, Nonlinear Ocean Waves and the Inverse Scattering Transform, volume 97 of International Geophysics, pages 779 – 794. Academic Press.
- [21]Perroud, P. H. (1957). The solitary wave reflection along a straight vertical wall at oblique incidence. PhD thesis, Univ. California.
- [22]Rayleigh, J. W. S. (1876). On waves. Philos. Mag., 5(1):257–79.
- [23]Redor, I. (2019). Étude expérimentale de la turbulence intégrable en eau peu profonde. PhD thesis, Univ. Grenoble Alpes.DOI
- [24]Redor, I., Barthélémy, E., Mordant, N., and Michallet, H. (2020). Analysis of soliton gas with large-scale video-based wave measurements. Exp. Fluids, 61(10).DOI
- [25]Renouard, D.P., Seabra-Santos, F.J., and Temperville, A.M. (1985). Experimental study of the generation, damping, and reflection of a solitary wave. Dyn. Atmos. Oceans, 9:341–356.DOI
- [26]Richard, G. L. (2021). An extension of the Boussinesq-type models to weakly compressible flows. Eur. J. Mech.-B/Fluids, 89:217–240.DOI
- [27]Russell, J. S. (1845). Report on Waves: Made to the Meetings of the British Association in 1842-43.
- [28]Ryskamp, S., Maiden, M. D., Biondini, G., and Hoefer, M. A. (2021). Evolution of truncated and bent gravity wave solitons: the Mach expansion problem. J. Fluid Mech., 909:A24.DOI
- [29]Tanaka, M. (1993). Mach reflection of a large-amplitude solitary wave. J. Fluid Mech., 248:637–661.DOI
- [30]Wang, C. and Pawlowicz, R. (2012). Oblique wave-wave interactions of nonlinear near-surface internal waves in the strait of georgia. Journal of Geophysical Research: Oceans, 117(C6).DOI
- [31]Zabusky, N. J. and Kruskal, M. D. (1965). Interaction of “solitons” in a collisionless plasma and the recurrence of initial states. Phys. Rev. Lett., 15(6):240.DOI
- [32]Zakharov, V. (1971). Kinetic equation for solitons. Sov. Phys. JETP, 33(3):538–540.




