Hierarchy of Higher Dimensional Integrable System
Extracted source text ยท arXiv solv-int/9802005
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% %
% Hierarchy of Higher Dimensional Integrable System %
% %
% %
% Format: LaTeX %
% Version: 30 January 1998 %
% %
% %
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\documentstyle{ioplppt}
\begin{document}
\def\T{{\cal T}}
\jl{1}
\title{Hierarchy of Higher Dimensional Integrable System}
\author{YU Song-Ju\ftnote{1}{fpc30017@bkc.ritsumei.ac.jp},
Kouichi TODA\ftnote{2}{sph20063@bkc.ritsumei.ac.jp} and
Takeshi FUKUYAMA\ftnote{4}{fukuyama@bkc.ritsumei.ac.jp}}
\address{Department of Physics, Ritsumeikan University, Kusatsu, Shiga 525
JAPAN}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{abstract}
Integrable equations in ($1 + 1$) dimensions have their own higher order
integrable equations, like the KdV, mKdV and NLS hierarchies etc. In this
paper we consider whether integrable equations in ($2 + 1$) dimensions
have also the analogous hierarchies to those in ($1 + 1$) dimensions.
Explicitly is discussed the Bogoyavlenskii-Schiff(BS) equation. For the
BS hierarchy, there appears an ambiguity in the Painlev\'e test.
Nevertheless, it may be concluded that the BS hierarchy is integrable.
\end{abstract}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\maketitle
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\section{Introduction}
The inverse scattering transform(IST) method is a powerful tool for the
investigation of ($1 + 1$) dimensional nonlinear differential equations.
The extent of applicability of this method, however, depends on the
dimension of space. The well known ($2 + 1$) dimensional integrable
equations are the Kadomtsev-Pettviashivili(KP), Davey-Stewarson(DS)
equation and some new integrable equations which has been constructed by
B.G.Konopelchenko and V.G.Dubrovsky\cite{kd}. In this paper we consider
a new integrable equation in (2+1) dimensions, the Bogoyavlenskii-Schiff
equation (BS equation) \cite{b1,jschiff}. This equation is given by
\begin{equation}
u_t + \Phi(u) u_z = 0,
\label{bs}
\end{equation}
where
\begin{equation}
\Phi(u) \equiv \frac{1}{4} \partial^2_x + u + \frac{1}{2} u_x
\partial^{-1}_x
\label{recursion}
\end{equation}
is the recursion operator of KdV equation\cite{ff}.
Potential form of this equation takes the form,
\begin{equation}
\phi_{xt} + \frac{1}{4} \phi_{xxxz} + \phi_x \phi_{xz} +
\frac{1}{2} \phi_{xx} \phi_z = 0 \hspace{0.5cm} (u \equiv \phi_x).
\label{pbs}
\end{equation}
The BS equation has an infinite number of conservation laws\cite{b3}, the
Painlev\'e property\cite{jschiff,b4} and $N$ soliton
solutions\cite{ytsf}. That is, the BS equation is the ($2 + 1$)
dimensional integrable equation. For ($1 + 1$) dimensional case($z = x$),
i.e. KdV equation, we can construct the higher order equations of the
KdV(the KdV hierarchy) with Lax operators and the conserved
quantities\cite{l,zf}. In this paper, we consider whether this
construction of hierarchy is also valid for ($2 + 1$) dimensional
integrable equation or not. Namely the purpose of the present paper is to
construct the BS hierarchy and to check its integrability through the
Painlev\'e test.
In Sec.2, we review the hierarchy of an integrable equation in ($1 + 1$)
dimensions and derive the Lax pair of the BS hierarchy. In Sec.3, we
investigate the integrability of the BS hierarchy obtained in the previous
section. Sec.4 is devoted to discussion mainly concerning with integrable
equations in ($3 + 1$) dimensions.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\section{The Lax Representation of The BS Hierarchy}
The potential KdV equation
\begin{equation}
\phi_{xt} + \frac{1}{4} \phi_{4x} + \frac{3}{2} \phi_x \phi_{xx} = 0 ,
\label{pkdV}
\end{equation}
admits the Lax representation\cite{LNPm10}
\begin{equation}
[L,T] = 0.
\label{laxeq1}
\end{equation}
$L$, $T$ operators have the form
\begin{eqnarray}
& &L(x,t) = \partial_x^2 + \phi_x(x,t), \label{L} \\
& &T(x,t) = \biggl(L(x,t)^{\frac{3}{2}}\biggr)_+ + \partial_t =
\partial_x L(x,t) + T'(x,t) + \partial_t,
\label{T}
\end{eqnarray}
where $(~~)_+$ means the part of $(~~)$ with non-negative powers of
$\partial_x$ and
\begin{eqnarray}
L(x,t)^{\frac{1}{2}} &=& \partial_x + \frac{1}{2} \phi_x \partial_x^{-1}
- \frac{1}{4} \phi_{xx} \partial_x^{-2} + \frac{1}{8} (\phi_{3x} -
\phi_x^2) \partial_x^{-3}
\nonumber \\
& & + \frac{3}{8} \biggl(\phi_x \phi_{xx} - \frac{1}{2} \phi{4x}\biggr)
\partial_x^{-4} + \cdots,
\label{L1/2} \\
T'(x,t) &=& \frac{1}{2} \phi_x \partial_x - \frac{1}{4} \phi_{xx}.
\label{T'}
\end{eqnarray}
The second equality in equation (\ref{T}) is given for the later use of
the BS equation and others.
An infinite set of equations,
\begin{equation}
[L,T_n] = 0 \hspace{0.5cm} (n = 1, 2, 3, \cdots),
\label{laxhieeq1}
\end{equation}
constitute the potential KdV hierarchy and the equation for $n$ is called
the $(2 n + 1)$th order potential KdV equation. The operator $T_n$ has
the form
\begin{equation}
T_n(x,t) = \biggl(L(x,t)^{\frac{2 n + 1}{2}}\biggr)_+ + \partial_t =
\partial_x L(x,t)^n + T'_n(x,t) + \partial_t.
\label{Tn}
\end{equation}
For $n = 2$ and $n = 3$, equation(\ref{laxhieeq1}) represents the $5$th
and the $7$th order potential KdV equations, respectively:
\begin{equation}
\fl \phi_{xt} + \frac{1}{16} \phi_{6x} + \frac{5}{8} \biggl(\phi_x^3 +
\frac{1}{2} \phi_{xx}^2 + \phi_x \phi_{xx}\biggr)_x = 0
\label{5kdv}
\end{equation}
and
\begin{equation}
\fl T'_2 = \frac{1}{2} \phi_z \partial_x^3 - \frac{1}{4} \phi_{xx}
\partial_x^2 + \biggl(\frac{1}{8} \phi_{3x} + \frac{7}{8} \phi_x^2\biggr)
\partial_x - \frac{1}{16} \phi_{4x} - \frac{1}{8} \phi_x \phi_{xx}
\label{T'2}
\end{equation}
for $n = 2$.
\begin{equation}
\fl \phi_{xt} + \frac{1}{64} \phi_{8x} + \frac{7}{64} \biggl(5 \phi_x^4
+ 10 \phi_x (\phi_{xx}^2 + \phi_x \phi_{3x}) + 2 \phi_x \phi_{5x} +
4 \phi_{xx} \phi_{4x} + 3 \phi_{3x}^2\biggr)_x = 0
\label{7kdv}
\end{equation}
and
\begin{eqnarray}
\fl & & T'_3 = \frac{1}{2} \phi_x \partial_x^5 - \frac{1}{4} \phi_{xx}
\partial_x^4 + \biggl(\frac{1}{8} \phi_{3x} + \frac{11}{8}
\phi_x^2\biggr) \partial_x^3 - \biggl(\frac{1}{16} \phi_{4x} -
\frac{9}{8} \phi_x \phi_{xx}\biggr) \partial_x^2 + \biggl(\frac{1}{32}
\phi_{5x} \nonumber \\
\fl & &- \frac{11}{32} \phi_{xx}^2 + \frac{16}{31} \phi_x \phi_{3x} +
\frac{19}{16} \phi_x^3\biggr) \partial_x - \frac{1}{64} \phi_{6x} -
\frac{9}{32} \phi_x \phi_{4x} - \frac{7}{16} \phi_{xx} \phi_{3x} +
\frac{9}{32} \phi_x^2\phi_{xx}
\label{T'3}
\end{eqnarray}
for $n = 3$. Equations (\ref{5kdv}) and (\ref{7kdv}) are
integrable\cite{matsuno}.
The Bogoyavlenskii's method is to modify the form of $T$
operator(\ref{T}) in ($1 + 1$) dimensions to in ($2 + 1$) dimensions as
\begin{equation}
T(x,z,t) = \partial_z L(x,z,t) + \partial_t + T'(x,z,t),
\label{BogoT}
\end{equation}
where
\begin{equation}
T'(x,z,t) = \frac{1}{2} \phi_z \partial_x - \frac{1}{4} \phi_{xz}.
\label{BogoT''}
\end{equation}
Namely we replace $\partial_x L$ in equation(\ref{T}) by $\partial_z L$.
Here $L$ has the same form as equation(\ref{L}), though $\phi$, in this
case, is a function of $x$, $z$ and $t$. Then, the Lax
equation(\ref{laxeq1}) becomes the BS equation (\ref{pbs})\cite{b1}. We
can construct the ($2 n + 1$)th order BS hierarchy analogously to the
potential KdV hierarchy:
\begin{equation}
[L,T_n] = 0,
\label{Bogolaxhie}
\end{equation}
where $T_n$ have the forms
\begin{equation}
T_n(x,z,t) = \partial_z L(x,z,t)^n + T'_n(x,z,t) + \partial_t.
\label{tn}
\end{equation}
We could construct the same equations for the Lax
equation(\ref{Bogolaxhie}) with other forms of the operator
$T_n(x,z,t) = \underbrace{L \cdots L}_{\mbox{$m - 1$ times}}
(\partial_z L) \underbrace{L \cdots L}_{\mbox{$n - m$ times}} +
T''_n(x,z,t) + \partial_t$. Here the operators $T''_n(x,z,t)$ is
different from $T'_n(x,z,t)$, though we have the same equation after
all. For $n = 2$ and $n = 3$, equations(\ref{Bogolaxhie}) represents the
$5$th and the $7$th order BS equations respectively:
\hspace{-0.55cm}For $n = 2$, equation(\ref{Bogolaxhie}) becomes
\begin{eqnarray}
\fl &\phi_{xt}& + \frac{1}{16} \phi_{xxxxxz} + \frac{1}{2} \phi_x
\phi_{xxxz} + \frac{3}{4} \phi_{xx} \phi_{xxz} + \frac{1}{2} \phi_{xxx}
\phi_{xz} + \frac{1}{8} \phi_{xxxx} \phi_z + \phi_x^2 \phi_{xz}
\nonumber \\
\fl &+& \frac{3}{4} \phi_x \phi_{xx} \phi_z + \frac{1}{8} \phi_{xx}
\partial_x^{-1} (\phi_x^2)_z = 0,
\label{Bogo5th}
\end{eqnarray}
and
\begin{eqnarray}
\fl T'_2 &=& \frac{1}{2} \phi_z \partial_x^3 - \frac{1}{4} \phi_{xz}
\partial_x^2 + \biggl(\frac{1}{8} \phi_{xxz} + \frac{3}{4} \phi_x \phi_z
+ \frac{1}{8} \partial_x^{-1} (\phi_x^2)_z\biggr) \partial_x -
\frac{1}{16} \phi_{xxxz} \nonumber \\
\fl & & - \frac{1}{2} \phi_x \phi_{xz} + \frac{3}{8} \phi_{xx} \phi_z.
\label{BogoT'2}
\end{eqnarray}
\hspace{-0.55cm}In this case, the above equation(\ref{Bogo5th}) is
equivalent to that obtained by Bogoyavlenskii\cite{b1,b2}. For $n = 3$,
equation(\ref{Bogolaxhie}) takes the form
\begin{eqnarray}
\fl &\phi_{xt}& + \frac{1}{64} \phi_{xxxxxxxz} + \frac{3}{16} \phi_x
\phi_{xxxxxz} + \frac{15}{32} \phi_{xx} \phi_{xxxxz} + \frac{5}{8}
\phi_{xxx} \phi_{xxxz}
\nonumber \\
\fl &+& \frac{15}{32} \phi_{xxxx} \phi_{xxz} + \frac{3}{16} \phi_{xxxxx}
\phi_{xz} + \frac{1}{32} \phi_{xxxxxx} \phi_z + \frac{9}{4} \phi_x
\phi_{xx} \phi_{xxz}
\nonumber \\
\fl &+& \frac{9}{8} \phi_{xx}^2 \phi_{xz} + \frac{3}{4} \phi_x^2
\phi_{xxxz} + \frac{3}{2} \phi_x \phi_{xxx} \phi_{xz} + \frac{5}{16}
\phi_x \phi_{xxxx} \phi_z + \frac{5}{8} \phi_{xx} \phi_{xxx} \phi_z
\nonumber \\
\fl &-& \frac{1}{32} \phi_{xx} \partial_x^{-1} (\phi_{xx}^2)_z +
\frac{1}{16} \phi_{xxxx} \partial_x^{-1} (\phi_x^2)_z + \phi_x^3
\phi_{xz} + \frac{15}{16} \phi_x^2 \phi_{xx} \phi_z
\nonumber \\
\fl &+& \frac{1}{16} \phi_{xx} \partial_x^{-1} (\phi_x^3)_z +
\frac{3}{16} \phi_x \phi_{xx} \partial_x^{-1} (\phi_x)_z = 0,
\label{Bogo7th}
\end{eqnarray}
and
\begin{eqnarray}
\fl T'_3 &=& \frac{1}{2} \phi_z \partial_x^5 - \frac{1}{4} \phi_{xz}
\partial_x^4 + \biggl(\frac{1}{8} \phi_{xxz} + \frac{5}{4} \phi_x
\phi_{z} + \frac{1}{8} \partial_x^{-1} (\phi_x^2)_z\biggr) \partial_x^3
\nonumber \\
\fl & & - \biggl(\frac{1}{16} \phi_{xxxz} + \frac{3}{4} \phi_x \phi_{xz}
- \frac{15}{8} \phi_{xx} \phi_z\biggr) \partial_x^2 + \biggl(\frac{1}{32}
\phi_{xxxxz} - \frac{5}{16} \phi_{xx} \phi_{xz}
\nonumber \\
\fl & & + \frac{3}{8} \phi_x \phi_{xxz} - \frac{1}{32} \partial_x^{-1}
(\phi_{xx})_z + \frac{25}{16} \phi_{xxx} \phi_{z} + \frac{15}{16}
\phi_x^2 \phi_z + \frac{1}{16} \partial_x^{-1} (\phi_x^3)_z
\nonumber \\
\fl & & + \frac{3}{16} \phi_x \partial_x^{-1} (\phi_x^2)_z\biggr)
\partial_x - \frac{1}{64} \phi_{xxxxxz} - \frac{3}{16} \phi_x
\phi_{xxxz} - \frac{1}{16} \phi_{xx} \phi_{xxz} - \frac{3}{8} \phi_{xxx}
\phi_{xz}
\nonumber \\
\fl & & + \frac{15}{32} \phi_{xxxx} \phi_z - \frac{3}{4} \phi_x^2
\phi_{xz} + \frac{15}{16} \phi_x \phi_{xx} \phi_z + \frac{3}{32}
\phi_{xx} \partial_x^{-1} (\phi_x^2)_z.
\label{BogoT'3}
\end{eqnarray}
Equations(\ref{Bogo5th}) and (\ref{Bogo7th}) are, of course, reduced to
the $5$th and $7$th order potential KdV equations in the case of $z = x$,
respectively. The calculations for $n \ge 4 $ can be performed similarly,
but the procedure to determine $T'_n$ operators for larger $n$ becomes
more complicated.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\section{Application of The Painlev\'e Test}
We have constructed the BS hierarchy in the previous section. In this
section we proceed to check the integrability of the BS
hierarchy(\ref{Bogolaxhie}) through the Painlev\'e test in the sense of
WTC\cite{wtc,t}. For the BS hierarchy, there appears an ambiguity in this
test. Hence we give the detail of calculation, which may help to
understand what it is. In order to eliminate the integral operator, we
rewrite the $5$th order BS equation(\ref{Bogo5th}) in the form of coupled
systems,
\begin{eqnarray}
&\phi_{xt}& + \frac{1}{16} \phi_{xxxxxz} + \frac{1}{2} \phi_x
\phi_{xxxz} + \frac{3}{4} \phi_{xx} \phi_{xxz} + \frac{1}{2} \phi_{xxx}
\phi_{xz}
\nonumber \\
& & + \frac{1}{8} \phi_{xxxx} \phi_z - \rho_{xx} \phi_{xz} + \frac{3}{4}
\phi_x \phi_{xx} \phi_z - \frac{1}{8} \phi_{xx} \rho_{xz} = 0,
\label{Bogo5th1} \\
&\rho_{xx}& + \phi_x^2 = 0.
\label{Bogo5th2}
\end{eqnarray}
We now consider a local Laurent expansion in the neighborhood of a
non-characteristic singular manifold $\gamma(x,z,t) = 0$,
($\gamma_x \ne 0$). Let us assume that the leading orders of the
solutions of equations(\ref{Bogo5th1}) and (\ref{Bogo5th2}) have the
forms
\begin{equation}
\phi = \phi_0 \gamma^{\alpha}, \hspace{0.5cm} \rho = \rho_0
\gamma^{\beta}.
\label{leadingsols}
\end{equation}
Here $\phi_0$ and $\rho_0$ are some analytic functions. Substituting
(\ref{leadingsols}) into (\ref{Bogo5th1}) and (\ref{Bogo5th2}), and
equating the powers of the most dominant terms, we obtain
\begin{equation}
\alpha = -1, \hspace{0.5cm} \beta = -2
\label{leadingorder}
\end{equation}
with
\begin{equation}
\phi_0 (\phi_0 - 6 \gamma_x) (\phi_0 - 2 \gamma_x) = 0, \hspace{0.5cm}
6 \rho_0 + \phi_0^2 = 0.
\label{leadingterm}
\end{equation}
To find the resonances we now substitute the full Laurent expansion of
the solutions
\begin{equation}
\phi = \sum_{j = 0} \phi_j \gamma^{j - 1}, \hspace{0.5cm} \rho =
\sum_{j = 0} \rho_j \gamma^{j - 2},
\label{Painlevesols}
\end{equation}
into equations(\ref{Bogo5th1}) and (\ref{Bogo5th2}). Rearranging
equation(\ref{Bogo5th1}) into terms of $\gamma^{j - 7}$ and the other
higher powers of $\gamma$ ($f_j$), we obtain recurrence relations for
$\phi_j$, $\rho_j$,
\begin{eqnarray}
& & \biggl(\bigl[(j - 1) (j - 2) (j - 3) (j - 4) (j - 5) (j - 6)
\gamma_x^5 \gamma_z
\nonumber \\
& & \hspace{0.6cm} - 10 (j - 1) (j - 6) (j^2 - 7 j + 16) \gamma_x^4
\gamma_z \phi_0
\nonumber \\
& & \hspace{0.6cm} - 108 (j - 1) (j - 2) \gamma_x^3 \gamma_z \rho_0 +
12 (j - 1) (j - 6) \gamma_x^3 \gamma_z \phi_0^2\bigr] \phi_j
\nonumber \\
& & \hspace{0.6cm} - 36 (j - 2) (j - 3) \gamma_x^3 \gamma_z \phi_0
\rho_j\biggr) \gamma^{j - 7} = f_j.
\label{gammaj-71}
\end{eqnarray}
Similarly rearranging equation(\ref{Bogo5th2}) into terms of
$\gamma^{j - 4}$ and higher powers of $\gamma$ ($g_j$), we have
\begin{equation}
\bigl(- 2 (j - 1) \gamma_x^2 \phi_0 \phi_j + (j - 2) (j - 3) \gamma_x^2
\rho_j\bigr) \gamma^{j - 4} = g_j.
\label{gammaj-72}
\end{equation}
Here $f_j$ and $g_j$ are given in terms of $\phi_l$ and $\rho_l$
($0 \le l \le j - 1$). Then we get three types of resonances:
\begin{eqnarray}
& & j = 0, 1, 1, 2, 3, 4, 5, 6,
\label{resonance1} \\
& & j = -1, 1, 2, 2, 3, 5, 6, 8
\label{resonance2}
\end{eqnarray}
and
\begin{equation}
j = -3, -1, 1, 2, 3, 6, 8, 10.
\label{resonance3}
\end{equation}
These correspond to the choices of solutions of equations
(\ref{leadingterm}),
\begin{eqnarray}
& & \phi_0 = 0, \hspace{0.5cm} \rho_0 = 0,
\label{sol1} \\
& & \phi_0 = 2 \gamma_x, \hspace{0.5cm} \rho_0 = - (2/3) \gamma_x^2
\label{sol2}
\end{eqnarray}
and
\begin{equation}
\phi_0 = 6 \gamma_x, \hspace{0.5cm} \rho_0 = - 6 \gamma_x^2,
\label{sol3}
\end{equation}
respectively. Here the first choice (\ref{sol1}) does not include the
universal resonance $j = - 1$ and is discarded. If the recurrence
relations are consistently satisfied at the resonances then the
differential equations is said to possess the Painlev\'e property.
Succeeding coefficients $\phi_j$ and $\rho_j$ are determined from
equations (\ref{gammaj-71}) and (\ref{gammaj-72}). However, from the
consistency condition, they must include arbitrary functions at the
resonances. To simplify the calculations, we use the reduced manifold
ansatz of Kruskal\cite{kjh}. That is, $\gamma(x,z,t) = x + \delta(z,t)$
and $\phi_j$, $\rho_j$ are function of $z$ and $t$.
Equations (\ref{gammaj-71}) and (\ref{gammaj-72}) must be satisfied in
the respective powers of $\gamma$. The lowest powers of $\gamma$,
($\gamma^{-7}$,$\gamma^{-4}$) in equations((\ref{gammaj-71}),
(\ref{gammaj-72})) give
\begin{equation}
\phi_0 (\phi_0 - 2) (\phi_0 - 6) = 0, \hspace{0.5cm} \rho_0 = -
\frac{1}{6} \phi_0^2,
\label{gammam7m4}
\end{equation}
which constitutes a set of equations(\ref{sol1}), (\ref{sol2}), and
(\ref{sol3}). Higher powers of $\gamma$,($\gamma^{-7 + k}$,
$\gamma^{-4 + k}$ with positive integer $k$) in
equations((\ref{gammaj-71}),(\ref{gammaj-72})) lead us to the following
consistency conditions.
\begin{equation}
\fl (\gamma^{-6}, \gamma^{-3}) :~~~~~~\phi_0 \rho_1 \delta_z = 0,
\hspace{0.5cm} \rho_1 = 0.
\label{gammam6m3}
\end{equation}
Hence $\phi_1$ is arbitrary.
\begin{equation}
\fl (\gamma^{-5}, \gamma^{-2}) :~~~~~~\phi_0 \bigg[(\phi_0 - 2)
\phi_{1,z} + 2 (\phi_0 - 5) \phi_2 \delta_z\bigg] = 0,
\hspace{0.5cm} \phi_0 \phi_2 = 0,
\label{gammam5m2}
\end{equation}
These equations imply that $\phi_0$ is fixed as $2$ and $\rho_2$ is
arbitrary, which contradicts equation(\ref{sol3}). So we must choose
(\ref{resonance2}).
\begin{equation}
\fl (\gamma^{-4}, \gamma^{-1}) :~~~~~~(3 \phi_0 - 20) \phi_3 \delta_z = 0,
\hspace{0.5cm} \phi_3= 0,
\label{gammam4m1}
\end{equation}
and $\rho_3$ is arbitrary.
\begin{equation}
\fl (\gamma^{-3}, \gamma^0) :~~~~~~6 (13 \phi_4 - 3 \rho_4) \delta_z +
8 \delta_t - \rho_{3,z} = 0, \hspace{0.5cm} \rho_4 = 6 \phi_4.
\label{gammam30}
\end{equation}
\begin{equation}
\fl (\gamma^{-2}, \gamma^1) :~~~~~~9 (8 \phi_5 - 3 \rho_5) \delta_z +
(6 \phi_4 - \rho_4)_z = 0, \hspace{0.5cm} \rho_5 = \frac{8}{3} \phi_5
\label{gammam21}
\end{equation}
and
\begin{equation}
\fl (\gamma^{-1}, \gamma^2) :~~~~~~36 (5 \phi_6 - 3 \rho_6) \delta_z +
(8 \phi_5 - 3 \rho_5)_z = 0, \hspace{0.5cm} \rho_6 = \frac{5}{3} \phi_6.
\label{gammam12}
\end{equation}
It follows from equations (\ref{gammam21}) and (\ref{gammam12}) that one of
the two variables must be arbitrary in both pairs ($\phi_5$, $\rho_5$) and
($\phi_6$, $\rho_6$).
\begin{equation}
\fl (\gamma^0, \gamma^3) :~~~~~~12 (13 \phi_7 - 15 \rho_7) \delta_z -
(5 \phi_6 + 4 \rho_6)_z - 6 \phi_5 \phi_{1,z} = 0, \hspace{0.5cm}
\rho_7 = \frac{6}{5} \phi_7.
\label{gamma03}
\end{equation}
\begin{eqnarray}
\fl & & (\gamma^1, \gamma^4) :~~~~~~\bigg[(504 \phi_8 - 540 \rho_8) +
72 \phi_4^2\bigg] \delta_z + 3 \phi_4 (- 18 \rho_4 \delta_z +
8 \delta_t - \rho_{3,z}) \nonumber \\
\fl & & \hspace{2.5cm} + 2 (6 \phi_7 - 5 \rho_7)_z = 0, \nonumber \\
\fl & & \hspace{2.2cm} \rho_8 = \frac{14}{15} \phi_8 -
\frac{3}{10} \phi_4^2.
\label{gamma14}
\end{eqnarray}
That is, one of the two variables ($\phi_8$, $\rho_8$) must be arbitrary.
Problematic is the following point: $\phi_2$ at double root $j = 2$ has
been fixed as $0$ and the system lacks one more arbitrary function.
However, $\phi_j$ and $\rho_j$ are determined consistently and, thus,
dependent variables $\phi$ and $\rho$ have only movable poles. The
apparently similar situation occurred in the system of coupled non-liner
Sch\"odinger equation(NLS)\cite{stl}. Starting from NLS leaving the
constant coefficients as free parameters in \cite{stl}, Sahadevan et al.
these parameters so as to include $N$ arbitrary functions at $N$-tuply
degenerate resonance. However, there seems to exist an essential
difference between their NLS and our $5$th order BS equation since NLS
does not allow the consistent Laurent expansion except for the special
choices of constant coefficients above mentioned. In the $5$th order BS
equation there is no such room to adjust the number of arbitrary function,
though the $5$th order BS equation has no inconsistency. We want to add
one remark here. The coupled system of equations (\ref{Bogo5th1}) and
(\ref{Bogo5th2}) is reduced to the $5$th order potential KdV equation in
the case of $z = x$. Even in that case, the resonance structures (the
positions and the number of arbitrary functions etc.) are quite the same
as the $5$th order BS equation. The $5$th order potential KdV equation is
manifestly integrable. From these facts we may conclude that the $5$th
order BS equation is also integrable.
Analogously, we have checked that the $7$th order BS equation is integrable
in the same sense.
Higher order equation can be constructed in several different ways. All
the higher order BS equations(\ref{Bogolaxhie}) are rewritten as
\begin{equation}
u_t + \Phi(u)^n u_z = 0
\label{recBogolaxhie}
\end{equation}
by the use of the recursion operator of KdV equation(\ref{recursion}).
Equation(\ref{recBogolaxhie}) was conjectured to be
integrable\cite{jschiff} because the integrability arose from the
existence of the recursion operator\cite{fs}. However, the recursion
operator in ($2 + 1$) dimensions is not so clear as in the case of
($1 + 1$) dimensions. Explicit calculation in this section has uncovered
a new feature and has shown that the BS hierarchy is integrable.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\section{Discussion}
In this paper, we have constructed the BS hierarchy(\ref{Bogolaxhie})
using the Bogoyavlenskii's method. This hierarchy is represented by the
recursion operator(\ref{recursion}) and reduced to the potential KdV
hierarchy by setting $z = x$. The BS hierarchy(\ref{Bogolaxhie}),
(\ref{recBogolaxhie}) has been proved to be integrable through the
Painlev\'e test. The BS equation is an extension of the KdV equation
like the KP equation is. However their extensions differ in algebraic
structure, which is schematically depicted in Figure.\ref{extensions}.
\begin{figure}[h]
\begin{center}
\begin{picture}(100,80)
\put(-100,70){\framebox(25,10){KdV}}
\put(-60,76){\vector(2,0){70}}
\put(28,70){\framebox(25,10){BS}}
\put(68,76){\vector(2,0){70}}
\put(155,70){\dashbox{2.0}(25,10){X3}}
\put(-120,50){\makebox(25,10){(\ref{KPL})}}
\put(-88,65){\vector(0,-1){25}}
\put(-38,60){\makebox(25,10){(\ref{BogoT})}}
\put(40,65){\vector(0,-1){25}}
\put(-100,26){\framebox(25,10){KP}}
\put(-60,32){\vector(2,0){70}}
\put(28,26){\dashbox{2.0}(25,10){X1}}
\put(-88,21){\vector(0,-1){25}}
\put(-100,-18){\dashbox{2.0}(25,10){X2}}
\end{picture}
\end{center}
\caption{Scheme of extensions of the KdV equation. There are two
directional routes of extensions: One leads us to the BS equation and the
other does to the KP equation. X equations are possible ($3 + 1$)
dimensional integrable equations obtained from further extensions.}
\label{extensions}
\end{figure}
\hspace{-0.41cm} The potential KP equation is obtained from the potential
KdV equation by the replacement,
\begin{eqnarray}
L(x,t) \longrightarrow L'(x,y,t) &=& \partial_x^2 + \phi_x(x,y,t) +
\partial_y \nonumber \\
&\equiv& L(x,y,t) + \partial_y,
\label{KPL}
\end{eqnarray}
The form of $T(x,y,t)$ is
\begin{equation}
T(x,y,t) = \partial_x L(x,y,t) + T'(x,y,t) + \partial_t,
\label{KPT}
\end{equation}
where
\begin{equation}
T'(x,y,t) = \frac{1}{2} \phi_x \partial_x - \frac{1}{4} \phi_{xx} -
\frac{3}{4} \phi_y.
\label{KPT'}
\end{equation}
Here we have denoted an extended spatial coordinate as y. These
extensions are contrasted with those of the BS equation(\ref{BogoT})
in which explicit $y$ (in that case $z$) dependence has appeared not in
$L$ but in $T$.
These two extensions may be performed in the different spatial
dimensions (i.e. $y \ne z$), which is described by X1 in
Figure.\ref{extensions}. For instance, we consider the Lax
equation(\ref{laxeq1}) with the operators $L'$ and $T$ having the form
\begin{eqnarray}
& &L'(x,y,z,t) = \partial_x^2 + \phi_x(x,y,z,t) + \partial_y
\equiv L(x,y,z,t) + \partial_y,
\label{newL} \\
& &T(x,y,z,t) = \partial_z L(x,y,z,t) + T'(x,y,z,t) + \partial_t,
\label{newT}
\end{eqnarray}
where
\begin{equation}
T'(x,y,z,t) = \frac{1}{2} \phi_z \partial_x - \frac{1}{4} \phi_{xz}
- \frac{1}{4} \partial_x^{-1} \phi_{yz} + c \partial_z^2
\label{newT'}
\end{equation}
and $c$ is a constant. However, the Lax equation(\ref{laxeq1}) with
operators(\ref{newL}) and (\ref{newT}) in ($3 + 1$) dimensions is reduced
to ($2 + 1$) dimensional equation,
\begin{equation}
\phi_{xt} + \frac{1}{4} \phi_{xxxz} + \phi_x \phi_{xz} + \frac{1}{2}
\phi_{xx} \phi_z + c^2 \partial_x^{-1} \phi_{zzz} = 0
\label{neweqfty}
\end{equation}
with the condition,
\begin{equation}
z = 2 c y.
\label{reducedcondition}
\end{equation}
This condition comes from the requirement that the coefficient of
$\partial_z$ should be vanished in the Lax equation and, therefore, seems
to be indispensable. Equation(\ref{neweqfty}) in ($2 + 1$) dimensions is
the same as that obtained by Bogoyavlenskii\cite{b1}. We have checked
that equation(\ref{neweqfty}) passes the Painlev\'e test. By the
dependent variable transformation $\phi = 2 (\log \tau)_x$, equation
(\ref{neweqfty}) is transformed into the trilinear form,
\begin{equation}
(36 \T_x^2 \T_t + \T^4_x \T^{\ast}_z + 8 \T_x^3 \T^{\ast}_x \T_z
+ 9 \T_z^3) \tau \cdot \tau \cdot \tau = 0.
\label{trineweqfty}
\end{equation}
The operators $\T$, $\T^{\ast}$ are defined by\cite{grh,hgr}
\begin{equation}
\fl \T_z^n f(z) \cdot g(z) \cdot h(z) \equiv (\partial_{z_1} +
j \partial_{z_2} + j^2
\partial_{z_3})^nf(z_1)g(z_2)h(z_3)|_{z_1=z_2=z_3=z}, \label{Tdef}
\end{equation}
where $j$ is the cubic root of unity, $j = \exp(2 \i \pi/3)$.
$\T^{\ast}_z$ is the complex conjugate operator of $\T_z$ obtained
by replacing $(\partial_{z_1} + j \partial_{z_2} + j^2
\partial_{z_3})$ by $(\partial_{z_1} + j^2 \partial_{z_2} + j
\partial_{z_3})$. Equation(\ref{trineweqfty}) coincides with the
trilinear form coming from the quite different approach\cite{hgr}.
Namely, Hietarinta, Grammaticos and Ramani derived this equation from the
singularity analysis of trilinear equations. We have obtained $N$ soliton
solutions of equation(\ref{trineweqfty}) by the computer program
{\it Mathematica} up to $N = 5$. However, we can not construct the
hierarchy of equation(\ref{neweqfty}). Namely we can not determine
$T'(x,y,z,t)$ operator in
\begin{equation}
T(x,y,z,t) = \partial_z L(x,y,z,t)^n + T'(x,y,z,t) + \partial_t,
\label{newThie}
\end{equation}
so as to constitute Lax equation(\ref{laxeq1}).
A little bit different formulation to reach X1 candidate was proposed by
us\cite{ytsf},
\begin{equation}
\biggl(u_t + \Phi(u) u_z\biggr)_x + \frac{3}{4} u_{yy} = 0.
\label{newoldeq}
\end{equation}
Unfortunately, equation (\ref{newoldeq}) has a movable logarithmic
branch point.
One of the examples for ($3 + 1$) dimensional X3 equations in
Figure.\ref{extensions} is obtained also by Bogoyavlenskii\cite{b4}.
However, it is not affirmative unless it is proved to be integrable.
The respective routes in Figure.\ref{extensions} are not unique.
Different formulations at each path are possible, though we can not
specify which is correct at the present stage. Such specification
enforces us to clarify the algebraic structures and integrability of the
equations depicted in Figure.\ref{extensions} and their hierarchies
furthermore.
\ack
We would like to thank Professor T. Kawahara for useful discussions.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\section*{References}
\begin{thebibliography}{99}
\bibitem{kd} B. G. Konopelchenko and V. G. Dubrovsky 1984 {\it Phys. Lett.}
{\bf 102A} 15
\bibitem{b1} O. I. Bogoyavlenskii 1990 {\it Math. USSR. Izv} {\bf 34} 245
\bibitem{jschiff} J. Schiff 1992 {\it Painlev\'e Transendents, Their
Asymtopics and Physical Applications} (Plenum) 393
\bibitem{ff} B. Fuchssteiner and A. S. Fokas 1981 {\it Physica} {\bf 4D} 47
\bibitem{b3} O. I. Bogoyavlenskii 1991 {\it Math. USSR. Izv} {\bf 36} 129
\bibitem{b4} O. I. Bogoyavlenskii 1991 {\it Math. USSR. Izv} {\bf 37} 475
\bibitem{ytsf} Yu. S, K. Toda, N. Sasa and T. Fukuyama solv-int/9801003
\bibitem{l} P. D. Lax 1968 {\it Comm. Pure Appl. Math.} {\bf 21} 467
\bibitem{zf} V. E. Zakharov and L. D. Faddeev 1971 {\it Func. Anal. Appl.}
{\bf 5} 280
\bibitem{LNPm10} J. Hoppe 1992 {\it Lec. Notes in Phys.} {\bf m 10} {\it
Lectures on Integrable Systems} (Springer-Verlag) 73
\bibitem{matsuno} Y. Matsuno 1980 {\it J. Phys. Soc. Jpn.} {\bf 48} 2138
\bibitem{b2} O. I. Bogoyavlenskii 1990 {\it Math. USSR. Izv} {\bf 35} 245
\bibitem{wtc} J. Weiss, M. J. Tabor and G. Carnevale 1983 {\it J. Math.
Phys.} {\bf 24} 522
\bibitem{t} M. J. Tabor 1988 {\it Chaos and Integrability in Nonlinear
Dynamics} (John Wiley \& Sons) 322
\bibitem{kjh} M. D. Kruskal, N. Joshi and R. Halburd solv-int/9710023
\bibitem{stl} R. Sahadevan, K. M. Tamizhmani and M. Lakshmanan 1986 {\it
J. Phys. A: Math. Gen.} {\bf 19} 1783
\bibitem{fs} A. S. Fokas and P. M. Santini 1990 {\it Solitons in Physics,
Mathematics and Nonlinear Optics} (Springer-Verlag) 79
\bibitem{grh} B. Grammaticos, A. Ramani and J. Hietarinta 1994 {\it Phys.
Lett. A} {\bf 190} 65
\bibitem{hgr} J. Hietarinta, B. Grammaticos and A. Ramani solv-int/9411003
\end{thebibliography}
\end{document}