## Abstract

Repeated oblique impacts and rebounds of a solid body or bodies on horizontal shallow water are investigated through mathematical modelling. The inclinations from the horizontal are supposed small as the skimming evolves, for a thin typical body shape. The new formulation aimed at improved prediction as well as the background involved is presented together with nonlinear analysis and computation. Comparatively fast or slow collisions and rebounds are found to be of special interest over short time-scales.

## 1. Introduction

Collisions, bouncing and skimming (skipping) in fluid–body or fluid–fluid impacts arise in numerous applications, whether of a serious or playful nature. Special mention should be made of phenomena in aircraft icing, storms, engine intakes, ship-slamming and meteor impacts. There is a wide background of recent literature [1–15], major aspects of which we summarize later in the introduction.

There are also connections to two long-standing distinguished themes of research, namely that of boundary layers and viscous–inviscid interactions stemming from [16–22] and that of impacts, splashing, rebounds and interfacial interactions stemming especially from [6] (see also [7,8,23–25]). The common features between the themes include their original industrially based motivations and their mechanisms of ‘stability, separation and close body interactions’ as well as their modern uses (for example in separation, transition, turbulence, stall on the one hand and in icing, storms and the like on the other hand) and their mechanisms of inner–outer dependence, system reduction, thin-layer dynamics and subtle interactions. Both benefit from analysis, computation, experimental links and improved physical understanding.

Three areas influence the present investigation. Concerning collisions, first, air–water interactions are addressed in [3–5]. In [5], in particular, near-impact behaviour is investigated for a solid body approaching another solid body with two immiscible incompressible viscous fluids occupying the gap in between. The fluids have viscosity and density ratios which are extreme, the most notable combination being water and air, such that either or both the bodies are covered by a thin film of water. Air–water interaction and the commonly observed phenomenon of air trapping are of concern. The subcritical regime is of most practical significance here in terms of the Reynolds number compared with a critical value that depends on the two fluid ratios, and it leads physically to the effect of inviscid water dynamics coupling with a viscous-dominated air response locally. This physical mechanism [3] induces touchdown (or an approach to touchdown), which is found to occur in the sense that the scaled air-gap thickness shrinks towards zero within a finite-scaled time according to analysis performed hand in hand with computation. A global influence on the local touchdown properties is also identified. Comparisons with computations prove favourable. Air trapping is produced between two touchdown positions, at each of which there is a pressure peak; an oblique approach would not affect the finding, unless the approach itself is extremely shallow. The mechanism of air–water interaction leading to air trapping is suggested as a wide ranging result.

Second, free-surface impacts without air effects are studied in [11,12]. Here, Ellis *et al.* [11] consider the effect of surface roughness. If a surface is sufficiently rough, then models of droplet impact need to include the possibility of many touchdowns between the fluid and the solid. The subsequent simultaneous motion of several contact points is therefore described. In particular, results are presented for the contact point motion during the impact of a sheet of water on to a periodic rough surface. A model for multiple contact point motion is also analysed for cases of deep surface roughnesses. The analysis is complemented by a direct numerical simulation using the volume of fluid method. The effect of surface roughness is to reduce the rate at which the droplet spreads. Comparisons between various roughness shapes are considered numerically and analytically. Possible applications are then discussed, especially in the context of aircraft icing. Follow-up study to allow for ice accretion is in [12].

The third area of direct relevance is fluid–body interactions [13–15]. Interactions between a finite number of bodies and the surrounding fluid are investigated in [13]. The bodies or modelled grains are thin solid bodies free to move in a nearly parallel formation within a quasi-inviscid fluid. The investigation involves numerical and analytical studies and comparisons. The three main features are a linear instability about a state of uniform motion, a clashing of the bodies (or of a body with a side wall) within a finite-scaled time when nonlinear interaction takes effect and a continuum-limit description of the body–fluid interaction holding for many bodies. Solid–solid and solid–fluid clashing, skimming and bouncing are the concern in [14]. A theoretical study is presented on fluid–body interaction in which the motion of the body and the fluid again influence each other nonlinearly. The clashing refers to solid–solid impacts arising from fluid–body interaction in a channel, whereas the skimming refers to another area where a thin body impacts obliquely upon a fluid surface. Bouncing usually then follows in both areas. The main new contribution concerns the influences of thickness and camber which lead to a distinct general form of clashing and hence bouncing. In [15], conditions are investigated under which a body lying at rest or rocking on a solid horizontal surface can be removed by hydrodynamic forces or instead continues rocking. The investigation is motivated by recent observations on Martian dust movement as well as other small- and large-scale applications. The nonlinear theory of fluid–body interaction here has unsteady motion of an inviscid fluid interacting with a moving thin body. Various shapes of body are addressed together with a range of initial conditions. The relevant parameter space is found to be subtle as evolution and shape play substantial roles coupled with scaled mass and gravity effects. Lift-off of the body from the surface generally cannot occur without fluid flow but it can occur either immediately or within a finite time once the fluid flow starts up: parameters for this are found and comparisons are made with Martian observations.

The new work here is directed at the challenge of understanding and predicting repeated oblique impacts and rebounds of a solid body or bodies on shallow liquid, typically water. Such skimming, colliding and bouncing induces splash jets at the leading edge (the unknown moving front of the wetted surface on the body) and can also induce jets at the trailing edge or not, depending on the body shape, on the initial conditions and on the evolution of the complete incompressible fluid–body interaction among other factors. The effects of the major interaction parameters and of different body shapes of concern are to be described by means of nonlinear analysis and computation with a view to connecting with experiments. With many temporal and spatial scales being active, the excitement in the modelling here is in seeking justified simplification wherever possible subject to the theoretical model remaining physically realistic.

Section 2 describes the framework for the new work which begins with figure 1, whereas §3 presents analysis. Interactive properties for enhanced and reduced downward speeds at impact are addressed in §§4 and 5, respectively. Section 6 adds further comments including predictions for repeated impacts and rebounds.

## 2. Background model

Experimental results covering collisions, rebounds and skimming in qualitative form are quite plentiful as described in [1,26], whereas justifiable mathematical models seem few. The present contribution is associated closely with the Hicks–Smith model [1] of skimming in unsteady two-dimensional interactions. Such exciting interactions which offer an example of fluid–body interaction and make several perhaps bold assumptions as delineated later are governed by a scaled nonlinear evolutionary system for the unknown functions *Y*,*θ*,*x*_{1},*D*,*γ*_{0},*γ*_{1},*γ*_{2},*γ*_{3}. Here, in scaled terms, *Y* is the height of the centre of mass of the body, *θ* is essentially the body's angle of inclination from the horizontal, *x*_{1} is the moving contact point that forms in effect the leading edge for the wetted part of the body at any instant, *D* is a fluid velocity contribution and *γ*_{0},*γ*_{1},*γ*_{2},*γ*_{3} are pressure coefficients. Furthermore, *x* is the horizontal coordinate, *y* is the vertical coordinate and *t* denotes time, again in appropriately scaled terms.

The Hicks–Smith model takes a reference frame fixed in the body and a non-dimensional form such that the horizontal velocity, body half-length and typical convective time are unity; the shapes of body and water layer involved are thin and nearly horizontal. The nonlinear shallow-water equations, i.e. unsteady inviscid boundary-layer equations, apply to the fluid flow nominally, because the Reynolds number and Froude number are large in practice, as is the Weber number, and therefore to the leading order the viscous, gravitational and surface tension effects are negligible. If in addition the vertical scale of the water layer exceeds that of the body by a factor 1/Δ say, then the following modification holds
2.1
where Δ≪1, and the terms with tildes are of order unity. The quantities *h*,*U*,*P*,*T* denote, respectively, the height of the body under-surface measured from the bottom of the liquid layer at *y*=0, the induced flow velocity, the fluid pressure and the body half-thickness in scaled format. Capitals *U*,*V*,*P* signify fluid-flow responses as opposed to *u*,*v* used later for essentially the body-motion responses. The trailing edge of the body is taken to be sharp rather than smooth and the liquid free surface detaches from the body there (figure 1). The unknown leading-edge position *x*_{1} is of order unity generally. The body shape implies
2.2
where *x*_{m} is the horizontal position of the centre of mass; assuming the body has uniform density this position is taken to be zero in our coordinate system. The shallow-water equations give now
2.3a,b
At the known trailing edge *x*_{0}, the equi-pressure Kutta condition to take account of viscous effects owing to the sensitive laminar or turbulent boundary layers at the trailing edge implies , whereas, in the wake region *x*>*x*_{0}, we have being identically zero (atmospheric); the trailing edge position *x*_{0} is normalized to be one. At the unknown leading edge *x*_{1}, jump conditions as in [1,27] yield
2.4a,b

In addition, the vertical momentum balance for the body simplifies to
2.5
the horizontal balance confirms that the body moves with constant horizontal velocity unity to leading order, and angular momentum requires
2.6
Here, *M*,*I* are the scaled mass and moment of inertia of the body in turn. The liquid in the wetted region [*x*_{1},*x*_{0}] is of primary interest.

For a body with parabolic shape with constants *A*,*B*,*C*, and so from (2.2)
2.7
is also quadratic. In this region, (2.3) indicates that is at most linear in *x*. Therefore, we can write
2.8
where *γ*_{n} for *n*=0,1,2,3 are unknown functions of time. The system for zero *A*,*B*,*C*, i.e. a body with negligible thickness is then of the nonlinear form: trailing-edge pressure condition,
2.9
two conditions at the leading edge,
2.10
and
2.11
where *D* is an unknown function of time representing the horizontal velocity of the fluid at the centre of mass, *x*=0; three momentum-balance effects,
2.12
2.13
2.14
linear and angular momentum of the body,
2.15
and
2.16
hence giving eight equations for eight unknowns at each time level in effect. The main assumptions made in the theory include the straight flat geometry of the body, the two-dimensionality of the entire fluid–body interaction, the neglect of air effects, the incompressibility of the quasi-inviscid fluid (water), the shallowness of the water layer and the smallness of the flow angles induced during the motion.

Numerical solutions and certain analytical properties are presented in [1]. The current new work begins with the framework of (2.9)–(2.16) and then moves on to study wider applications for more general body shapes and all-round configurations.

## 3. Fast responses

Figure 2 presents our results for the moving contact position, vertical displacement and body-motion velocities from the above framework but in a parameter range rather distinct from previously, because the typical timescale is quite short. The results still show the rapid change of leading-edge velocity just prior to lift-off [1,2] of the body from the water at time *t* of approximately 0.45, by the way. With regard to the present investigation and novel formulation computations such as in figure 2 point to a simplification occurring for fast behaviour over a time-scale of order *E* say with the constant *E* being small and positive, for small mass factors *M* of order *E*^{3}. At first sight, the left-hand side of (2.15) might be taken to be negligible when *M* is small, but instead the timescale shrinks to achieve balance there and this then leaves the contribution of the left-hand side in (2.16) as outstanding, a point discussed further near the end of the paper. The solution expands as
3.1
with are initial values and it is found that . The prime denotes . The most significant behaviour in this fast-time response clearly occurs spatially near the incident trailing-edge location. Here, and I are of order unity. It is found also that the model can then be reduced to two coupled nonlinear differential equations for ,
3.2a
and
3.2b
where *ω*_{0} is a prescribed constant and . Similarly, for small moments of inertia *I* of order *E*^{3} with *M* of order unity a pair of nonlinear equations holds for , at leading order. If both *M*,*I* are of order *E*^{3}, then the same approach leaves the three interactive equations
3.3a
3.3b
3.3c
controlling the nonlinear evolution of with scaled time. The following investigation is on the regime of small mass as in (3.2a) and (3.2b) primarily. It is noted however that computed solutions for all three of the simpler forms just described agree well with the full solutions of (2.9)–(2.16) at small *M* or *I* values as exemplified by figure 2 and in more detail in [2].

The reduced system (3.2a) and (3.2b) in which we now set as for convenience can be normalized by putting
3.4
along with , where *α* is the initial angle . Here, *β* is the initial entry position *x*_{0}; the transformation accounts for the influence , identifies the central dynamics and also expects *α* to be negative because of the incident downward motion. Thus, the system becomes
3.5a–d
subjected to initial conditions which correspond to knowing the location and downward speed of the impact at first,
3.6a,b
Primes in the above represent . The result for stems from properties at small times . In effect, above all, the original system can take *α*,*M*,*β* as −1,1,0 in turn without loss of generality, and we are left with the task of solving (3.5) and (3.6) in terms of the single parameter which represents the scaled downwards approach speed.

Numerical results obtained from a straightforward time-marching scheme and checked as regards grid effects are given in figure 3 for a range of values of . The increasing downward plunge as is increased makes physical sense, and we observe the subsequent recovery leading to lift-off in every case calculated. Insight into these fast responses is provided by the analyses in the next two sections, and an ensuing study of the physical mechanisms involved.

## 4. Increased downward speed

With more downwards entry speed, the parameter is large and negative. Two or three distinct successive stages are now found to come into operation (figure 4*a*–*c*).

First, over very small time-scales, the relevant expressions for are
4.1a–c
with *x**,*Y* * of order unity. Substitution into (3.5) then yields the dominant mechanisms as
4.2a–d
subjected to (*x**,*Y* *,*u**,*v**)(0)=(0,0,−3/2,−1). The large effective velocities provoked here as the wetting and downward distance both increase quite rapidly are responsible for the reduction in form. Hence, the solutions during this first-wetting stage are as depicted in figure 4*a*. Of concern next is what happens to the reduced nonlinear system (4.2) as time *t** becomes large and positive. The downward motion tends to dominate as the wetted area ‘saturates’ then. The results coupled with a large-time study indicate that the asymptotic features
describe the behaviour then, and indeed substitution into (4.2) shows consistency of these features provided that the relation holds, which acts to determine the coefficient *β*_{2} in essence. The coefficients *α*_{1},*α*_{2},*β*_{0} remain arbitrary in the large-*t** features and are believed to depend on overall global properties of the system (4.2) applying for all *t** of order unity. These asymptotic features lead into the structure of the next stage.

The second stage then occurs a little later and has the expansions
4.3a–c
where to lowest order *α*_{1} represents the most forward position achieved by the wetting process. Hence, the main governing equations become simply
4.4a–d
with the relatively deep plunge here associated with and being the major effects accompanying the deceleration in wetting. These lead to the solution constant, where matching requires *c*_{3}=−*α*_{2} and fixes as *β*_{0}. The terms in and 1 in are the novel leading-order contributions in this stage, pointing to a minimum of which determines the maximum of the wetted interval and to the depth reached continuing to increase linearly with time ( in this stage; figure 4*b*).

In the third and final stage, the time-scale rises to be of order unity, and the trend reversion to lift-off takes place in full. Now
4.5a–c
and more interactive nonlinear influences come back into play. Substituting into (3.5) shows that the dominant balance now has
4.6a–d
Again, the influence of the deepened plunge is evident through comparison with (3.5)–(3.6). The matching with the earlier stage provides , for , as the initial conditions in effect. The present stage then takes the body into an underlying upward or outward phase which leads at a finite positive value of time to becoming zero which in turn signifies the advent of lift-off. Figure 4*c* shows results for a suitable range of values of the constant *α*_{1} and confirms that the appropriate solution is relatively simple at leading order, namely for all time , thus determining the lift-off time as −1/*α*_{1} to the first approximation. The correction term for can be shown to be of order and proportional to , in line with the second stage earlier on.

The tendencies suggested by this increased-speed analysis tie in well with the numerical results described in the previous section. The presence of the short timescale in (4.1) explains the relatively rapid response seen in the previous numerical results, in line with physical expectations that more downward speed should provoke a quicker and perhaps deeper plunge overall. The trend reversal countering the downward plunge is then really seen to begin during the second stage over a slightly longer timescale, prior to the full nonlinear interaction largely reasserting its presence in linearized analytical form during the third stage and forcing the eventual lift-off.

## 5. Decreased downward speed

Lessened downward entry speed corresponds to the parameter being small and negative in which case the whole process occurs within the single-order-unity time-scale. So now in view of the initial values in (3.6) as well as the dynamics in (3.5), we have the expressions
5.1a–b
applying and as a result the governing equations and constraints (3.5)–(3.6) simplify to the form
5.2a–d
with the prime denoting differentiation with respect to , at leading order. The influences of horizontal acceleration effects owing to the movement of the contact point in the balance (3.5) are secondary in this regime, whereas the corresponding vertical acceleration effects in (3.5) remain primary ones. The values (*x*_{L},*Y* _{L},*u*_{L},*v*_{L})(0)=(0,0,−1,−1/2^{1/2}) are the initial conditions.

The solution of the small nonlinear system (5.2) is expressible in terms of which produces the third-order linear system
5.3
from (5.2). Hence, with powers *m*_{n} denoting for *n*=1,2,3 the solution is simply
5.4
where the complex constants *a*_{n} are determined by the initial conditions which require *Z*(0)=*Z*^{′}(0)=0,*Z*^{′′}(0)=1. Thus, the values
5.5
are obtained for the constants, whence can be found from (5.4), and the test then determines the complete wetting process up to the occurrence of lift-off where *x*_{L} returns to zero (figure 5). The trends in (5.1)–(5.5) again agree qualitatively with those observed in the earlier numerical study as is reduced in magnitude, and as a quantitative check the value of just over 5 for the scaled time at lift-off ties in with the trend observed in figure 3 as decreases.

## 6. Further comments

The recent works on collisions, rebounds and skimming described in the introduction cover a diversity of applications which are felt to be of much interest and challenge. These produce in particular spin-off suggestions on tackling anew multi-body problems as in [13–15] and are even linked through the dynamics of fluid–body interactions to predictions and observations on the movement of dust on the planet Mars [15]. They can also include viscous effects as in the theory of air cushioning in [3], which in turn leads on to air-pocket properties in [5] and most recently to Hicks *et al.* [28], who show close and encouraging connections between theory and experimental measurements. (The air–water interaction mentioned earlier comprising interplay of viscous and inviscid effects is potentially relevant throughout the current investigations.) Experimental studies concerning the skimming and skipping of a thin body on a shallow stream of water with fixed velocity are conducted in [26], the results demonstrating interesting phenomena that are qualitatively consistent with our model predictions, in particular the ‘super-elastic’ exit of a skimming body where its vertical velocity is greater on the way out than in. A further connection with experiments in the current broad area is shown by Elliott & Smith [12], who investigate theoretically the accretion of ice on a solid surface when a super-cooled droplet impacts upon the surface and compare with experimental measurements. The agreement in qualitative terms or in orders of magnitude is quite close and the work also tends to add weight to the view that theory here can provide a means to explore the parameter space over a substantial range of parameter values.

By the same token, the use of nonlinear evolution here connects with the present issue's theme of ‘stability, separation and close body interactions’. Comparison might suggest indeed that it would be beneficial to address viscous effects near a departure (separation) point of a fluid–fluid interface (e.g. between air and water) from a fixed solid surface, first of all in steady flow perhaps and later in unsteady flow. Building into the overall theory the important local influence of the viscous effects there would seem to represent another considerable step forward in understanding.

As well as broadly looking back and looking forward as above, our prime concern in the study here is to follow through with the research described in §2 and in [2]. We should add straight away that apart from the technological applications such as in aircraft safety and icing, food-particle sorting and sports applications there are also fun applications (ducks and drakes) and possible relevance to meteor impacts, to cleansing and to forensic examinations from different parts of the research. Again, the scope of the specific analysis in §§3–5 may be seen more clearly by means of the original partial differential system in §2, in order to indicate the main physical features involved. Thus, under the small-mass scalings described at the beginning of §3 but accompanied by 6.1 with during the wetting the effective shape is altered to 6.2 and the angular momentum balance becomes merely 6.3 to leading order, which points to being identically. Apart from that the original system (2.2)–(2.6) remains intact in scaled quantities provided the integral in (2.5) is taken over the wetted area of course. It then follows however that the pressure response in general is quadratic in the local coordinate instead of cubic as in (2.8) and this can be shown to produce the simplified interaction of §3 exactly. Further alterations within the above framework lead to the cases of comparatively large and small downward speeds considered in §§4 and 5. Exploiting this structure for fast collisions and rebounds takes us into future work. The study of fast responses in §§3–5 is concentrated effectively near the trailing edge of the thin impacting body. More generally or more usefully, that can be regarded as having the centre of mass of the body positioned relatively far away from the contact area of wetted surface. This view might well be relevant in turn to the classical problem of describing analytically the skimming of a bluff smooth body over a liquid surface as opposed to a thin body skimming as in the present configuration.

Future studies should address the following issues and challenges. First, there is much interesting work to be carried out to relate the in-water phase (impact and rebound) quantitatively to the in-air phase (usually lasting longer) as far as skimming of a body over water is concerned. Some initial effort on the in-air part is in [2] with gravity included and incorporating the in-water phase as delineated in §§2–5, over several repeated impacts and rebounds. Clearly, the initial and end conditions of the latter largely nonlinear phase associated with touchdown and lift-off, respectively, interact with those of the largely linear in-air motion. A sample result is shown in figure 6: see also [2]. Second, there is a considerable challenge inherent in rationally modelling many bodies rebounding as in a storm in reality, especially given that the real-world situation is in three dimensions. It may well be the case that a reasonable aim for an in-water description in this context of complexity is one which is more readily calculable; if so, the approaches in §§4 and 5 indicate a promising possibility there. The question of how much structure or physics must be captured for realistic predictions in any reduced system remains in the background throughout. Third, there is a need to check in detail on trends similar to those investigated in this paper but for relatively small moments of inertia or for small mass and moment of inertia together, as anticipated in §3. Fourth, the theory and reduced-system calculation can also probably enlarge its scope: to other shapes of body with a view to including more realism, starting with non-zero thickness coefficients A, B, C introduced in §2; to consecutive impacts by multiple bodies; to non-shallow water; to smooth bodies, about which similar work is being done in [2] (some smooth bodies can provoke splash jets at both the leading edge, i.e. the moving front contact point, and the trailing edge, i.e. the rear contact point, during the early stages of impact [2,29]); to flexible bodies [30]; to three-dimensional interactions. Fifth, and partly to repeat, more useful understanding requires more efforts in handling three spatial dimensions, in handling viscous–inviscid separation of two fluids and in handling more widely air effects in the presence of water, with the objective to link to experiments and observations throughout.

## Acknowledgements

We thank numerous colleagues and contacts for their help in this venture, especially Roger Gent, Richard Moser, David Hammond, Peter Hicks, Sasha Korobkin and Tatiana Khabakhpasheva.

## Footnotes

One contribution of 15 to a Theme Issue ‘Stability, separation and close body interactions’.

- © 2014 The Author(s) Published by the Royal Society. All rights reserved.