1. Introduction
An unsteady plane potential flow of an ideal incompressible fluid with a free boundary is considered. The external forces and capillarity are assumed to be absent. At the initial time, the flow domain and flow velocity are prescribed; the evolution of the domain occupied by the fluid and the flow velocity at subsequent times are to be determined. Problems of this kind have a long history. The first exact solutions that describe flows with a linear velocity field were found long ago by Dirichlet (Reference Dirichlet1861). Later on, exact solutions, also with a linear velocity field, were derived by Ovsiannikov (Reference Ovsiannikov1967) and Longuet-Higgins (Reference Longuet-Higgins1972). A new class of solutions, i.e. solutions with a ballistic free boundary were found comparatively recently by Karabut & Zhuravleva (Reference Karabut and Zhuravleva2014) and Karabut, Zhuravleva & Zubarev (Reference Karabut, Zhuravleva and Zubarev2020). The term ‘ballistic free boundaries’ was introduced in the review paper by Liu & Pego (Reference Liu and Pego2021) and is applied to flows with a zero acceleration at the free boundary. Nevertheless, quite a few exact solutions that describe the flow of an ideal fluid with free boundaries were found. The primary reason is the necessity of satisfying complex nonlinear conditions at the free boundary. The authors of all publications mentioned above tried to simplify the problem by using either a linear velocity field or a zero acceleration at the free boundary. One more possibility of unsteady problem simplification is the use of self-similar variables. This option made it possible to obtain exact solutions, e.g. John (Reference John1953), Longuet-Higgins (Reference Longuet-Higgins1976) and Zakharov (Reference Zakharov2020). In the present study, we also use self-similar variables.
The next important issue to be mentioned is the use of power series in the present study. For unsteady problems with free boundaries, a natural parameter inherent in any problem is the time
$t$
; therefore, it seems natural to seek the solution in the form of a power series with respect to time. Apparently, the idea of using power series with respect to time comes back to Stoker (Stoker Reference Stoker1992). Later on, this procedure was successfully used, e.g. by Makarenko et al. (Reference Makarenko and Kostikov2013) and Baumel et al. (Reference Baumel, Burley, Freeman, Gammel and Nuttall1982). Normally, several initial terms of the series are found, and hence a solution asymptotically suitable at small values of
$t$
is obtained. A specific feature of the present study is the exact summation of power series.
2. Formulation of the problem
Let us consider the following formulation of the problem. Let an ideal incompressible fluid at the initial time
$t = 0$
in the plane
$x, y$
occupy a wedge with an angle
$\alpha$
at the apex, as is shown in figure 1. One boundary of the wedge is a free boundary, while the other one is a solid wall at
$y = 0$
. The solid wall is added here for the sake of symmetry. It can be removed; in this case, the flow in a fluid wedge with an apex angle
$2 \alpha$
can be considered.
Initial shape of the flow domain.

We consider an inertial flow, and the governing factor here is the initial velocity field. An exact solution that describes the fluid flow in a wedge with an initially quadratic velocity field was obtained in Karabut et al. (Reference Karabut and Zhuravleva2020). In the present study, we construct a solution for the initial velocity field of the form
Here,
$z = x+ \textrm {i} y$
is the complex variable,
$U = u - \textrm {i} v$
is the complex velocity and
$u(x,y,t),$
$v(x,y,t)$
are the components of the fluid velocity vector. Note that the method used to derive the solutions imposes certain limitations. In particular, the complex velocity function at the origin of the coordinate system has to be analytical. The case with
$n=1$
does not fit our scheme. In situations where
$n$
is a real number that does not belong to the set of natural numbers, the method cannot be applied, in particular, because of the singularity of the complex velocity function at the origin. We will discuss the possibility of constructing a solution in these cases later.
One of the difficulties arising in solving problems with a free boundary is the fact that the boundary-value problem has to be solved in a domain that has to be determined in the course of problem solution. The use of conformal mappings allows the situation to be significantly simplified. For this purpose, we introduce an auxiliary plane
$\zeta = \xi + \textrm { i} \eta$
, choose a fixed domain
$D$
in this plane and seek a conformal mapping of the domain
$D$
onto the domain occupied by the fluid in the physical plane
$z = Z(\zeta ,t).$
Thus, we obtain two sought functions: the complex velocity
$U(\zeta , t)$
and conformal mapping
$Z(\zeta ,t).$
It was shown by Karabut et al. (Reference Karabut and Zhuravleva2023) that the kinematic and dynamic conditions imposed on the free boundary take the following form in the case of using the conformal mapping:
\begin{equation} \begin{aligned} \textrm {Im} \left [\frac {\partial Z}{\partial s} \left ( \overline {\frac {\partial Z}{\partial t}} - U \right ) \right ] = 0, \\ \textrm {Re} \left [\frac {\partial Z}{\partial s} \frac {\partial U}{\partial t} - \frac {\partial U}{\partial s} \frac {\partial Z}{\partial t} + \frac {\partial U}{\partial s} \bar {U} \right ] = 0. \end{aligned} \end{equation}
Here
$s$
is an element of the length of the free boundary of the domain
$D$
in the auxiliary plane
$\zeta$
.
In the problem of fluid wedge deformation considered here, we choose the domain
$D$
to be the domain occupied by the fluid at the initial time, i.e. a wedge with an apex angle
$\alpha$
. We also require that the wedge apex
$\zeta = 0$
should transform to the cusp point of the free boundary located at the origin of the coordinate system
$z = 0.$
Therefore, the boundary conditions (2.2) have to be satisfied on the ray
$\zeta = s e^{ \textrm {i} \alpha }, \ (s \geqslant 0).$
The solid wall
$\zeta = s \ (s \geqslant 0)$
is subjected to the no-slip condition, which corresponds to the following conditions in our case:
3. Self-similar solution
We assume that the only dimensional constant in our problem is the constant
$A$
(with the dimension
$[A] = [t]^{-1} [z]^{1-n}).$
In this case, instead of two independent variables
$z$
and
$t$
, we consider the self-similar variable
$z \boldsymbol{\cdot }t^\gamma$
, where
$\gamma = 1/(n-1).$
The sought after function in the physical plane is the complex velocity function
$U(z,t)$
. It is sought in the form of a time series. From the consideration of the dimension, this series has the form
Obviously,
$U(0,t)=0.$
This means that the fluid particle initially located at the wedge apex always remains there, which means that the free surface always passes through the origin of the coordinate system.
The sought after functions in the auxiliary plane
$\zeta$
are those of the complex velocity and conformal mapping. It follows from the dimensional considerations that they can be presented as
Thus, instead of two unknown functions
$U$
and
$Z$
of two variables
$\zeta$
and
$t$
, we have two unknown functions
$f(\zeta )$
and
$g(\zeta )$
of one variable
$\zeta$
. Substituting (3.2) into conditions (2.2) and (2.3), we obtain the boundary conditions for the sought after functions
$f(\zeta )$
and
$g(\zeta )$
\begin{equation} \begin{aligned} \textrm {Im} \big [e^{\textrm {i} \alpha } g'(\zeta ) \left (\gamma \,\bar {g}(\zeta ) + f(\zeta )\right ) \big ] = 0, \quad \mbox{where } \zeta = s e^{\textrm {i} \alpha } ,\\\textrm {Re} \big [e^{\textrm {i} \alpha } \left (\gamma f'(\zeta ) g(\zeta ) - (\gamma + 1) g'(\zeta ) f(\zeta ) + f'(\zeta ) \bar {f}(\zeta ) \right ) \big ] = 0, \quad \mbox{where } \zeta = s e^{\textrm {i}\alpha }, \\\textrm {Im} f = 0, \quad \textrm {Im} g = 0, \quad \mbox{where } \textrm {Im} \zeta = 0. \end{aligned} \end{equation}
Here, the prime means the derivative with respect to
$\zeta .$
There is no time in the new boundary-value problem (3.3). Therefore, there arises a question of in which form the solution of the boundary-value problem (3.3) has to be sought for this solution to correspond to the solution in the form of the power series (3.1). It follows from series (3.1) that the power series in powers of
$t$
is simultaneously a power series in powers of
$z$
in the vicinity of the point
$z=0.$
As the point
$z=0$
with the chosen normalisation of the conformal mapping corresponds to the point
$\zeta = 0,$
we arrive at an idea of using power series in powers of
$\zeta .$
We seek the solution in the form
It follows from the third condition in system (3.3) that all
$f_j$
and
$g_j$
have to be real. The second series in (3.4) starts from the
$n$
th power. This is a consequence of having
$U \approx A Z^n$
or
at small times. Thus, for small values of
$\zeta$
, we have
$f(\zeta ) \approx g^n(\zeta ).$
Therefore, it should be assumed that
$f_n = g_1^n.$
For determining the remaining coefficients
$f_j$
and
$g_j$
, series (3.4) have to be substituted into the two first boundary conditions of system (3.3). Equating the coefficients at identical powers of
$s$
, we obtain recurrent formulas for consecutively finding the coefficients of the power series (3.4). However, no matter how many coefficients we find, this is only a partial sum of the series, which means that this is an approximate solution. In the present study, we managed to derive an exact solution, i.e. the sum of these series.
For summation of the series, we apply the technique described in Karabut et al. (Reference Karabut, Zhuravleva and Zubarev2020). The basic idea of this technique is the analytical continuation of the unknown function outward from the domain of its definition. After multiple rounds over the wedge apex, different branches of the sought after function turn out to be related by an infinite system of ordinary differential equations. Let us demonstrate that the system acquires a finite character if
$\alpha /\pi$
is a rational number.
Let us introduce the notation
$\omega = e^{\textrm {i} 2 \alpha }$
and consider an infinite set of functions
Note that each pair of functions
$P_j (\zeta )$
and
$Q_j (\zeta )$
consists of the functions
$g(\zeta )$
and
$f(\zeta )$
, but in a coordinate system rotated with respect to the initial coordinate system by the angle
$2 \alpha (j-1)$
.
Statement 1. The functions
$P_j (\zeta )$
and
$Q_j (\zeta )$
satisfy the infinite system of ordinary differential equations
\begin{equation} \begin{aligned} \frac {{\rm d} P_{j+1}}{{\rm d}\zeta } \big(\gamma P_{\!j} + Q_{j+1} \big) &= \frac {{\rm d} P_{\!j}}{{\rm d}\zeta } \big(\gamma P_{j+1} + Q_{\!j} \big) \quad (j \geqslant 1) ,\\(\gamma + 1) \left (\frac {{\rm d} P_{j+1}}{{\rm d}\zeta } Q_{j+1} + \frac {{\rm d} P_{\!j}}{{\rm d}\zeta } Q_{\!j} \right ) &= \frac {{\rm d} Q_{j+1}}{{\rm d}\zeta } \left (\gamma P_{j+1} + Q_{\!j} \right ) \\&\quad + \frac {{\rm d} Q_{\!j}}{{\rm d}\zeta } \left (\gamma P_{\!j} + Q_{j+1} \right ) \quad (j \geqslant 1). \end{aligned} \end{equation}
To prove statement 1, we consider the boundary conditions (3.3). On the ray
$\zeta = s e^{\textrm {i} \alpha }$
, we have
In addition, the equality of the imaginary part of a certain expression to zero means that this expression coincides with the corresponding conjugate expression. The equality of the real part of a certain expression to zero means that this expression is opposite to the corresponding conjugate expression. Therefore, the first boundary condition of (3.3) can be written as
As it follows from the third boundary condition of (3.3), the functions
$g(\zeta )$
and
$f(\zeta )$
are real at the real axis. Therefore, using the principle of symmetry, we can eliminate conjugation in (3.9) by using the formulas
Then we use formulas (3.6) for the values of
$\zeta$
lying on the ray
We obtain
Therefore, equality (3.9) on the ray (3.11) can be rewritten as
Using formula (3.8), we obtain the first equation of (3.7), which is valid only along the ray (3.11). Then we use the principle of analytical continuation, which reads that, if a certain analytical function is equal to zero on a segment of a certain straight line, then it is identically equal to zero. Therefore, the functions
$P_j (\zeta )$
and
$Q_j (\zeta )$
satisfy the first equation of (3.7) everywhere. To prove that they also satisfy the second equation of (3.7), we have to apply similar considerations to the second condition of system (3.3). Statement 1 is proved.
Statement 2. If
$\alpha = \pi k/m,$
where
$k$
and
$m$
are natural numbers, then
To prove this statement, it is sufficient to note that the functions
$f(\zeta )$
and
$g(\zeta )$
are initially assumed to be holomorphic because they are sought in the form of power series with respect to
$\zeta$
; therefore, the functions
$P_j (\zeta )$
and
$Q_j (\zeta )$
are also assumed to be holomorphic at the point
$\zeta =0.$
Therefore, equality (3.14) directly follows from formulas (3.6).
Thus, we can use the following statement for summation of the series.
Statement 3. If
$\alpha = \pi k/m,$
where
$k$
and
$m$
are natural numbers, then, to find the analytical functions
$f(\zeta )$
and
$g(\zeta )$
, it is sufficient to solve the system of ordinary differential equations
\begin{equation} \begin{aligned} \frac {{\rm d} P_{j+1}}{{\rm d}\zeta } \big(\gamma P_{\!j} + Q_{j+1} \big) &= \frac {{\rm d} P_{\!j}}{{\rm d}\zeta } \big(\gamma P_{j+1} + Q_{\!j} \big) \quad (j = \overline {1, m}) ,\\(\gamma + 1) \left (\frac {{\rm d} P_{j+1}}{{\rm d}\zeta } Q_{j+1} + \frac {{\rm d} P_{\!j}}{{\rm d}\zeta } Q_{\!j} \right ) &= \frac {{\rm d} Q_{j+1}}{{\rm d}\zeta } \big (\gamma P_{j+1} + Q_{\!j} \big ) \\&\quad + \frac {{\rm d} Q_{\!j}}{{\rm d}\zeta } \big (\gamma P_{\!j} + Q_{j+1} \big ) \quad (j = \overline {1, m}), \\P_{m+1} &=P_1 \quad Q_{m+1} = Q_1 ,\end{aligned} \end{equation}
and then assume that
Using statement 3, one can always find the functions
$f(\zeta )$
and
$g(\zeta )$
satisfying the boundary conditions (3.3). However, it cannot be guaranteed that the resultant functions are holomorphic everywhere in the domain occupied by the fluid.
Let us now consider the case with a cubic initial velocity field in more detail and then extend it to
$U(z, 0) = A z^n$
.
4. Cubic initial velocity field
Let us assume that the flow domain at the initial time is a wedge with an apex angle
$\alpha$
. The initial velocity field is a cubic field, i.e.
$U(z,0) = A z^3,\ (A\gt 0).$
We have to determine the flow domain and flow velocity at subsequent times. Following the previous considerations, we seek the complex velocity and conformal mapping functions in the form
The functions
$f(\zeta )$
and
$g(\zeta )$
are sought in the form of series (3.4), where
$n =3.$
We consider two values of the apex angle of the fluid wedge:
$\alpha = \pi /3$
and
$\alpha = \pi /6.$
If
$\alpha = \pi /3$
, then system (3.15) in the case with
$m = 3$
and
$\gamma =1/2$
consists of six nonlinear differential equations, with the method of their integration being not obvious. However, calculating several first coefficients
$f_j$
and
$g_j$
via direct substitution of series (3.4) into the first two boundary conditions of system (3.3) and equating the coefficients at identical powers of
$s$
, we can note some regularities, namely,
These features suggest an idea of introducing new unknown functions
which are related with the initial functions as
For six new unknown functions, we have six differential equations; however, owing to the regularities observed, we can assume that there exists a solution in which
$S_2 = R_1 =R_2 = 0,$
$S_3 = R_3.$
In this case, system (3.15) for the case with
$m = 3$
and
$\gamma =1/2$
is equivalent to the equation
We resolve this equation with respect to
$R_3$
for an arbitrary function
$S_1.$
We obtain a family of solutions, where we are interested only in two functions
Here,
$S_1$
is an arbitrary function of the form
$S_1 = g_1 \zeta + g_4 \zeta ^4 + g_7 \zeta ^7+\ldots ,$
$(g_j \in \mathbb{R}, g_1\gt 0).$
The arising uncertainty is associated with non-uniqueness of the conformal mapping of the wedge in the plane
$\zeta$
onto the domain occupied by the fluid. It should be noted, however, that the solution in the physical plane
$z$
is determined uniquely. Let us demonstrate this fact. According to statement 3,
$g(\zeta ) = P_1,$
$f(\zeta ) = Q_1.$
Therefore,
$g(\zeta ) = S_1 + S_1^3$
and
$f(\zeta ) = S_1^3.$
Taking into account (4.1), we obtain
Let us consider (4.9) as a system and eliminate the function
$S_1$
from it. We obtain the equation that implicitly defines the dependence of the complex velocity on
$z$
and
$t$
Thus, the sought after function of the complex velocity in the plane
$z$
is independent of the choice of the function
$S_1.$
The form of
$S_1$
affects only the conformal mapping. For certainty, let us consider the simplest possible variant and assume that
$S_1 = \zeta .$
In this case, the conformal mapping function has the form
Assuming in (4.11) that
$Z = x + \textrm {i} y$
and
$\zeta = s e^{\textrm {i}\pi /3},$
we obtain a parametrically defined equation of the free boundary in the physical plane (parameter
$s \gt 0$
)
For the function
$U(z,t)$
defined by formula (4.10) to describe a certain fluid flow in reality, it should be analytical in the domain occupied by the fluid. Let us study the singular points of this function. Differentiating the implicitly defined function (4.10), we obtain
Therefore, the singular points of the complex velocity are determined by the system of equations
\begin{equation} \begin{cases} 1+3 A t(z -t U)^2 =0, \\ U - A(z - U t)^3 =0. \end{cases} \end{equation}
Studying system (4.14), we find that there are two singular points in the physical plane:
$z_1^{\ast } = 2 \textrm {i}/(3 \sqrt {3 A t})$
and
$z_2^{\ast } = -2 \textrm {i}/(3 \sqrt {3 A t})$
, which move along the imaginary axis from infinity to the origin of the coordinate system with an increase in
$t.$
However, this information is insufficient to understand whether the singular points are located inside or outside the flow domain. It is more reasonable to consider the positions of the singular points in the auxiliary plane
$\zeta .$
For this purpose, we note that system (4.14) yields
Here,
$z^{\ast }$
is a solution of system (4.14), i.e. a singular point. Assuming in (4.9) that
$S_1 = \zeta$
and taking into account (4.15), we obtain
$\zeta ^{\ast } = 3 \sqrt {A t} z^{\ast }/2.$
Therefore, there are two singular points in the plane
$\zeta$
:
$\zeta ^{\ast } = \pm \textrm {i}/\sqrt {3},$
which are located at the imaginary axis. As the domain occupied by the fluid in the plane
$\zeta$
is fixed, remains unchanged with time and is shaped as a wedge as shown in figure 1 with an angle
$\alpha = \pi /3$
, we can conclude that the singular points are outside the flow domain. This means that function (4.10) is analytical in the domain occupied by the fluid and describes the flow velocity in the fluid wedge. The solution found for
$A =1$
is shown in figure 2. Here, the solid wall at
$y=0$
is replaced by the line of symmetry, and the fluid wedge’s evolution is presented for the wedge with an apex angle of
$120^\circ$
. The domain occupied by the fluid is marked by the grey colour, and there are streamlines in this domain. The singularities are shown by dots. Thus, we found a flow that transforms a wedge with an apex angle of
$120^\circ$
into a plane with a cut within an infinite time.
Evolution of the fluid wedge with an apex angle of
$120^\circ$
.

Evolution of the fluid wedge with an apex angle of
$60^\circ$
.

Let us consider a situation where the apex angle of the fluid wedge is twice smaller, i.e.
$\alpha = \pi /6$
for the same cubic initial velocity field. In this case, system (3.15) consists of 12 nonlinear differential equations. However, the analysis of the first coefficients of series (3.4) obtained by direct substitution of the series into the boundary conditions reveals a large number of zero coefficients. Repeating the procedure proposed for the case with
$\alpha = \pi /3,$
we obtain the following formula for implicit definition of the complex velocity:
It differs only slightly from formula (4.10), which describes the fluid flow in a wedge with an apex angle
$\alpha = \pi /3.$
However, owing to these minor differences, the situation becomes significantly different. The singular points of the function defined by (4.16) have the coordinates
$\zeta ^\ast = \pm 1/\sqrt {3}$
, i.e. one of them is on the solid wall or directly in the flow domain if the solid wall is replaced by the line of symmetry. Thus, we cannot say that formula (4.16) defines an analytical function and describes the fluid flow in the wedge. Nevertheless, the situation can be improved by changing the initial direction of the velocity vector. Indeed, if we set
$U(z,0) = -A z^3$
$(A \in \mathbb{R}, \ A\gt 0)$
at the initial time, the complex velocity is implicitly defined by the equation
In this case, the singular points
$\zeta ^\ast = \pm \textrm {i}/\sqrt {3}$
are again at the imaginary axis, i.e. outside the flow domain. The flow that corresponds to formula (4.17) and describes the evolution of the fluid wedge symmetric with respect to the real axis with an apex angle of
$60^\circ$
is shown in figure 3. Here,
$U(z,0) = - z^3.$
The domain occupied by the fluid is marked by the grey colour, and there are streamlines in this domain. The singularities are shown by dots.
Thus, we found two exact solutions. One of them describes the evolution of the fluid wedge with an apex angle of
$120^\circ$
into a plane with a cut along the half-axis
$y = 0, x \lt 0.$
The second solution transforms the fluid wedge with an apex angle of
$60^\circ$
into a half-plane. The initial velocity has the form
$U(z,0) = A z^3$
in the first case and
$U(z,0) = -A z^3$
$(A \in \mathbb{R}, \ A\gt 0)$
in the second case.
5. General case
Let us demonstrate that flows of this kind also exist in the general case for the initial velocity of the form of (2.1). Let the fluid at the initial time occupy a wedge with an apex angle
$\alpha = \pi /n$
. By means of direct substitution, we can easily show that functions of the form
satisfy the boundary conditions (2.2) at the free boundary
$\zeta = s e^{\textrm {i} \pi /n}.$
Eliminating
$\zeta$
from system (5.1), we obtain an equation for implicit definition of the complex velocity
Let us study the singular points of the function
$U(z,t)$
defined by (5.2). Clearly, they are solutions of the system
\begin{equation} \begin{cases} 1 + n A t(z -t U)^{n-1} =0, \\ U - A(z - U t)^n =0. \end{cases} \end{equation}
Let us express
$(z - t U)^{n-1}$
from the first equation of system (5.3) and substitute into the second equation. We find that
at the singular point. Substituting (5.4) into the first equation of system (5.3), we obtain an expression for determining the positions of the singular points in the physical plane
It follows from system (5.1) that
$\zeta = (n z \sqrt [n-1]{A t})/(n-1).$
Taking into account (5.5), we obtain an equation for determining the positions of the singular points in the auxiliary plane
$\zeta$
Thus, we find that the plane
$\zeta$
contains
$(n-1)$
singular points of function (5.2). All of them are located on a circumference of radius
$(\sqrt [n-1]{n})^{-1}$
and have the arguments
The smallest argument is
$\pi /(n-1).$
As the flow domain in the plane
$\zeta$
corresponds to a wedge with an apex angle
$\pi /n$
, all singularities are outside the fluid; hence, the function
$U(z,t)$
defined by (5.2) is holomorphic in the domain occupied by the fluid and defines an exact solution that describes the transformation of a fluid wedge with an apex angle
$2 \pi /n$
into a plane with a cut along the real half-axis owing to inertial motion caused by the initial velocity field
$U(z ,0)= A z^n.$
Qualitatively, these flows coincide with the flow shown in figure 2 for the case with
$n = 3.$
It can be shows in a similar manner that there exists an exact solution defined by the formula
which describes the evolution of a fluid wedge with an apex angle
$\pi /n$
into a half-plane owing to inertial motion caused by the initial velocity field
$U(z, 0) = - A z^n \ (A\gt 0).$
Qualitatively, these flows coincide with the flow shown in figure 3 for the case with
$n = 3.$
Thus, we derived exact solutions that allow the fluid wedge with an apex angle
$\pi /n$
to be transformed into a plane with a cut or a half-plane by changing the initial velocity field.
It should be noted that all solutions obtained in the study have an interesting feature: the initial velocity field is arranged in such a way that one of the velocity vector coordinates at the free boundary is equal to zero. However, this condition is not sufficient for the exact solution to exist. As an opposite situation, we can consider the case discussed above, where
$U(z,0) = A z^3,$
$\alpha = \pi /6.$
In this case, we have
Therefore, for points at the free boundary
$z = s e^{\textrm {i}\pi /6}$
, the velocity vector has the form
$(0, - A s^3),$
but it was shown above that no solution exists in this case because the domain occupied by the fluid contains singular points of the complex velocity function.
It should be noted that, although the method cannot be applied to construct a solution for a linear initial velocity field, it is still possible to construct an exact solution for the fluid flow in the angle at
$U(z, 0 ) = Az$
. From the dimensional considerations, the velocity field for this solution remains linear over
$z$
for any value of
$t$
:
$U(z, t) = A(t) z.$
The free boundary remains a ray:
$y(t) = \tan (\alpha (t)) x.$
It can be easily shown that the sought after functions
$A(t)$
and
$\alpha (t)$
satisfy the following system of ordinary differential equations:
\begin{equation} \begin{cases} \dfrac {{\rm d} A}{{\rm d} t} \cos (2\alpha ) + A^2 =0, \\[10pt] \dfrac {{\rm d} \alpha }{{\rm d} t} + A \sin (2\alpha )=0,\\[8pt] A(0)= A_0,\\ \alpha (0)=\alpha _0. \end{cases} \end{equation}
Studying system (5.10), we obtain two different types the flow depending on the initial angle
$\alpha _0$
.
-
(i) If
$0\lt \alpha _0 \lt \pi /4,$
the value of the angle decreases with time
$\alpha (t) \rightarrow 0$
as
$t \rightarrow \infty$
, and
$A(t) \sim ({1}/{t})$
in this case. The solution exists for any
$t$
. -
(ii) If
$\pi /4\lt \alpha _0 \lt \pi /2$
the solution is broken within a finite time owing to an infinite increase in the flow velocity. In this case, we have
$\alpha (t) \rightarrow \pi /4.$
As noted above, in situations where
$n$
is a real number that does not belong to the set of natural numbers, the method proposed in the paper cannot be applied. Nevertheless, an exact solution can be obtained using the idea proposed in Karabut et al. (Reference Karabut, Zhuravleva, Zubarev and Zubareva2022). Let the fluid at the initial time occupy an angle
$\alpha = \pi /n$
and
$U(z,0) =z^n.$
For the chosen relationship between the initial angle and initial velocity, the ray corresponding to the free boundary in the physical plane
$z$
is mapped onto the negative semi-axis of the real axis (
$v=0, u\lt 0$
) in the hodograph plane. Thus, the domain occupied by the fluid at the initial time is mapped onto the upper half-plane of the hodograph plane. As was shown in Karabut & Zhuravleva (Reference Karabut and Zhuravleva2014), if the complex velocity satisfies the Hopf equation
$U_t + U U_z =0,$
then the dynamic and kinematic conditions are satisfied at the line
$v =0$
. Therefore, the flow domain in the hodograph plane is fixed for any time instant. The complex velocity satisfies the equation
For the thus-defined function
$U$
to be analytical, the following condition has to be satisfied (Karabut et al. Reference Karabut, Zhuravleva, Zubarev and Zubareva2022):
$U^{1/n} \neq -n t U.$
As can be easily shown, we have
${\rm Im} (U^{1/n}) \lt 0$
for
$n\gt 2$
, which means that the equality cannot be reached for
$v\gt 0.$
Therefore, the function
$U$
is analytical in the domain occupied by the fluid. Let us express
$z$
from formula (5.11) by using the principal branch of the logarithm. We obtain a conformal mapping of the flow domain in the hodograph plane onto the physical plane
$z = U t + U^{1/n}$
. In this case, the free boundary (
$v=0, \ u\lt 0$
) in the plane
$z$
is defined by the equation
$z = - s t + s^{1/n} e^{\textrm {i} \pi /n}, (s\gt 0).$
Qualitatively, this flow resembles the solution shown in figure 2.
6. Conclusions
It is necessary to note some features of the solutions found. A specific feature associated with the fact that the pressure in the fluid is negative was noted both in our publications (Karabut et al. Reference Karabut, Zhuravleva and Zubarev2020) and by other researchers (Liu & Pego Reference Liu and Pego2019).
As the flow domain is unbounded, all solutions found are carrying infinite energy. To avoid this drawback, it is necessary to constrain the flow domain, as it was done, e.g. in Karabut et al. (2022). It should be noted that the known Dirichlet solutions with a linear velocity field also have diverging asymptotic solutions at infinity. Nevertheless, this fact did not prevent their use for modelling the unsteady evolution of gravity waves.
Some researchers say that the solutions found by us are too exotic, the problem solved is too simple and numerical methods allow one to consider more complicated problem formulations with due allowance for, e.g. surface tension. However, exact solutions have their own area of applicability. In addition to decorating our lives, as Landau said, they can be also used for verification of numerical methods; in this case, the ‘exotic’ nature of these solutions can serve as an advantage rather than a drawback. Moreover, although the problem is simple, it has not been yet finally solved; studying available analytical solutions, one can achieve much progress in proving integrability of the Euler equations for problems with a free boundary (Dyachenko et al. Reference Dyachenko, Dyachenko, Lushnikov and Zakharov2019). One of the key issues is the study of singularities located outside the flow domain (Dyachenko et al. Reference Dyachenko, Dyachenko, Lushnikov and Zakharov2021). It should be noted that the singular points of the complex velocity in the physical plane for solutions found in the present study move along the rays fixed in time. This curious fact was noted in all flows of an ideal fluid with a free boundary, as far as the authors are aware.
We would like to mention one more possible application of the solutions found. It was assumed in Longuet-Higgins (Reference Longuet-Higgins1980) that the free surface curvature needed for wave breakdown modelling can be obtained with the use of a complex potential expressed as a polynomial of the complex variable
$z$
of the fourth and higher powers. In the solutions found in the study, the complex velocity function can be a polynomial of the third and higher powers. Thus, we may hope that these solutions will be useful for wave breakdown modelling.
Declaration of interests
The authors report no conflict of interest.




