J . Fluid Mech. (1991), vol. 222, p p .

21,5230 215
Printed in Great Britain

Incompressible water-entry problems at small

deadrise angles
By S. D. HOWISON', J. R. OCKENDON' A N D S. K. WILSON2
Mathematical Institute, 24-29 St Giles, Oxford OX 1 3LB, UK
School of Mathematics and Physics, University of East Anglia, Norwich NR4 7TJ,UK

(Received 26 June 1989 and in revised form 11 May 1990)

This paper summarizes and cxtends some mathematical results for a model for a class
of water-entry problems characterized by the geometrical property that the
impacting body is nearly parallel to the undisturbed water surface and that the
impact is so rapid that gravity can be neglected. Explicit solutions for the pressure
distributions are given in the case of two-dimensional flow and a variational
formulation is described which provides a simple numerical algorithm for three-
dimensional flows. We also pose some open questions concerning the well-posedness
and physical relevance of the model for exit problems or when there is an air gap
between the impacting body and the water.

1. Introduction
The aim of this note is to review and extend the mathematical techniques available
for analysing high-velocity entry flows into a half-space of inviscid fluid (water) in
the absence of surface tension in cases where there is a small 'deadrise angle ' between
the impacting body, be it liquid or solid, and the undisturbed free surface. In these
cases the effect of gravity is small over most of the flow, and the contact region
between the impacting body and the water half-space expands rapidly. The principal
theoretical goal is to find that part of the contact region over which appreciable
hydrodynamic forces are exerted.
An excellent review of the subject has been given by Korobkin & Pukhnachov
(1988). Also the pioneering but intuitive work of von Kirman (1929) and Wagner
(1932) has recently been put on a firmer theoretical basis using matched asymptotic
expansions (Cointe & Armand 1987; Wilson 1989; Cointe 1989). This approach forms
the basis for our treatment of this class of problems, the first part of which can be
regarded as an extension of the results of Cointe & Armand to non-self-similar
motion. In addition t o this, we will mention the extension of the theory to include
three-dimensional impact, air-cushion effects and some questions of stability.

2. Impact by rigid bodies

Mathematically, the best-studied entry problem is the case of a two-dimensional
self-similar geometry in which the impacting body is a wedge, gravity and
compressibility are neglected throughout, and the effects of any cushioning fluid,
such as air, between the wedge and the water are ignored. Theoretical studies of this
case have been made by Wagner (1932),Garabedian (1953) and Mackie (1969) among
others. The situation is too idealized to be of much practical value but it is
216 S. D . Howison, J . R . Ockendon and S . K . Wibon
(4 ZA

E=-, @=O
az
- 4f)

VP@ = 0

FIGURE
1 ( a , b ) .For caption see facing page.

susceptible to numerical algorithms and experiments (see e.g. Dobrovol’skaya 1969;

Hughes 1972 ; Greenhow 1987) and can give clues about more general cases. Of most
interest for our purposes is the small ‘deadrise’ limit ( E $ 1 in figurela) in which the
numerical and experimental evidence of Greenhow (1987) both suggest the formation
of thin ‘jets ’ running up the sides of the impacting wedge. Indeed, the principal new
phenomenon observed in this configuration is the existence of a very small region of
high pressure on the body which enables a precise description of the jet formation
mechanism to be given. This high pressure region is implicit in the work of Wagner
(1932) and it has also been discussed in a more modern framework of matched
asymptotic expansions by Watanabe (1986), Cointe & Armand (1987) and Wilson
(1989).
The wedge-entry problem is special in that it has no lengthscale, but, by exploiting
the smallness of E , we will be able to construct an approximate solution not just for
a wedge with a small deadrise angle but for any impacting body z /L = f ( s x / L )whose
slope is small for 1x1 x O(L).By rescaling time, our approach can be modified to
describe the initial stages of impact of a general smooth body for times such that the
penetration depth is much smaller than the radius of curvature of the body a t the
 / // / / / / / // / / .'/ / // / / / / / / / / / / / // / / / // / /

-=o
ad I
I
8
az I

%
al + u-au = 0
ax
ah a(uh)
%+-
ax = o
FIGURE
1 . (a) Fluid entry geometry of a wedge at small deadrise angle. (b) Outer flow region
(c) Inner flow region. ( d ) Jet region.

point of impact. I n such a case it is sufficient to approximate the body by a parabola,

or a paraboloid in three dimensions. In these geometries the linearized problem (but
not the full problem) has a similarity solution which is discussed by Cointe & Armand
(1987) and Armand (1989). In this paper we will solve the linearized problem for a
general uniformly shallow body.
We will consider a non-dimensional model in which distances are made
dimensionless with the penetration distance L , velocity with a typical impact
velocity V,, time with L / & , velocity potential with V,L and pressure with the liquid
dynamic head corresponding to 5. Also, we will begin by taking V, to be constant
and the initial free surface to be flat. The basic idea of the theories presented by
Cointe & Armand (1987) and Wilson (1989) is that the flowfield decomposes into the
three regions shown in figure 1 ( a ) .
There is a large outer region I in which, by resealing x = X / e , z = Z/E, and the
velocity potential q5 = @ / e , we find that the liquid responds, to lowest order in e, to
the normal impact of a flat plate 1x1
< d ( t ) moving in the negative 2-direction with
unit speed (figure l b ) ; the corresponding pressure is O ( E - ~and ) the free-surface
elevation is of the order of the penetration distance. There is also an inner region 11,
218 S. D. Howison, J . R. Ockendon and S. K . Wilson
in which I X k d l = 0 ( e 2 ) which
, is a high pressure Kelvin-Helmholtz cavity flow with
velocity O ( E - ~pressure
), 0 ( c 2 )and
, extent O ( B )as in figure 1 (c). Finally region 111
is an equally high-velocity jet of length O(s-'), and thickness O ( E ) ,but at a low
pressure O ( E )as in figure 1 ( d ) .
The size d/E of the equivalent flat plate is determined, as suggested in Wagner
(1932), by the condition that the leading-order free-surface elevation as X J. d in the
outer flow should equal the wedge elevation as X t d to within the O(e2)lengthscale
of region 11. I n terms of the leading-order outer velocity potential @ this elevation
is

@ ( X , O , t )= 0,
I

(V@I+ 0 as X2+Z2 a, --f ( 2d )

and t = 0 is the time at which the impacting body meets Z = 0 . i Hence if we write
the wedge as Z = ef(X) a t t = 0,

and, since @ = Re { - ( Z + ( d 2 ( t ) - ( X + iZ)2)h)>,(3) is simply

Then, by inversion (Tollmien 1934),

d-'(X) = - (4)

We would clearly expect this result to apply t o more general symmetric impacting
shapes than wedges, and we will say more about this later. Meanwhile we emphasize
that the wetted area extends to the tip of the jet and thus exceeds 1x1
= d ; it is only
for 1x1
< d in region I and in region I1 that pressures of O(6-l) or greater are
discernible. Indeed, the lowest-order surface pressure in the outer solution is

1x1> d .
t It can be shown that (1) and (2) are equivalent to saying that, to lowest order in E , the volume
of fluid displaced by the body (i.e. the area of the body below the undisturbed waterline Z = 0) is
equal to the volume of fluid above the undisturbed waterline as computed from the model for
region I ; in other words, the flux into the jet is small compared to the net volume displaced.
 Incompressible water-entry problems at small deadrise angles 219

-
2E
Dimensionless distance
FIGURE
2. Typical pressure distribution on a wedge y = ~1x1.

As in Cointe & Armand (1987), we can find the lowe$-order solution in region I1
by writing X--d = s2z, 2-s(f(d)-t) = s22, @ - d X = @ to obtain the free-boundarx
problem shown in figure 1(c). The boundary conditions on the potential function @
are such that its normal derivative vanishes and its tangential derivative has
modulus d on the free boundary, and that a & / d = 0 on 2 = 0. Matching with the
first term in region I also gives that, as 2 +- ao,
a&/ar? -i a6/a& - - d + ( -d/2(r? +i2)):

-
and that the lower branch of the free boundary is described asymptotically by
2 - (2dz)$/d.We note that is invariant under a translation
of r?. Again, as in Cointe & Armand (1987) we can use standard conformal mapping
methods (Birkhoff & Zarantonello 1957) to obtain & + i!? = dw@+ i&) where, up to
an arbitrary function o f t added to w,

where w' = dw/d(r?+i@ and h,(t) is the asymptotic jet thickness as we approach
region 111. Matching with region I gives

h, = xd/8d2 (7)
and the pressure distribution on the body in region I1 is given parametrically to
lowest order by
d2
p = -[l-[gf(E2-l)q2] (8)
2s2

where
8 FLM 222
320 S . D. Howison, J . R. Ockendon and S. K . Wilson

xt
-
2

Dimensionless time

21 A
w

(21):

IA
Aa
(d(t)l- a2)i--+
2

a sin-' (g) =-
At

Dimensionless time

TABLE1. Some explicit outer solutions for different f

for 5 < - 1, > + 1 respectively ; 161 = co corresponds to the relative stagnation point
2 = 0 where the maximum pressure occurs.
I n (8) we have arbitrarily chosen the relative stagnation point to be at 2 = 0. The
problem of calculating the O(e)correction to the distance of this point from X = d has
been raised in Cointe & Armand (1987) and Wilson (1989). As in similar shock
location problems in gasdynamics (Lardner 1986), its resolution seems to require a
necessarily complicated second-order analysis of the outer solution.
The jet region I11 is described t o lowest order by zero-gravity shallow-water
theory and hence the tangential velocity u(X, t ) / e and thickness sh(X,t ) are such
that
au au ah a(&)
-+u-=o, -+-- - 0.
at ax at ax
The fact that the mass flow in the jet is only 0(1) as E + O confirms the mass
conservation argument leading t o (1).
 Incompressible water-entry problems at small deadrise angles 22 1
Experiment Theory

I60
140
120
100
80
60
40
20
0
0

k8 20
0
B

20 - B

20 C
0 /-

0 50 100 150 200 250 0 SO 100 150 200 250

Time after impact (ms) Time after impact (ms)

FIGURE 3. Comparison of experimental and theoretical pressure histories. Experimental data from
Nethercote et al. (1986) with L = 12 ft, V = 20 ft/s and E = 0.707 so that the dimensionless hull
shape is y = ( 0 . 7 0 7 ~ )At
~ . the keel the theory predicts an infinite pressure and the oscillations
observed experimentally are probably due t o air entrapment.

Matching with region I1 gives u = 28, h = h, at X = d , and hence for the wedge

To lowest order the jet is only affected by the shape of the impacting body through
d ( t ); however, the pressure on the body in the jet is determined by a higher-order
analysis and depends on the body curvature K through

p - -€KhU2, (11)
where K has the sign off”. Also, from (4), it is easy to show that shocks (hydraulic
jumps) will develop in the solution of (9a) if d < 0. For curved f, the jet pressure
takes its lowest value on the body whenever it is convex and this might provide a
criterion for whether or not the jet separates ; however, surface tension and gravity
8-2
222 S . D . Howison, J . R. Ockendon and S. K . Wilson

Body shape Cross-section in plane Equivalent flat plate

z = E ( f ( X ,Y ) - t ) z=o t = w ( X ,Y )
Circular cone Circle Concentric circle
f ( X ,Y ) = (X2+ y2)t x2+ P = t2 w ( X , Y ) = iX(X2+ ye)&
Elliptic paraboloid Ellipse 'Confocal ellipse
+
f ( X , Y ) = i(k , X z k , P) +
k,X2 k, P = 2t w ( X , Y ) = )k,(2-e2-az)XI
Korobkin & Pukhnachov (1988)
a= ( -2r (."+A)
TABLE2. Some known explicit solutions

may have a controlling effect on this phenomenon (see Vanden-Broeck & Keller
1989).
Neglecting the jet pressure, the uniformly valid composite expansion for the
pressure exerted on the wedge can now be constructed as in figure 2. The total
dimensionless force exerted on the wedge is O(E?) ; the leading-order term, which just
results from region I is, from (5),ndd/s2.Some explicit outer solutions for different f
are given in table 1, which also lists schematically the corresponding pressure
distributions in 0 < X < d ( t ) as t increases.
As mentioned above these results hold formally for arbitrary rigid body impact of
a surface 2 = sf(X) whose deadrise angle is everywhere small and also, with slight
modification, t o the case when v(X, , is O(1) or when V, is
0) is non-zero but q X ( X 0)
non-constant and varies on a timescale L/V,. Indeed the small-time oblique impact
of a circular cylinder on an initially circular free boundary has been analysed in this
way by Cointe (1989) as a model for the interaction of spilling breakers with marine
structures. However we can in general only treat cases in whichf(X) and y(X,0) are
even.
Figure 3 compares the leading-order dimensional composite pressure (again
neglecting the jet) exerted on a parabolic impacting body with pressure histories
measured by Nethercote, Mackay & Menon (1986) for a hull of nearly parabolic cross-
section. The agreement is good except on the keel, and this disparity is discussed
further in $4, The use of the composite pressure to compute the force on a circular
cylinder for small times after impact has been carried out more comprehensively, and
compared with experiments, by Cointe & Armand (1987). These authors give a.
helpful account of the way in which the force in this case falls from its initial, so-
called Wagner value (namely xdd/s2)towards the value which would be obtained if
d were set equal t o the semichord in which the x-axis meets the cylinder (the so-called
von KBrmBn value).
We can also write down formally analogous results for the impact of a three-
dimensional body z = f(m, BY) except that the inversion (4) is no longer available in
general unless there is axial symmetry. Some of the known explicit solutions are
listed in table 2, where y = Y / s and the boundary of the equivalent flat 'disk' in
2 = 0 is denoted by t = w(X, Y ) . Analytical progress is easiest when there is axial
symmetry, in which case the leading-order force on the body is 4+r2/e2, where r / e is
the radius of the equivalent disk. For a paraboloid with unit curvature the force is
 Incompressible water-entry problems at small deadrise angles 223
6 1/3ti which is in fair agreement with the measurements of Moghisi & Squire (1981)
for the initial stages of the impact of a sphere, where the force was approximately
8.209. Further details of this are given in Wilson (1989).

3. Variational formulation
The three-dimensional version of the model described in $2 can in general only be
solved numerically. An easily-implemented algorithm for carrying this out has been
suggested by Korobkin (1982), and the transformation he uses has the added bonus
of providing a framework in which t o discuss the existence, uniqueness and
regularity properties of the weak solution of the model, although we will not discuss
these aspects here.
I n the spirit of the so-called ‘Baiocchi transformation ’ in the theory of variational
inequalities (Lions & Stampacchia 1967), we define a displacement potential

@*(X,Y ,2,t ) =

a simple calculation shows that

s, @(X,Y ,Z , 7)d7 ;

VW*=O, z<o (13)

with @*(X,
Y ,0, t ) = 0 for t c w ( X ,Y ) . (14a)
Also for t -co,

but, for t > o,

a@*
-(X,
azY ,0, t ) =
(1L) + 3

= ?/qX,Y,o)-(t-w)
7

= f ( X ,Y )- t ,
from the three-dimensional generalization of (1). Hence as t + o+ 0,

+-t+f(X, Y).
az 2-0

Thus the great advantage of working with 9*instead of @ is that IV@*l is bounded
on t = w and so @* satisfies the ‘complementarity problem’ (equivalent t o a
variational inequality)
VW*=O, z<o (164

This provides both a mathematical basis for the expanding plate model (13),(14)and
a minimization algorithm for numerical calculations (see for example Elliott &
224 S . D . Howison, J . R . Ockendon and S. K . Wilson

(in- l ) f

I
FIGURE
4. Comparison of x , simple finite-element calculation and -, exact solution of the
outer free-surface elevation during a two-dimensional wedge impact at t = 0.1.

Ockendon 1982). A rough finite-element discretization for the case of wedge impact
is given in figure 4 for comparison with the exact solution which is available
analytically from (4) and table 1.
We note that difficulties may arise with this procedure when the curve t = w ( X , Y )
extends to infinity, as may be the case for an approximate model of the impact of a
long ship with small deadrise angle a t a small angle of attack. However it seems likely
that if the angle of attack is much smaller than the deadrise angle the solution may
be approximated by a sequence of two-dimensional solutions such as those shown in
table 1.

4. Generalized impact problems

4.1. Air-cushion eSfects
The above model neglects several mechanisms which are important in practical
problems. Apart from those already mentioned, the most likely explanation for the
discrepancy in the pressure readings illustrated in figure 3 for the evolution of a
parabolic impact (Nethercote et al. 1986) is that air pressure is not negligible in the
cushion between the solid and liquid.
The simplest configuration in which t o discuss this mechanism is that of a flat-
bottomed wedge approaching an initially horizontal free surface as in figure 5. This
geometry has been studied experimentally by Driscoll & Lloyd (1982) and
theoretically by Verhagen (1967) and Asryan (1972) and we assume the wedge itself
has a large enough deadrise angle t o allow us to neglect air pressure in 1x1 > 1. If we
also neglect air compressibility in 1x1 < 1 then the pressure gradient in the air is
+
proportional to (aV/at) V(aV/ax)where V is its velocity, which is nearly in the x-
direction except near the stagnation point.
Now for a small surface elevation 7, the surface water pressure gradient is
proportional to the Hilbert transform
 Incompressible water-entry problems at small deadrise angles 225

FIGURE
5. Air cushioning geometry

Hence if we write 7 = ye71 and 77 = yd where y %- 1 is the initial aspect ratio of the
cushion and 0 is the air to liquid density ratio, the continuity of pressure on z = 0
requires

The second relation between v" and i j results from mass conservation in the air layer,
namely

for times t such that the air gap thickness is much less than unity. This model is
mathematically intractable but as yB+O, we can solve (18) for v" and (17) for f to
obtain
v"-- X q 2[2+xlog(G)]logltl
1-x
t' R
as t t 0 .
The fact that 71" -+ 00 as 1x1+ 1suggests the initiation of the profile sketched in figure
5, but whether this profile is attained in the presence of the nonlinear terms in (18)
is unclear. However, the possibility of air escaping through narrow gaps at 1x1 = 1
suggests that air compressibility effects will first become significant there ; such
effects are discussed in Lewison (1970).
4.2. Liquid-liquid and liquid-solid impacts
Several of the ideas mentioned above apply to liquid-liquid and liquid-solid impacts
at small deadrise angles. For example, the impact of two identical liquid cylinders
along a common generator is identical to the impact of a solid plane on one of the
cylinders and can be solved as described before table 1.
For the case of an asymmetric impact with speeds V,, V, as in figure 5, it is easy to
226 8.D. Howison, J . R . Ockendon and S . K . Wilson

FIGURE6. Fluid-fluid impact geometry. V, and V, denote the dimensionless speeds of the impacting
fluid masses and the dashed curves are the positions each would have reached in the absence of the
other.

see by considering the flow in a frame moving in the 2-direction with speed $(&- K),
in which the contact line is a t rest, that the jet angle a is O ( E only
) if the high pressure
peaks at X = d,(t) are within O ( B )from each other. However the jet angle a can only
be obtained by carrying out the analysis in region I1 to second order.

5. Stability and exit problems

We will only consider a stability analysis of the model for region I locally in space
and time near the ‘free boundary’ X = d . We thus only consider (1)-(3) with a semi-
infinite plate 2 = 0,X < d .
In the two-dimensional solution, @ has a square-root singularity at the free
boundary which means the local velocity is so large that we can neglect the impact
speed, but not aT/at, in (2) compared to a@/aZ. When there are variations in the
Y-direction, the approximate local model is thus
V2@=0, z<o, (20a)

z = 0,

with 7 + 0 as X J. d ; also, to match with a two-dimensional solution,

@ - A Re (-X-iZ)i (21)
+
as X 2 y2 + Z2+ 00 where, because we are considering temporal variations which are
short on the outer timescale, A is a constant.
 Incompressible water-entry problems at small deadrise angles 227
This approximate model has an exact travelling wave solution independent of Y
in the form
@ = ARe(d-X-iZ)i (=a)
d = Vt, 7 = -(A/V)(X-d)i, X > d, (22b)
where V is arbitrary. Like A, V may also be chosen to match with the outer solution
and A and V will in fact be related in any particular problem. The precise relationship
is unimportant for our analysis but we note that A and V have the same sign and
A > 0 corresponds to a plate expanding in an impact problem. Also 7 is negative
because the water surface is depressed relative to the X-axis in our local coordinate
system.
It is convenient to change to moving axes E = X- Vt, which changes (20d) into

and we denote the local boundary X = d by

6 = 6cos(ny)eat+O(S2), n > 0 ;
here 6 is a prescribed small number and a is the growth rate we are seeking. A naive
expansion in powers of 6 yields
B eat
+
= Ari sin (p) 6cos (ny)- -sin
ri
e-nr + o ( P ) ,
(9)
7 = --E
A
t+6~os(ny)~,(~)e~~+O(~~),
V
where r = (E2+Z2)i,0 are polar coordinates and B is an undetermined
These expansions satisfy (20a-c) and then (23) implies that

so that

after integration by parts. Here C is another constant, and at the moment CT is

arbitrary. Now if a / V > 0, we can only satisfy the condition that
7, = o(& as [+ 00
ds = - ( m / ~ + n ) t n f ,
if
C = -(g/V+n)
(om e-(u/v+n)s-
St

whereas if a / V < 0,this condition is satisfied even if C remains arbitrary.

However, the expansion (25) is invalid for r = O(S),and this necessitates an inner
expansion in which E = SC,Z = 6 2 ,@ = &@', 7 = &7', and (23) becomes

The inner expansion then proceeds

0' A Re (cos (ny)eut- 5'- i 2 ) i+ . ..
- (284
N

7' -( A/V)(~-cos(ny)eat)~ + C > cos(ny)eUt,

..., (28b)
228 S . D . Howison, J . R.Ockendon and 8.K . Wilson
the next terms being of O(S)apart from any extra terms generated by matching with
the outer expansi0n.t A further matching of the two-term outer with the one-term
inner expansion for @ shows that B = $4.
Now the inner expansion of (26b) takes the form

The first term in (29) automatically matches the term of O(f'-i) in the expansion of
(28b) as f'+ co but the second term generates a constant of O(d)with which the next
term in (28b) has to match. This would lead to a potential problem for the second
term in the expansion for W which would have to satisfy homogeneous Neumann
and Dirichlet data on 2' = 0, f' < cos (ny)e"t, f' > cos (ny)eUtrespectively and be
such that its normal derivative on 2' = 0, 5' < cos(ny)euthad finite integral. The
boundary conditions mean that such a function would have to be a combination of
square roots of f ' + i Z and no such function has the required finite integral. We
conclude that the term &4C in (29) is zero; hence if u/V > 0, g/V+n = 0 which is
impossible. Thus u and V must have opposite signs, which suggests that an
expanding plate problem is stable but that an 'exit ' problem when V < 0 is unstable.
The fact that we are unable to write down a dispersion relation between u and n
is related to the behaviour of the solution of the linear initial-value problem for (20)
in which the initial condition 7 = SZ(t),say, at t = 0 is imposed instead of (24). We
will not pursue the details here but we note that (23) now yields a first-order partial
g.
differential equation for the difference between T,I and - ( A / V ) The characteristics
are (+ Vt = constant, but we need to solve in ( 2 0 and data is prescribed at t = 0;
hence the problem cannot be solved without some continuation process involving SZ
if V < 0.
The above analysis suggests there may be marked differences between exit and
entry problems. Even though any solution of our expanding plate model can be
reversed in time to give the solution of an exit problem with initial conditions
identical to those encountered in the evolution of an entry problem, this tells us very
little about the evolution of the exit problem for arbitrary initial data. In addition
to the stability argument, differences between the two situations are suggested by
the facts that
(i) time reversal in an entry problem reverses the sign of the pressure in region I ;
hence, from table 1, large negative pressures are predicted which could invalidate the
model by causing the boundary of the contracting plate to break up ;
(ii) the smoothing transformation (12) can still be applied to the exit problem but
we can no longer derive the crucial boundary condition (14)in the region traversed
by t = o(X, Y ) . This difficulty also prevents us writing down an integral equation
corresponding to (3) without some extra physical assumption about the nature of 7
in this region.
Finally, we note that a formal solution to any exit problem in the absence of
gravity is that the body instantaneously loses contact at all points and q(x,t ) G ~ ( z0).
,
However, further work is needed to see if there is a mathematical interpretation
which is in better agreement with the observations of Greenhow (1988).
t The expansion (28) is itself invalid near 5' = 2' = 0 because repeated inner expansions are
needed to determine the higher-order terms in the free-boundary position (24).
 Incompressible water-entry problems at small deadrise angles 229

6. Conclusion
We have presented a brief account of the theory of small deadrise impacts between
solids and liquids, assuming for the most part that surface tension, gravity, viscous
and compressible effects can be neglected. This enables explicit solutions t o be
calculated for the pressures exerted on the impacting body in two-dimensional and
axisymmetric cases. Moreover, general three-dimensional impacts can be reduced to
a variational formulation which is suitable for numerical discretization, and which
avoids explicit reference t o the boundary of that part of the impacted region where
high pressures are exerted.
Although the results of these calculations have been shown to agree quite well with
experimental evidence, several of the solutions we have described suggest that some
or all of the neglected effects may be important in localized regions. I n particular
more realistic analyses of the jet tip and of jet separation may be needed. Another
important open question concerns the formulation of exit problems which, even in
the absence of gravity, do not seem to be well described in the small deadrise limit
as time reversals of entry problems.

We are very grateful Mr D. Chalmers and Mr J. Clarke for suggesting this research
and t o Drs J. Byatt-Smith, M. Greenhow and P. Wilmott for discussions concerning
the modelling of splashing and stability respectively. We also acknowledge some very
helpful referees’ comments, and would like to thank Professor H. Peregrine for
pointing out the work of Cointe & Armand. S. K. W. acknowledges financial support
from the SERC.

REFERENCES
ASRYAN, N. G. 1972 Solid plate impact on surface of incompressible fluid in the presence of a gas
layer between them. Isv. Akad. Nauk. Arm. SSR Mekh. 25, 3 2 4 9 .
BIRKHOFF,G. & ZARANTONELLO, E. H. 1957 Jets, Wakes and Cavities. Academic.
COINTE,R. 1989 Two-dimensional water-solid impact. ASME J . Offshore Mech. Arc. Engng 111,
109-1 14.
COINTE,R. & ARMAND, J.-L. 1987 Hydrodynamic impact analysis of a cylinder. ASME J . Offshore
Mech. Arc. Engng 109, 237-243.
DOBROVOL’SKAYA, Z. N. 1969 On some problems of similarity flow of fluid with a free surface.
J . Fluid Mech. 36, 805-829.
DRISCOLL, A. & LLOYD,A. 1982 Slamming experiments - description of facilities and details of
impact pressure results. Rep. AMTE (H) R82002.
ELLIOTT,C. M. & OCKENDON, J. R. 1982 Weak and Variational Methods for Free and Moving
Boundary Problem. Pitman Research Notes in Mathematics, no. 59.
GARABEDIAN, P. R. 1953 Oblique water entry of a wedge. Commun. Pure Appl. M a t h 6 , 157-165.
GREENHOW, M. 1987 Wedge entry into initially calm water. Appl. Ocean Res. 9, 214-223.
GREENHOW, M. 1988 Water entry and exit of a horizontal circular cylinder. Appl. Ocean Res. 10,
191-198.
HUGHES,0 . F. 1972 Solution of the wedge entry problem by numerical conformal mapping.
J . Fluid Mech. 56, 173-192.
KARM~N T., VON 1929 The impact of sea plane floats during landing. NACA TN 321.
KOROBKIN, A. A. 1982 Formulation of penetration problem as a variational inequality. Din.
Sploshnoi Sredy 58, 73-79.
KOROBKIN, A. A. & PUKHNACHOV, V. V. 1988 Initial stage of water impact. Ann. Rev. FluidMech.
20, 159-185.
LARDNER, R. W. 1986 Third-order solutions of Burgers’ equation. &. Appl. M a t h 64, 293-301.
230 8. D . Howison, J . R . Ockendon and S . K . Wilson
LEWISON, G. R. G. 1970 On the reduction of slamming pressures. Trans. RZNA 112, 285-306.
LIONS,J.-L. & STAMPACCHIA, G. 1967 Variational Inequalities. Commun. Pure Appl. Maths 20,
493-519.
MACKIE, A. G. 1969 The water entry problem. &. J . Mech. Appl. Maths 22, 1-17.
MOGHISI,M. & SQUIRE,P. T. 1981 An experimental investigation of the initial force of impact on
a sphere striking a liquid surface. J. Fluid Mech. 108, 13-146.
NETHERCOTE, W. C. E., MACKAY,M. & MENON, B. 1986 Some warship slamming investigations.
DREA Tech. Mem. 86f 206.
TOLLMIEN, W. 1934 Zum LandestoB von Seeflugzeugen. 2.Angew. Math. Mech. 14, 251.
VANDEN-BROECK, J. M. & KELLER,J. B. 1989 Pouring flows with separation. Phys. Fluids A 1 ,
156-159.
VERHAGEN, J. H. G. 1967 The impact of a flat plate on a water surface. J . Ship Res. 11,211-223.
WAQNER, H. 1932 Uber Stoss-und Gleitvorgange a n der Oberflache von Flussigkeiten. 2.angew.
Math. Mech. 12, 193-215. (Trans. Phenomena associated with impacts and sliding on liquid
surfaces. NACA Trans. 1366.)
WATANABE, I. 1986 Analytical expression of hydrodynamic impact pressure by matched
asymptotic expansion technique. Trans. West Japan Soc. Naval Arch. 71, 77-85.
WILSON,S. K. 1989 The mathematics of ship slamming. D. Phil. thesis, Oxford University.