A Classification of Integrable Hydrodynamic Type Chains Using The Haantjes Tensor

Classification of integrable hydrodynamic
chains using the Haantjes tensor.
D.G. Marshall.
Department of Mathematical Sciences

Loughborough University
Loughborough, Leicestershire LE11 3TU
United Kingdom
1
Keywords 2
Keywords
Integrability
Nijenhuis tensor
Haantjes tensor
Hydrodynamic chains
Benney Chain
Diagonalizability
Generating function of conservations laws
Abstract 3
Abstract
The integrability of an m-component system of hydrodynamic type, ut = v(u)ux ,

by the generalized hodograph method requires the diagonalizability of the m × m
matrix v(u). The diagonalizability is known to be equivalent to the vanishing of
the corresponding Haantjes tensor. This idea is applied to hydrodynamic chains —
infinite-component systems of hydrodynamic type for which the ∞ × ∞ matrix v(u)
is ‘sufficiently sparse’. For such ‘sparse’ systems the Haantjes tensor is well-defined,
and the calculation of its components involves only a finite number of summations.
The calculation of the Haantjes tensor is done by using Mathematica to perform
symbolic calculations. Certain conservative and Hamiltonian hydrodynamic chains
are classified by setting Haantjes tensor equal to zero and solving the resulting sys-
tem of equations. It is shown that the vanishing of the Haantjes tensor is a necessary
condition for a hydrodynamic chain to possess an infinity of semi-Hamiltonian hy-
drodynamic reductions, thus providing an easy-to-verify necessary condition for the
integrability of such sysyems. In the cases of the Hamiltonian hydrodynamic chains
we were able to first construct one extra conservation law and later a generating
function for conservation laws, thus establishing the integrability.
Acknowledgements 4
Acknowledgements
I would like to express my gratitude to the following people for their assistance
during my work on this thesis:
Prof. E. V. Ferapontov and Dr K. R. Khusnutdinova, who gave me such superb and

enthusiastic supervision along with unwavering support and encouragement. Thank
you for all the time you have given me and all the patience you have shown me
throughout this work.
All the of staff of the Mathematical Sciences Department at Loughborough Univer-

sity for all their help.
Loughborough University, particularly the Mathematical Sciences Department, for

financial support.
All my friends for their moral support and understanding.

Contents 5
Contents
1 Introduction 7
2 Hydrodynamic type systems in 1+1 dimensions 15

2.1 Riemann invariants . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.2 Commutativity Condition . . . . . . . . . . . . . . . . . . . . . . . . 22
2.3 Conservation Laws . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.4 Semi-Hamiltonian property . . . . . . . . . . . . . . . . . . . . . . . . 26
2.5 Generalized hodograph method . . . . . . . . . . . . . . . . . . . . . 28
2.6 The Haantjes tensor . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.7 Reciprocal Transformations . . . . . . . . . . . . . . . . . . . . . . . 34
3 Hydrodynamic Chains 38
3.1 The Benney chain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.2 Hydrodynamic chains and the Haantjes tensor . . . . . . . . . . . . . 43
3.3 Another chain considered by Benney . . . . . . . . . . . . . . . . . . 53
3.4 Hydrodynamic reductions and diagonalizability . . . . . . . . . . . . 54
4 Conservative chains 65
4.1 Egorov case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.2 Method of hydrodynamic reductions . . . . . . . . . . . . . . . . . . 67
4.3 Integration of the Egorov case . . . . . . . . . . . . . . . . . . . . . . 72
4.4 More general conservative chains . . . . . . . . . . . . . . . . . . . . 76
4.5 Integration of the general conservative chains . . . . . . . . . . . . . . 79
5 Kupershmidt’s brackets 89
5.1 Integration of the system of integrability conditions . . . . . . . . . . 91
5.2 The vanishing of the Haantjes tensor . . . . . . . . . . . . . . . . . . 93
Contents 6
5.3 Conservation laws . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96

5.4 Generating functions for conservation laws of Hamiltonian chains cor-
responding to Kupershmidt’s bracket . . . . . . . . . . . . . . . . . . 98
5.4.1 Generating functions in the general case . . . . . . . . . . . . 100
5.4.2 Generating functions in the case β = 0 . . . . . . . . . . . . . 100
5.4.3 Generating functions in the case α + 2β = 0 . . . . . . . . . . 101
6 Manin-Kupershmidt’s bracket 103

6.1 Integration of the system of integrability conditions . . . . . . . . . . 106
6.2 Conservation laws for Manin-Kupershmidt’s bracket . . . . . . . . . 117
6.3 Generating functions for conservation laws of Hamiltonian chains with
Manin-Kupershmidt’s bracket . . . . . . . . . . . . . . . . . . . . . . 119
7 Conclusions and further work 125

Introduction 7
1 Introduction
This thesis is a study in the area of integrable systems. It is devoted to a classification

of integrable hydrodynamic chains of the type that orginally appeared in the context
of fluid mechanics, in particular with reference to the Benney chain [3]. Integrable
systems have been actively investigated since the discovery of the Inverse Scattering
Transform of the KdV equation [21]. There have been significant developments for
1 + 1 dispersive equations, see [2], and 1 + 1 dimensional dispersionless equations
[9]. Classification of integrable systems has been an important theme of much of
the recent research.
In this thesis we are concerned with the integrability of special infinite component
dispersionless systems known as hydrodynamic chains. Our approach towards clas-
sification of such systems is based on the use of the so-called Haantjes tensor [27].
Let us begin by defining the form of finite component hydrodynamic type systems
ut = v(u)ux (1)
where u = (u1 , u2 , ..., un )t is an n-component column vector and v(u) is an n × n

matrix. In what follows we consider the strictly hyperbolic case when the eigenval-
ues of the matrix v(u), also called the characteristic speeds of the system (1), are
real and distinct.
Some systems of the form (1) are diagonalizable, i.e. reducible to the Riemann
invariant form, more details follow in Sect. 2. The diagonalizability of (1) is a
necessary condition for integrability via the generalized hodograph method [61].
There exists an efficient tensor criterion of the diagonalizability which does not
require the computation of eigenvalues and eigenvectors of the matrix v(u). This is
Introduction 8
useful because calculating the eigenvalues and eigenvectors can be troublesome in

practice. Let us first calculate the Nijenhuis tensor [27] of the matrix v(u) = vji ,
i
Njk = vjp ∂up vki − vkp ∂up vji − vpi (∂uj vkp − ∂uk vjp ), (2)
and introduce the Haantjes tensor [27]
i i p r p i r p i r p i r
Hjk = Npr vj vk − Njr vp vk − Nrk vp vj + Njk vr vp . (3)
For strictly hyperbolic systems, i.e. when the eigenvalues are positive and distinct,
the condition of diagonalizability is given by the following theorem:
Theorem 1 [27] A hydrodynamic type system with mutually distinct characteris-

tic speeds is diagonalizable if and only if the corresponding Haantjes tensor (3) is
identically zero.
The focus of this work is hydrodynamic chains, these are infinite component systems
of form (1). More formally a hydrodynamic chain is,
ut = V (u)ux
where V (u) is a square matrix of infinite dimension, while u = (u1 , u2 , . . . , )t is

an infinite-component vector. If the infinite matrix V (u) is ’sufficiently sparse’ then
both tensors (2) and (3) make sense. The formal definition of ’sufficiently sparse’ is
Definition 1 [20] An infinite matrix V (u) is said to belong to the class C (chain
class) if it satisfies the following two properties:
(a) each row of V (u) contains finitely many nonzero elements;
(b) each matrix element of V (u) depends on finitely many variables ui .
Introduction 9
All the chains considered throughout the paper are of the class C. Below is the
Benney chain written in matrix form, the matrix is of class C, it has only two
nonzero elements in every row apart from the first where there is only one. The sole
element in the first row is constant while of the two elements on every other row
one is constant and the other is only a function of one variable. The Benney chain
has the form
    
A0 0 1 0 0 0 0 . A0
    
 1   0  1 
 A   A 0 1 0 0 0 .  A 
 
   
    
 A2  +  2A1 0 0 1 0 0 .   A2  = 0.
    
    
 3 
 3A2
 3 
 A  0 0 0 1 0 .  A 

    
· . . . . . . . ·
t x
The Benney chain is a classical example of a hydrodynamic chain. This chain was
derived in [3] and was shown to consist of an infinite number of conservation laws
for the system of equations that describe fluid under the action of gravity. The
derivation of the Benney chain is shown in Sect. 3.1.
In Sect. 3 we propose that a hydrodynamic chain is said to be diagonalizable if all

components of the corresponding Haantjes tensor (3) are zero. The work in Sect.
3 goes on to show that the vanishing of the Haantjes tensor is a necessary (and in
some cases — sufficient) condition for a hydrodynamic chain to possess an infinity
of finite-component diagonalizable hydrodynamic reductions. The main advantage
of our approach is that the classification does not require any extra objects such
as commuting flows, Hamiltonian structures, Lax pairs, etc. The vanishing of the
Haantjes tensor turns out to be an easy to calculate classification criterion. As an
Introduction 10
example let us consider the chain
unt = un+1
x + u1 unx + cn un u1x
where cn = const. It can be verified that the vanishing of the Haantjes tensor implies
the relation cn+2 = 2cn+1 − cn . Since the form of this chain is relatively simple it
is straightforward to calculate the Haanjtes tensor by hand though it is quicker to
using symbolic computations. Throughout this thesis the Mathematica [43] package
has been used, the three files contained in the Appendices can be downloaded from
http://www-staff.lboro.ac.uk/%7emakk/Marshall.html. The PDF file of this theis
is also available online at the fore mentioned address.
Definition 2 A hydrodynamic chain from the class C is said to be integrable if,

for any m, it possesses infinitely many m-component semi-Hamiltonian reductions
parametrised by m arbitrary functions of a single variable.
The method of hydrodynamic reductions is illustrated in Sect. 3.4 by considering

the Benney chain.
In Sect. 3.4 the following theorem is proved:
Theorem 2 The vanishing of the Haantjes tensor H is a necessary condition for

the integrability of hydrodynamic chains from the class C.
If the spectrum of the infinite matrix V is simple, that is, for any λ there exists a
unique eigenvector ξ such that V ξ = λξ, the following stronger result is obtained:
Theorem 3 In the simple spectrum case the vanishing of the Haantjes tensor H is
necessary and sufficient for the existence of two-component reductions parametrized
by two arbitrary functions of a single variable.
Introduction 11
Theorem 3 provides an easy-to-verify necessary condition for testing the integrabil-

ity of hydrodynamic chains. For finite component systems of hydrodynamic type, it
is known that any diagonalizable semi-Hamiltonian (38) system possesses infinitely
many conservation laws and commuting flows of hydrodynamic type, and can be
solved by the generalized hodograph method [61]. At this point it must be empha-
sized that the vanishing of the Haantjes tensor is not sufficient for the integrability in
general: one can construct examples of diagonalizable chains which possess infinitely
many diagonal reductions none of which are semi-Hamiltonian (38) (see the example
in Sect. 3.4). To eliminate these cases let us recall that for finite-component systems
(1) there exists a tensor object which is responsible for the semi-Hamiltonian prop-
s
erty [48]. This is a (1, 3)-tensor Pkij (see Appendix 1 for explicit formulas in terms
of the matrix vji ). Similarly to the Haantjes tensor H, the tensor P is well-defined
for hydrodynamic chains that are of class C. Thus the following conjecture is made
Conjecture 1 The vanishing of both tensors H and P is necessary and sufficient

for the integrability of hydrodynamic chains from the class C.
The necessary part of this conjecture, the claim that the integrability implies that
both H and P vanish identically, is proved in Sect. 3.4. The sufficiency is a much
more difficult property to show, and this is not yet established in general. We point
out that the vanishing of H (in fact, the vanishing of the first few components of
H), is already sufficiently restrictive and implies the integrability in many cases (e.g.
for conservative chains, Hamiltonian chains, etc).
In Sect. 4 the Haantjes tensor criterion is used to classify conservative chains.

Initially we restrict the flux of the first equation to be only u2 , this is known as the
Egorov case. The chain is
u1t = u2x , u2t = g(u1 , u2 , u3 )x , u3t = h(u1 , u2 , u3 , u4 )x , ... . (4)

Introduction 12
The chain (4) was investigated in [47], where a class of new integrable examples
was found, see also [6], based on the symmetry approach. These papers provide a
classification of conservative chains of the form (4). It turns out that the condi-
1
tions Hjk = 0 are already sufficiently restrictive and give an over-determined system
expressing all second order partial derivatives of h(u1 , u2 , u3 , u4 ) in terms of g, see
(103). The consistency conditions of these equations lead to a system of equations
expressing all third order partial derivatives of g in terms of its lower order deriva-
tives, see (104). Computer algebra was used to calculate the Haantjes tensor and to
verify the involutivity by calculating the consistency conditions. It must be empha-
sized that the same system of equations for g was found in [47] using the symmetry
approach, as well as in [17] based on the method of hydrodynamic reductions. So,
for conservative chains (4) the condition of diagonalizability is equivalent to the
i
integrability. The requirement of the vanishing of other components Hjk , i ≥ 2,
imposes no additional constraints on h and g: these conditions reconstruct the re-
m
maining equations of the chain (4). Furthermore, the conditions Hjk = 0 specify
the right hand side of the mth equation um
t = ..., etc.
Also contained in Sect. 4 is the classification of conservative chains where the flux
of the first equation is f (u1 , u2 ), i.e. chains of the type
u1t = f (u1 , u2 )x , u2t = g(u1 , u2 , u3 )x , u3t = h(u1 , u2 , u3 , u4 )x , ... . (5)
1
As in the Egorov case, the conditions Hjk = 0 lead to expressions for all second
order partial derivatives of h in terms of g and f . The consistency conditions of
these equations result in a system of equations expressing all third order partial
derivatives of g and f in terms of lower order derivatives.
Sect. 5 and Sect. 6 are devoted to the classification of Hamiltonian chains which
Introduction 13
are of the form

d d ∂h
ut = B + Bt , (6)
dx dx ∂u
where u = (u1 , u2 , u3 , ...)t is an infinite-component column vector of the dependent

d d
variables, and B dx + dx B t is known as the Hamiltonian operator. Chains of the form
(6) are known as Hamiltonian chains, h is called the Hamiltonian density. We are
going to look at the problem of the classification of the densities for two particular
Hamiltonian operators:
• Kupershmidt’s bracket: B ij = (α(i − 1) + β)ui+j−1 , h = h(u1 , u2 ), see [34];
• Manin-Kupershmidt’s bracket: B ij = (i − 1)ui+j−2 , h = h(u1 , u2 , u3 ), see [33];
so that the corresponding Hamiltonian system is integrable. The Kupershmidt’s

bracket will be dealt with in Sect. 5, while the work on the Manin-Kupershmidt’s
bracket will be in Sect. 6.
According to the results of Tsarev [61], the vanishing of the Haantjes tensor is nec-
essary and sufficient for the integrability of finite-component Hamiltonian systems
of hydrodynamic type by the generalized hodograph method. Thus, we formulate
our main conjecture regarding Hamiltonian hydrodynamic chains:
Conjecture 2 [20] The vanishing of the Haantjes tensor is a necessary and suffi-
cient condition for the integrability of Hamiltonian hydrodynamic chains. In par-
ticular, it implies the existence of infinitely many Poisson commuting conservation
laws, and infinitely many hydrodynamic reductions.
The necessity part of this conjecture follows from Theorem 2, which is proved in
Sect 3.4, which gives the vanishing of the Haantjes tensor as a necessary condition
Introduction 14
for the integrability of hydrodynamic chains (not necessarily Hamiltonian). The suf-
ficiency is more difficult to establish. The conjecture is supported by all examples
of integrable Hamiltonian chains known to us. Since components of the Haantjes
tensor can be calculated using computer algebra, this approach provides an effective
classification criterion. We are going to demonstrate that the conjecture is indeed
true for Hamiltonian chains of the type (6) with Kupershmidt’s bracket and Manin-
Kupershmidt’s bracket, respectively.
We also investigate the existence of higher order (than the hamiltonian, h) con-
servation laws for the two types of Hamiltonian systems in question; recall that
the existence of infinitely many conservation laws is one of the main features of
the integrability. We show that the requirement of the existence of one additional
conservation law of the form,
• p(u1 , u2 , u3 )t = q(u1 , u2 , u3 , u4 )x for Hamiltonian systems with Kupershmidt’s

bracket;
• p(u1 , u2 , u3 , u4 )t = q(u1 , u2 , u3 , u4 , u5 )x for Hamiltonian systems with Manin-

Kupershmidt’s bracket;
1
leads to the same relations as requiring Hjk = 0 for the respective systems. Further-
1
more the system of equations obtained from setting Hjk = 0, imply the existence
of a generating function of conservation laws and, hence, an infinity of conservation
laws. This establishes the integrability of all examples constructed in Sect. 5 and
Sect. 6.
The main results of this thesis were published in [18], [20].

Hydrodynamic type systems in 1+1 dimensions 15
2 Hydrodynamic type systems in 1+1 dimensions
An m-component system of hydrodynamic type has the form,
ut = v(u)ux , (7)
where v(u) is a m×m matrix and u is a column vector with components u1 , u2 , u3 , . . . ,

um . The integrability of these systems by the generalized hodograph method re-
quires that the matrix v(u) is diagonalizable, or in other words possesses Riemann
invariants. This section contains information on hydrodynamic type systems in 1+1
dimension. We are interested only in systems of hyperbolic type, this means all
eigenvalues of matrix v in (7) must be real and distinct.
2.1 Riemann invariants
Consider (7), such a system possesses Riemann invariants if there exists a change of
variables
u1 , . . . , un → R1 (u), . . . , Rn (u) (8)
so that (7) goes to

Rti = λi (R)Rxi . (9)
If (7) possesses Riemann invariants then it is said that the matrix v(u) is diagonal-
izable.
To find Riemann invariants of a 2x2 system, proceed as follows:

1. Write the equations in the form (7) and find the eigenvalues of matrix v(u),
det v(u) − λi I = 0.

2. Require the gradients of the Riemann invariants are left eigenvectors of matrix
v corresponding to eigenvalues λ1 , λ2 .
grad Ri v(u) − λi I = 0.

3. Solve the resulting two pairs of PDE’s to obtain the Riemann invariants
R1 , R2 ,
∂R1 ∂R1 1

( , ) v(u) − λ I = 0,
∂u1 ∂u2
∂R2 ∂R2
) v(u) − λ2 I = 0.

( 1, 2
∂u ∂u
An example follows to further highlight the steps to calculate Riemann Invariants.
Example Calculating Riemann invariants for the equations of gas dynamics
ρt + ux ρ + uρx = 0, ut + uux + γργ−2 ρx = 0. (10)
Write (10) in matrix form
    
 ρ   u ρ  ρ 
  +    = 0. (11)
γ−2
u γρ u u
t x
Comparing (11) to (9) we see that
     
 u1   ρ  u ρ 
 =  , vji =  .


u2 u γργ−2 u
In step one we find the eigenvalues of vji ,
λ1,2 = u ± (γργ−1 )1/2 . (12)
In step two we find the Riemann invariants by requiring,

 
 u − λ1 ρ

Rρ1 Ru1  = 0,


γ−2
γρ u − λ1
the same is done for λ2 , R2 . In the third step, multiply out and solving we see that
1 2γ 1/2 ρ(γ−1)/2
R =u+ .
γ−1
We obtain R2 by the same method, ultimately,
γ−1 γ−1
1 2γ 1/2 ρ 2 2 2γ 1/2 ρ 2
R =u+ , R =u− . (13)
γ−1 γ−1
So we have equation (9) for i = 1, 2, thus
Rt1 + λ1 Rx1 = 0, Rt2 + λ2 Rx2 = 0. (14)
Note at this point that from (13) we have
1 γ−1
R1 + R2 1 4γ 2 ρ 2
2
= u, R −R = . (15)
2 γ−1
Substitute (15) into (12) to obtain
R1 + R2 (γ − 1)(R1 − R2 ) R1 + R2 (γ − 1)(R2 − R1 )
λ1 = + , λ2 = + . (16)
2 4 2 4
If (16) is substituted into (14) then we have
R1 + R2 (γ − 1)(R1 − R2 )

Rt1+ + Rx1 = 0,
2 4
1 2
(γ − 1)(R2 − R1 )

2 R +R
Rt + + Rx2 = 0. (17)
2 4
One can verify that (11) indeed goes to (17) under the change of variables
1 γ−1
R1 + R2 1 4γ 2 ρ 2
2
= u, R −R = .
2 γ−1
Thus, R1 and R2 are Riemann invariants of (10).
Example [38] This is a much more involved example of determining the Riemann
invariants for a given system. Equations of Chromatography are
uix + ai (u)t = 0, i = 1, · · · , n. (18)
Different models are obtained via the choice of ai (u), called isotherms of adsorption.
In this example we are going to obtain the Riemann invariants for the Langmuir
isotherm, thus,
n
i Γi ui X
a (u) = , ∆=1+ Γs us . (19)
∆ s=1
First, we calculate the eigenvalues of matrix A where
uix + Aij ujt = 0, (20)
It was shown in [38] that the characteristic equation, det(A − λI) = 0, has the form,
n
!
∆Γ1 − Γ21 u1 − α1 ∆2 X Γ2s un (∆λ1 − Γ1 )
det(A − λI) = +
∆2 s=2
∆2 (Γs − ∆λs )
Γ2 − ∆λ2 Γn − ∆λn

...
∆ ∆
n
! n
Γ1 − ∆λ1 (∆λ1 − Γ1 ) X Γ2s un Γp − ∆λp
Y
= + +
∆ ∆2 Γ − ∆λs
s=1 s p=2
∆
n n n
Γp − ∆λp 1 Y Γs − ∆λs X Γ2s us
Y
= − ,
p=1
∆ ∆ s=1 ∆ s=1
Γs − ∆λs
so,
n
X Γ2s us
∆= , (21)
s=1
Γs − ∆λs
since, det(A − λI) = 0. Thus, substituting into (19) we obtain,
n n
X
s
X Γ2s us
1+ Γs u = (22)
s=1 s=1
Γs − ∆λ
The eigenvalues, λi , of A are the solutions of (22).
We now show that Riemann invariants are given by Ri = λi ∆. It is required (9) is

satisifed modulo Ri = λi ∆. Take R = λ∆ in (22), calculate Rx , Rt in terms of ujt ,
R
and then substitue to show Rx + R
∆ t
= 0.
X Γs R Γs us p upt Γp Γs us
P
t
Rx = − 2
− ,
s
Γs − R ∆ ∆
P Γs ust
s Γs −R
Rt = −R P up Γ2p
.
p (Γp −R)2
Substitue into (9) to obtain,
n n n
X Γ2s ust X p
X Γs ust
− + Γp ut + R = 0, (23)
Γ − R p=1
s=1 s
Γ −R
s=1 s
R
one can see that coefficients at ukt cancel out. Thus, Rx + ∆ Rt = 0 modulo R = λ∆,
as required.
Now, we show the diagonal form of (20) is
Qn
Rs i
Rxi i
+ R Qs=1
n Rt = 0. (24)
s=1 Γs
Using the result that Ri = λi ∆ in (22) we have
n
X Γ2s us
= ∆,
s=1
Γs − R
write as a polynomial
n
Y n
Y n
Y
2 1 2 n
∆ (Γs − R) − Γ1 u (Γs − R) − · · · − Γn u (Γs − R) = 0.
s=1 s6=1 s6=n
The coefficient of Rn is (−1)n ∆. The coefficient of R0 is
n n
! n
Y X Y
i
∆ Γs − Γi u Γs = 0,
s=1 i=1 s=1
so, the polynomial can be written in the from,
Q
Γs
Rn + · · · + (−1)n s
= 0.
∆
It can also be written in the form (R − R1 ) · · · (R − Rn ) = 0, so by Viet’s theorem,

equating the free term gives
n Q
n
Y
s n Γs
s
(−1) R = (−1) ,
s=1
∆
Γs
Q
thus, ∆ = s Rs , so we have (24) as required.
Now it is claimed that the formulas that connect Riemann invariants Ri with vari-
ables ui are,
Γi
Q
1 s6=q Rs
−1
ui = Q Γi
. (25)
Γi s6=i Γs
−1
Indeed, this can be shown to be true as follows. From (24) we know that
n
i Ri i
Y Rs
λ = =R .
∆ s=1
Γs
Qn Γs
This implies that ∆ = s=1 Rs . Next, take (21) and substitute (25) in to eliminate
uk , obtaining an expression for ∆, in terms of Γ’s and R’s:
n n
Γi s ( RΓis − 1)
Q
X Γ2s us X
∆ = =
Γ − Rk (Γi − Rk ) s6=i ( ΓΓsi − 1)
Q
s=1 s i=1
n Qn n Q s
Γi − Rs s6=k Γi − R
Y Γs X Y Γs X
s=1Q
= =
Rs i−1 (Γi − Rk ) ns6=i (Γi − Γs )
Q
s s
Rs i=1 s6=i (Γi − Γs )
Y Γs
= s
.
s
R
So, (25) is consistent as it gives the correct expression for ∆ when substituted into
(21). Thus, it has been shown that the system (19) has Riemann invariants and the
diagonal form is given in (24).
2.2 Commutativity Condition
For two systems to commute means that they are consistent with each other. Take
two systems of type (7), ut = V (u)ux , uy = W (u)ux , for them to commute means
uty = uyt .
Theorem 4 Let us consider two systems of the form (9),
Rti = λi (R)Rxi , Ryi = µi (R)Rxi , i = 1 . . . n, (26)
here t and y are the corresponding ’times’. It is claimed that for these equations to
be consistent then the following condition must be true,
∂j λi ∂j µi
= , i 6= j. (27)
(λj − λi ) (µj − µi )
∂
Where ∂j = ∂Rj
.
Proof For the equations to be consistent it is required that,
i i
Rty = Ryt . (28)
So, let us calculate the necessary condition for (28) to be satisfied.
i
Rty = (λi Rxi )y = ∂j λi Ryj Rxi + λi Rxy
i
= ∂j λi µj Rxj Rxi + λi (µi Rxi )x = ∂j λi µj Rxj Rxi + λi ∂j µi Rxj Rxi + λi µi Rxx

i
; (29)
i
Ryt = (µi Rxi )t = ∂j µi Rtj Rxi + µi Rxt
i
= ∂j µi (λj Rxj )Rxi + µi (λi Rxi )x
= ∂j µi λj Rxj Rxi + µi ∂j λi Rxj Rxi + µi λi Rxx

i
. (30)
From substituting (30) and (29) into (28) one obtains
∂j λi µj Rxj Rxi + λi ∂j µi Rxj Rxi + λi µi Rxx

i
= ∂j µi λj Rxj Rxi + µi ∂j λi Rxj Rxi + µi λi Rxx
i
,
i
coefficients of Rxx cancel, leaving
∂j λi (µj − µi ) = ∂j µi (λj − λi ).
Which gives (27). (27) shall be referred to as the commutativity condition [61]. Fur-
thermore, if (27) is satisified then the systems (26) are said to be commuting flows.
Example Let us calculate commuting flows for the system,

Rt1 = R2 Rx1 ,
Rt2 = R1 Rx2 ,
in the form,
Ry1 = µ1 Rx1 , (31)
Ry2 = µ2 Rx2 . (32)
Substituting into (27) gives
1 ∂2 µ1 1 ∂1 µ2
= , = . (33)
R1 − R2 µ2 − µ1 R2 − R1 µ1 − µ2
Look for µ1 , µ2 in the form,
µ1 (R1 , R2 ) = Af1 + Bf + Cg, µ2 (R1 , R1 ) = Dg2 + Ef + F g. (34)
A, B, C, D, E, F are certain functions of R1 , R2 to be determined, while f = f (R1 ), g =

g(R2 ) are arbitrary functions.
Thus the general solution is, µ1 = (R2 − R1 )f 0 + f − g and µ2 = (R2 − R1 )g 0 + f − g,
which can be verified by substituting into (33).
2.3 Conservation Laws
Consider a diagonal system of type (9). A conservation law is a relationship
[f (R)]t = [g(R)]x (35)

which must hold identically by the virtue of (9). The quantities f (R) and g(R) are
called the conserved density and the flux, respectively. By substituting (9) into (35)
we see that
∂i f λi Rxi = ∂i gRxi
so that ∂i f λi = ∂i g for any i. For consistency we need
∂j ∂i g = ∂i ∂j g.
From this condition we obtain
∂j λi ∂i λj
∂i ∂j f = ∂i f + ∂j f, i 6= j.
λj − λi λi − λj
Example Consider the system
Rt1 = R2 Rx1 , Rt2 = R1 Rx2 , (36)
here λ1 = R2 , λ2 = R1 . The equation (36) for conserved densities takes the form
1 1 ∂1 f − ∂2 f
∂2 ∂1 f = ∂1 f + 2 ∂2 f = 1 . (37)
R1 −R 2 R −R 1 R − R2
Its general solution is

p(R1 ) − q(R2 )
f= ;
R1 − R2
which can be verified by a straightfoward differentiation. Here p(R1 ) and q(R2 ) are
arbitrary functions of one variable.
Remark The condition (35) is equivalent to the 1-form f dx+gdt being closed. This
can be seen by applying the differential d to dF = f dx + gdt,
d2 F = fx d2 x + ft dxdt + gx dtdx + gt d2 t
d2 F = (ft − gx )dxdt.
Note, d2 F is identically zero, thus ft = gx .
2.4 Semi-Hamiltonian property
The system (9) is said to be semi-Hamiltonian [61] if
∂j λi ∂k λi

∂k = ∂j (38)
λj − λi λk − λi
∂j λi
for any i 6= j 6= k 6= i. Introducing aij = λj −λi
one can rewrite the previous equation
in the simpler form
∂k aij = ∂j aik . (39)
From equation (39) it can be shown that
∂k aij = aij ajk − aij aik + aik akj . (40)
Proposition [61]. If (38) is satisfied then commuting flows depend on n arbitrary

functions of one argument.
Proof Consider commuting flows of (9),
∂j µi = aij (µj − µi ), ∂k µi = aik (µk − µi ). (41)

Work out ∂j ∂k µi and ∂k ∂j µi and use (40) and (41) to eliminate ∂n µm and ∂m anp ,
∂k ∂j µi = (aij ajk − aij aik + aik akj )(µj − µi )
+aij (ajk (µk − µj ) − aik (µk − µi )), (42)
∂j ∂k µi = (aik akj − aik aij + aij ajk )(µk − µi )
+aik (akj (µj − µk ) − aij (µj − µi )). (43)
Now subtract (43) from (42) and collect coefficients of µi , µj , µk . It is straightfoward

to determine that the coefficients of µi , µj , µk are equal to 0. We know all mixed
partial derivatives from (41), we also now know that all the partial derivatives are
consistent identically in µi , µj , µk . Thus commuting flows depend on n arbitrary
functions of one argument, the functions of integration. Indeed, µ1 can be defined
arbitrarily on the R1 - axis, since we know ∂j µ1 ∀j 6= 1, µ2 can be defined arbitrarily
on the R2 -axis, etc.
Proposition [61]. If (38) is satisfied then conserved densities depend on n arbitrary

functions of one argument.
Proof Consider
∂j ∂i f = aij ∂i f + aji ∂j f,
∂k ∂i f = aik ∂i f + aki ∂k f, (44)
compute ∂k ∂j ∂i f and ∂j ∂k ∂i f , substitute (44) to eliminate all terms of the form

∂m ∂n f . One obtains
∂k ∂j ∂i f = ∂k aij fi + aij (aik fi + aki fk ) + ∂k aji fj + aji (ajk fj + akj fk ),
∂j ∂k ∂i f = ∂j aik fi + aik (aij fi + aji fj ) + ∂j aki fk + aki (akj fk + ajk fj ). (45)

Set these last two equations equal to each other and collect coefficients of fi , fj , fk .
The coefficients of fi immediately cancel to give 0, while the coefficients of fj , fk are
0 due to (40). So we know all mixed partial derivatives of all f , we also know that
all these partial derivatives are consistent. Thus conserved densities depend on n
arbitrary functions of one variable, indeed one can define f arbitrarily on any of the
coordinate lines.
2.5 Generalized hodograph method
The generalized hodograph method allows the general solution of a system of the
form (7) to be found. First we will demonstrate with a scalar example and then give
a more general outline of the generalized hodograph method. Consider the scalar
Hopf equation
Rt = RRx . (46)
It is known that the general solution of (46) is given by the implicit formula
µ(R) = x + Rt (47)
where f (R) is an arbitaray function. Indeed, if we differentiate (47) by x and t

respectively and solve for Rx and Rt :
1
µ0 R x = 1 + R x t ⇒ R x = ,
µ0
−t
R
µ0 R t = R + R t t ⇒ R t = 0 . (48)
µ −t
By substituting (48) into (46) we obtain the identity. Thus, (47) is the general
solution.
Theorem 5 [61] The general solution of the diagonal system
Rti = λi (R)Rxi (49)
is given by
µi (R) = x + λi (R)t. (50)
where λi (R) satisifes (38) and µi (R) are characteristic speeds of commuting flows:
∂j µi ∂j λi
= j i 6= j. (51)
µj − µi λ − λi
Proof Differentiate (50) by x and t respectively,
∂j µi Rxj + ∂i µi Rxi = ∂j λi Rxj t + ∂i λi Rxi t,
∂j µi Rtj + ∂i µi Rti = ∂j λi Rtj t + ∂i λi Rti t + λi . (52)
By substituting (50) into (51) and rearranging one obtains
∂j µi = ∂j λi t. (53)
Now substitute (53) into (52) to obtain
1 λi
Rxi = , Rti = . (54)
∂i µ − ∂i λi t
i ∂i µi − ∂i λi t
Thus by substituing (54) into (49) we obtain the identity so (50) is the general so-
lution. General solution means that (50) has the same amount of freedom as the
system (46) being solved. Considering the initial value problem, t = 0, of (49), all
Ri can be defined arbitrarily at t = 0, thus we have the freedom of n arbitrary
functions of one variable in the system being solved. In (51) all derivatives of the
solution, µi , are defined apart from the ith one, thus µi can be defined arbitrarily on
the Ri − axis, thus there are n arbitrary functions of one variable.
2.6 The Haantjes tensor
The Haantjes tensor (3) is used to verify the diagonalizability of the system (7).
Calculating the Haantjes tensor using computer algebra is relatively straightforward
for a well defined matrix and thus is an efficient way to find out if the system is
diagonalizable whereas calculating the eigenvalues and eigenvectors, in some cases,
is extremely difficult. As defined previously the Nijenhuis tensor and the Haantjes
tensor are,
i
Njk = vjp ∂p vki − vkp ∂p vji − vpi ∂j vkp + vpi ∂k vjp .
Hjk = Npr vj vk − Njr vp vk − Nrk vp vj + Njk vr vp ,
∂
where ∂s = ∂us
. Both the Nijenhuis tensor and the Haantjes tensor are skew sym-
metric in lower indices, this means reversing the two lower indices changes the sign,
i i i
Hjk = −Hkj . A direct corollary of this is Hjj = 0 for j = 1 · · · n.
i
Theorem 6 The matrix v(u) from (7) can be diagonalized if and only if Hjk ≡ 0.
Proof To prove this we first derive the necessary condition for the Haanjtes tensor
to be identically zero. Then we obtain the condition for v(u) to be diagonalizable
and see that they are equivalent. Evaluated on pairs of vector fields X, Y , the
Nijenhuis tensor and the Haantjes tensor can be written in the form,
N (X, Y ) = [vX, vY ] + v 2 [X, Y ] − v[X, vY ] − v[vX, Y ], (55)
H(X, Y ) = N (vX, vY ) + v 2 N (X, Y ) − vN (X, vY ) − vN (vX, Y ). (56)
j j
where [Z, Y ] = (Z i δY
δui
− Y i δZ ) δ . Let Xi , Xj be eigenvectors of v, thus,
δui δuj
vXi = λi Xi , (57)
using this upon direct calculation one obtains,
N (Xi , Xj ) = [λi Xi , λj Xj ] + v 2 [Xi .Xj ] − v[Xi , λj Xj ] − v[λi Xi , Xj ]
= λi λj [Xi , Xj ] + λi λji Xj − λj λij Xi + v 2 [Xi , Xj ] − V λj [Xi , Xj ] − vλji Xj
−vλi [Xi , Xj ] − vλij Xi (58)
= (λi λji − λj λji )Xj + (λi λij − λj λij )Xi + (λj λi + v 2 − (λi + λj )v)[Xi , Xj ].
Then using (58) calculating the Haantjes tensor and collecting terms at Xi , Xj , [Xi , Xj ],
we obtain
H(Xi , Xj ) = (λi λj + v 2 − v(λi + λj ))2 [Xi , Xj ] + (λi λj + v 2 − v(λi + λj ))
(λi − λj )λij Xi + (λi − λj )λji Xj

= (λi λj + v 2 − (λi + λj )v)2 [Xi , Xj ].

The coefficients of Xi , Xj cancel identically due to the fact vXk = λk Xk . Next, let
the expansions of commutators [Xi , Xj ] be nk=1 Cijk Xk . Thus,
P
H(Xi , Xj ) = (λi λj + v 2 − (λi + λj )v)2 [Xi , Xj ],
= (λi λj + v 2 − (λi + λj )v)2 Cijk Xk ,
= (λi λj + v 2 − (λi + λj )v)2 Cijk Xk ,

= (λi λj + (λk )2 − (λi + λj )λk )2 Cijk Xk ,

= (λk − λi )2 (λk − λj )2 Cijk Xk .
Hence H(Xi , Xj ) = 0 if and only if
Cijk = 0, i 6= j, i 6= k, k 6= j. (59)
The matrix v is diagonalizable if and only if each distribution < Xi , Xj > is inte-
grable, i.e [Xi , Xj ] ∈ span(Xi , Xj ), which is exactly (59). For an explanation let
us consider 3D space. Take X1 (u1 , u2 , u3 ), X2 (u1 , u2 , u3 ), X3 (u1 , u2 , u3 ) as eigenvec-
tors. We are looking for a change of variables of type (8). The Riemann invariant
R3 (u1 , u2 , u3 ) is defined by the equations:
X1 R3 = 0, X2 R3 = 0,
the compatibility condition gives X2 X1 R3 −X1 X2 R3 = 0, which is precisely, [X1 , X2 ]R3 =

0. Similarly, the Riemann invariant R2 (u1 , u2 , u3 ) is defined by:
X1 R2 = 0, X3 R2 = 0,
the compatibility condition gives [X1 , X3 ]R2 = 0 if C12

3
= 0. Therefore v possesses
Riemann Invariants. This idea can be extended to n dimensions.
Example Consider the Benney-like system

    
 a   v a 0  a 
    
 v  +  0 v 1   v  = 0. (60)
    
    
w q(a) 0 0 w
t x
Is (60) diagonalizable and under what restrictions on q(a)? To answer this ques-
tion we calculate all components of the Haanjtes tensor for the system (60). First
i i i
calculate Njk . Note that Njk = −Nkj .
1 1 1 2
N12 = 0, N13 = −1, N23 = 0, N21 = 0,
2
N23 2
= −1, N13 3
= 0, N31 3
= 0, N32 3
= 0, N12 = −aq 0 (a). (61)
i i i
Now we can calculate Hjk . Note that Hjk = −Hkj
1
H23 1
= 0, H13 1
= 0, H12 = −aq 0 (a), H13
2
= −aq 0 (a), H23
2
= 0,
2
H21 = −avq 0 (a), H12
3
= av 2 q 0 (a), H32
3
= a2 q 0 (a), H31
3
= avq 0 (a).
i
Thus, Hjk ≡ 0 if and only if q 0 (a) = 0. So, q(a) must be a constant. Thus (60) is
diagonalizable if q(a) is constant.
2.7 Reciprocal Transformations
Let B(u)dx + A(u)dt and N (u)dx + M (u)dt be two conservation laws of the system
(7), equivalently we can view them as 1-form closed by virtue of (7). One can
introduce new variables X and T defined by
dX = Adx + Bdt, dT = M dx + N dt
Lets apply the transformation to equation (7). First we need to calculate ux , ut in

terms of the new variables X, T :
ut = N uT + BuX , ux = M uT + AuX , (62)
substitue these into (7) and write in the same form we get,
uT = w(u)uX ,
where w(u) = (Av − BI)(N I − M v)−1 . These transformations originate from gas
dynamics.
Example Take the diagonal form of Chromatography equations (24). Introduce

the new independent variable, X, T, by the formulae
(Rs + αβ
Q
β Rs
Q
n+1
dT = α Q s dt + Q s dx ,
R α Γ
Q s δ
(R + γ δ Rs
Q Y
n+1
dX = γ Q s dt + Q s dx Γs ,
R γ Γ
αRi +β
and the new dependent variable R̃ = γRi +δ
. So, (24) becomes
!
s β Y Rs Y Rs + β
δ nR + α
Y
i n s i i
RX γ δ (R + + RT βα Q +R RTi αn+1 α
s
γ Γs s
Γs s
Rs
!
Y (Rs + γδ ) Y
i n+1
+RX γ Γs = 0.
Rs s
After simplification,
! !
Y Y Ri
(δ + γRi ) (γRs + δ) RX
i
+ (αRi + β) (αRs + β) Q T = 0.
s s
Γs
s
Q
Divide by s (γR + δ) to get
Y R̃s
i i i
(δ + γR )RX + (αR + β) RTi = 0.
s
Γs
Dividing by (δ + γRi ) gives
Y R̃s
i
RX + R̃i RTi = 0.
s
Γs
αδ−γβ
Multiply both sides by (γRi +δ)2
to obtain
Y R̃s
i
ṘX + R̃i R̃Ti = 0.
s
Γs
Proposition The semi-Hamiltonian property (38) is preserved under reciprocal

transformations
Proof Applying the reciprocal transformation,
dX = Adx + Bdt,
dT = M dx + N dt
to (9) we obtain
i λi A − B i
RX = R
N − λi M T
Thus we need to show µi satisfies (38) where,
λi A − B

i
µ = .
N − λi M
µij

We will start by expanding µ −µi
j and using the fact that Bj = λj Aj , Nj = λj Mj ,
k
µij

=
µj − µi k
(N − λi M )(λij A + (λi − λj )Aj ) − (λi A − B)(λij M + (λi − λj )Mj )
(N − λj M )
(N − λi M )(λj − λi )(AN − M B)
λij λij M (N − λi M )Aj − (λi A − B)Mj N − λj M
= − −
N − λi M N − λi M AN − M B N − λi M
i i i
λj Nj − λ j M − λ M j Nj − λ i M j
= + − −
N − λi M N − λi M N − λi M
((N − λi M )Aj − (λi A − B)Mj ) (N − λj M )
(AN − M B)(N − λi M )
λij (AN )j − (BM )j
+ ln(N − λi M ) j −

= i
N −λM AN − BM
i i
!
λj

N −λM
= + ln .
λj − λi AN − M B j
k
So, by symmetry we have
µij µik

=
µj − µi k µk − µi j
as required.
Hydrodynamic Chains 38
3 Hydrodynamic Chains
A hydrodynamic chain is of the form (7),
ut = V (u)ux
where V (u) is a square matrix of infinite dimension, while u = (u1 , u2 , · · · , )t is

an infinite-component vector. We assume that V (u) belongs to class C as defined
in Def. 1 in the Introduction. A well known example of a hydrodynamic chain is
the Benney chain [3].
3.1 The Benney chain
The Benney chain is,
A0t + A1x = 0,
A1t + A2x + A0 A0x = 0,
A2t + A3x + 2A1 A0x = 0, (63)
A3t + A4x + 3A2 A0x = 0,
A4t + A5x + 4A3 A0x = 0,
...
or, in matrix form,
    
0 0
A 0 1 0 0 0 0 . A
    
 1   0  1 
 A   A 0 1 0 0 0 . 
 A 
 
  
    
 A2  +  2A1
  A  = 0.
0 0 1 0 0 .   2 
  
    
 3 
 3A2
 3 
 A  0 0 0 1 0 .  A 

    
· . . . . . . . ·
t x
Let us take a moment to look at how and where the (63) first appeared. This chain
was derived in [3] and was shown to consist of an infinite number of conservation
laws for the system of equations that describe long waves on shallow ideal fluid
under the action of gravity. These equations are,
ux + vy = 0, (64)
ut + uux + vuy + ghx = 0, (65)
v = 0, y = 0, (66)
ht + uhx − v = 0, y = h. (67)
Here h(x, t, ) is the height of the water, u(x, y, t), v(x, y, t) are the horizontal and
vertical velocities of the water, g is gravity. The conservation laws of mass, momen-
tum and energy are all straightforward to derive. Equation (67) can be rewritten
as
Z h
∂
ht + udy = 0; (68)
∂x o
(64) is used to eliminate v. This is the mass conservation law. Integrate (65) w.r.t
y from 0 to h to obtain the momentum conservation law:
h h
gh2
Z Z
∂ ∂ 2
udy + u dy + = 0; (69)
∂t o ∂x o 2
(64), (65), (67) have each been used. Finally multiply (65) by u and integrate w.r.t
y from 0 to h to obtain the energy conservation law:
Z h
∂ 1 h 3
Z h
gh2
Z
∂ 1 2
u dy + + u dy + gh udy = 0. (70)
∂t 2 o 2 ∂x 2 o o
Note that the identity

!
Z r(x) Z r(x)
∂
udy = ux dy + u|y=r rx − u|y=s sx
∂x s(x) s(x)
has been used. Benney went on to show that there are an infinte number of such
conservation laws as follows. Let us introduce the notation
Z h
An = un (x, y, t)dy,
0
Rh
so A0 = h, A1 = 0
u(x, y, t)dy, · · · , etc, then by multiplying (65) by un−1 and then
integrating from y = 0 to y = h. Using (64), (65), (66) we obtain an unclosed set of
equations for the moments:
∂An ∂An+1 ∂A0

+ + nAn−1 = 0, n = 1, 2, . . . (71)
∂t ∂x ∂x
These equations are not conservative and nor is it in any way clear that they can
be brought into such a form. The existence of an infinity of conservation laws was
demonstrated in [3, 32]. In our discussion we follow [22] which extends these ideas.
Instead of dealing directly with (71) introduce a generating function
u1 u2 u3
λ=p+ + 2 + 3 + ... (72)
p p p
which was shown in [22] to satisfy, by virtue of (63), the relation
λt − pλx = λp pt − (p2 /2 + u1 )x .

(73)
Let us show that (72) does indeed satisfy (73) modulo (63). Using (73) and differ-
entiating we obtain,
u1t u1 pt u2 2u2 pt uk kuk pt

λt = p t + − 2 + 2t − 3 + · · · + kt − k+1 + . . . . (74)
p p p p p p
u1x u1 px u2x 2u2 px ukx kuk px

λx = p x + − 2 + 2 − + · · · + − k+1 + . . . . (75)
p p p p3 pk p
u1 2u2 3u3 kuk

λp = 1 − − − − · · · − − .... (76)
p2 p3 p4 pk+1
Substituting (74), (75), (76) into (73) ,
λt − pλx = λp pt − (p2 /2 + u1 )x ,

uk kuk pt uk kuk px kuk

pt + kt − k+1 − p px + kx − k+1 pt − (p + u1x ) ,

= 1 − k+1
p p p p p
where k = 1, 2, 3, . . . , now collect t derivatives on the left hand side and x derivatives
on the right hand side,
uk kuk pt kuk ukx kuk px kuk

pt + kt − k+1 − 1 − k+1 pt = p px + k − k+1 − 1 − k+1 px ,
p p p p p p
after cancelation we are left with,
ukt uk+1
x kuk u1x
= + , k = 1, 2, 3, . . .
pk pk pk+1
upon equating powers of p we finally have,
ukt = uk+1
x + (k − 1)uk−1 u1x , k = 1, 2, 3, . . .
which is precisely (63) as expected. This relation provides an infinity of conserved

densities Hn (u) defined by the equation
pt = (p2 /2 + u1 )x
where one has to substitute the expression for p(λ) obtained from (72): p = λ −
H1 H2 H3
λ
− λ2
− λ3
− .... Explicitly, one gets H 1 = u1 , H 2 = u2 , H 3 = u3 + (u1 )2 , etc:
Ht1 = (H 2 )x ,
H1
Ht2 = (H 3 − )x , (77)
2
Ht3 = (H 4 − H 2 H 1 )x ,
(H 2 )2

4 5 3 1
Ht = H −H H − .
2 x
3.2 Hydrodynamic chains and the Haantjes tensor
Both the tensors (2) and (3) make perfect sense for infinite matrices which are of
class C as stated in the introduction. For matrices from the class C all contractions
in the expressions (2) and (3) reduce to finite summations so that each particular
i
component Hjk is a well-defined object which can be effectively computed. More-
over, for a fixed value of the upper index i there exist only finitely many non-zero
i
components Hjk .
We propose the following
Definition 3 [20] A hydrodynamic chain from the class C is said to be diagonaliz-

able if all components of the corresponding Haantjes tensor (3) are zero.
Although we have stated the above definition it must be noted that the Riemann
invariants of a hydrodynamic chain cannot be calculated because the chain is infinite.
Example In this example we illustrate how the Nijenhuis tensor and the Haantjes
tensor can be calculated for a general hydrodynamic chain of class C. First, calculate
the Nijenhuis tensor as this is required to calculate the Haantjes tensor. Then
1
calculate the Haanjtes tensor, in this example the upper index is fixed at 1, Hjk .
Consider the infinite system
u1t + u2x + v11 (u1 )u1x = 0,
u2t + u3x + v22 (u1 , u2 )u2x + v12 (u1 , u2 )u1x = 0, (78)
u3t + u4x + v33 (u1 , u2 , u3 )u3x + v23 (u1 , u2 , u3 )u2x + v13 (u1 , u2 , u3 )u1x = 0,
...
Or, in matrix form,
    
1
u v11 1 0 0 0 0 . u 1
    
 2   2  2 
 u 
 
 v1
 v22 1 0 0  u 
0 .   
    
 u3  +  v 3 v23 v33   u  = 0.
1 0 0 .   3 
   1
    
 4   4
v24 v34 v44
 4 
 u   v1 1 0 .  u 
    
· ··· ··· ··· ·
t x
(79)
Note that vji = vji (u1 , . . . ui ). It is easily seen that

v i (u1 , . . . , ui ) k <i+1


 k


vki = 1 k =i+1



 0

k > i + 1.
i
To calculate Njk the index i will be considered fixed, this implies that we need only
i
consider j ≤ i + 1, k ≤ i + 1, 1 ≤ p ≤ i + 1. Similarly when calculating Hjk with
i fixed, implies that we need only consider j ≤ i + 4, k ≤ i + 4, 1 ≤ p, r ≤ i + 3.
In general if vki = 0 ∀k > i + l, then to calculate Njk
i
we only need to consider
i
j ≤ i + l, k ≤ i + l, 1 ≤ p ≤ i + l. To calculate Hjk we only need to consider
j ≤ i + 4l, k ≤ i + 4l, 1 ≤ p, r ≤ i + 3l.
i ∂
Calculate Njk , i ≤ 3, j, k ≤ 5, 1 ≤ p ≤ 4. The notation ∂j = ∂uj
is used in the
following evaluation.

∂1 v11 + ∂1 v22 − ∂2 v12 j = 2, k = 1





1
Njk = −∂1 v11 − ∂1 v22 + ∂2 v12 j = 1, k = 2



 0

else




 (v22 − v11 )∂1 v22 + ∂1 v12 + ∂1 v23 − v12 ∂2 v22 − ∂2 v13 j = 2, k = 1


∂1 v33 + ∂2 v12 − ∂3 v13 j = 3, k = 1







(v 1 − v22 )∂1 v22 − ∂1 v12 − ∂1 v23 + v12 ∂2 v22 + ∂2 v13 j = 1, k = 2


 1


2
Njk = ∂2 v22 + ∂2 v33 − ∂3 v23 j = 3, k = 2



−∂1 v33 − ∂2 v12 + ∂3 v13 j = 1, k = 3










 −∂2 v22 − ∂2 v33 + ∂3 v23 j = 2, k = 3



 0 else

∂1 (v13 + v24 ) + (v33 − v11 )∂1 v23 + v23 ∂1 v22 + (v22 − v33 )∂2 v13






−∂2 (v12 v23 ) − ∂2 v14 + v23 ∂3 v13 − v13 ∂3 v23




 j = 2, k = 1


(v33 − v11 )∂1 v33 + ∂1 v34 + ∂2 v13 − v12 ∂2 v33 − v13 ∂3 v33 − ∂3 v14 j = 3, k = 1







∂1 v44 + ∂3 v13 − ∂4 v14 j = 4, k = 1






(v11 − v33 )∂1 v23 − v23 ∂1 v22 − ∂1 (v24 + v13 ) + ∂2 (v12 v23 )







+∂2 (v33 − v22 ) + ∂2 v14 + v13 ∂3 v23 − v23 ∂3 v13 j = 1, k = 2







−∂1 v33 + ∂2 v23 + (v33 − v22 )∂2 v33 + ∂2 v34 − v23 ∂3 v33 − ∂3 v24 j = 3, k = 2





3
Njk = ∂2 v44 + ∂2 v23 − ∂4 v24 j = 4, k = 2



(v11 − v33 )∂1 v33 − ∂1 v34 + v12 ∂2 v33 − ∂2 v13 + v13 ∂3 v33 + ∂3 v14 j = 1, k = 3










 ∂1 v33 + v22 ∂2 v33 − ∂2 v23 − v33 ∂2 v33 − ∂2 v34 + v23 ∂3 v33 + ∂3 v24 j = 2, k = 3


∂3 (v44 + v33 ) − ∂4 v34



 j = 4, k = 3



−∂1 v44 − ∂3 v13 + ∂4 v14 j = 1, k = 4










 −∂2 v44 − ∂3 v23 + ∂4 v24 j = 2, k = 4


−∂3 (v44 + v33 ) + ∂4 v34



 j = 3, k = 4



0 otherwise


1
Now Hjk can be calculated,



 v11 ∂1 v23 − v13 (∂2 v22 − 2∂3 v23 + ∂2 v33 ) + v23 (2∂2 v12 −



∂1 v22 − 2∂3 v13 + ∂1 v33 ) − v22 ∂2 v13 + v33 ∂2 v13 +






v12 ∂1 v23 − ∂1 v13 − v33 ∂1 v23 + ∂2 v14 − ∂1 v24




 j = 1, k = 2









v11 ∂1 (v33 − v22 ) + v22 (∂1 (v22 − v33 ) − ∂2 v12 + ∂3 v13 )+






v12 (2∂2 v33 − ∂3 v23 ) + v33 (∂2 v12 − ∂3 v13 ) + ∂1 (v12 + v23 − v34 )







−2∂2 v13 + ∂3 v14 + v13 ∂3 v33 j = 1, k = 3













−∂1 v44 − 2∂3 v13 + ∂4 v14 + ∂2 v12 + ∂1 v33




 j = 1, k = 4









v23 (−2∂2 v12 + ∂1 v22 + 2∂3 v13 − ∂1 v33 ) + v13 (∂2 v22 +





1
Hjk = ∂2 v33 − 2∂3 v23 ) + v22 ∂2 v13 − v33 ∂2 v13 − v12 ∂2 v23 + ∂1 v13 − v11 ∂1 v23 +



v33 ∂1 v23 − ∂2 v14 + ∂1 v24 j = 2, k = 1













v11 (−∂2 v22 − ∂2 v33 + ∂3 v23 ) + v33 (∂2 v22 − ∂3 v23 )







+v23 ∂3 v33 ) + ∂1 (−v11 − v22 + 2v33 ) + ∂2 (2v12 − v23 − v34 )










 +v22 ∂2 v33 + ∂3 (v24 − v13 ) j = 2, k = 3









∂2 (v22 + v33 − v44 ) − 2∂3 v23 + ∂4 v24 j = 2, k = 4













v11 ∂1 (v22 − v33 ) + v12 (∂3 v23 − 2∂2 v33 ) + v22 (∂1 (v33 − v22 ) + ∂2 v12 −







∂3 v13 ) + v33 (∂3 v13 − ∂2 v12 ) + ∂1 (v34 − v12 − v23 ) + 2∂2 v13







 −∂3 v14 − v13 ∂3 v33 j = 3, k = 1




 v11 (∂2 (v22 + v33 ) − ∂3 v23 ) + v33 (−∂2 v22 + ∂3 v23 ) − 2∂1 v33 +


∂1 (v11 + v22 ) + ∂3 (v13 − v24 ) − v22 ∂2 v33 + ∂2 (v23 − v34 − 2v12 )







−v23 ∂3 v33 j = 3, k = 2













∂3 (−v44 − v33 ) + ∂4 v34 j = 3, k = 4












1
Hjk = ∂1 v44 + 2∂3 v13 − ∂4 v14 − ∂2 v12 − ∂1 v33 j = 4, k = 1













 ∂2 (v44 − v22 − v33 ) + 2∂3 v23 − ∂4 v24 j = 4, k = 2









∂3 (v44 + v33 ) − ∂4 v34 j = 4, k = 3














 0 otherwise.
1
It is a necessary condition that each component of Hjk is identically zero for the
matrix to be diagonalizable.
Example In this example we show how to calculate every component of the Haantjes
tensor for the Benney chain (63). As in the previous example the Nijenhuis tensor
is first calculated. Since the system under consideration is infinite, it means that in
theory the summation for Nijenhuis and Haantjes tensor may be infinite. However,
the matrix is of the class C so the summation is in fact finite, as is now shown. It
is clear that
m
vm+1 = 1, v1m = (m − 1)Am−2 ,
all other vnm = 0. Now we calculate the Nijenhuis tensor, first consider N1k
i
,
i
N1k = v1p ∂p vki − vkp ∂p v1i − vpi ∂1 vkp + vpi ∂k v1p .
Let us consider each term separately for a moment, the first term:

 (i − 1)(i − 2)Ai−3

k=1
p i
v1 ∂p vk =
 0
 otherwise
The second term:


(i − 1)(i − 2)Ai−3 k = 1





vkp ∂p v1i = i−1 k=i



 0

otherwise
The third term:


 1

i=k=1
i p
vp ∂1 vk =
 0
 otherwise
The fourth term:


 k

i=k
i p
vp ∂k v1 =
 0
 otherwise
Combining all the terms we have
 
 0 0 0 0 0 0 . . . 
 

 0 1 0 0 0 0 . 

 
0 0 1 0 0 0 . 
  

 1 i = k, i 6= 1
  
i
N1k = = . .
 
. 1
 0
 otherwise 




 . . . 

 

 . . . 

 
. . .
Now we calculate the Nijenhuis tensor with j 6= 1, by the skew symmetry property,
i i
Njk = −Nkj , it is clear that,

1 j = 1, i = k, i 6= 1





i
Njk = 1 k = 1, i = j, i 6= 1



 0

otherwise
Next we must calculate the Haantjes tensor, first consider
H1k = Npr v1 vk − N1r vp vk − Nrk vp v1 + N1k vr vp .
Work on each of the terms separately, the first term:


 −(i − 1)Ai−2

k=2
i p r
Npr v1 vk =
 0
 otherwise
The second term:


iAi−1 k=1





p i r
N1r vp vk = 1 k =i+2



 0

otherwise
The third term:


 −iAi−1

k=1
p i r
Nrk vp v1 =
 0
 otherwise
The fourth term:


(i − 1)Ai−2 k=2





p i r
N1k vr vp = 1 k =i+2



 0

otherwise
i
Thus, H1k = 0, ∀i, k. Now work on
H2k = (Npr v2 vk ) − (N2r vp vk ) − (Nrk vp v2 ) + (N2k vr vp ).
Again work on each of the terms separately, the first term:


(i − 1)Ai−2 k=1





i p r
Npr v2 vk = 1 k = i + 1, i 6= 1



 0

otherwise
The second term:


 −1

i = 1, k = 2
p i r
N2r vp vk =
 0
 otherwise
The third term:


 1

k =i+1
p i r
Nrk vp v2 =
 0
 otherwise
The fourth term:


 −(i − 1)Ai−2

k=1
p i r
N2k vr vp =
 0
 otherwise
i
Thus, H2k = 0, ∀i, k. Now work on
Hjk = (Npr vj vk ) − (Njr vp vk ) − (Nrk vp vj ) + (Njk vr vp ), j 6= 1, 2.
Again work on each of the terms separately, the first term:


 −1

i = j − 1, k = 2
i p r
Npr vj vk =
 0
 otherwise
The second term:


 −1

i = j − 1, k = 2
p i r
Njr vp vk =
 0
 otherwise
The third term:


 −1

i = j − 2, k = 1
p i r
Nrk vp vj =
 0
 otherwise
The fourth term:


 −1

i = j − 2, k = 1
p i r
Njk vr vp =
 0
 otherwise
i i i
Thus, Hjk = 0, ∀ i, k, j ≥ 3. We have previously shown that H1k = 0, H2k = 0.
i
Thus, Hjk = 0, ∀ i, j, k.
3.3 Another chain considered by Benney
An extension to (63) is considered in [4],
∂Cn ∂Cn+1 ∂C0

+ an + bn Cn−1 = 0, n≥0 (80)
∂t ∂x ∂x
such a system only posesses an infinity of conservation laws if
an bn+1
= constant. (81)
n+1
Calculating the Haanjtes tensor of system (80) and setting each component equal
to zero implies that
an bn+1 − 2an+1 bn+2 + an+2 bn+3 = 0. (82)

This implies
an bn+1 = αn + β (83)
where α, β are constants. It is clear that (81) just (83) with α = β = 1, thus
requiring that the Haantjes tensor is zero has produced a slightly more general
result than obtained in [4]. Under the change of variables An → a0 · · · an−1 C n , the
Benney chain, (63), is transformed to (80) modulo (83).
3.4 Hydrodynamic reductions and diagonalizability
To illustrate the method of hydrodynamic reductions we consider the Benney chain

(63),
u1t = u2x ,
u2t = u3x + u1 u1x ,
u3t = u4x + 2u2 u1x ,
u4t = u5x + 3u3 u1x ,
etc. Following the approach of [24, 25] let us seek solutions in the form ui =
ui (R1 , . . . , Rm ) where the Riemann invariants R1 , . . . , Rm solve the diagonal system
(9),
Rti = λi (R)Rxi .
Substituting this ansatz into the Benney equations and equating to zero coefficients
at Rxi we arrive at the following relations:
∂i u2 = λi ∂i u, (84)
∂i u3 = ((λi )2 − u)∂i u, (85)
∂i u4 = ((λi )3 − uλi − 2u2 )∂i u, (86)
∂i u5 = ((λi )4 − u(λi )2 − 2u2 λi − 3u3 )∂i u, (87)
etc. Here u = u1 , ∂i = ∂Ri , i = 1, ..., m (no summation!) The consistency conditions

of the first three relations (84)–(86) imply
∂j λi ∂i λj
∂i ∂j u = ∂i u + ∂j u,
∂j λi ∂i u + ∂i λj ∂j u = 0,
λi ∂j λi ∂i u + λj ∂i λj ∂j u + ∂i u∂j u = 0,
respectively. Solving these equations for ∂j λi we arrive at the Gibbons-Tsarev sys-

tem
∂j u ∂i u∂j u
∂j λi = , ∂i ∂j u = 2 . (88)
λj − λi (λi − λj )2
It is a truly remarkable fact that all other consistency conditions (e.g., of the relation
(87), etc), are satisfied identically modulo (88). Moreover, the semi-Hamiltonian
property (38) is also automatically satisfied. Thus, the system (88) governs m-
component reductions of the Benney chain. Up to reparametrizations Ri → f i (Ri )
these reductions depend on m arbitrary functions of a single variable. Solutions
arising within this approach are known as multiple waves, or nonlinear interactions
of planar simple waves.
The main result of this section is the proof of Theorem 2 formulated in the intro-
duction: The vanishing of the Haantjes tensor H is a necessary condition for the
integrability of hydrodynamic chains from the class C [20].
Two different proofs of this statement will be given.
Proof This proof is computational. Writing down the equations of the chain in the
form um m n i i 1 m
t = Vn ux and substituting the ansatz u = u (R , ..., R ) we arrive at an
infinite set of relations

Vnm ∂i un = λi ∂i um ;
we point out that all summations here and below involve finitely many nonzero
terms. Applying the operator ∂j , j 6= i, we obtain
m
Vn,k ∂i un ∂j uk + Vnm ∂i ∂j un = ∂j λi ∂i um + λi ∂i ∂j um . (89)
Interchanging the indices i and j and subtracting the results we arrive at the ex-
pression for ∂i ∂j um in the form
m m
∂j λi ∂i λj Vn,k − Vk,n
∂i ∂j um = ∂i u m
+ ∂j u m
+ ∂i un ∂j uk .
λj − λi λi − λj λi − λj
Substituting this back into (89) we arrive at a simple relation
m
Nnk ∂i un ∂j uk
∂j λi ∂i um + ∂i λj ∂j um =
λi − λj
where N is the Nijenhuis tensor of V . This can be rewritten in the invariant form
N (∂i u, ∂j u)
∂j λi ∂i u + ∂i λj ∂j u = (90)
λi − λj
which implies the following four relations:

V 2 N (∂i u, ∂j u)
(λi )2 ∂j λi ∂i u + (λj )2 ∂i λj ∂j u = , (91)
λi − λj
V N (V ∂i u, ∂j u)
(λi )2 ∂j λi ∂i u + λi λj ∂i λj ∂j u = , (92)
λi − λj
V N (∂i u, V ∂j u)
λi λj ∂j λi ∂i u + (λj )2 ∂i λj ∂j u = , (93)
λi − λj
N (V ∂i u, V ∂j u)
λi λj ∂j λi ∂i u + λi λj ∂i λj ∂j u = . (94)
λi − λj
The first relation can be obtained by applying the operator V 2 to (90) and using
V ∂i u = λi ∂i u. Similarly by applying V λi , V λj , λi λj respectively, we can obtain
(92),(93),(94). Thus combining all four relations (91), (92), (93), (94),
V 2 N (∂i u, ∂j u) − V N (V ∂i u, ∂j u) − V N (∂i u, V ∂j u) + N (V ∂i u, V ∂j u) = 0. (95)
Relation (95) can be rewritten in the form H(∂i u, ∂j u) = 0 where H is the Haantjes
tensor, indeed, a coordinate-free form of the relation (3) is
H(X, Y ) = V 2 N (X, Y ) − V N (V X, Y ) − V N (X, V Y ) + N (V X, V Y )
where X, Y are arbitrary vector fields. Keeping in mind that
• ∂i u and ∂j u are eigenvectors of the matrix V corresponding to the eigenvalues

λi and λj ;
• λi and λj can take arbitrary values;
we conclude that H(X, Y ) = 0 for any two formal eigenvectors of the matrix V .
By formal eigenvectors we mean vectors X defined by AX = λX. So, for the Benney
chain (63) we get, in matrix form, the relations,
 
0 1 0 0 0    
1 1

 1

 x  x 
 u 0 1 0 0 

   
x2  x2 

  
 u2 = λ .
   
 0 0 1 0 
  

 3

 x3 
 
 3 
x 
 u 0 0 0 1    
  . .
. . . . .
Assuming that eigenvectors of V span the space of dependent variables u (this is

true for all examples discussed in this paper), we obtain H = 0. In more detail,
let X(λ) = (ξ 1 (λ), ξ 2 (λ), ξ 3 (λ), ...)t be a formal eigenvector of the matrix V corre-
sponding to the eigenvalue λ. Let us assume that these eigenvectors span the space
of dependent variables, that is, that there exist no non-trivial relations of the form
ci ξ i (λ) = 0 with finitely many nonzero λ-independent coefficients ci . In other words,
ξ i (λ) are linearly independent as polynomials in λ.
The condition H(X(λ), X(µ)) = 0, written in components, takes the form

i j
Hjk ξ (λ)ξ k (µ) = 0; recall that, since the matrix V belongs to the class C, these sums
contain finitely many terms for any fixed value of the upper index i. Taking into
account the linear independence of ξ j (λ) and ξ k (µ), one readily arrives at Hjk
i
= 0.
As an illustration, let V be the matrix corresponding to the Benney chain (63).
V X = λX,
    
0 1 0 0 0 0 . x1 x1
    
 1    
 u 0 1 0 0 0 .   x2   x2 
    
    
 u2
 0 0 1    
.
0 0 .   x3  = λ  x3 
    
 3
 u 0 0 0 1 0 .   x4   x4 
   
    
. . . . . . . . .
Then setting x1 = 1 we obtain the equations
x2 = λx1 ,
x3 + u1 x1 = λx2 ,
x4 + u2 x1 = λx3 ,
...
Then, writing each xi in terms of λ
x1 = 1,
x2 = λ,
x 3 = λ2 − u 1 ,
x 4 = λ3 − u 1 λ − u 2 ,
...
Thus we arrive at X(λ) = (1, λ, λ2 − u1 , λ3 − u1 λ − 2u2 , ...)t , so that each ξ i (λ) is a

polynomial in λ of the degree i − 1. These eigenvectors span the space of dependent
variables since polynomials ξ i (λ) are manifestly linearly independent.
Notice that we have proved a more general result, namely, that the existence of
sufficiently many two-component reductions already implies the vanishing of the
Haantjes tensor H. Indeed, nothing changes in the proof if we set i = 1, j = 2
in the formula (90). As demonstrated in the Theorem 5 below, one can further
strengthen the result by reversing the above proof under the additional assumption
of the simplicity of the spectrum of the matrix V . This will require the relations
N (V ∂i u, ∂j u) − V N (∂i u, ∂j u) N (V ∂j u, ∂i u) − V N (∂j u, ∂i u)
∂i λj ∂j u = i j 2
, ∂j λi ∂i u = ,
(λ − λ ) (λi − λj )2
which can be obtained by applying V to both sides of (90) and solving for ∂i λj ∂j u
and ∂j λi ∂i u.
Now, the second proof is given.

Proof 2 Our first remark is that for the chains from the class C one needs to know
only finitely many rows of the matrix V (u) to calculate each particular component
of the Haantjes tensor. Let us fix the values of indices i, j, k and denote by C(i, j, k)
i
the maximal number of rows needed to calculate Hjk (counting from the first row).
i
We need to show that Hjk = 0. Let us consider an m-component diagonal re-
duction ui (R1 , . . . , Rm ), i = 1, 2, .... Choosing the first m variables u1 , . . . , um as
independent, we can represent the reduction explicitly as
um+1 = um+1 (u1 , . . . , um ), um+2 = um+2 (u1 , . . . , um ), . . . ,
etc. Substituting these expressions into the first m equations of the chain we ob-
tain an m-component system Sm for u1 , . . . , um , while the remaining equations will
be satisfied identically (by the definition of a reduction). Notice that the Haantjes
tensor of the reduced system Sm is identically zero since the reduction is diagonal-
izable. Let us now choose the number m sufficiently large so that the first C(i, j, k)
equations of the chain do not contain variables um+1 , um+2 , . . . (one can always do
so since any equation of the chain depends on finitely many u’s, and m can be ar-
bitrarily large). Then the first C(i, j, k) equations of the reduced system Sm will
be identical to the first C(i, j, k) equations of the original infinite chain. Hence, the
i
corresponding components Hjk for the reduced system and for the infinite chain will
also coincide. This proves that all components of the Haantjes tensor of the chain
must be zero.
A straightforward modification of the second proof allows one to show that the ex-
istence of an infinity of semi-Hamiltonian reductions implies the vanishing of the
tensor P . This establishes the necessity of the conjecture formulated in the Intro-
duction.
We emphasize that the condition of diagonalizability alone is not sufficient for the
integrability in general. This can be seen as follows.
Example Let us consider the chain
u1t = u2x + p(u1 )u1x ,
u2t = u3x + p(u1 )u2x + u1 u1x ,
u3t = u4x + p(u1 )u3x + 2u2 u1x ,
u4t = u5x + p(u1 )u4x + 3u3 u1x ,
etc, which is obtained from the Benney chain (63) ut = V (u)ux by the transfor-
mation V → V + p(u1 )E where E is an infinite identity matrix and p is a function
of u1 . One can verify that the corresponding Haantjes tensor is zero (which is not
at all surprising since the addition of a multiple of the identity does not effect the
diagonalizability). A simple calculation shows that hydrodynamic reductions of this

chain are governed by exactly the same equations as in the Benney case, the only
difference is that now the Riemann invariants Ri solve the equations
Rti = (λi (R) + p(u1 ))Rxi .
The semi-Hamiltonian property is satisfied if and only if p00 = 0. Thus, we have con-
structed examples which possess infinitely many diagonal hydrodynamic reductions
none of which are semi-Hamiltonian (if p00 6= 0).
We now prove Theorem 3 from the Introduction: in the simple spectrum case the
vanishing of the Haantjes tensor H is necessary and sufficient for the existence
of two-component reductions parametrized by two arbitrary functions of a single
argument.
Proof The necessity part is contained in the first proof of the previous theorem. To
establish the sufficiency one has to show that the vanishing of the Haantjes tensor
implies the solvability of the equations
V ∂1 u = λ1 ∂1 u, V ∂2 u = λ2 ∂2 u, (96)
m m
m ∂2 λ1 m ∂1 λ2 m
Vn,k − Vk,n
∂1 ∂2 u = 2 1
∂1 u + 1 2
∂2 u + 1 2
∂1 un ∂2 uk (97)
λ −λ λ −λ λ −λ
and
N (V ∂1 u, ∂2 u)−V N (∂1 u, ∂2 u)
∂1 λ2 ∂2 u = (λ1 −λ2 )2
,
(98)
N (V ∂2 u, ∂1 u)−V N (∂2 u, ∂1 u)
∂2 λ1 ∂1 u = (λ1 −λ2 )2
,
which govern two-component reductions. Our first observation is that the vanishing
of the Haantjes tensor implies that the vectors
N (V ∂1 u, ∂2 u) − V N (∂1 u, ∂2 u), N (V ∂2 u, ∂1 u) − V N (∂2 u, ∂1 u) (99)
are the eigenvectors of V with the eigenvalues λ2 and λ1 , respectively. Let us show
that N (V ∂1 u, ∂2 u) − V N (∂1 u, ∂2 u) is indeed the eigenvector of eigenvalue λ2 .
V X = λX (100)
V (N (V ∂1 u, ∂2 u) − V N (∂1 u, ∂2 u)) = λ2 (N (V ∂1 u, ∂2 u) − V N (∂1 u, ∂2 u))
V N (V ∂1 u, ∂2 u) − V 2 N (∂1 u, ∂2 u) = λ2 N (V ∂1 u, ∂2 u) − λ2 N (∂1 u, ∂2 u)
V N (V ∂1 u, ∂2 u) − V 2 N (∂1 u, ∂2 u) = N (V ∂1 u, V ∂2 u) − V N (∂1 u, V ∂2 u)
(100) is satisfied modulo Haantjes tensor identically zero. The same process can
be carried out for the other eigenvector in question. By the assumption of the
simplicity of the spectrum, that is the eigenvectors are distinct, it is known that
(99) is proportional to ∂2 u and ∂1 u. Thus, equations (98) reduce to a pair of first
order PDEs for λ1 and λ2 ,
k1 k2
∂1 λ2 = , ∂2 λ1 = , (101)
(λ1 − λ2 )2 (λ1 − λ2 )2
here k1 and k2 are the corresponding coefficients of proportionality. The relations

(96) allow one to reconstruct all components u2 , u3 , ... of the infinite vector u in terms
of its first component u1 . Finally, the equations (97), which are the consistency
conditions of (96), reduce to a single second order PDE for the first component u1
of the infinite vector u,
1 1
1 Vn,k − Vk,n
∂1 ∂2 u1 = k ∂
1 2 u 1
− k ∂
2 1 u 1
+ ∂1 un ∂2 uk . (102)
(λ1 − λ2 )3 λ1 − λ2
Thus, relations (96) – (98) reduce to a pair of first order equations (101) plus one
second order PDE for u1 . Up to reparametrizations R1 → f 1 (R1 ), R2 → f 2 (R2 ),
their general solution depends on two arbitrary functions of a single variable.
Conservative chains 65
4 Conservative chains
In this section we classify chains of form (5),
u1t = f (u1 , u2 )x ,
u2t = g(u1 , u2 , u3 )x ,
u3t = h(u1 , u2 , u3 , u4 )x ,
... .
All known integrable hydrodynamic chains can be written in conservative form, for
example we showed that the Benney chain, (63), can be written in the conservative
form (77).
4.1 Egorov case
The Egorov case is when f (u1 , u2 ) = u2 in (5) thus becoming (4). There are ten
i
nonzero components of Hjk and the requirement that they vanish leads to all second
order partial derivatives of h(u1 , u2 , u3 , u4 ) in terms of first order partial derivatives
of h and g:
2g1 g12 − g2 g11 + 2h1 g13

h11 = ,
g3
g22 g1 + g11 + g13 h2 + g23 h1
h12 = ,
g3
g2 g22 + 2g12 + 2g23 h2
h22 = ,
g3
g13 (h3 − g2 ) + g23 g1 + g12 g3 + g33 h1
h13 = , (103)
g3
g13 + h3 g23 + h2 g33
h23 = g22 + ,
g3
g33 (g2 − 2h3 )
h33 = 2g23 − ,
g3
h4 g13 h4 g23 h4 g33
h14 = , h24 = , h34 = , h44 = 0.
g3 g3 g3
The consistency conditions for the equations (103) lead to expressions for all third
order partial derivatives of the function g(u1 , u2 , u3 ) in terms of its lower order
derivatives:
2
2g33 2g13 g33 2g23 g33
g333 = , g133 = , g233 = ,
g3 g3 g3
2
2g13 2g13 g23 2g 2
g113 = , g123 = , g223 = 23 ,
g3 g3 g3
2 2

g222 = g2 g23 + g 23 (g3 g22 + 2g13 ) − g 33 (g2 g22 + 2g12 ) ,
g32
2 2

g122 = 2
g1 g23 + g13 (g3 g22 + g13 ) − g33 (g1 g22 + g11 ) , (104)
g3
2
g112 = (g33 (g2 g11 − 2g1 g12 ) − g13 (g2 g13 − 2g3 g12 ) − g23 (g3 g11 − 2g1 g13 )) ,
g32
2
g111 = (g1 + g22 )g13
2
+ g12 g23
2
+ g32 (g12
2
− g11 g22 ) − g22 g33 g12
g32
+g13 g3 (g11 + 2(g1 g22 − g2 g12 )) + 2g23 (g2 (g3 g11 − g1 g13 ) − g1 g3 g12 )
− g33 ((g1 + g22 )g11 − 2g1 g2 g12 ) .

This systems general solution depends on ten integration constants, indeed, the
values of g and its partial derivatives up to the second order can be prescribed
arbitrarily at any fixed point u10 , u20 , u30 . The system (104) was first derived in [47]
from the requirement that the chain (4) is embedded into a hierarchy of commuting
hydrodynamic chains of Egorov’s type. Exactly the same equations for g were
obtained in [17] by applying the method of hydrodynamic reductions to the (2+1)-
dimensional PDE
utt = g(uxx , uxt , uxy ) (105)
which is naturally associated with the chain (4); here the function g is the same
as in (4), (104). Thus, for hydrodynamic chains of the type (4) the condition of
diagonalizability is necessary and sufficient for the integrability. Let us summarize
the work referred to in [17].
4.2 Method of hydrodynamic reductions
First, introduce the notation uxx = a, uxt = b, uxy = c. Thus (105) becomes
utt = G(a, b, c). (106)
Now we write the PDE (106) in quasilinear form, that is in the form,
at = bx ,
ay = cx ,
by = c t , (107)
bt = G(a, b, c)x .
Look for hydrodynamic reductions in the form a = a(R1 , · · · , Rn ), b = b(R1 , · · · , Rn ),

c = c(R1 , · · · , Rn ), where the Riemann invariants satisfy the equations (26). We ar-
rive at the following identities
∂i b = λi ∂i a, (108)
∂i c = µi ∂i a, (109)
(λi )2 = Ga + Gb λi + Gc µi . (110)
Now, differentiate (110) w.r.t j, to obtain,
2λi λij = aj Gaa + λj Gab + λij Gb + λi aj gab + λj Gbb + µj Gbc + µij Gc +

aj µi Gac + λj Gbc + µj Gcc ,

(111)
rearrange (111) to make λij the subject. Use (26) to eliminate µij , and (110) to
eliminate µi . Thus (111) becomes
aj
λij = i j
[Gaa Gc + (λi + λj )Gab Gc + λi λj Gbb Gc +
λ −λ
(Gac + λi Gbc )((λj )2 − λj Gb − Ga )
Gcc j 2
((λi )2 − λi Gb − Ga )(Gac + λj Gbc + ((λ ) − λj Gb − Ga ))]. (112)
Gc
The compatibility conditions of (108), (109), that is (bi )j = (bj )i and (ci )j = (cj )i
respectively, imply that
λij λji
∂i ∂j a = j ∂i a + i ∂j a. (113)
λ − λi λ − λj
The consistency conditions of (112), (λij )k = (λik )j , are of the form
X∂j a∂k a.
The coefficient X is some rational expression in λi , λj , λk , the coefficients depend

on partial derivatives of G(a, b, c) up to the third order. To obtain the integrability
conditions only three component reductions need be considered, thus set i = 1, j =
2, k = 3. Requiring that X vanishes identically we obtain expressions for all third
order partial derivatives of G(a, b, c). The compatibility conditions of equations
(113), that is, ∂j (∂i ∂k a) = ∂k (∂i ∂j a), take the form
Y ∂i a∂j a∂k a (114)
The coefficient Y is rational in λi , λj , λk . Upon equating Y to zero one obtains

the same conditions as when we required X is identically zero. These conditions are
equivalent to (104).
Returning to the the Haantjes tensor, it can be shown that the vanishing of other
components of the Haantjes tensor does not impose any additional constraints on the
derivatives of g and h. Thus, writing the fourth equation of the chain (4) in the form
u4t = s(u1 , u2 , u3 , u4 , u5 )x and setting Hjk
2
= 0, one obtains the expressions for all
second order partial derivatives of s in terms of h and g, which are analogous to (103).
The consistency conditions are satisfied identically modulo (103), (104). Similarly,
3
the condition Hjk = 0 specifies the right hand side of the fifth equation of the chain,
etc. Although we know no direct way to show that this recursive procedure works in
general, there exists an alternative direct approach to the reconstruction of a chain
from the function g(u1 , u2 , u3 ). To illustrate this procedure we consider a simple
example g = u3 − 12 (u1 )2 which automatically satisfies (104). The corresponding h,
as specified by (103), is given by h = µ + αu1 + βu2 + γu3 + δu4 − u1 u2 , which can
be reduced to a canonical form h = u4 − u1 u2 by redefining u4 appropriately (this
transformation freedom allows one to absorb arbitrary integration constants arising
at each step of the construction). Thus, the first three equations of the chain are
1
u1t = u2x , u2t = (u3 − (u1 )2 )x , u3t = (u4 − u1 u2 )x , ..., (115)
2
etc. Equations (115) are nothing but the first three equations in the conservative
representation of the Benney chain (63). The recostruction of the remaining equa-
tions of the chain consists of three steps.
(i) One introduces the corresponding PDE (105). To find the required PDE we
proceed as follows, introduce the commuting chain:
u1y = u3x , u2y = · · · (116)
substitute (116) into (115) to eliminate u3x , thus obtaining two equations depending
only on u1 , u2 , that depend on independent variables x, t, y,
u1t = u2x ,
u2t = u1y − u1 u1x . (117)
Make the substitution u1 = vx , u2 = vt , in (117), thus obtaining
1 2
vtt = vxy − vxx , (118)
2
make the substitution u = vx to obtain the dispersionless KP equation,
1
utt = uxy − u2xx . (119)
2
It was demonstrated in [17] that the general PDE (105) is integrable by the method
of hydrodynamic reductions if and only if the function g satisfies the relations (104).
(ii) One constructs a dispersionless Lax pair for the PDE (105), the procedure is as
follows. Look for functions R and Q such that,
ψt = R(uxx , uxy , uxt , uyy , uyt , ψx ), (120)
ψy = Q(uxx , uxy , uxt , uyy , uyt , ψx ) (121)
Differentiate both equations w.r.t x and then make the substitution ψx = p, thus
we have
p t = Rx , (122)
py = Qx (123)
the consistency conditions pty = pyt are satisfied identically modulo (105), or more
specifically in our example (119). The existence of such Lax pairs was established
in [47, 17] for any equation (105) provided g satisfies the compatibility conditions
(104). In our example it takes the form
1 1
pt = ( p2 + uxx )x , py = ( p3 + uxx p + uxt )x ;
2 3
(iii) One looks for p as an expansion in the auxiliary parameter λ,
u1 u2 u3
p=λ− − 2 − 3 − ....; (124)
λ λ λ
the substitution of this ansatz into the first equation of the Lax pair implies an
infinite hydrodynamic chain for the variables ui . The first three equations of this
chain identically coincide with (115). The substitution into the second equation of
the Lax pair produces a commuting chain (one has to set u1 = uxx , u2 = uxt ). Both
chains possess infinitely many hydrodynamic reductions since this is the case for
the generating equation (105). Thus, the Haantjes tensor will automatically vanish.
Notice that (124) does not work in general, and a more complicated dependance of
p on λ is required.
4.3 Integration of the Egorov case
Now let us return to the work in question, solving the system (104). To explicitly
calculate g(u1 , u2 , u3 ) we will follow [47]. The main observation is that the first six
equations in (104) imply that the function 1/g3 is linear, 1/g3 = α + βu1 + γu2 + δu3 .
If δ 6= 0 then, up to a linear change of variables, one can assume that 1/g3 = u3 .
Similarly, if δ = 0, γ 6= 0, one can set 1/g3 = u2 . If δ = γ = 0, β 6= 0 one has
1/g3 = u1 . The last possibility is 1/g3 = 1. Thus, we have four cases to consider:
u3 u3
g = u3 + p(u1 , u2 ), g= + p(u1 , u2 ), g= + p(u1 , u2 ), g = ln u3 + p(u1 , u2 );
u1 u2
here the function p(u1 , u2 ) can be recovered after the substitution into the remaining
four equations (104). In each of these cases the resulting equations for p(u1 , u2 )
integrate explicitly, see [47], leading to four different forms:
1 1
g = u3 + (Au2 + 2Bu1 )2 + Ce−Au , (125)
4A
u3 B2

1 A B 2 1
g = 1+ 1
− 1 2
(u 2 2
) + 1 2
u − 1 2
− Ce−A/u , (126)
u u 4(u ) (u ) A(u )
3
u 1
g = 2 + η(u1 )(u2 )2 , (127)
u 6 Z
3 1 2
1
g = ln u − ln σ u , u − η(u1 )du1 . (128)
4
Here η(u1 ) is a solution to the Chazy equation [7],
2
η 000 + 2ηη 00 − 3η 0 = 0, (129)
and σ(u1 , u2 ) is an elliptic sigma function in the variable u2 whose dependence on

u1 is governed by the Chazy equation (see Case IV below). Details of the derivation
of canonical forms (125) - (128) can be summarized as follows.
Case I. Substituting the ansatz g = u3 + p(u1 , u2 ) into (104) one arrives at the
equations
p111 = 2(p212 − p11 p22 ),
p112 = p122 = p222 = 0.
The last three equations imply
1
p = A(u2 )2 + (Bu1 + D)u2 + q(u1 ),
4
and the substitution into the first equation results in the linear ODE q 000 +Aq 00 = 2B 2 .
Up to a transformation of the form u3 → u3 + αu2 + βu1 + γ this leads to (125).
Case II. Substituting the ansatz g = u3 /u1 + p(u1 , u2 ) into (104) one arrives at the
equations
2 2
p111 = 2(p212 − p11 p22 ) + 1 2
(p1 + p22 ) − 1 (p11 + 2p1 p22 − 2p2 p12 ),
(u ) u
2 4
p112 = − 1 2
p2 − 1 p12 ,
(u ) u
2 2
p122 = 1 2 − 1 p22 , p222 = 0.
(u ) u
B2

1 A 2 2 D B
p= 1
− (u ) + + 1 2 u2 − + q(u1 ),
u 4(u1 )2 u 1 (u ) A(u1 )2
and the substitution into the first equation results in the linear ODE (u1 )3 q 000 +
u1 (6u1 − A)q 00 + (6u1 − 2A)q 0 = 0 whose basis of solutions consists of 1, 1/u1
1
and e−A/u . Up to a transformation of the form u3 → u3 + αu2 + βu1 + γ this
implies (126). It was observed in [47] that the cases I and II are reciprocally re-
lated: under the change from x, t to the new independent variables X, T defined as
dX = u1 dx+u2 dt, T = t, and the introduction of the new dependent variables U 1 =
1 2 3
u1
, U 2 = − uu1 , U 3 = − uu1 , etc, the chains from the case I transform to the chains
from the case II, and vice versa. On the level of the corresponding equations this
1 2 (u2 )2 (u2 )2
means that the change of variables U 1 = u1
, U 2 = − uu1 , P = u1
− p, G = u1
−g
transforms the equations for p from the Case I to the equations for p from the Case
II. Equivalently, (125) goes to (126). The details of such a change of variables will
be demonstrated in a more general context later on.
Case III. Substituting the ansatz g = u3 /u2 + p(u1 , u2 ) into (104) one arrives at
the equations
2 2 2 4
p111 = p 1 + 2(p 12 − p 11 p 22 ) + (p1 p12 − p2 p11 ),
(u2 )2 u2
2 2
p112 = 2 p11 , p122 = 2 2 p1 ,
u (u )
2 2
p222 = 2 2 p2 − 2 p22 .
(u ) u
B + Cu1 1
p=A+ 2
+ η(u1 )(u2 )2 ,
u 6
and the substitution into the first relation results in the Chazy equation (129) for
η. Elimination of the constants A, B, C leads to (127).
Case IV. Substituting the ansatz g = ln u3 + p(u1 , u2 ) into (104) one arrives at the
equations
p111 = 2(p212 − p11 p22 ) + 2(p1 + p22 )p11 − 4p1 p2 p12 + 2p21 p22 ,
p112 = 4p1 p12 − 2p2 p11 ,
p122 = 2p11 + 2p1 p22 ,
p222 = 4p12 + 2p2 p22 .
The general solution of the fourth equation can be represented in the form
Z
1 2 1
η(u1 )du1 ;

p = − ln σ u , u −
4
here σ solves the heat equation 4σ1 = σ22 and η is a function of u1 . It is convenient
to introduce the new variable v(u1 , u2 ) by the formula v = −(ln σ)22 . Taking into
account the heat equation for σ one has
1 1 1
v1 = v22 − v 2 + (ln σ)2 v2 .
4 2 2
Rewritten in terms of v, the third equation for p implies
v22 = 6v 2 − 4vη − 4η 0 , (130)
0
≡ d/du1 , the second equation is satisfied identically and the first takes the form
8
v22 = 4v 3 − 4v 2 η − 8vη 0 − η 00 . (131)
3
This shows that v is a shift of the Weierstrass elliptic function in the variable u2 .
Since v = −(ln σ)22 , the function σ is the corresponding sigma function. Notice that
(130) can be obtained as a result of differentiation of (131) by u2 . Thus, we have

two equations for v:
1
v1 = v 2 − vη − η 0 + (ln σ)2 v2 ,
2
8
v2 = 4v − 4v η − 8vη 0 − η 00 .
2 3 2
3
The condition of their consistency leads to the Chazy equation for η.
4.4 More general conservative chains
In this section we examine chains of the form (5),
u1t = f (u1 , u2 )x , u2t = g(u1 , u2 , u3 )x , u3t = h(u1 , u2 , u3 , u4 )x , ...,
These chain are an extension of those considered in the previous section. The general
chain has been altered by only the first flux, f = f (u1 , u2 ), which is now a function
of u1 , u2 , rather than just u2 . The conditions Hjk
1
= 0 lead to expressions for all
second order partial derivatives of h in terms of g and f . The consistency conditions
of these equations result in a system expressing all third order partial derivatives
of g and f in terms of lower order derivatives. Appendix 2 shows the Mathematica
1 1
file used to calculate the components of Hjk . The condition Hjk = 0 implies the
following formulae for second order partial derivatives of h:
h4 g13 h4 g23 h4 g33

h14 = , h24 = , h34 = , h44 = 0,
g3 g3 g3
f22 g1 g13 (h3 − g2 ) + g23 g1 + g12 g3 + g33 h1
h13 =− + ,
f2 g3
(f1 − g2 )f22 (h3 − f1 )g23 + f2 g13 + h2 g33
h23 = g22 − f12 + + ,
f2 g3
f22 g3 g33 (f1 + g2 − 2h3 )
h33 = 2g23 − − , (132)
f2 g3
f22 (f12 − 2f1 g2 + g22 + f2 g1 ) (g2 − f1 )(g22 − 2f12 ) + f2 (2g12 − f11 ) + 2g23 h2
h22 =− + ,
g3 f2 g3
f22 g1 (f1 − g2 ) g22 g1 + f2 g11 + g13 h2 + g23 h1 − 2g1 f12
h12 = + ,
g3 f2 g3
g 2 f22 2g1 g12 + (f1 − g2 )g11 + 2h1 g13 − g1 f11
h11 =− 1 + .
g3 f2 g3
By calculating the consistency conditions for the above equations we obtain the ex-
pressions for all third order partial derivatives of g and f .
Equations for g:
2
2g33 2g13 g33 2g23 g33
g333 = , g133 = , g233 = ,
g3 g3 g3
2g 2 2g13 g23 2g 2
g113 = 13 , g123 = , g223 = 23 ,
g3 g3 g3
f11 (−g33 g2 g1 + g3 (g23 g1 + g2 g13 − g3 g12 )) + f1 f11 (g33 g1 − g3 g13 )

g111 =
f2 g32
f 2 (g 2 − g33 g11 ) + 2f1 ((−g33 g1 + g3 g13 )g12 + g23 (g1 g13 − g3 g11 )
+2 1 13
f2 g32
2 2 2
g2 (−g13 + g33 g11 )) + g23 g1 + g22 (−g12 g33 + 2g3 g1 g13 − g32 g11 ) + g13
2
(g22 + f2 g1 )
+2
f2 g32
g12 (2g33 g2 g1 − 2g2 g3 g13 + g32 g12 ) + g11 (−g33 g22 − f2 g33 g1 + f2 g3 g13 )
+2
f2 g32
(133)
−2g23 (g1 g3 g12 + g2 (g1 g13 − g3 g11 ))f12 (g33 g12 + g3 (−2g1 g13 + g3 g11 ))
+2 , (134)
f2 g32
(g33 g12 + g3 (g3 g11 − 2g1 g13 ))f22 + f2 (f11 (g33 g1 − g3 g13 ) − 2g13 (g2 g13 − 2g3 g12 ))
g112 =
f2 g32
2
g33 (2g1 g12 − g2 g11 ) + g23 (g3 g11 − 2g1 g13 ) + f1 (g33 g11 − g13 )
−2 ,
g32
(g1 g2 g33 + (g3 g13 − g1 g33 )f1 + (g3 g12 − g23 g1 − g2 g13 )g3 )
g122 = f22
f2 g32
2(g 2 g1 − g33 g22 g1 + g3 g22 g13 + f2 g13
2
+ (g33 g1 − g3 g13 )f12 − f2 g33 g11 )
+ 23 2
,
g3
(((f1 − 2g2 )g33 + 2g3 g23 )f1 + (g22 + f2 g1 )g33 + (g3 g22 − 2g2 g23 − f2 g13 )g3 )f22
g222 =
f2 g32
2 2
−f1 (f12 g33 + g23 − g33 g22 ) + f12 (g33 g2 − g3 g23 ) + g3 g23 g22 + g2 (g23 − g22 g33 )
+2 2
g3
f2 (g33 (f11 − 4g12 ) + 4g23 g13 )
+ .
g32
Equations for f :
f11 (g33 (f12 + g22 + f2 g1 ) − 2g2 g3 g23 + g32 g22 − f2 g3 g13 ) − 2f1 f11 (g33 g2 − g3 g23 )
f111 = −
f2 g32
g32 f11 + 2(g1 g2 g33 + f1 (−g33 g1 + g3 g13 ) + g3 (−g23 g1 − g2 g13 + g3 g12 ))
+f12
f2 g32
f22 (g33 g12 − 2g1 g3 g13 + g32 g11 )
− , (135)
f2 g32
f22 f11 g32 − f2 (f1 f11 g33 + f11 (g3 g23 − g33 g2 ) + 2f12 (g33 g1 − g3 g13 ))
f112 = ,
f2 g32
−f22 f11 g33 + f22 (f12 g32 + f2 (−g33 g1 + g3 g13 ))

f122 = ,
f2 g32
2 2
f22 g3 − 2f22 f12 g33 + f2 f22 ((f1 − g2 )g33 + g3 g23 )
f222 = .
f2 g32
We have verified that the system (134), (135) is in involution. We claim that the
functions f and g contain all the necessary information about conservative chains
(5). In particular, equations (132) allow one to reconstruct the function h and, hence,
the right hand side of the third equation of the chain. Similarly, the requirement
2
Hjk = 0 reconstructs the fourth equation (i.e., provides an involutive second order
system for the right hand side of the fourth equation), etc. Although we have verified
directly that our procedure is non-obstructed up to order 5, we could not establish
this property in general.
4.5 Integration of the general conservative chains
Notice that the first six equations in (134) are exactly the same as when (104) was
being investigated. Thus, there are four essentially different subcases to consider:
g = u3 + p(u1 , u2 ), g = u3 /u1 + p(u1 , u2 ), g = u3 /u2 + p(u1 , u2 ), g =

ln u3 + p(u1 , u2 ).
Case I: g = u3 + p(u1 , u2 ). Substituting this ansatz into (134), (135) one arrives at
two cases to consider,
Subcase I1 :
2 2
f = s(u1 )eu , p = q(u1 )eu .
The substitution of these expressions into the remaining equations for p111 and f111
leads to a system of coupled ODEs for s(u1 ) and q(u1 ):
2((q 0 )2 − qq 00 ) − q 0 s00 + 2q 00 s0
q 000 = ,
s
s00 s0 + 2s0 q 0 − qs00 − q 00 s
s000 = .
s
Setting q = −s0 , the second equation will be satisfied identically while the first
implies a fourth order ODE
s0000 s + 3(s00 )2 − 4s0 s000 = 0
whose general solution is an elliptic sigma-function: s = σ(u1 ), here (ln σ)00 =

−℘, (℘0 )2 = 4℘3 − c (notice that g2 = 0, g3 = c). Thus, as a particular case we have
2 2
f = σ(u1 )eu , p = −σ 0 (u1 )eu .
Subcase I2 :
1
f = (au1 + b)u2 + s(u1 ), p = A(u2 )2 + Bu1 u2 + q(u1 ),
2
a, b, A, B=const. The substitution of these expressions into the remaining equations

for p111 and f111 leads to linear ODEs for s(u1 ) and q(u1 ): One needs to consider
two different cases: a = 0, b = 1 and a = 1, b = 0. If a = 0, b = 1 then, up to
unessential integration constants, we have
1 B2 1 2 B 1 1
s = αe−Au , q= (u ) − α e−Au + βe−2Au .
2A A
The case a = 1, b = 0 leads to
B B 2 (A − 2) 1 2 αB
s= (u1 )2 + α(u1 )3−A , q= 2
(u ) + (u1 )3−A + β(u1 )2(2−A) .
A−1 2(A − 1) 1−A
Case II: g = u3 /u1 +p(u1 , u2 ). Substituting this ansatz into (134), (135) one arrives
at the following set of relations.
equations for p:
p22 + f12 + f2 p1 − 4p2 f1 (f1 − p2 )(f11 − 4p12 ) + 4p1 (f12 − p22 )

p111 = 2 + +
f2 (u1 )2 u1 f2
p12 (2p12 − f11 ) + 2p11 (f12 − p22 ) p11
−2 1,
f2 u
2 f11 − 4p12 2p1 f22 f22 p11
p112 = 1 2 (f1 − p2 ) + + 1 + ,
(u ) u1 u f2 f2
f2 2 f22 (p2 − f1 ) f22 p12
p122 = 2 1 2 + 1 (f12 − p22 ) + + ,
(u ) u u1 f2 f2
f22 p22 f22
p222 = + 1;
f2 u
equations for f :
f11 f2 f12 f22

f111 = (f12 − p22 − 1 ) + 1
(2p2 − 2f1 + p12 u1 ) − (2p1 + p11 u1 ),
f2 u f2 u f2 u1
f11 f22 f12 f12 f22 f22 f2
f112 = − 2 1 , f122 = − 1 , f222 = 22 .
f2 u f2 u f2
The cases I and II are reciprocally related (we thank Maxim Pavlov for point-
ing out this equivalence): under the change from x, t to the new independent
variables X, T defined as dX = u1 dx + f dt, T = t, and the introduction of
1 2 3
the new dependent variables U 1 = u1
, U 2 = − uu1 , U 3 = − uu1 , etc, the chains
from the case I transform to the chains from the case II, and vice versa. At
the level of the corresponding equations this means that the change of variables
2 u2
U1 = 1
u1
, U 2 = − uu1 , F = − uf1 , P = u1
f − p transforms the equations for p, f
from the Case I to the equations for P, F from the Case II. Lets now demonstrate
that the change of variables described previously does indeed link the cases I and
II. Introduce the new independent variables X, T, described by
dX = u1 dx + f dt,
T = t. (136)
where f is from the first equation of (5). Also, introduce the new dependent vari-
ables,
1 u2 u3
U1 = , U 2
= − , U 3
= − , ···
u1 u1 u1
using (136) uix , uit become,
dT dX
uix = uiT + uiX
dx dx
= u1 uiX , (137)
dT dX
uit = uiT + uiX
dt dx
= uiT + f uiX . (138)
Let us consider the first two equations of the chain (5),
u1t = f (u1 , u2 )x ,
u2t = g(u1 , u2 , u3 )x ,
make the change of variables described by (136) by substituting (137), (138) and
proceed to show that u1t = f (u1 , u2 )x becomes UT1 = F (U 1 , U 2 )X ,
u1t = f (u1 , u2 )x ,
rewrite in terms of X, T by substituting (137), (138),

u2

1 −1 1 2
u1T + f u1X = f1 (− 1 2 ) + f2 1 2 u1 u1X + f2 u uX ,
(u ) (u ) u1
collect T derivatives on the left hand side, X derivatives on the right, cancel u1
when it is contained as a common factor in both the numerator and denominator.
Collect terms on the right hand side so it starts to resemble a total derivative,
u2 1

f1
u1T =− +f u1X + f2 2
u − uX ,
u1 u1 X
1
multiply the whole equation by a factor (u1 )2
so that the left hand side becomes a
total derivative,
u1T u1X u2 1 u2X

f1 f2
= +f + 1 u − ,
(u1 )2 u1 1
(u ) 2 u (u1 )2 X u1
2
u1T u1X u
using the fact that UT1 = (u1 )2
, UX1 = (u1 )2
and UX2 =− u1
we have,
X

f f2 2
UT1 =− +f UX1 − U ,
u1 u1 X
1
eliminate u1 using the fact that U 1 = u1
,
UT1 = − f1 U 1 + f UX1 − f2 U 1 UX2 ,

now, it is clear to see that the right hand side is a total derivative
UT1 = − f U 1

X
which can be written,
UT1 = F (U 1 , U 2 )X ,
where F (U 1 , U 2 ) = − (f U 1 ) .
Similaraly, let us show the second equation in Case I, u2t = (u3 + p(u1 , u2 ))x , can be
3
expressed in the form UX2 = UU 1 + P (U 1 , U 2 , U 3 )X .
X
Starting with
u2t = u3 + p(u1 , u2 ) x ,

rewrite in terms of X, T by substituting (137), (138),
u2

1 p2 1 2
u2T + f u2X = −p1 1 2 + p2 1 2 u1 u1X − u uX + u1 u3X ,
(u ) (u ) u1
u2 u1T
collect T derivatives on the left hand side, X derivatives on the right. Subtract u1
1
from each side of the equation and multiply the equation by a factor u1
, this makes
the left hand side a total derivative,
u2T u2 u1T u2 u2 u1T f u2X

1 p2 2
− = −p1 1 2 + p2 1 2 u1X − u + u 3
− − ,
u1 (u1 )2 (u ) (u ) u1 X X
(u1 )2 u1
now work on the right hand side, start to arrange terms so that it looks like a total
derivative, this is slight more complicated that the first equation due to the UT1 term,
u2 −u1 u2 1 u2X f u2X u2 1

= p1 1 X2 + p2 u − 1 + u3X − − u ,
u1 T (u ) (u1 )2 x u u1 (u1 )2 T
eliminate UT1 using the first equation in the form UT1 = − (f U 1 )X ,
U3

−UT2 = p1 Ux1 + p2 UX2 − − f U 1 u2X − u2 (f U 1 )X ,
U1 X
it can now be seen that the left hand side is indeed a total derivative,
U3 u2

−UX2 =− + p−f 1 ,
U1 X u X
which takes the form,
U3

UX2 = + P (U 1 , U 2 )X
U1 X
2
where P (U 1 , U 2 ) = p − f uu1 . Thus it has been shown that cases I and II are
reciprocally related. So, this case need not be discussed any further here.
Case III: g = u3 /u2 + p(u1 , u2 ). Substituting this ansatz into (134), (135) one
arrives at the following set of relations for f and p.
Equations for p:
2p21 p1 (4p12 − f11 ) + 4p11 (f1 − p2 ) p12 (2p12 − f11 ) + 2p11 (f12 − p22 )
p111 = 2 2
+ + ,
f2 (u ) u2 f2 f2
f22 p11 p11 2f2 p1 + u2 f22 (p1 + u2 p12 )
p112 = + 2 2 , p122 = ,
f2 u f2 (u2 )2
f22 (p2 − f1 ) − f2 p22 p22 f22 f12 2
p222 =2 2
+ + 2 2 − 2 2 (f1 − p2 );
u f2 f2 u (u )
Equations for f :
f11 f12 2 f22 p11

f111 = 2
(2(f1 − p2 ) − p22 u2 ) + 2
(u (f11 + 2p12 ) + 2p1 ) − ,
f2 u f2 u f2
f22 f11 f11 f22 f12 f2 f22
f112 = + 2 , f122 = , f222 = 22 − 2 .
f2 u f2 f2 u
The last three equations for f and the last three equations for p lead to the two
essentially different possibilities:
Subcase III1 :
f = s(u1 )(u2 )k , p = q(u1 )(u2 )k+1 ,
k=const. The substitution of these expressions into the remaining equations for p111
and f111 leads to the coupled system of ODEs for s(u1 ) and q(u1 ):
k+2
q 000 = (2(k + 2)(q 0 )2 − 2(k + 1)qq 00 − q 0 s00 + 2q 00 s0 ),
ks
k + 2 00 0
s000 = (s s − k(k + 1)s00 q + 2ks0 q 0 ) − (k − 1)q 00 .
ks
Notice that under the substitution s = k = 1 the equation for q reduces to the
Chazy equation (129) for q(u1 ) = 61 η(u1 ).
Subcase III2 :
f = (au1 + b) ln u2 + s(u1 ), p = q(u1 )u2 + a(u2 ln u2 − u2 ),
a, b=const. The substitution of these expressions into the remaining equations for
p111 and f111 leads to the coupled system of ODEs for s(u1 ) and q(u1 ):
8(q 0 )2 − 4qq 00 − 2q 0 s00 + 4q 00 s0

q 000 = ,
au1 + b
000 s00 (s0 − q) + 2aq 0
s =2 1
+ q 00 .
au + b
Case IV: g = ln u3 +p(u1 , u2 ). Substituting this ansatz into (134), (135) one obtains
equations for p:
p22 (2p21 − 2p11 ) + p11 (p2 (2p2 − 4f1 ) + 2(f12 + f12 + f2 p1 ))

p111 =
f2
p12 (2p12 − f11 + 4p1 (f1 − p2 )) − p1 (2p1 f12 + (f1 − p2 )f11 )
+ ,
f2
f22 (p11 − p21 )
p112 = + p1 (4p12 − f11 ) + 2p11 (f1 − p2 ),
f2
f22
p122 = (p1 (f1 − p2 ) + p12 ) + 2p1 (p22 − f12 ) + 2f2 p11 ,
f2
f22
p222 = (p22 + f1 (2p2 − f1 ) − p22 ) + 2(p22 − f12 )(p2 − f1 ) − f22 p1 + f2 (4p12 − f11 );
f2
equations for f :
f11 2 f12
f111 = (f1 + p22 + f2 p1 − p22 − 2f1 p2 + f12 ) + 2 (p1 (f1 − p2 ) + p12 ) +
f2 f2
f22 2
(p − p11 ),
f2 1
f22 f11 f22 f12
f112 = + f11 (f1 − p2 ) + 2f12 p1 , f122 = + f2 f11 + f22 p1 ,
f2 f2
f2
f222 = 22 + 2f2 f12 + f22 (p2 − f1 ).
f2
Although this system is in involution and reduces to the corresponding previously

discussed where f = u2 , we were not able to integrate it in general. Let us just
2
mention that the last three relations for f imply the equation f11 f22 −f12 = a(u1 )f22 ,
this is an equation of Monge-Ampere form,
2

θ fxx fyy − fxy + ψfxx + ηfxy + σfyy + Θ = 0.
Where θ(x, y, f, fx , fy ) = 1, ψ = η = Θ = 0 and σ = a(u1 ). This suggests a

separable ansatz f = s(u1 )r(u2 ). A simple analysis leads to the two possibilities:
(i) f = s(u1 )(u2 )k , p = 1
1−k
s0 (u1 )(u2 )k+1 − ln u2 , where s satisfies the equation
ss00 − k
k−1
(s0 )2 = 0. Without any loss of generality one can take s = (u1 )1−k .
2 2
(ii) f = s(u1 )eu , p = −s0 (u1 )eu , where s satisfies the equation ss00 − (s0 )2 = 0. Up
1
to a linear transformation of u1 one has s = eu .
Kupershmidt’s brackets 89
5 Kupershmidt’s brackets
We consider Hamiltonian systems of the form

d d ∂h
ut = B + Bt (139)
dx dx ∂u
where h = h(u1 , u2 ), u = (u1 , u2 , u3 , ...)t is an infinite-component column vector of

d d
the dependent variables, and the matrix B of the Hamiltonian operator B dx + dx Bt
is defined as B ij = (α(i − 1) + β)ui+j−1 . Explicitly, B ij is,
 
1 2 3
 βu βu βu .... 
 
 (α + β)u2 (α + β)u3 (α + β)u4 .... 
B= .
 
 (2α + β)u3 (2α + β)u4 (2α + β)u5 .... 
 
 
.... .... .... ....
Written in component form, (139) is
u1t = 2βu1 h11 + (α + 2β)u2 h12 + βh1 ) u1x + 2βu1 h12 + (α + 2β)u2 h22 +

(α + β)h2 ) u2x ,
u2t = (α + 2β)u2 h11 + (2α + 2β)u3 h12 u1x + (α + 2β)u2 h12 +

(2α + 2β)u3 h22 + βh1 u2x + ((α + β)h2 ) u3x .

(140)
u3t = (2α + 2β)u3 h11 + (3α + 2β)u4 h12 u1x + (2α + 2β)u3 h12 +

(3α + 2β)u4 h22 u2x + (βh1 ) u3x + ((α + β)h2 ) u4x ,

...
where α and β are arbitrary constants. We impose the restraint that the components
1
Hjk of the Haantjes tensor are equal to zero. There are only finitely many non-
zero components, in fact there are just enough non-zero coefficients to provide the
expressions for all third order partial derivatives of h(u1 , u2 ) in terms of lower order
derivatives,
(2α + 3β)h222
h222 = ,
(α + β)h2
(2α + 3β)h22 h12

h122 = ,
(α + β)h2
(α + 2β)h212 + (α + β)h22 h11

h112 = , (141)
(α + β)h2
−α(α + 2β)u2 h312 + (α + β)((2α + 3β)h2 + 3(α + 2β)u2 h22 )h11 h12
h111 =
(α + β)h2 4
2β(α + 3β)u h12 h11 + 2αβu1 h22 h211
1 2
+ ;
(α + β)h2 4
here 4 = (α + β)h2 + (α + 2β)u2 h22 + 2βu1 h12 . We have verified the consistency of
this system. There are two ‘trivial’ cases which are to be excluded from the further
analysis:
(i) α + β = 0. In this case the Haantjes tensor vanishes identically for any Hamilto-
nian density h(u1 , u2 ), indeed, the first two equations for u1 , u2 form an independent
subsystem, while the remaining equations become strictly lower-triangular.
(ii) 4 = (α + β)h2 + (α + 2β)u2 h22 + 2βu1 h12 = 0. In this case the first equation
of the chain decouples from the rest, taking the form u1t = λ(u1 )u1x where λ(u1 ) is a
function of u1 . The Haantjes tensor is also identically zero.
5.1 Integration of the system of integrability conditions
A complete list of integrable Hamiltonian densities is obtained in [18] by solving

equations (141) for h(u1 , u2 ). In the general case (β 6= 0, α + 2β 6= 0) we have three
essentially different examples,
β α+2β α+2β
• h(u1 , u2 ) = (u2 + f (u1 )) α+2β , where f (u1 ) = a(u1 + c) β + b(u1 − c) β .
α+β
• h(u1 , u2 ) = (u1 + c)− β u2 + a(u1 − c)2+α/β (u1 + c)−1−α/β .
• h(u1 , u2 ) = u2 + c(u1 )2 .
Note that in all three examples a, b, c are arbitrary constants.
In the case β = 0 we have two subcases h22 = 0 and h22 6= 0. First if h22 6= 0 then
we have
h(u1 , u2 ) = ln(u2 + f (u1 )), f 00 = cf. (142)
If h22 = 0, then we obtain,
1 1
h(u1 , u2 ) = ecu u2 + e2cu , (143)
or,
h(u1 , u2 ) = u2 + c(u1 )2 . (144)
Finally, the case α + 2β = 0 again has two subcases, h22 = 0 and h22 6= 0. Beginning
with h22 6= 0, the densities obtained are,
2
h(u1 , u2 ) = f (u1 )eau , (145)
1
h√ 1
i √b
2 b2 +1
here f (u1 ) = (b2 + 1 − c2 (u1 )2 ) 2 √b +1+cu 2
.
b +1−cu1
2
The case h22 = 0 leads to one of the following,
u1 + c
h(u1 , u2 ) = (u1 + c)u2 + b(u1 + c) ln , (146)
u1 − c
or
c
h(u1 , u2 ) = u1 u2 + . (147)
u1
The more restricted case h22 = h12 = 0, again leads to (5.1).
In Sect. 5.2 it is proved that all remaining components of the Haantjes tensor of the
corresponding Hamiltonian chains vanish identically by virtue of (141). In Sect. 5.3
we demonstrate that the requirement of the existence of an additional conservation
law of the form p(u1 , u2 , u3 )t = q(u1 , u2 , u3 , u4 )x leads to the same relations (141).
As shown in Sect. 5.4, equations (141) imply the existence of a generating function
of conservation laws and, hence, the infinity of conservation laws. This establishes
the integrability of all examples constructed in this Section.
Remark 1. Since u1 is the density of the momentum, that is when h(u1 , u2 ) = u1 ,

the hydrodynamic chain (140) becomes
u1t = βu1x , u2t = βu2x , u3t = βu3x , u4t = βu4x , . . . ,
the addition of terms linear in u1 to the Hamiltonian density h effects neither the
integrability of the chain, nor the vanishing of the Haantjes tensor because the
addition of the identity martrix to a diagonalizable matrix does not affect the diag-
onalizability. Thus, the classification below is carried out up to transformations of
the form
h → ah + bu1 + c (148)
where a, b, c are arbitrary constants.
5.2 The vanishing of the Haantjes tensor
In this section we demonstrate that the relations (141), which were obtained from
1
the requirement of the vanishing of the first few components Hjk of the Haantjes
tensor, are already sufficiently restrictive and imply the vanishing of all other com-
ponents. The proof utilizes an important property of Hamiltonian chains (139),
namely, the existence of finite-component reductions for any (not necessarily inte-
grable) Hamiltonian density h [52]. Following [52], let us parametrize ui in terms of
finitely many ‘moments’ v a , a = 1, ..., n, as follows:
n
1 X α
k
u = α (v a )2+ β (k−1) ,
2 + β (k − 1) 1
k = 1, 2, 3, .... (we consider the generic case when all expressions in the denominators
are non-vanishing, see [52] for a discussion of the exceptional cases). Thus,
n n
1X a 2 1 X α
1
u = (v ) , 2
u = α (v a )2+ β , (149)
2 1 2+ β 1
etc. One can verify that under this substitution the infinite chain (139),
ukt = (α(k − 1) + β)uk (h1 )x + (α(k − 1) + β)uk+1 (h2 )x + β(uk h1 )x + (α + β)(uk+1 h2 )x ,

reduces to an n-component ‘symmetric’ conservative system for v a ,

a 1+ α
vta a
= β v h1 + (v ) β h2 , (150)
x
a = 1, ..., n, see [53]. Here h(u1 , u2 ) is an arbitrary Hamiltonian density, not necessar-
ily satisfying the integrability conditions (141), and u1 , u2 are given by (149). Notice
that the system (150) is Hamiltonian: vta = β(∂h/∂v a )x . We have the following
Theorem 7 [18] For a Hamiltonian density h(u1 , u2 ) satisfying the integrability

conditions (141), the Haantjes tensor of any finite-component reduction (150) is
identically zero.
Proof The proof follows from the explicit formulas for the Nijenhuis tensor,

1 α α α α
N23 = βv v v ((v ) − (v ) ) 2β 2 u1 (h11 h212 − h211 h22 ) + β(2 + )(αu2 − 2β(v 1 ) β u1 )
1 2 3 2 β 3 β
β
α α α
(h11 h22 h12 − h312 ) + (α + β)(2(1 + )u3 − β(2 + )2 (v 1 ) β u2 )(h11 h222 − h212 h22 )
β β
α α α α
+(α + β)h2 ([(α + 2β)h11 h12 + (v 1 v 2 v 3 ) β h222 ) + ((v 1 v 2 ) β + (v 1 v 3 ) β + (v 2 v 3 ) β )h12 h22 +
α α α α α α α
((α + β)((v 1 ) β + (v 2 ) β + (v 3 ) β ) − (2α + β)((v 1 ) β )h11 h22 + (β((v 1 ) β + (v 2 ) β + (v 3 ) β )
α
i
+(2α + β)(v 1 ) β )h212 )] ,
and

1 α α α α
N12 = β (v ) v ((v ) − (v ) ) (2 + )(αu2 − 2β(v 1 ) β u1 )(h11 h22 h12 − h312 )+
2 1 2 2 1 β 2 β
β
α α α
2β 2 u1 (h11 h212 − h211 h22 ) + ((2 + )(α − 2β)(v 1 ) β u2 − 2(α + β)(1 + )u3 )(h212 h22 −
β β

α α α α
h222 h11 ) − β(α + β)h2 v 2 (v 1 v 2 ) β (2(α + β)(1 + )u3 + (α + 2β)((v 2 ) β − (v 1 ) β )

β
α α α α α α
(v 1 )2+ β )h222 + (−α(v 1 ) β (−(2 + )u2 − 2(v 2 ) β u1 + 2((v 1 ) β − (v 2 ) β )(v 1 )2 )+
β
α 1 α α α α α α α
β((2 + )((v ) β + (v ) β )u + 2(v 1 v 2 ) β u1 + 2(1 + )u3 + (v 1 )2+ β (2(v 2 ) β − 3(v 1 ) β )+
2 2
β β
2 α α α α α
(v ) β (v ) ))h12 + 2βu h11 + (2((α + β)(v ) β + β(v 2 ) β )u1 + β(2 + )(((v 2 ) β −
1 2 2 1 2 1
β
α α α α α α
(v 1 ) β )(v 1 )2 + u2 ))h11 h12 + (2(1 + )((α + β)(v 1 ) β + β(v 2 ) β )u3 + (2 + )(v 1 ) β
β β
α 2α 2α α
i
(2(v 2 ) β (α + β)u3 + β(2(v 2 ) β − (v 1 ) β − (v 1 v 2 ) β )(v 1 )2 ))h12 h22 − (α + β)2 h2 v 2
h α α α α α α
((α − β)(v 1 ) β + β(v 2 ) β )h11 + ((α + β)(v 2 ) β − β(v 1 ) β )(v 1 v 2 ) β h22 + (2α(v 1 v 2 ) β +
i
2 2α 1 2α
β((v ) − (v ) ))h12 ,
β β
which were obtained using the integrability conditions (141). Notice that, since (150)
1 1
is invariant under permutations of v’s, it is sufficient to specify N23 and N12 only.
Note that that these expressions do not explicitly depend on the size n of the reduc-
tion (150): this dependence is contained in the variables u1 , u2 , u3 . The vanishing of
the Haantjes tensor (3) is now a straightforward algebraic calculation. Thus, for the
Hamiltonian densities h(u1 , u2 ) which satisfy the integrability conditions (141), the
corresponding hydrodynamic chains possess diagonalizable n-component reductions
(150) for any value of n. According to the results of [20], this implies that the full
Haantjes tensor of the Hamiltonian chain vanishes identically.
5.3 Conservation laws
For any Hamiltonian density h(u1 , u2 ) the system (139) necessarily possesses two
conservation laws, namely,
u1t = (2βu1 h1 + (α + 2β)u2 h2 − βh)x
and
ht = ((α + β)u3 h22 + (α + 2β)u2 h1 h2 + βu1 h21 )x ,
which correspond to the conservation of the momentum and the Hamiltonian, re-
spectively. Let us require that there exists an ‘extra’ conservation law of the form
p(u1 , u2 , u3 )t = q(u1 , u2 , u3 , u4 )x . (151)
Theorem 8 The integrability conditions (141) are necessary and sufficient for the
existence of an additional conservation law of the form (151).
Proof The proof is computational: substituting in for u1t , u2t , u3t , u4t into the left
hand side of (151), we collect coefficients at u1x , ..., u4x and equate them to zero. This
results in a system of first order partial differential equations for the flux q, note
that q is linear in u4 ,
q1 = (3α + 2β)u4 p3 h12 + (α + 2β)u2 (p1 h12 + p2 h11 ) + 2(α + β)u3 (p3 h11 + p2 h12 ) +
βh1 p1 + 2βu1 p1 h11 ,
q2 = (3α + 2β)u4 h22 p3 + (α + β)(2u3 (h12 p3 + h22 p2 ) + h2 p1 ) + (αu2 p2 + 2β(u1 p1 +
u2 p2 ))h12 + u2 (α + 2β)p1 h22 + βh1 p2 ,

q3 = βh1 p3 + (α + β)h2 p2 ,
q4 = (α + β)h2 p3 .
Calculating the consistency conditions for the flux, qij = qji , we obtain all second
order partial derivatives of the density p,
2(2α + β)((α + β)u3 h212 + βu1 h211 + (α + 2β)u2 h12 h11 )p3 + (α + β)2 p2 h2 h11
p11 = ,
(α + β)2 h22
1
p12 = (α + β)2 h2 h12 p2 + (2α + β)((α + β)(2u3 h12 + h2 h22 ) + 2βh11 h12 +
(α + β)2 h22
(α + 2β)(h212 + h11 h22 ))p3 ,

1
p22 = 2
(2α + β)((α + β)(2u3 p3 h222 + 2h12 h2 p3 ) + 2((α + 2β)u2 h22 +
(α + β) h2
βu1 h12 )p3 h12 ) + (α + β)2 h22 h2 p2 ,

(2α + β)h12 p3
p13 = ,
(α + β)h2
(2α + β)h22 p3
p23 = ,
(α + β)h2
p33 = 0.
These equations imply that p is linear in u3 , and q is linear in u4 . Upon imposing

the consistency condition (pij )k = (pik )j we get a system of third order partial dif-
ferential equations for the density h(u1 , u2 ), which is identical to (141). This finishes
the proof.
Next, it will be demonstrated in Sect. 5.4 that equations (141) imply the existence
of a generating function for conservation laws. Thus, we can claim that the exis-
tence of one additional conservation law is already very restrictive and implies the
existence of an infinity of conservation laws, thus confirming the integrability.
5.4 Generating functions for conservation laws of Hamilto-
nian chains corresponding to Kupershmidt’s bracket
In this section we present a theorem for the existence of generating functions of

conservation laws for hydrodynamic type systems of the form (139).
Theorem 9 The integrability conditions (141) are necessary and sufficient for the
existence of a generating function for conservation laws of the Hamiltonian systems
of the type (139).
Proof The structure of reductions (150) suggests that a generating function for
conservation laws should be sought in the form of
λ = λ(p, u1 , u2 , u3 , ...) (152)
so that the following Gibbons-type relation holds:
α

1+ α
λt − β h1 + (1 + α/β)p h2 λx = λp pt − β(ph1 + p
β β h2 )x ; (153)
this relation is required to be satisfied identically modulo (139). We will demonstrate

the existence of a generating function of this form for any Hamiltonian density
satisfying the integrability conditions (141). Suppose the relation (153) is already
established. Then, setting λ = const, one has λt = λx = 0, so that the relation
(153) takes the form
α

pt − β ph1 + p1+ β h2 = 0.
x
This provides an infinite sequence of conserved densities after one expands p as a

series in λ by virtue of (152). The method of generating functions is standard, and
can be traced back to [3, 22, 32, 33, 63], see also [52] for recent developments. A
detailed analysis of the relation (153) reveals that the dependence of λ on u2 , u3 , u4 ,
etc, is fixed uniquely,
∞
X β α
λ= q 1−2 α −k uk + s(u1 , q), q = pβ , (154)
k=2
while the function s(u1 , q), which specifies the dependence of λ on u1 , satisfies the
equations
β 1 −1−2 β
s1 = q −1−2 α F, sq = q α G. (155)
α
Here F and G are the following rational expressions in q depending on the Hamil-
tonian density h:
(h11 +qh12 )((α+β)qh2 −(α+2β)u2 h12 )+(α+2β)u2 h11 (h12 +qh22 )

F = (α+β)h2 [q 2 h22 +2qh12 +h11 ]+(2βqu1 −(α+2β)u2 )[h212 −h11 h22 ]
,
(α+β)2 qh2 +2β(α+2β)u1 u2 (h212 −h11 h22 )+(α+β)h2 [(α+2β)qu2 h22 −2βu1 h11 ]
G= (α+β)h2 [q 2 h22 +2qh12 +h11 ]+(2βqu1 −(α+2β)u2 )[h212 −h11 h22 ]
.
With the ansatz (154), all terms in (153) containing pt , px , u3x , u4x , ..., cancel identi-
cally, while the requirement of cancellation of the coefficients at u1x and u2x results
in (155). The functions F and G satisfy the relations
G1 = αFq − (α + 2β)q −1 F, F2 = G2 = 0, (156)
which are the consistency conditions of the equations (155). These relations are sat-
isfied identically modulo the integrability conditions (141). Conversely, the relations
(156) imply the integrability conditions (141). Note that, although the variable u2
is seemingly present in the expressions for F and G, the right hand sides of (155)
do not depend on u2 . This lack of dependance on u2 can be seen from the fact that
F2 = G2 = 0. For each of the cases arising in the classification in Sect. 5.1, the
equations (155) for s(u1 , q) can be solved explicitly. Thus, the generating function
λ can be reconstructed in closed form. This is mainly due to a simple dependence
of the derivative sq on q: integrating it with respect to q we obtain a closed form

expression which, in most of the cases, automatically solves the equation for s1 .
We consider the forms (5.1)–(147) individually below.
5.4.1 Generating functions in the general case
Here we consider densities (5.1)–(5.1). For the Hamiltonian density (5.1) the func-
tion s(u1 , q) is given by
β 2 q 1−2β/α βq(u1 − c)−α/β

h
1 1 2− α
s(u , q) = a(c − u ) βF
4abc2 (α2 − 4β 2 ) 2bc(α + 2β)
βq(u1 + c)−α/β
i
1 2− α
+b((c + u ) β F − + q −2β/α u1 ;
2ac(α + 2β)
here F (t) = 2 F1 (1 − 2β/α, 1, 2 − 2β/α, t) is the hypergeometric function of Gauss.

In the case (5.1) one has
β 2 q 1−2β/α (u1 − c)2−α/β βq(u1 − c)−α/β

1
s(u , q) = F + q −2β/α u1 ;
4ac2 (α2 − 4β 2 ) 2ac(α + 2β)
here F is the same hypergeometric function as above. Finally, the case of (5.1) leads
to

α+β
1
s(u , q) = 1
u + q q −2β/α . (157)
2c(α − 2β)
5.4.2 Generating functions in the case β = 0
Let us now show the generating functions for the case β = 0, densities (142)–(144).
Here the relation (153) takes the form
λt − ep h2 λx = λp (pt − (h1 + ep h2 )x )
where
∞
X
λ= q 1−k uk + s(u1 , q), q = ep .
k=2
The function s(u1 , q) satisfies the equations
h2 (h11 +qh12 )+u2 (h11 h22 −h212 )

s1 = u2 (h11 h22 −h212 )+h2 (h11 +2qh12 +q 2 h22 )
,
h2 (u2 h22 +h2 )

sq = u2 (h11 h22 −h212 )+h2 (h11 +2qh12 +q 2 h22 )
,
which are consistent and define s(u1 , q) explicitly: for the Hamiltonian density (142)
one has
q + f0

1 1
s(u , q) = √ Arctanh √ ;
c cf
recall that f 00 = cf . The density (143) leads to
1 1 1
s(u1 , q) = u1 + log(2cecu + q).
2 2c
The density (144) gives

s(u1 , q) = u1 + q/2c;
notice that this expression can be obtained from (157) by setting β = 0.
5.4.3 Generating functions in the case α + 2β = 0
Let’s look at densities when α + 2β = 0, (145)–(147). Here the relation (153) takes
the form (set β = 1, α = −2):
λt − h1 − p−2 h2 λx = λp pt − (h1 p + p−1 h2 )x

where
∞
X
λ= q 2−k uk + s(u1 , q), q = p−2 .
k=2
The function s(u1 , q) satisfies the equations
s1 = −qh2 (h11 +qh12 )(2u1 h12 −hh11 +qh12

1
2 )−(h12 +qh22 )(2u h11 +qh2 )
,
1
2u h11 +qh2
sq = − h22 (h11 +qh12 )(2u1 h12 −h 1
2 )−(h12 +qh22 )(2u h11 +qh2 )
,
which are consistent and define s(u1 , q) explicitly: for the Hamiltonian density (145)
one has
1
2q(1+b2 −c2 (u1 )2 )2 −cb(b−cu1 )2 −cb+2c2 u1
bArctanh √cu
1+b2
bArctanh √
c 1+b2 (1+(b−cu1 )2 )
s(u1 , q) = − √ + √
2 1 + b2 4 1 + b2
1
+ log −4cbq(1+b2 )+4q 2 (1+b2 )2 −4c3 bq(u1 )2 +4c4 q 2 (u1 )4 −c2 +8c2 qu1 (1+b2 )(1−qu1 ) .

8
The density (146) leads to
1 b
s(u1 , q) = q(u1 + c)2 + 1 − log(q(u1 − c)2 + 2cb) .

4c 2
In the case of (147) one has
1 2 1 4
s(u1 , q) = qu1 + q (u ) .
8c
Finally, the case of h22 = h12 = 0 gives
1 2
s(u1 , q) = qu1 + q ;
8c
this follows from (157) when β = 1, α = −2.

Manin-Kupershmidt’s bracket 103
6 Manin-Kupershmidt’s bracket
In this section we shall be classifying chains of the form

d d ∂h
ut = B + Bt , (158)
dx dx ∂u
here h = h(u1 , u2 , u3 ), while B ij = (i − 1)ui+j−2 or, in matrix form,
 
 0 0 0 0 0··· 
 

 u1 u2 u3 u4 u5 · · · 
 
2u2 2u3 2u4 2u5 2u6 · · · 
 

 
B ij =  3u3 3u4 3u5 3u6 3u · · ·  .
7
 
 
 

 . . . . . 

 

 . . . . . 

 
. . . . .
In component form (158) is,
u1t = um h1m + nun h1(n+1) + h2 u1x + 2um+1 h1m + nun h2(n+1) + 2h3 u2x

+ 3um+2 h1m + nun h3(n+1) u3x ,

u2t = um h2m + nun+1 h1(n+1) u1x + 2um+1 h2m + nun+1 h2(n+1) + h2 u2x

+ 3um+2 h2m + nun+1 h3(n+1) + 2h3 u3x ,

(159)
u3t = um h3m + nun+2 h1(n+1) u1x + 2um+1 h3m + nun+2 h2(n+1) u2x

+ 3um+2 h3m + nun+2 h3(n+1) + h2 u3x + 2h3 u4x ,

u4t = nun+3 h1(n+1) u1x + nun+3 h2(n+1) u2x + nun+3 h3(n+1) u3x + h2 u4x + 2h3 u5x ,

...
where m is a summation index over 1, 2, 3 and n is a summation index over 1, 2. Also

∂2h
note that terms of the form hp(q) in (159) should be read as ∂up ∂uq
. The Benney
chain (63) corresponds to h = (u3 + (u1 )2 )/2, see [33]. We have found a broad
family of new integrable Hamiltonian densities [20], thus extending the results of
[34]. Further integrable examples can be constructed by looking at Hamiltonian
densities in the form h = u3 + p(u1 , u2 ), and imposing the constraint Hjk
1
= 0. This
implies the relations
p111 (2 + u1 p22 ) = p11 p22 − p212 , p112 = p122 = p222 = 0,
which, up to a natural equivalence h → αh + au1 + bu2 + c, lead to the Hamiltonian

densities
h = u3 + α(u1 )2 + βu1 u2 + γ(u2 )2 + δ(u1 )3 , (160)
where the constants α, β, γ, δ satisfy a single relation β 2 −4αγ+12δ = 0. These densi-

ties were first introduced in [34]. Appendix 3 contains the Mathematica programme
1
that was used to calculate Hjk with h = u3 + α(u1 )2 + βu1 u2 + γ(u2 )2 + δ(u1 )3 .
i
One can prove that all other components Hjk , i ≥ 2, of the Haantjes tensor are
identically zero. The aim of this section is to characterize all densities of the form
h(u1 , u2 , u3 ) such that the Haantjes tensor of the Hamiltonian chain (158) is zero.
1
The conditions Hjk = 0 provides ten linearly independent relations which allows
expressions for all of the ten third order partial derivatives of h to be found in terms
of lower order derivatives, the simplest six of them being
5h233 5h13 h33 5h23 h33

h333 = , h133 = , h233 = ,
2h3 2h3 2h3
3h213 + 2h33 h11
h113 = , (161)
2h3
3h13 h23 + 2h33 h12
h123 = ,
2h3
3h2 + 2h33 h22
h223 = 23 ;
2h3
the remaining expressions for h111 , h112 , h122 , h222 are not written out explicitly due to
their complexity (we have used Mathematica to manipulate with these expressions).
It was verified by using Mathematica that the system for h is in involution. It was
demonstrated in [26] that the relations (161) imply the vanishing of all remaining
components of the Haantjes tensor, that is,
1 i
Hjk = 0 =⇒ Hjk = 0 for any i ≥ 2.
In fact, we have the following stronger result.
Theorem 10 [20] The relations (161) imply the existence of a generating function
which gives rise to an infinity of conservation laws. To be more precise, we claim
that for any n there exist n linearly independent conserved densities which are func-
tions of the first n coordinates ui . All higher conservation laws are polynomial in
u4 , u5 , u6 , ....
The proof of this statement is given in Sect. 6.3.

6.1 Integration of the system of integrability conditions

5/2
The first three equations in (161) imply that h33 = sh3 , s = const. Thus, there
are two cases to consider, s = 0 and s 6= 0. Let us start with the case s = 0.
Case I: s = 0. In this case h is linear in u3 and the equations (161) imply
u3
h= 1 2 2
+ p(u1 , u2 ).
(c + au + bu )
This ansatz can be simplified by utilizing the transformations [36],
1 3 1 4
ũ1 = λu1 , ũ2 = u2 , ũ3 = u, ũ4 = u ,..., (162)
λ λ2
etc, λ=const, and
ũ1 = u1 , ũ2 = u2 + su1 , ũ3 = u3 + 2su2 + s2 u1 , . . . , (163)
etc, s=const. Both transformations preserve the Poisson bracket specified by (158).
The transformations do not alter the chain. Hence, they can be used to simplify the
Hamiltonian. Suppose, for instance, that b 6= 0. Then, up to the second canonical
transformation, one can assume the ansatz h = u3 /(c + u2 )2 + p(u1 , u2 ). If b = 0
then h = u3 /(c + u1 )2 + p(u1 , u2 ). Thus, there are two subcases:
Subcase I1 : b = 0. Then h = u3 /(c + u1 )2 + p(u1 , u2 ). The substitution of this

ansatz into the four remaining equations for h gives:
16(u2 )2 + (c + u1 )3 (8(u2 )2 p22 + 4(3c + u1 )u1 p12 ) + 2(c + u1 )4 (5c + u1 )p11

p111 = − +
c(c + u1 )5 (2 + u1 (c + u1 )2 p22 )
(c + u1 )(c2 p212 − c(c − 3u1 )p22 p11 )
,
c(2 + u1 (c + u1 )2 p22 )
2u2 − c(c + u1 )3 p12 2 + c(c + u1 )2 p22

p112 = 2 , p122 = − , p222 = 0.
c(c + u1 )4 c(c + u1 )3
1+αc(c+u1 ) βu2
The last three equations lead to p(u1 , u2 ) = c(c+u1 )2
(u2 )2 + c+u1
+ q(u1 ) where α
and β are arbitrary constants. The substitution into the first equation gives a linear
ODE for q,
c2 β 2
(αc2 − 3u1 αc − 4)q 00 − (c + u1 )(1 + cu1 α)q 000 = .
2(c + u1 )3
Without any loss of generality one has
m n
q(u1 ) = 1 2
+
(c + u ) c + u1
where the constants m, n satisfy a single relation 2(1 − c2 α)n − 6cαm + 21 c2 β 2 = 0.

Finally, we have Hamiltonian densities
1 + αc(c + u1 ) 2 2 βu2 m n
h = u3 /(c + u1 )2 + 1 2
(u ) + 1
+ 1 2
+ ,
c(c + u ) c+u (c + u ) c + u1
where 2(1 − c2 α)n − 6cαm + 21 c2 β 2 = 0.
Subcase I2 : b 6= 0. Then h = u3 /(c + u2 )2 + p(u1 , u2 ). The substitution of this

ansatz into the four remaining equations for h implies that,
(u1 )3 (c + u2 )p111 − 2c(u1 )2 p11 − 2cu1 (c − u2 )p12 − (5c − 3u2 )(c − u2 )2 p22
p222 = ,
(c − u2 )3 (c + u2 )
(u1 )2 (c + u2 )p111 − 2cu1 p11 − 2(2c2 − 3cu2 + (u2 )2 )p12
p122 = ,
(c − u2 )2 (c + u2 )
u1 (c + u2 )p111 + (u2 − 3c)p11
p112 = ,
(c − u2 )(c + u2 )
along with one more relation for p111 which we do not write out due to its complexity.
u1 u2 −αu2 − αc +(c−u2 )3 g(η)
The first three relations imply p(u1 , u2 ) = 3
(c+u2 )2
where g(η) is an
u1
arbitrary function of a single variable η = c−u2
. The result of the substitution of
this ansatz into the remaining equation for p111 factors into a product of two terms,
leading to the cases (i) and (ii) below:
(i) The function g satisfies a first order ODE
4c2 η 2 g 0 − 12c2 ηg − 1 − αη + 2cη 2 = 0,
η α 1
the solution of this equation is g(η) = µη 3 + 4c
− 12c2
− 16c2 η
. This results in the
Hamiltonian densities of the form
u3 u1 u2 u2 + c/3 (u1 )3
h= + −α +µ +
(c + u2 )2 (c + u2 )2 (c + u2 )2 (c + u2 )2
1 (c − u2 )3
1
c − u2

u α
− − .
4c (c + u2 )2 c − u2 3c 4cu1
(ii) The function g solves a third order ODE
1
(4c2 η 2 g 0 −12c2 ηg−1−αη+2cη 2 )g 000 +(α−4cη+12c2 g+4c2 ηg 0 −2c2 η 2 g 00 )g 00 +(4c−8c2 g 0 )g 0 = .
2
Remarkably, this complicated equation trivializes after being differentiated by η

once, taking the form
(4c2 η 2 g 0 − 12c2 ηg − 1 − αη + 2cη 2 )g 0000 = 0.
Since the possibility when the coefficient at g 0000 equals zero was considered above in
the case (i), we conclude that g must be a cubic polynomial,
g = µ + νη + γη 2 + δη 3 ,
where the constants satisfy a single relation 12δ − 8cν + 16c2 (ν 2 − 3γµ) − 4γα + 1 = 0
which can be obtained by substituting back into the original ODE. This leads to
Hamiltonian densities of the form
αc
u3 u1 u2 − αu2 − 3
+ µ(c − u2 )3 + νu1 (c − u2 )2 + γ(u1 )2 (c − u2 ) + δ(u1 )3
h= + .
(c + u2 )2 (c + u2 )2
Notice that the particular case c = 0 results (up to obvious equivalence transforma-
tions and relabeling of constants) in Hamiltonian densities of the form
u3 (u1 )2 u1 1 (u1 )3
h= + α + β + γ + δ (164)
(u2 )2 u2 u2 u2 (u2 )2
where α, β, γ, δ satisfy a single relation β 2 − 4αγ + 12δ = 0.
Remark. The apparent similarity of the cases (160) and (164) is not accidental and
highlights a reciprocal invariance of the class of Hamiltonian chains (158). Recall
that the conservation of momentum reads
u2t = (u1 h1 + 2u2 h2 + 3u3 h3 − h)x .

Let us change from x, t to the new independent variables X, T where
dX = u2 dx + (u1 h1 + 2u2 h2 + 3u3 h3 − h)dt, T = t.
It is known that reciprocal transformations of this type preserve Poisson brackets of

the form (158), see e.g. [12] (it is crucial that u2 is the momentum of the correspond-
ing Poisson bracket). One can verify directly that performing the above change of
independent variables and introducing
h 1 u1 2 1 3 u3 n un
H= , U = , U = , U = , . . . , U = , (165)
u2 u2 u2 (u2 )3 (u2 )n
one arrives at the system which takes the original form (158) in the variables
X, T, U n , H. Thus, the above reciprocal transformation is canonical. One can verify
directly that the change of variables (165) indeed identifies (160) and (164).
Case II: s 6= 0. In this case the elementary integration gives
h = (γu3 + p(u1 , u2 ))1/3 + q(u1 , u2 ),
and the substitution into the last three equations (161) implies that q is linear. Up
to the equivalence h → αh + au1 + bu2 + c we thus have
h = (u3 + p(u1 , u2 ))1/3 , (166)
the substitution of this ansatz into the remaining equations for h implies a compli-
cated system of third order PDEs for p(u1 , u2 ). A useful observation is that this
system is invariant under a 3-parameter group of point symmetries which is gener-
ated by the two canonical transformations (162), (163) and the reciprocal transfor-
mation (165). We want to find a change of variables which simplifies the system
(161) so it is easier to work with. To find this change of variables we first find the
three infinitesimal generators of (162), (163) and (165), we then look for a change
of variables to simplify the infinitesimal generators, it is change of variables we use
on (161). Let us calculate the infinitesimal generators of these symmetries, starting
with the canonical transformation (162). From (162) we have
1 3
ũ1 = λu1 , ũ2 = u2 , ũ3 = u , ... ,
λ
making the substitution λ = er so that the change of variables is a group over addi-
tion not multiplication. For the form of (166) to be preserved in the new variables
ũi then we must have
h = (u3 + p(u1 , u2 ))1/3
= (er ũ3 + p(u1 , u2 ))1/3
= ((er ũ3 + er p̃(u1 , u2 )1/3 ,
thus we have p̃ = e−r p. We now have all the necessary information for the infinites-
imal generator
∂ ũ1 ∂ ∂ ũ2 ∂ ∂ p̃ ∂
X1 = |r=0 1 + |r=0 2 + |r=0
∂r ∂u ∂r ∂u ∂r ∂p
= u1 ∂u1 − p∂p , (167)
∂
where ∂K = ∂K
.
Now carry out the same procedure for the second canonical transformation. From
(163) we have
ũ1 = u1 , ũ2 = u2 + su1 , ũ3 = u3 + 2su2 + s2 u1 , . . . .

Applying this change of variables to (166) and insisting that the form is preserved
then p must transform as
p̃ = p − 2su2 − s2 u1 ,
now we can calculate the infinitesimal generator
∂ ũ1 ∂ ∂ ũ2 ∂ ∂ p̃ ∂
X2 = |r=0 1 + |r=0 2 + |r=0
∂r ∂u ∂r ∂u ∂r ∂p
= u1 ∂u2 − 2u2 ∂p . (168)
The method of how to calculate the infinitesimal generator of the reciprocal transfor-
mation (165) is different. We proceed as follows, substitute the change of variables
(165) into (167), (168), this should provide us with the third infinitesimal generator
up to a linear combination of (167) and (168). Making the change of variables
X1 = u1 ∂u1 − p∂p
= ũ1 ∂ũ1 − p̃∂p̃ . (169)
X2 = u1 ∂u2 − 2u2 ∂p
= −(ũ1 )2 ∂ũ1 − ũ1 ũ2 ∂ũ2 − (3p̃ũ1 + 2(ũ2 )2 )∂p̃ . (170)
The change of variables on X2 has provided the third generator. Thus, we have the
three required infinitesimal generators, which are,
X1 = u1 ∂u1 − p∂p , X2 = u1 ∂u2 − 2u2 ∂p ,
X3 = (u1 )2 ∂u1 + u1 u2 ∂u2 + (3pu1 + 2(u2 )2 )∂p ; (171)

it is straightforward to calculate the commutativity relations, for example calculating

[X1 , X2 ],
[X1 , X2 ] = X1 X2 − X2 X1
= [u1 ∂u1 − p∂p ][u1 ∂u2 − 2u2 ∂p ] − [u1 ∂u2 − 2u2 ∂p ][u1 ∂u1 − p∂p ]
= u1 ∂u2 + (u1 )2 ∂u1 u2 − 2u1 u2 ∂u1 p − pu1 ∂pu2 + 2pu2 ∂pp
− (u1 )2 ∂u1 u2 − u1 p∂u2 p − 2u2 u1 ∂u1 p + 2u2 ∂p + 2u2 p∂pp

= u1 ∂u2 − 2u2 ∂p = X2 ,
the same can be done for [X1 , X3 ] and [X1 , X3 ] to obtain the relations
[X1 , X2 ] = X2 , [X1 , X3 ] = X3 , [X2 , X3 ] = 0.
These symmetries suggest a change of variables which considerably simplifies the

equations for p. The idea is to choose new coordinates such that the symmetry
u2
generators assume the simplest possible form. Introducing ξ = − u11 , η = u1
, s=
pu1 +(u2 )2
(u1 )4
, we write the equations (171) in the new variables. We show how the
operators transform and then substitute into (171),
∂ ∂ ∂ξ ∂ ∂η ∂ ∂s
1
= 1
+ 1
+
∂u ∂ξ ∂u ∂η ∂u ∂s ∂u1
u2 ∂ 4(u2 )2 ∂

1 ∂ 3p
= − − + ,
(u1 )2 ∂ξ (u1 )2 ∂η (u1 )4 (u1 )5 ∂s
∂ ∂ ∂ξ ∂ ∂η ∂ ∂s
2
= 2
+ 2
+ (172)
∂u ∂ξ ∂u ∂η ∂u ∂s ∂u2
1 ∂ 2u2 ∂
= + ,
(u1 )−2 ∂ξ (u1 )4 ∂s
∂ ∂ ∂ξ ∂ ∂η ∂ ∂s
= + +
∂p ∂ξ ∂p ∂η ∂p ∂s ∂p
1 ∂
= .
(u1 )3 ∂s
Substituting (172) into (171)1 we obtain
X1 = u1 ∂u1 − p∂p

1 1 −2 ∂ 2 1 −2 ∂ 1 −4 2 2 1 −5 ∂ 1 ∂

= u (u ) − u (u ) − 3p(u ) + 4(u ) (u ) −p 1 3
∂ξ ∂η ∂s (u ) ∂s
u2 ∂ 4(u2 )2 ∂

1 ∂ 4p
= 1 − 1 − +
u ∂ξ u ∂η (u1 )3 (u1 )4 ∂s
= −ξ∂ξ − η∂η − 4s∂s , (173)
the same change of variables on the other two equations in (171) results in
X1 = −ξ∂ξ − η∂η − 4s∂s , X2 = ∂η , X3 = ∂ξ .
2 2
Setting s = s(ξ, η) we have p = s(ξ, η)(u1 )3 − (uu1) . In terms of s(ξ, η) the equations
for p assume a remarkable symmetric form
sξξη sηη − sηηη sξξ = 4sη , (174)
sξηη sξξ − sξξξ sηη = 4sξ , (175)
sηηη sξ + sξηη sη = sηη sξη , (176)
sξξξ sη + sξξη sξ = sξξ sξη , (177)

1
sηηη sξξ sη + sξηη sηη sξ + 4s2η = sηη (12s + s2ξη + sξξ sηη ), (178)
2
1
sξξξ sηη sξ + sξξη sξξ sη + 4s2ξ = sξξ (12s + s2ξη + sξξ sηη ); (179)
2
in the process of derivation of these equations we have assumed that sξ , sη and

sξξ s2η − sηη s2ξ are nonzero: these expressions appear as denominators in the interme-
diate calculations. Particular cases when some of these expressions vanish will be
discussed below.
To solve the equations (174) – (179) we proceed as follows. Differentiating (176),

(177) and using (174), (175) we obtain four relations among the fourth order partial
derivatives of s,
sξ sηηηη + sη sξηηη = 0, sξ sξηηη + sη (sξξηη − 4) = 0,
sξ (sξξηη − 4) + sη sξξξη = 0, sξ sξξξη + sη sξξξξ = 0,
which can be parametrized as
sηηηη = q, sξηηη = qr, sξξηη − 4 = qr2 , sξξξη = qr3 , sξξξξ = qr4 , (180)
r = −sξ /sη . The further analysis leads to the two possibilities.
Subcase II1 : q = 0. In this case s(ξ, η) is a polynomial of the form
s = ξ 2 η 2 + aξ 3 + bξ 2 η + cξη 2 + dη 3 + αξ 2 + βξη + γη 2 + µξ + νη + .
The substitution into the remaining equations for s implies the following relations
among the coefficients:
β = bc − 9ad, µ = αc − 3aγ, ν = bγ − 3αd, 12 + β 2 + 4αγ = 4(bν + cµ).
u2 (u2 )2
Notice that the corresponding p(u1 , u2 ) = s(− u11 , u1
)(u1 )3 − u1
will be a cubic
polynomial in u1 , u2 . A particular example from this class with the Hamiltonian
density h = (u3 + τ )1/3 was discussed in [34]. It corresponds to the case where
a = −τ and all other coefficients of s(ξ, η) are zero.
Subcase II2 : q 6= 0. Then the consistency conditions of (180) imply the relations
qξ = (qr)η , rξ = rrη . Taking into account that r = −sξ /sη we have sξξ s2η −sηη s2ξ = 0.
This case is discussed below.

There are two more possibilities one needs to consider to complete the classification
(notice that the equations (174) – (179) can no longer be used since they were
derived under the assumption that certain expressions do not vanish).
Subcase II3 : sξ = 0 or sη = 0. A simple analysis leads to Hamiltonian densities of

the form
1/3
(u2 )2

3 1 3
h= u + α(u ) − 1 ,
u
α =const.
s
Subcase II4 : sξξ s2η − sηη s2ξ = 0. Setting r = − sηξ one obtains
rξ = rrη , sξ = −rsη .
Calculating partial derivatives of s with the help of the above relations and sub-
1
stituting them into the conditions Hjk = 0 we obtain rηη = 0; without any loss of
generality one can set r = − ηξ . This implies ξsξ − ηsη = 0, therefore, s = s(z), z =
ξη = −u2 /(u1 )2 . For s(z) we obtain the ODE
8z 2 s000 s0 + 8zs00 s0 − 4z 2 (s00 )2 − (s0 )2 − 12s = 0
which linearizes after being differentiated by z once. Its general solution is given by
the formula
s(z) = z 2 + α + β(−z) + γ(−z)1/2 + δ(−z)3/2
where the constants α, β, γ, δ satisfy a single quadratic relation 12α + β 2 − 3γδ = 0.

This results in Hamiltonian densities of the form
1/3
h = u3 + α(u1 )3 + βu1 u2 + γ(u1 )2 (u2 )1/2 + δ(u2 )3/2 .
6.2 Conservation laws for Manin-Kupershmidt’s bracket
The result of this Section is to show that Hamiltonian chains of the type (158)
possess a conservation law of higher order than the Hamiltonian density, h. All
Hamiltonian chain (158) possess three conservation laws of the form
u1t = (u1 h2 + 2u2 h3 )x ,
u2t = (u1 h1 + 2u2 h2 + 3u3 h3 − h)x ,
and
ht = (u1 h1 h2 + 2u2 h1 h3 + u2 h22 + 3u3 h2 h3 + 2u4 h23 )x ,
which correspond to the Casimir (or Annihilator), Momentum and the Hamiltonian
respectively. Let us require the existence of an extra ‘higher’ conservation law of
the form
P (u1 , u2 , u3 , u4 )t = Q(u1 , u2 , u3 , u4 , u5 )x (181)
whose density P depends on the first four coordinates u1 , u2 , u3 , u4 . We have

obtained the following
Theorem 11 [20] The relations (161) are necessary and sufficient for the existence
of an extra conservation law of the form (181).
Proof Let us assume that there exists a conservation law of the form
P (u1 , u2 , u3 , u4 )t = Q(u1 , u2 , u3 , u4 , u5 )x .
Using (158) and collecting coefficients at uix we obtain the expressions for Qi =
∂Q/∂ui ,
2
X 3
X
5 4 1
Q1 = 5u P4 h13 + (4u P5−i h1,(i+1) + u P3−i h1,i ) + (3u3 P5−j h1,j + 2u2 P4−j h1,j ) +
i=1 j=1
P1 h2 ,
2
X 3
X
3 4 1
Q2 = 5u P4 h23 + (4u P5−i h2,(i+1) + u P3−i h2,i ) + (3u3 P5−j h2,j + 2u2 P4−j h2,j ) +
i=1 j=1
P2 h2 + 2P1 h3 ,
X 2 3
X
5 4 1
Q3 = 5u P4 h33 + (4u P5−i h3,(i+1) + u P3−i h3,i ) + (3u3 P5−j h3,j + 2u2 P4−j h3j ) +
i=1 j=1
P3 h2 + 2P2 h3 ,
Q4 = h2 P4 + 2h3 P3 ,
Q5 = 2h3 P4 .
The consistency conditions Qij = Qji imply the following expressions for second
order partial derivatives of P ,
6u4 P4 h213 + 9u3 P4 h12 h13 + 3u2 P4 (h212 + 2h13 h11 ) + 3u1 P4 h11 h12 + 2P3 h3 h11
P11 = ,
2h23
12u4 P4 h23 h13 + 9u3 P4 (h22 h13 + h23 h12 ) + 6u2 P4 (h12 (h22 + h13 ) + h23 h11 )
P12 = +
4h23
3u1 P4 (h212 + h22 h11 ) + 2h3 (3p4 h11 + 2p3 h12 )
,
4h23
12u4 P4 h13 h33 + 9u3 P4 (h23 h13 + h33 h12 ) + 6u2 P4 (h213 + h23 h12 + h33 h11 )
P13 = +
4h23
3u1 P4 (h13 h12 + h23 h11 ) + 2h3 (3P4 h12 + 2P3 h13 )
,
4h23
3P4 h13
P14 = ,
2h3
6u4 P4 h223 + 9u3 P4 h22 h23 + 3u2 P4 (h222 + 2h23 h12 ) + 3u1 P4 h12 h22
P22 = +
2h23
2h3 (3P4 h12 + P3 h22 )
2h23
12u4 P4 h23 h33 + 9u3 P4 (h223 + h22 h33 ) + 6u2 P4 (h23 (h22 + h13 ) + h33 h12 )
P23 = +
4h23
3u1 P4 (h13 h22 + h23 h12 ) + 2h3 (3P4 h13 + 2P3 h23 + 3P4 h22 )
,
4h23
3P4 h23
P24 = ,
2h3
6u4 P4 h233 + 3P4 (3u3 h33 h23 + u2 (h223 + 2h33 h13 ) + u1 h23 h13 ) + 2h3 (3P4 h23 + P3 h33 )
P33 = ,
2h23
3P4 h33
P34 = ,
2h3
P44 = 0.
By ensuring the consistency of this system we obtain the set of relations (161). We
believe that the same process applied to any higher order conservation law
P (u1 , . . . , um )t = Q(u1 , . . . , um+1 )x
will results in the same relations (161). The Mathematica program that was used
to collect coefficients and calculate the consistency conditions for this proof is con-
tained in Appendix 4.
6.3 Generating functions for conservation laws of Hamilto-
nian chains with Manin-Kupershmidt’s bracket
In this section we demonstrate that any Hamiltonian chain (158) whose density
h = h(u1 , u2 , u3 ) satisfies the system (161), possesses a generating function which
provides an infinity of conservation laws. In Sect. 3.1 we showed how to obtain
the generating function for conservation laws for the Benney chain, we now out-
line the process in general. Given an arbitrary Hamiltonian density h(u1 , u2 , u3 ), a
generating function is sought in the form
∞
1 2
X uk
λ = ψ(p, u , u ) + ; (182)
k=1
pk
notice that the dependence on higher moments u3 , u4 , u5 , etc, is exactly the same
as in (72). As demonstrated in [51], the generating function (182) has to satisfy the
Gibbons-type relation
λt − (2ph3 + h2 )λx = λp pt − (p2 h3 + ph2 + h1 )x ,

(183)
which reduces to (73) for the Benney Hamiltonian, h = (u3 + (u1 )2 )/2. Substituting
(182) into (183) and using (158), one obtains the following relations among the first
order partial derivatives ψ1 , ψ2 , ψp :
3u3 2
(ψp − 4
)(p h13 + ph12 + h11 ) + ψ1 (2u2 h13 + u1 h12 − 2ph3 ) + ψ2 (3u3 h13 +
p
1 3u3
2u2 h12 + u1 h11 ) + 3 (3u3 h12 + (2u2 + )h11 ) = 0,
p p
3u3
(ψp − 4 )(p2 h23 + ph22 + h12 ) + ψ1 (2u2 h23 + u1 h22 + 2h3 ) + ψ2 (3u3 h23 +
p
1 3u3
2u2 h22 + u1 h12 − 2ph3 ) + 3 (3u3 h22 + (2u2 + )h12 ) = 0, (184)
p p
3u3
(ψp − 4 )(p2 h33 + ph23 + h13 ) + ψ1 (2u2 h33 + u1 h23 ) + ψ2 (3u3 h33 + 2u2 h23 +
p
1 3u3
u1 h13 + 2h3 ) + 3 (3u3 h23 + (2u2 + )h13 − 2ph3 ) = 0.
p p
These relations uniquely specify the partial derivatives ψ1 , ψ2 , ψp :
ψ1 =
1 2
− 4p h3 (ph3 h33 + h23 (2h3 + u1 h13 ) − h33 (u1 h12 + u2 h22 ))
4
p(2h3 (3u3 (h33 h22 − h223 ) + 2h3 (h22 + h13 ) + u1 (h13 (2h22 + h13 ) − 2h12 h23 − h11 h33 ))
+(u1 )2 (h12 (h12 + h33 − h13 h23 ) + h23 (h11 h23 − h12 h13 ) + h22 (h213 − h11 h33 )))
h3 (6u3 (h13 h23 − h12 h33 ) − 2h3 h12 + u1 (h11 h23 − h12 h13 )) + 2u2 u1 (h12 (h12 h33 −
h13 h23 ) + h23 (h11 h23 − h12 h13 ) + h22 (h212 − h11 h33 )) ,

ψ2 =
1
− 4ph23 h33 p2 + 4h3 (h23 (h3 + u2 h23 + u1 h13 ) − h33 (u2 h22 + u1 h12 ))p+
4
(u1 )2 (h12 (h12 h33 − h13 h23 ) + h23 (h11 h23 − h12 h13 ) + h22 (h213 − h11 h33 )) +
2h3 (h13 (2h3 + 2u2 h23 + u1 (h13 + h22 )) − h12 (2u2 h33 + u1 h23 ) − u1 h11 h33 ) ,

ψp =
1 2
− 4h3 h33 u1 p2 + 4h3 p(2h3 (u2 h33 + u1 h23 ) + u2 u1 (h223 − h22 h33 ) − (u1 )2 (h13 h23
4
−h12 h33 )) + 2h3 (2h3 (3u3 h33 + 4u2 h23 + u1 (2h13 + h22 ) + (u1 )3 (h12 (h12 h33 − h13 h23 ) +
h23 (h23 h11 − h12 h13 ) + h22 (h213 − h11 h33 ))) + 4(u2 )2 (h223 − h22 h33 ) + 4u1 u2 (h13 h23 −
h12 h33 ) − 3u1 u3 (h223 − h22 h33 ) + (u1 )2 (h13 (2h22 + h13 ) − 2h23 h12 − h11 h33 )) .

Here
4=
4p4 h23 h33 + 4p3 h2 (2h3 h23 + u2 (h223 − h22 h33 ) + u1 (h13 h23 − h12 h33 )) + ((u1 )2 p2 +
2u2 u1 p)(2h3 (h13 (h13 + 2h22 ) − 2h12 h23 − h11 h33 ) + (h12 (h12 h33 − h13 h23 ) + h23 (h11 h23
−h12 h13 ) + h22 (h213 − h11 h33 ))) + 2h3 p2 (3u3 (h22 h33 − h223 ) + 2h3 (2h13 + h22 )) + 4h3 p
(3u3 (h13 h23 − h12 h33 ) − 2h3 h12 ) + 2h11 (h3 (2h3 + u1 h22 + 4u2 h23 ) + 2(u2 )2 (h223 −
h22 h33 )) − 3u3 ((h213 − h11 h33 )(2h3 + u1 h22 ) + u1 h23 (h12 h33 − h13 h23 ) + u1 h23 (h11 h23
−h12 h13 )).
Mathematica was used to verify the consistency of these equations modulo the sys-
tem (161). This proves Theorem 10: indeed, an infinite sequence of conservation
laws results from the equation
pt = (p2 h3 + ph2 + h1 )x
where one has to substitute the expansion for p in terms of λ obtained from (182).
The calculation of ψ(p, u1 , u2 ) can be summarized as follows:
— one first integrates the equation for ψp , which appears to be rational in p, with
respect to p. This defines ψ up to a function of u1 , u2 .
— one fixes this function by a substitution into the equations for ψ1 , ψ2 .
This procedure leads to the following explicit formulae for generating functions
for the Hamiltonian densities obtained in Sect. 6.1.
Linear case. This is the simplest generalization of the Benney Hamiltonian with
the density
h(u1 , u2 , u3 ) = u3 + α(u1 )2 + βu1 u2 + γ(u2 )2 + δ(u1 )3 ,
where the constants satisfy the condition β 2 − 4αγ + 12δ = 0. Even in this case the
function ψ is quite nontrivial,
" #
2 β + 2γp
ψ(p, u1 , u2 ) = 1 arctan 1 .
(4αγ − β 2 ) 2 (1 + γu1 )(4αγ − β 2 ) 2
The generating function is obtained by substituting this expression into (182).
Below we show some of the ψ functions obtained for some of the subcases from Sect
6.1.
Subcase I1 . In this case the Hamiltonian density takes the form
u3 1 + αc(c + u1 ) 2 2 βu2 m n
h(u1 , u2 , u3 ) = + (u ) + + + ,
(c + u1 )2 c(c + u1 )2 c + u1 (c + u1 )2 c + u1
where the constants satisfy the single relation, 2(1 − c2 α)n − 6cαm + 12 c2 β 2 = 0. The
corresponding function ψ is given by
3
" #
c2 2(c2 α − 1)u2 + c(2(c2 α − 1)p − βc) − cβu1
ψ(p, u1 , u2 ) = − 1 tanh−1 1 .
(3m) 2 2(3cm) 2 (1 + cαu1 )
Subcase I2 (i). For the Hamiltonian density
u3 + u1 u2 − α(u2 + 3c ) + µ(u1 )3 (c − u2 )3 u1 c − u2

1 2 3 α
h(u , u , u ) = + − − ,
(c + u2 )2 4c(c + u2 )2 c − u2 3c 4cu1
the function ψ is given by

" #
1 2 1 (u2 )2 − 2cu2 + 4cpu1 + c2
ψ(p, u , u ) = 1 arctan 1 .
(3µ) 2 4c(3µ) 2 (u1 )2
Subcase II3 : For the Hamiltonian density
(u2 )2 1
h(u1 , u2 , u3 ) = (u3 + α(u1 )3 − )3 ,
u1
the function ψ is given by

" #
1 pu1 − u2
ψ(p, u1 , u2 ) = 1 arctan 1 .
(3α) 2 (3α) 2 (u1 )2
Other examples obtained in Sect. 6.1 lead to more complicated expressions for ψ.
Conclusions and further work 125
7 Conclusions and further work
In this chapter we outline possibilities to extend the work presented in this thesis.
First, we summarize the ideas of this thesis.
We have given a simple and easy-to-verify necessary condition for the integrability
of hydrodynamic chains based on the vanishing of the Haantjes tensor; it must be
emphasized that the chain must be of class C. The utilization of Mathematica was
crucial in our ability to calculate the Haantjes tensor. We remember that vanishing
of the Haantjes tensor is not sufficient for the integrability in general, although zero
Haantjes tensor is already sufficiently restrictive and implies the integrability in the
cases of conservative chains and Hamiltonian chains. For finite-component systems
of hydrodynamic type, the following is a known result: any diagonalizable semi-
Hamiltonian system possesses infinitely many conservation laws and commuting
flows of hydrodynamic type, and can be solved by the generalized hodograph method
[61]. The work of this thesis has been to extend this idea to infinite-component
systems of hydrodynamic type, known as hydrodynamic chains. We have classified
conservative chains of the form,
u1t = f (u1 , u2 )x , u2t = g(u1 , u2 , u3 )x , u3t = h(u1 , u2 , u3 , u4 )x , . . .
first with f (u1 , u2 ) = u1 finding cases (125), (126), (127), (128) thus agreeing with
[47]. Then with f (u1 , u2 ) again finding four subcases for consideration.
We showed that for the two Hamiltonian systems (139), (158) that the conditions
imposed by zero Haantjes tensor, (141) and (161) respectively, not only leads to one
additional conservation law of higher order than the Hamiltonian density, but in
fact leads to a generating function which provides an infinity of conservation laws.
During the time spent working towards this thesis I have had two joint publications,
[18], [20].
There exists a whole variety of approaches to the classification of integrabile Hamil-

tonian hydrodynamic chains, based on seemingly different requirements, namely:
— the vanishing of the Haantjes tensor;
— the existence of infinitely many n-component hydrodynamic reductions for any n;
— the existence of one ‘extra’ conservation law;
— the existence of a generating function for conservation laws.
Each example known to us support the evidence that these approaches are equiv-
alent, by equivalent we mean they lead to the same integrability conditions and
thus the same classification results. The approach using the Haantjes tensor in our
experience has been relatively trouble free when using Mathematica, leading to the
required integrability conditions in a straightforward way.
Questions that have arisen during this work and are possible areas of further inves-
tigation include:
— Recurrent reconstruction of Conservative chains. In Sect. 4 it was found that

m
the requirement Hjk = 0 reconstructs the density of the (m + 2)th equation of (4).
This has been verified directly up to order 5, no obstructions were encountered. A
future investigation could be to verify that this iterative procedure will continue to
work and not fail at some later stage. Another area for future work is to attempt
to establish a way to reconstruct the right hand side of any equation in (4) without
doing it iteratively;
— Integrating the remaining examples in the case of the general conservative chain.
In Cases III and IV of Sect. 4.5 we were unable to complete the integration of the
systems of equations, thus we could not explicitly show the form of the chains.
Appendix 1 128
Appendix 1: the invariant formulation of the semi-
Hamiltonian property
For m-component systems (1) there exists a tensor object which is responsible for
the semi-Hamiltonian property. First of all one computes (1, 3)-tensors M and K,
s p q s p q p q p q
s
Mkij = Nkp vq Nij + Npq s
vk Nij − Npq Nik s
vj − Npq Nkj s
vi − Nkp Niqp vjq − Nkp
s p q
Nqj vi
and
s
Kkij = bsp ∂uk Nijp − bpk ∂up Nijs + Nijp ∂up bsk − Nkp
s
∂ui bpj + Nkp
s
∂uj bpi
p
+bsp ∂ui Njk − bpi ∂up Njk
s p
+ Njk ∂up bsi − Nips ∂uj bpk + Nips ∂uk bpj
p
+bsp ∂uj Nki − bpj ∂up Nki
s p
+ Nki ∂up bsj − Njp
s
∂uk bpi + Njp
s
∂ui bpk ;
here b = v 2 , that is, bij = vpi vjp . Using M and K one defines a (1, 3)-tensor Q as
Qskij = vkp Kpqj

s
viq + vkp Kpiq
s
vjq − vqp vkq Kpij
s s
− Kkpq vip vjq
+4vkp Mpij
s s
− 2Mkpj vip − 2Mkip
s
vjp .
Ultimately, one introduces a tensor P ,
s
Pkij = vps Qpkqj viq + vps Qpkiq vjq − vqs vpq Qpkij − Qskpq vip vjq . (185)
Theorem 12 [48] A diagonalizable system (1) is semi-Hamiltonian if and only if

the tensor P vanishes identically.
These objects can be calculated using computer algebra. Notice that they are well-
defined for hydrodynamic chains from the class C: all tensor operations will involve
Appendix 1 129
finite summations only. Invariant coordinate-free definitions of the above tensors

can be found in [48].
It was pointed out in [56] that any strictly hyperbolic conservative system
uit = f i (u)x , i = 1, ..., m (186)
with the zero Haantjes tensor is necessarily semi-Hamiltonian. The shortest proof
of this statement known to us can be summarized as follows. Let us first rewrite
(186) in the Riemann invariants, Rti = λi (R)Rxi , where λi 6= λj due to the strict
hyperbolicity. The conserved densities u(R) satisfy an over-determined system of
second order linear PDEs [61],
∂j λi ∂i λj
∂i ∂j u = ∂i u + ∂j u, i 6= j. (187)
The consistency conditions ∂k (∂i ∂j u) = ∂j (∂i ∂k u), i 6= j 6= k, lead to linear relations

P
among the first order derivatives of u of the form cn ∂n u = 0 where cn are certain
functions of λ’s. It is well-known that the vanishing of all coefficients cn , that
is, the involutivity of the linear system (187), is equivalent to the semi-Hamiltonian
property (38). In this case we have infinitely many conserved densities parametrized
by m arbitrary functions of a single variable. It turns out that the requirement of
the existence of m functionally independent solutions of the linear system (187) is
already sufficient to conclude that all coefficients cn must be zero. Indeed, if a
P
relation of the form cn ∂n u = 0 is satisfied by m functionally independent u’s, the
vector cn will be a zero eigenvector of the corresponding Jacobian matrix which is
non-degenerate. Thus, cn = 0 and we have shown that
(m conservation laws) + (Riemann invariants) ⇒ (∞ of conservation laws).

Appendix 1 130
It would be of interest to obtain a direct tensor proof of the above result by showing
that the vanishing of the Haantjes tensor H for a conservative system (186) implies
the vanishing of the tensor P . Such a proof would then generalize to hydrodynamic
chains, for which we have yet no rigorous definition of Riemann invariants.
Appendix 2 131
Appendix 2: Mathematica file for calculating the
Hanntjes tensor of a conservative type system

H*Mathematica Programme for calculating
the Haantjes tensor for a general conservative chain*L
H*Define the functions of the chain*L
Format@f@u1, u2DD := f
Format@g@u1, u2, u3DD := g
Format@h@u1, u2, u3, u4DD := h
Format@s@u1, u2, u3, u4, u5DD := s
H*Defining the matrix*L

Format@V11@u1, u2DD := V11;
Format@V12@u1, u2DD := V12;
Format@V21@u1, u2, u3DD := V21;

Format@V31@u1, u2, u3, u4DD := V31;

Format@V41@u1, u2, u3, u4, u5DD := V41;

Format@V51@u1, u2, u3, u4, u5, u6DD := V51;

Format@V61@u1, u2, u3, u4, u5, u6, u7DD := V61;

Format@V71@u1, u2, u3, u4, u5, u6, u7, u8DD := V71;

H*Defining the output form for the derivatives

of the functiof the chain and the elements of the matrix*L
Format@Derivative@i_, j_D@fD@a_, b_D D :=

d@ fD
;
du1 ^ i du2 ^ j
Format@Derivative@i_, j_, k_D@gD@a_, b_, c_D D :=

d@ gD
;
du1 ^ i du2 ^ j du3 ^ k
Format@Derivative@i_, j_, k_, l_D@hD@a_, b_, c_, d_D D :=

d@hD
;
du1 ^ i du2 ^ j du3 ^ k du4 ^ l
Format@Derivative@i_, j_, k_, l_, m_D@sD@a_, b_, c_, d_, e_D D :=
;
2 conservative chain for thesis.nb
d@ sD
;
du1 ^ i du2 ^ j du3 ^ k du4 ^ l du5 ^ m
Format@Derivative@i_, j_, k_, l_, m_, n_D@V51D@a_, b_, c_, d_, e_, f_D D :=
d@ V51D
;
du1 ^ i du2 ^ j du3 ^ k du4 ^ l du5 ^ m du6 ^ n
d@ V52D
;
d@ V53D
;
d@ V54D
;
d@ V55D
;
d@ V56D
;
d@ V57D
;
Format@Derivative@i_, j_, k_, l_, m_, n_, p_D@V61D@a_, b_, c_, d_, e_, f_, g_D D :=
d@ V61D
;
du1 ^ i du2 ^ j du3 ^ k du4 ^ l du5 ^ m du6 ^ n du7 ^ p
d@ V62D
;
d@ V63D
;
d@ V64D
;
d@ V65D
;
d@ V66D
;
conservative chain for thesis.nb 3
d@ V67D
;
d@ V68D
;
Format@Derivative@i_, j_, k_, l_, m_, n_, p_D@V71D@a_, b_, c_, d_, e_, f_, g_, q_D D :=
d@ V71D
;
du1 ^ i du2 ^ j du3 ^ k du4 ^ l du5 ^ m du6 ^ n du7 ^ pdu8 ^ q
d@ V72D
;
d@ V73D
;
d@ V74D
;
d@ V75D
;
d@ V76D
;
d@ V77D
;
d@ V78D
;
$HistoryLength = 3
3
4 conservative chain for thesis.nb
H* Matrix V *L
H**********************L
V = 88V11@u1, u2D , V12@u1, u2D, 0, 0, 0, 0, 0, 0, 0<,

8V21@u1, u2, u3D, V22@u1, u2, u3D, V23@u1, u2, u3D, 0, 0, 0, 0, 0, 0<,
8V31@u1, u2, u3, u4D, V32@u1, u2, u3, u4D,
8V41@u1, u2, u3, u4, u5D, V42@u1, u2, u3, u4, u5D, V43@u1, u2, u3, u4, u5D,
V33@u1, u2, u3, u4D, V34@u1, u2, u3, u4D, 0, 0, 0, 0, 0<,
8V51@u1, u2, u3, u4, u5, u6D, V52@u1, u2, u3, u4, u5, u6D,
V44@u1, u2, u3, u4, u5D, V45@u1, u2, u3, u4, u5D, 0, 0, 0, 0<,
V53@u1, u2, u3, u4, u5, u6D, V54@u1, u2, u3, u4, u5, u6D,
8V61@u1, u2, u3, u4, u5, u6, u7D, V62@u1, u2, u3, u4, u5, u6, u7D, V63@u1, u2, u3,
V55@u1, u2, u3, u4, u5, u6D, V56@u1, u2, u3, u4, u5, u6D, 0, 0, 0<,
u4, u5, u6, u7D, V64@u1, u2, u3, u4, u5, u6, u7D, V65@u1, u2, u3, u4, u5, u6, u7D,
8V71@u1, u2, u3, u4, u5, u6, u7, u8D, V72@u1, u2, u3, u4, u5, u6, u7, u8D,
V66@u1, u2, u3, u4, u5, u6, u7D, V67@u1, u2, u3, u4, u5, u6, u7D, 0, 0<,
V73@u1, u2, u3, u4, u5, u6, u7, u8D, V74@u1, u2, u3, u4, u5, u6, u7, u8D,
V77@u1, u2, u3, u4, u5, u6, u7, u8D, V78@u1, u2, u3, u4, u5, u6, u7, u8D, 0<<;
V11@u1, u2D = D@f@u1, u2D, u1D;

V12@u1, u2D = D@f@u1, u2D, u2D;
V21@u1, u2, u3D = D@g@u1, u2, u3D, u1D;

V22@u1, u2, u3D = D@g@u1, u2, u3D, u2D;
V23@u1, u2, u3D = D@g@u1, u2, u3D, u3D;
V31@u1, u2, u3, u4D = D@h@u1, u2, u3, u4D, u1D;

V41@u1, u2, u3, u4, u5D = D@s@u1, u2, u3, u4, u5D, u1D;
conservative chain for thesis.nb 5
MatrixForm@VD
d@fD d@fD
du1 du2
0 0 0 0 0 0 0
d@gD d@gD d@gD
du1 du2 du3
0 0 0 0 0 0
d@hD d@hD d@hD d@hD
du1 du2 du3 du4
0 0 0 0 0
d@sD d@sD d@sD d@sD d@sD
du1 du2 du3 du4 du5
0 0 0 0
V51 V52 V53 V54 V55 V56 0 0 0
V61 V62 V63 V64 V65 V66 V67 0 0
V71 V72 V73 V74 V75 V76 V77 V78 0
H*Setting up u@iD so it can be used in the summation to calulate the two tensors*L
u@1D = u1;
u@2D = u2;
u@3D = u3;
u@4D = u4;
u@5D = u5;
u@6D = u6;
u@7D = u7;
H* Nijenhuis tensor of the matrix V *L

H**************************************************************************************Ö
*****************L
Ne@i_, j_, k_D := CollectBSimplifyB‚ HV@@t, jDD D@V@@i, kDD, u@tDD -

i+1
t=1
V@@t, kDD D@V@@i, jDD, u@tDD -

V@@i, tDD HD@V@@t, kDD, u@jDD - D@V@@t, jDD, u@kDDLLF, u1F;
8i, 1, 4<, 8j, 1, i + 3<, 8k, 1, i + 3<D

Do@Print@StringForm@"i= ``, j= ``, k= ``, Ne= ``", i, j, k, Ne@i, j, kDDD,
H* Haantjes tensor of the matrix V *L

H************************************************************************L
H@i_, j_, k_D := SimplifyB ‚ ‚ HNe@i, t, wD V@@t, jDD V@@w, kDD -

i+2 i+3
t=1 w=1
Ne@t, j, wD V@@i, tDD V@@w, kDD -

Ne@t, w, kD V@@i, tDD V@@w, jDD +
Ne@t, j, kD V@@i, wDD V@@w, tDD L F;
H*Output of the Haantjes Tensor*L
8i, 1, 1<, 8j, 1, i + 4<, 8k, 1, i + 4<D

Do@Print@StringForm@"i= ``, j= ``, k= ``, T= ``", i, j, k, H@i, j, kDDD,
H**L
Appendix 3 137
Hanntjes tensor of a Hamiltonian system with the
Manin-Kupershmidt bracket
H*Mathematica Programme for calculating the Haantjes tensor for a hamiltonian chain*L
H*Define H ,the Hamiltonian density, as a function of three variables*L

Format@H@u1, u2, u3DD := H
Format@P@u1, u2DD := P
Format@Derivative@i_, j_, k_D@HD@a_, b_, c_D D :=

d@ HD
;
du1 ^ i du2 ^ j du3 ^ k
Format@Derivative@i_, j_D@PD@a_, b_D D :=

d@ PD
;
du1 ^ i du2 ^ j
H*Defining the matrix*L







Format@V81@u1, u2, u3, u4, u5, u6, u7, u8, u9DD := V81;
Format@V91@u1, u2, u3, u4, u5, u6, u7, u8, u9, u10DD := V91;
2 Hamiltonian Systems Classification. for thesis.nb
Format@V101@u1, u2, u3, u4, u5, u6, u7, u8, u9, u10, u11DD := V101;
$HistoryLength = 3
3
Hamiltonian Systems Classification. for thesis.nb 3
H* Matrix VL *L
H**********************L
V = 88V11@u1, u2, u3D , V12@u1, u2, u3D, V13@u1, u2, u3D, 0, 0, 0, 0, 0, 0, 0, 0, 0<,
8V21@u1, u2, u3D, V22@u1, u2, u3D, V23@u1, u2, u3D, 0, 0, 0, 0, 0, 0, 0, 0, 0<,
8V31@u1, u2, u3, u4D, V32@u1, u2, u3, u4D,

V33@u1, u2, u3, u4D, V34@u1, u2, u3, u4D, 0, 0, 0, 0, 0, 0, 0, 0<,
8V41@u1, u2, u3, u4, u5D, V42@u1, u2, u3, u4, u5D, V43@u1, u2, u3, u4, u5D,
V44@u1, u2, u3, u4, u5D, V45@u1, u2, u3, u4, u5D, 0, 0, 0, 0, 0, 0, 0<,
8V51@u1, u2, u3, u4, u5, u6D, V52@u1, u2, u3, u4, u5, u6D,
V53@u1, u2, u3, u4, u5, u6D, V54@u1, u2, u3, u4, u5, u6D,
V55@u1, u2, u3, u4, u5, u6D, V56@u1, u2, u3, u4, u5, u6D, 0, 0, 0, 0, 0, 0<,
8V61@u1, u2, u3, u4, u5, u6, u7D, V62@u1, u2, u3, u4, u5, u6, u7D,
V63@u1, u2, u3, u4, u5, u6, u7D, V64@u1, u2, u3, u4, u5, u6, u7D,
V65@u1, u2, u3, u4, u5, u6, u7D, V66@u1, u2, u3, u4, u5, u6, u7D,
V67@u1, u2, u3, u4, u5, u6, u7D, 0, 0, 0, 0, 0<,
8V71@u1, u2, u3, u4, u5, u6, u7, u8D,

V78@u1, u2, u3, u4, u5, u6, u7, u8D, 0, 0, 0, 0<,
8V81@u1, u2, u3, u4, u5, u6, u7, u8, u9D, V82@u1, u2, u3, u4, u5, u6, u7, u8, u9D,
V83@u1, u2, u3, u4, u5, u6, u7, u8, u9D, V84@u1, u2, u3, u4, u5, u6, u7, u8, u9D,
V89@u1, u2, u3, u4, u5, u6, u7, u8, u9D, 0, 0, 0<,
8V91@u1, u2, u3, u4, u5, u6, u7, u8, u9, u10D,
V92@u1, u2, u3, u4, u5, u6, u7, u8, u9, u10D, V93@u1, u2, u3, u4, u5,
u6, u7, u8, u9, u10D, V94@u1, u2, u3, u4, u5, u6, u7, u8, u9, u10D,
V95@u1, u2, u3, u4, u5, u6, u7, u8, u9, u10D, V96@u1, u2, u3, u4, u5,
u6, u7, u8, u9, u10D, V97@u1, u2, u3, u4, u5, u6, u7, u8, u9, u10D,
V98@u1, u2, u3, u4, u5, u6, u7, u8, u9, u10D, V99@u1, u2, u3, u4, u5, u6,
u7, u8, u9, u10D, V910@u1, u2, u3, u4, u5, u6, u7, u8, u9, u10D, 0, 0<,
8V101@u1, u2, u3, u4, u5, u6, u7, u8, u9, u10, u11D,
V102@u1, u2, u3, u4, u5, u6, u7, u8, u9, u10, u11D,
V103@u1, u2, u3, u4, u5, u6, u7, u8, u9, u10, u11D, V104@u1, u2, u3, u4, u5,
u6, u7, u8, u9, u10, u11D, V105@u1, u2, u3, u4, u5, u6, u7, u8, u9, u10, u11D,
u6, u7, u8, u9, u10, u11D, V108@u1, u2, u3, u4, u5, u6, u7, u8, u9, u10, u11D,
u6, u7, u8, u9, u10, u11D, V1011@u1, u2, u3, u4, u5, u6, u7, u8, u9, u10, u11D, 0<
<;
u@1D = u1;
u@2D = u2;
u@3D = u3;
u@4D = u4;
u@5D = u5;
u@6D = u6;
u@7D = u7;
u@8D = u8;
u@9D = u9;
u@10D = u10;
u@11D = u11;
H*defined h to be linear in u3*L

p = P@u1, u2D
h = Hu3 + pL ^ H1 ê 3L;
V11@u1, u2, u3D = D@h, u2D + u1 * D@h, u1, u2D + 2 * u2 * D@h, u1, u3D;
V12@u1, u2, u3D = u1 * D@h, u2, u2D + 2 * u2 * D@h, u2, u3D + 2 * D@h, u3D;
V13@u1, u2, u3D = u1 * D@h, u2, u3D + 2 * u2 * D@h, u3, u3D;
V21@u1, u2, u3D =

u1 * D@h, u1, u1D + u2 * D@h, u2, u1D + u3 * D@h, u3, u1D + u2 * D@h, u1, u2D + 2 * u3 * D@h, u1, u3D;
V22@u1, u2, u3D = u1 * D@h, u1, u2D + u2 * D@h, u2, u2D + u3 * D@h, u3, u2D +
D@h, u2D + u2 * D@h, u2, u2D + 2 * u3 * D@h, u3, u2D;
V23@u1, u2, u3D = u1 * D@h, u1, u3D + u2 * D@h, u2, u3D + u3 * D@h, u3, u3D +
u2 * D@h, u2, u3D + 2 * u3 * D@h, u3, u3D + 2 * D@h, u3D;
V31@u1, u2, u3, u4D = 2 * Hu2 * D@h, u1, u1D + u3 * D@h, u2, u1D + u4 * D@h, u3, u1DL +
2 * u4 * D@h, u1, u3D + u3 * D@h, u1, u2D;
2 * u4 * D@h, u2, u3D + u3 * D@h, u2, u2D;
2 * u4 * D@h, u3, u3D + u3 * D@h, u3, u2D + D@h, u2D;

V34@u1, u2, u3, u4D = 2 * D@h, u3D;
V41@u1, u2, u3, u4, u5D = 3 * Hu3 * D@h, u1, u1D + u4 * D@h, u2, u1D + u5 * D@h, u3, u1DL +
2 * u5 * D@h, u1, u3D + u4 * D@h, u1, u2D;
2 * u5 * D@h, u2, u3D + u4 * D@h, u2, u2D;
2 * u5 * D@h, u3, u3D + u4 * D@h, u3, u2D;

V44@u1, u2, u3, u4, u5D = D@h, u2D;
V45@u1, u2, u3, u4, u5D = 2 * D@h, u3D;
V51@u1, u2, u3, u4, u5, u6D = 4 * Hu4 * D@h, u1, u1D + u5 * D@h, u2, u1D + u6 * D@h, u3, u1DL +
2 * u6 * D@h, u1, u3D + u5 * D@h, u1, u2D;
2 * u6 * D@h, u2, u3D + u5 * D@h, u2, u2D;
2 * u6 * D@h, u3, u3D + u5 * D@h, u3, u2D;

V54@u1, u2, u3, u4, u5, u6D = 0;
V55@u1, u2, u3, u4, u5, u6D = D@h, u2D;
V56@u1, u2, u3, u4, u5, u6D = 2 * D@h, u3D;
V61@u1, u2, u3, u4, u5, u6, u7D = 5 * Hu5 * D@h, u1, u1D + u6 * D@h, u2, u1D + u7 * D@h, u3, u1DL +
2 * u7 * D@h, u1, u3D + u6 * D@h, u1, u2D;
2 * u7 * D@h, u2, u3D + u6 * D@h, u2, u2D;
2 * u7 * D@h, u3, u3D + u6 * D@h, u3, u2D;

V64@u1, u2, u3, u4, u5, u6, u7D = 0;
V65@u1, u2, u3, u4, u5, u6, u7D = 0;
V66@u1, u2, u3, u4, u5, u6, u7D = D@h, u2D;
V67@u1, u2, u3, u4, u5, u6, u7D = 2 * D@h, u3D;
V71@u1, u2, u3, u4, u5, u6, u7, u8D = 6 * Hu6 * D@h, u1, u1D + u7 * D@h, u2, u1D + u8 * D@h, u3, u1DL +
2 * u8 * D@h, u1, u3D + u7 * D@h, u1, u2D;
2 * u8 * D@h, u2, u3D + u7 * D@h, u2, u2D;
2 * u8 * D@h, u3, u3D + u7 * D@h, u3, u2D;

V74@u1, u2, u3, u4, u5, u6, u7, u8D = 0;
V75@u1, u2, u3, u4, u5, u6, u7, u8D = 0;
V76@u1, u2, u3, u4, u5, u6, u7, u8D = 0;
V77@u1, u2, u3, u4, u5, u6, u7, u8D = D@h, u2D;
V78@u1, u2, u3, u4, u5, u6, u7, u8D = 2 * D@h, u3D;
V78@u1, u2, u3, u4, u5, u6, u7, u8D = 2 * D@h, u3D;
7 * Hu7 * D@h, u1, u1D + u8 * D@h, u2, u1D + u9 * D@h, u3, u1DL +
V81@u1, u2, u3, u4, u5, u6, u7, u8, u9D =
2 * u9 * D@h, u1, u3D + u8 * D@h, u1, u2D;
V82@u1, u2, u3, u4, u5, u6, u7, u8, u9D =
2 * u9 * D@h, u2, u3D + u8 * D@h, u2, u2D;
V83@u1, u2, u3, u4, u5, u6, u7, u8, u9D =
2 * u9 * D@h, u3, u3D + u8 * D@h, u3, u2D;

V84@u1, u2, u3, u4, u5, u6, u7, u8, u9D = 0;
V85@u1, u2, u3, u4, u5, u6, u7, u8, u9D = 0;
V86@u1, u2, u3, u4, u5, u6, u7, u8, u9D = 0;
V87@u1, u2, u3, u4, u5, u6, u7, u8, u9D = 0;
V88@u1, u2, u3, u4, u5, u6, u7, u8, u9D = D@h, u2D;
V89@u1, u2, u3, u4, u5, u6, u7, u8, u9D = 2 * D@h, u3D;
V91@u1, u2, u3, u4, u5, u6, u7, u8, u9, u10D =
2 * u10 * D@h, u1, u3D + u9 * D@h, u1, u2D;
V92@u1, u2, u3, u4, u5, u6, u7, u8, u9, u10D =
2 * u10 * D@h, u2, u3D + u9 * D@h, u2, u2D;
V93@u1, u2, u3, u4, u5, u6, u7, u8, u9, u10D =
2 * u10 * D@h, u3, u3D + u9 * D@h, u3, u2D;

V94@u1, u2, u3, u4, u5, u6, u7, u8, u9, u10D = 0;
V95@u1, u2, u3, u4, u5, u6, u7, u8, u9, u10D = 0;
V96@u1, u2, u3, u4, u5, u6, u7, u8, u9, u10D = 0;
V97@u1, u2, u3, u4, u5, u6, u7, u8, u9, u10D = 0;
V98@u1, u2, u3, u4, u5, u6, u7, u8, u9, u10D = 0;
V99@u1, u2, u3, u4, u5, u6, u7, u8, u9, u10D = D@h, u2D;
V910@u1, u2, u3, u4, u5, u6, u7, u8, u9, u10D = 2 * D@h, u3D;
MatrixForm@VD
d@PD d@PD d@PD d@PD d@PD

4 u2 2
3 Hu3+PL 9 Hu3+PL 9 Hu3+PL 3 Hu3+PL2ê3

du2 du1 du2 du1 du1 du2
2ê3
- 5ê3
+ u1 - 5ê3
+
2J N
d@PD d@PD d@PD d@PD d@PD 2 d@PD
2 u3 2
3 Hu3+PL5ê3 9 Hu3+PL5ê3 3 Hu3+PL2ê3 9 Hu3+PL5ê3 3 Hu3+PL2ê3

du1 du2 du1 du1 du2 du1 du12
- + 2 u2 - + + u1 - +
2J N
d@PD d@PD d@PD d@PD d@PD d@PD d@PD d@PD d@PD 2 d@PD
4 u4 2 2 u4 2
9 Hu3+PL 9 Hu3+PL 3 Hu3+PL 9 Hu3+PL 9 Hu3+PL 3 Hu3+PL 9 Hu3+PL 3 Hu3+P

du1 du2 du1 du1 du2 du1 du2 du1 du1 du2 du1 du12
- 5ê3
+ u3 - 5ê3
+ 2ê3
+2 - 5ê3
+ u3 - 5ê3
+ 2ê3
+ u2 - 5ê3
+
2J N
4 u5 2 2 u5 2
9 Hu3+PL5ê3 9 Hu3+PL5ê3 3 Hu3+PL2ê3 9 Hu3+PL5ê3 9 Hu3+PL5ê3 3 Hu3+PL2ê3 9 Hu3+PL5ê3 3 Hu3+P

- + u4 - + +3 - + u4 - + + u3 - +
2J N
4 u6 2 2 u6 2

- 5ê3
+ u5 - 5ê3
+ 2ê3
+4 - 5ê3
+ u5 - 5ê3
+ 2ê3
+ u4 - 5ê3
+
2J N
4 u7 2 2 u7 2

- + u6 - + +5 - + u6 - + + u5 - +
2J N
4 u8 2 2 u8 2

- + u7 - + +6 - + u7 - + + u6 - +
2J N
4 u9 2 2 u9 2

- 5ê3
+ u8 - 5ê3
+ 2ê3
+7 - 5ê3
+ u8 - 5ê3
+ 2ê3
+ u7 - 5ê3
+
2J N
4 u10 2 2 u10 2

- + u9 - + +8 - + u9 - + + u8 - +
V101
H*We wish to obtain all third order derivatives of HHu1,u2,u3L*L
H* Nijenhuis tensor of the matrix V*L

H**************************************************************************************Ö
*****************L
Ne@i_, j_, k_D := SimplifyB‚ HV@@t, jDD D@V@@i, kDD, u@tDD -

i+2
t=1
V@@t, kDD D@V@@i, jDD, u@tDD -

V@@i, tDD HD@V@@t, kDD, u@jDD - D@V@@t, jDD, u@kDDLLF;
H* Haantjes tensor of the matrix V *L

H************************************************************************L
T@i_, j_, k_D := ‚ ‚ HNe@i, t, wD V@@t, jDD V@@w, kDD -

i+6 i+6
t=1 w=1
Ne@t, j, wD V@@i, tDD V@@w, kDD -
Ne@t, j, kD V@@i, wDD V@@w, tDD L ;

Ne@t, w, kD V@@i, tDD V@@w, jDD +
H*Output of the Haantjes Tensor*L
8i, 1, 1<, 8j, 1, i + 8<, 8k, 1, i + 8<D

Do@Print@StringForm@"i= ``, j= ``, k= ``, T= ``", i, j, k, Simplify@T@i, j, kDDDD,
H*Simplifying terms where lower indices are the same,

these should be idnetically zero *L
K11 = T@1, 1, 1D;
Simplify@K11D
K22 = T@2, 2, 2D;
Simplify@K22D
H*Finding condition for every component of Haanjtes to be zero*L

Se1 = Solve@T@1, 1, 2D ã 0, dD
::d Ø I-b2 + 4 a gM>>

1
12
d = Part@Part@Part@Se1, 1D, 1D, 2D
I-b2 + 4 a gM
1
12
H*Recalculating Haantjes with the above condition imposed*L
8i, 1, 1<, 8j, 1, i + 8<, 8k, 1, i + 8<D

Do@Print@StringForm@"i= ``, j= ``, k= ``, T= ``", i, j, k, Simplify@T@i, j, kDDDD,
Appendix 4 146
conditions imposed by the existence of an extra
conservation law of a Hamiltonian system with the
Manin-Kupershmidt bracket
H*Mathematica Programme for calculating the conditions imposed on a
particular chain when insisting that there exists one extra conservation law*L
H*Define the output form for functions of the chain*L
Format@H@u1, u2, u3DD := H
Format@P@u1, u2, u3, u4DD := P
Format@Q@u1, u2, u3, u4, u5DD := Q
Format@U1@x, tDD := U1
Format@Derivative@i_, j_, k_D@HD@a_, b_, c_D D :=

d@ HD
;
dU1 ^ i dU2 ^ j dU3 ^ k
Format@Derivative@i_, j_, k_, l_, m_D@QD@a_, b_, c_, d_, e_D D :=
d@QD
;
dU1 ^ i dU2 ^ j dU3 ^ k dU4 ^ l dU5 ^ m
Format@Derivative@i_, j_, k_, l_D@PD@a_, b_, c_, d_D D :=

d@PD
;
dU1 ^ i dU2 ^ j dU3 ^ k dU4 ^ l
Format@Derivative@i_, j_D@U1D@a_, b_D D :=

d@ U1D
dx ^ i dx ^ j

d@ U2D
dx ^ i dx ^ j

d@ U3D
dx ^ i dx ^ j

d@ U4D
dx ^ i dx ^ j

d@ U5D
dx ^ i dx ^ j

d@ U6D
dx ^ i dx ^ j
h = H@u1, u2, u3D;
p = P@u1, u2, u3, u4D;
q = Q@u1, u2, u3, u4, u5D;
u1 = U1@x, tD;
u2 = U2@x, tD;
u3 = U3@x, tD;
u4 = U4@x, tD;
u5 = U5@x, tD;
u6 = U6@x, tD;
H*Defining the Chain in question. Expressing

t derivative of u's interms of x derivatives of u's*L
H*u^k_t = \sum_8m=1<^83<Hk-1Lu^mHh_mL_x+Hu^kh_2+2u^8k+1<h_3L_x*L
Derivative@0, 1D@U1D@x, tD = Simplify@D@u1 * D@h, u2D + 2 * u2 * D@h, u3D, xDD;
Derivative@0, 1D@U2D@x, tD = u1 * D@D@h, u1D, xD +

u2 * D@D@h, u2D, xD + u3 * D@D@h, u3D, xD + D@u2 * D@h, u2D + 2 * u3 * D@h, u3D, xD;
Derivative@0, 1D@U3D@x, tD = 2 * u2 * D@D@h, u1D, xD +
2 * u3 * D@D@h, u2D, xD + 2 * u4 * D@D@h, u3D, xD + D@u3 * D@h, u2D + 2 * u4 * D@h, u3D, xD;
2 Extra conservation law .nb

H*Insist on conservation law and equate coefficeints of uî_x to zero

to obtain an involutive system of the first order derivaitves of Q*L
D@p, tD
d@U5D d@HD d@U4D d@HD d@U3D d@HD d@U2D d@HD d@U1D d@HD
2 + + 5 U5 + + +
dx dU3 dx dU2 dx dU3 2 dx dU2 dU3 dx dU1 dU3
d@U3D d@HD d@U2D d@HD d@U1D d@HD
4 U4 + + +
dx dU2 dU3 dx dU2 2 dx dU1 dU2
d@U3D d@HD d@U2D d@HD d@U1D d@HD d@PD
3 U3 + + +
dx dU1 dU3 dx dU1 dU2 dx dU1 2 dU4
2 + + 4 U4 + + +
3 U3 + + +
2 U2 + + +
2 + + 3 U3 + + +
2 U2 + + +
U1 + + +
d@U1D d@HD d@U3D d@HD d@U2D d@HD d@HD d@HD
+ U1 + 2 + 2 U2 + U1 +
dx dU2 dx dU2 dU3 dx dU3 dU2 dU3 dU22
2 U2 + + U1
dx dU32 dx dU1 dU3 dx dU1 dU2 dU1
D@q, xD
d@U5D d@QD d@U4D d@QD d@U3D d@QD d@U2D d@QD d@U1D d@QD
+ + + +
dx dU5 dx dU4 dx dU3 dx dU2 dx dU1
H*Eqn1 ã0 for an extra conservation law to exist*L

Extra conservation law .nb 3
Eqn1 = Simplify@D@p, tD - D@q, xDD
2 + + 5 U5 + + +
4 U4 + + +
3 U3 + + +
dx dU1 dU3 dx dU1 dU2 dx dU12 dU4
2 + + 4 U4 + + +
3 U3 + + +
2 U2 + + +
dx dU1 dU3 dx dU1 dU2 dx dU12 dU3
d@U3D d@HD d@HD d@HD d@HD
2 + 3 U3 + 2 U2 + U1 +
dx dU3 dU3 2 dU2 dU3 dU1 dU3
d@U2D d@HD d@HD d@HD d@HD
+ 3 U3 + 2 U2 + U1 +
dx dU2 dU2 dU3 dU22 dU1 dU2
d@U1D d@HD d@HD d@HD d@PD
3 U3 + 2 U2 + U1 +
dx dU1 dU3 dU1 dU2 dU1 2 dU2
d@U1D d@HD d@U3D d@HD d@U2D d@HD d@HD d@HD
+ U1 + 2 + 2 U2 + U1 +
dx dU2 dx dU2 dU3 dx dU3 dU2 dU3 dU22
2 U2 + + U1 -
dx dU3 2 dx dU1 dU3 dx dU1 dU2 dU1
d@U5D d@QD d@U4D d@QD d@U3D d@QD d@U2D d@QD d@U1D d@QD
- - - -
dx dU5 dx dU4 dx dU3 dx dU2 dx dU1
H*Looking at each coefficient of uî_x so

the conditions that need to be imposed can be seen*L
Simplify@Coefficient@Eqn1, D@u1, xDDD;
H*Equating the coefficient of uî_x to zero and solving to find

the first order derivatives of q, starting with the most simple *L
Solve@Simplify@Coefficient@Eqn1, D@u5, xDDD ã 0, D@q, u5DD
Derivative@0, 0, 0, 0, 1D@QD@u1, u2, u3, u4, u5D =

Part@Part@Part@Solve@Simplify@Coefficient@Eqn1, D@u5, xDDD ã 0, D@q, u5DD, 1D, 1D, 2D;




H*We obtain the complete system of 2nd order

derivatives of p by insisting the flux q is consistent, *L
q1 = Simplify@D@q, u1DD;
H*Ensuring consistency of the first order derivatives of q,

we obtain the 2nd order derivaitves of p*L
Simplify@D@q1, u2D - D@q2, u1DD

H*Sove these 10 consistency equations to find p_HijL*L
Simplify@D@q4, u5D - D@q5, u4DD;
Solve@Simplify@D@q4, u5D - D@q5, u4DD ã 0, D@p, u4, u4DD;
Derivative@0, 0, 0, 2D@PD@u1, u2, u3, u4D =

Part@Part@Part@Solve@Simplify@D@q4, u5D - D@q5, u4DD ã 0, D@p, u4, u4DD, 1D, 1D, 2D
0

d@HD d@PD
3
dU32 dU4
d@HD
2 dU3

d@HD d@PD
3 dU2 dU3 dU4
d@HD
2 dU3


d@HD d@PD
3 dU1 dU3 dU4
d@HD
2 dU3

Simplify@Part@Part@Part@Solve@Simplify@D@q3, u4D - D@q4, u3DD ã 0, D@p, u3, u3DD, 1D, 1D, 2DD
2
1 d@HD d@PD
I dU3 M
6 U4 +
d@HD 2 dU3 2 dU4
2
2
d@HD d@HD d@HD d@HD d@HD d@HD d@HD d@PD
3 3 U3 + U1 + U2 +2 +
dU3 2 dU2 dU3 dU2 dU3 dU1 dU3 dU2 dU3 dU3 2 dU1 dU3 dU4
d@HD d@HD d@PD d@HD d@PD
2 3 +
dU3 dU2 dU3 dU4 dU32 dU3

1
I dU3 M
d@HD 2
4
2
d@HD d@HD d@PD d@HD d@PD d@HD d@HD d@PD d@HD d@HD
12 U4 + 6 U2 +6 + 6 U2
dU3 2 dU1 dU3 dU4 dU1 dU3 dU4 dU3 dU1 dU2 dU4 dU2 dU3 dU1 dU2
d@PD d@HD d@HD d@PD d@HD d@HD d@HD d@HD d@PD
+ 3 U1 + 9 U3 + +
dU4 dU1 dU3 dU1 dU2 dU4 dU2 dU3 dU1 dU3 dU32 dU1 dU2 dU4
d@HD d@HD d@PD d@HD d@HD d@PD d@HD d@HD d@PD
6 U2 + 3 U1 +4
dU3 2
dU1 2 dU4 dU2 dU3 dU12 dU4 dU3 dU1 dU3 dU3


1 d@HD d@HD d@PD d@HD d@HD d@PD
4I M
12 U4 +6 +
d@HD 2
dU3 2 dU2 dU3 dU4 dU3 dU22 dU4
dU3
2
d@HD d@HD d@PD d@HD d@HD d@HD d@PD
6 U2 + 9 U3 + +
dU2 dU3 dU22 dU4 dU2 dU3 dU32
dU2 2 dU4
6 + 6 U2 + 3 U1 +
dU3 dU1 dU3 dU4 dU2 dU3 dU1 dU3 dU4 dU22 dU1 dU3 dU4
6 U2 + 3 U1 +4


1 d@HD d@HD d@HD d@HD d@HD d@PD
4I M J2 N
27 U32 - + -
d@HD 2 d@HD d@HD d@HD dU2 dU3 dU2 2 dU1 dU3 dU2 dU3 dU1 dU2 dU4
dU3 dU3
+ 2 U2 dU2 dU3
+ U1 2
dU2
2 2
d@HD d@HD d@HD d@HD d@PD
3 U3 6 U2 - -
dU22 dU1 dU3 dU2 dU3 dU1 2 dU4
2
d@HD d@HD d@HD d@HD d@PD d@HD d@HD d@HD d@PD d@HD d@PD
3 U1 - +2 2 - +
dU2 dU3 dU1 dU2 dU2 2
dU1 2 dU4 dU3 dU1 dU3 dU22 dU3 dU3 dU22
d@HD d@HD d@PD d@HD d@PD d@HD d@PD
-3 -2 +2 +
dU2 dU3 dU1 2 dU4 dU1 dU2 dU3 dU3 dU1 dU2
2 2
d@HD d@HD d@HD d@HD d@HD d@HD d@HD d@HD d@HD
2 -6 U22 - + -
dU1 dU3 dU1 dU2 dU2 dU3 dU1 dU3 dU12 dU2 dU3 dU1 dU2 dU22 dU12
2 2 2
d@PD d@HD d@HD d@HD d@HD d@HD d@HD d@PD
+ 2 3 U4 2 + U1 - +
dU4 dU3 dU1 dU3 dU2 2 dU1 dU3 dU2 dU3 dU1 2 dU4
2
d@HD d@HD d@PD d@PD
2 + U1 +
dU3 dU1 2 dU3 dU22
d@HD d@HD d@HD d@HD d@HD d@PD
U2 -12 U4 - + -
2 2
d@HD d@HD d@HD d@HD d@PD d@HD d@HD d@PD
3 U1 - + -6 -
dU1 2 dU4 dU3 dU1 dU2 dU4
d@HD d@HD d@PD d@HD d@HD d@PD d@HD d@PD d@HD d@PD
4 +2 3 +3 +2 +
dU1 dU3 dU1 dU2 dU3 dU1 2
dU2 2 dU4 dU1 dU3 dU4 dU2 dU3 dU3
2
d@HD d@HD d@PD d@HD d@HD d@PD
4 + - +
dU3 dU1 dU2 dU22 dU2 2 dU1 dU3 dU1 dU2


1
4I M J2 U2 N
d@HD 2 d@HD d@HD
dU3
+ U1 dU2 dU3
dU32
2 2
d@HD d@HD d@HD d@PD d@HD d@HD d@PD d@HD d@HD d@PD
54 U32 + 12 U2 + 12 U22 +
dU3 2 dU2 dU3 dU22 dU4 dU3 dU2 2 dU4 dU2 dU3 dU2 2 dU4
2
d@HD d@HD d@PD d@HD d@HD d@HD d@HD d@HD
6 U1 U2 + 12 U4 2 + 3 U3 +
dU2 2 dU1 dU3 dU4 dU2 dU3 dU3 dU2 dU3 dU3 2 dU2 dU3
2
2 U2 + 2 U2 + 2 U1 +
2 2
d@HD d@HD d@PD d@HD d@HD d@HD d@PD d@HD d@HD d@PD
24 + 48 U2 + 24 U22 +
d@HD d@HD d@HD d@PD d@HD d@HD d@HD d@PD
12 U1 + 12 U22 +
dU2 2 dU1 dU2 dU4
18 U1 U2 + 12 U1 +
dU2 dU3 dU22 dU1 dU2 dU4 dU3 dU1 dU3 dU1 dU2 dU4
12 U22 + 18 U1 U2 +
dU3 2 dU1 dU3 dU1 dU2 dU4 dU2 dU3 dU1 dU3 dU1 dU2 dU4
2
d@HD d@HD d@HD d@PD d@HD d@HD d@PD
6 U12 + 6 U1 U2 +
dU2 2 dU1 dU3 dU1 dU2 dU4 dU3 2 dU1 dU2 dU4
2
d@HD d@HD d@PD d@HD d@HD d@HD d@PD d@HD d@HD d@HD d@PD
3 U12 + 12 U2 + 6 U1 +
dU2 dU3 dU1 dU2 dU4 dU3 dU3 2
2
d@HD d@HD d@HD d@PD d@HD d@HD d@PD d@HD d@HD d@HD d@PD
12 U22 + 6 U1 U2 + 6 U1 U2 +
dU3 2 dU2 dU3 dU1 2 dU4 dU2 dU3 dU1 2 dU4 dU32 dU22 dU12 dU4
2
d@HD d@HD d@HD d@PD d@HD d@HD d@PD d@HD d@HD d@HD d@PD
3 U12 +8 + 8 U2 +
dU2 dU3 dU22 dU12 dU4 dU3 dU2 2 dU3 dU3 dU2 dU3 dU22 dU3
d@HD d@HD d@HD d@PD d@HD d@HD d@HD d@PD d@HD d@HD d@HD d@PD
4 U1 + 8 U2 + 4 U1 -
dU3 dU22 dU1 dU3 dU3 dU3 dU32 dU1 dU2 dU3 dU3 dU2 dU3 dU1 dU2 dU3
3 2 2
d@HD d@PD d@HD d@HD d@PD d@HD d@HD d@PD
8 - 8 U2 - 4 U1 +
dU3 dU2 2 dU3 dU2 dU3 dU22 dU3 dU1 dU3 dU22
2
3 U3 3 2 U2 2 + + +3 +
dU2 dU3 dU2 2
dU3 2
dU2 2
dU2 2 dU1 dU3 dU3 2 dU2 dU3 dU1 dU2
2
d@HD d@HD d@HD d@HD d@HD d@HD d@HD d@HD d@PD d@HD
U1 3 + +2 +4
dU2 dU3 dU22 dU1 dU3 dU2 dU3 dU1 dU2 dU3 2
dU2 2 dU1 dU2 dU4 dU3
2
d@HD d@HD d@PD d@HD d@HD d@PD d@HD d@PD d@HD d@HD d@PD
3 + 3 + -4
dU2 dU3 dU22 dU4 dU3 2 dU1 dU2 dU4 dU2 2 dU3 dU3 dU32 dU22


2 2
1 d@HD d@PD d@HD d@HD d@PD d@HD d@PD
I dU3 M
6 U4 + 9 U3 + 3 U2 +
d@HD 2 dU2 dU3 dU4 dU2 dU3 dU22 dU4 dU22 dU4
2
d@HD d@HD d@PD d@HD d@HD d@PD d@HD d@HD d@PD d@HD d@HD d@PD
6 + 6 U2 + 3 U1 +2
dU3 dU1 dU2 dU4 dU2 dU3 dU1 dU2 dU4 dU22 dU1 dU2 dU4 dU3 dU22 dU3
p11 = Simplify@D@p, u1, u1DD;

H*Ensure consistency of the flux, p_8ijk<=p_8ikj<,

thus obtaining a system of equations defining all 3rd order derivatives of h*L
Simplify@D@p11, u2D - D@p12, u1DD
I dU3 M
-
d@HD 3
8
2 2
d@HD d@HD d@HD d@HD d@HD d@HD d@HD d@HD
3 -2 U2 + 2 U2 +2 + 6 U2
dU2 2 dU1 dU3 dU1 dU2 dU1 dU3 dU1 dU2 dU3 dU1 dU2 dU2 dU3
2 2
+ U1 - 4 U2 -2 -
4 U2 - 10 - 2 U2 -
3 U1 + 2 U1 - 4 U2 -
d@HD d@HD d@HD d@HD d@HD d@HD d@HD
2 U1 + 4 U2 + 8 U4
dU3 dU1 dU22 dU12 dU3 dU1 dU2 dU12 dU3 dU32
- + - + +
dU1 dU3 dU1 dU2 dU2 dU3 dU1 2 dU3 dU1 dU3 dU1 dU2 dU3 dU2 dU3 dU12 dU3
d@HD d@HD d@HD d@HD d@HD d@HD
4 U2 - 4 U2 -
dU3 dU2 2 2
2
3 U3 2 - + + +
dU3 2 dU1 dU2 dU3 dU1 dU2 dU1 dU2 dU3 dU1 dU3 dU1 dU22
2 2
2 + +2 -2 -
dU2 dU3 dU12 dU22 dU1 dU3 dU32 dU12 dU3 dU12 dU3
d@HD d@HD d@HD d@HD d@HD
3 +2 +
dU2 dU3 dU1 dU3 dU1 dU2 dU3 dU12 dU2
2
d@HD d@HD d@HD d@HD d@HD d@HD d@HD d@HD d@PD
4 + 4 U2 + 2 U1
dU3 dU1 3 dU3 dU2 dU3 dU13 dU3 dU22 dU13 dU4
3
1 d@HD d@HD d@HD d@HD
8I M
- 3 2 U2 -4 +
d@HD 3 dU1 dU3 dU3 dU1 dU3 dU1 dU2
dU3
2 2
2 U2 + U1 + 2 U2 -
2
4 U2 -6 - 4 U2 -
dU3 dU1 dU2 dU1 dU2 dU3 dU3 dU2 dU3 dU1 2 dU2 dU3 dU12
2 U2 - 3 U1 - 4 U2 +
dU3 2 dU1 dU3 dU1 2 dU2 dU3 dU1 dU3 dU1 2 dU3 dU1 dU3 2
dU12
2 U1 - 2 U1 - 2 U1 -
dU3 2 dU1 dU2 dU1 2 dU3 dU1 dU2 dU3 dU1 2 dU3 dU1 dU2 dU12 dU3
2 2
8 U4 + - + +
dU3 2
dU1 2
2
4 + 4 U2 + 2 U1 +
dU3 2
2
3 U3 -2 + + -
4 +2 + 3 +2 +
dU3 2
dU1 2 dU3 dU1 dU3 2
dU3 2 dU1 dU3 dU1 dU2 dU3 dU12 dU2
d@HD d@HD d@HD d@HD d@HD d@HD d@PD
4 U2 + 2 U1
dU3 dU32 dU13 dU3 dU2 dU3 dU13 dU4
3 J3 I dU1 dU3 M + 2 N dU4

d@HD 2 d@HD d@HD d@HD d@HD d@PD
-2 dU3 dU12 dU3
dU32 dU12
4I M
d@HD 2
dU3
2 2
1 d@HD d@HD d@HD d@HD d@HD d@HD d@HD
8I M
- 3 -2 U2 + 4 U2 + 2 U2 -
d@HD 3
dU2 2 dU1 dU3 dU2 2 dU1 dU3 dU2 dU3 dU2 2 dU1 dU2
dU3

4 U2 -4 - 6 U2 -
2 2
2 U1 - 4 U2 - U1 +
4 U2 + 8 U4 - +
- + +
dU3 dU2 dU3 dU1 dU3
2 dU2 dU3 dU1 dU2 dU3
4 U2 - 4 U2 +
dU3 dU22 dU1 dU22 dU3 dU1 dU3 dU1 dU22
2
U3 -3 -6 + - +
dU2 dU3 dU1 dU2 dU3 dU2 3 dU1 dU3 dU2 dU3 dU1 dU2
2
dU2 2 dU1 dU2 dU3
3 +2 -6 +
dU2 dU3 dU2 2 dU1 dU3 dU3 dU1 dU2 2 dU3 dU2 dU3 dU12
2
2 U2 + 4 U2 + 3 U1 -
dU2 dU3 dU1 2
dU3 2
dU2 2
2
4 U2 - 2 U1 +4 +
dU3 dU2 dU3 dU1
2 2 dU3 dU2 3
dU1 2 dU3 dU12 dU2
4 U2 + 2 U1
dU3 dU2 dU3 dU12 dU2 dU3 dU22 dU12 dU2 dU4
2
1 d@HD d@HD d@HD d@HD d@HD d@HD d@HD d@HD d@HD
8I M
- 3 -2 + 4 U2 +8 + 2 U2
d@HD 3 dU3 dU2 2 dU1 dU3 dU2 dU3 dU2 2 dU1 dU3 dU3 dU1 dU3 dU2 dU3
dU3
2 2 2
+ 2 U1 -8 - 6 U2 +
dU1 dU3 dU2 2 dU1 dU3 dU3 dU2 dU3 dU1 dU2 dU2 dU3 dU1 dU2
2 U2 - 2 U2 - 3 U1 -
dU3 2
dU2 2 dU1 dU2 dU3 2 dU1 dU3 dU1 dU2 dU2 dU3 dU1 dU3 dU1 dU2
2
4 U2 + U1 - 4 U2 +
dU3 dU1 dU3 2 dU1 dU2 dU3 2 dU1 dU2 dU3 dU2 2 dU1 dU2 dU3
4 U2 - 2 U1 - 8 U4
2
d@HD d@HD d@HD d@HD d@HD d@HD d@HD d@HD d@HD d@HD
+ - + +
dU3 2 dU1 dU2 dU3 2 dU2 dU3 dU1 dU3 dU3 dU1 dU2 dU3
2
4 + 4 U2 + 2 U1 +
dU3 dU1 dU2 2 dU3 dU2 dU3 dU1 dU2 2 dU3 dU1 dU3 dU1 dU22
2
3 U3 -2 -4 +
dU2 dU3 dU1 dU3 dU3 dU22 dU1 dU32 dU3 2 dU2 dU3 dU1 dU2
2
3 +2 +2 - U1 +
dU3 2
dU2 2 dU1 dU3 dU3 dU1 dU22 dU3 dU32 dU12 dU2 dU3 dU12
2
U1 -4 - 4 U2 -
dU3 2
dU2 2
dU1 2 dU3 dU1 dU3 2 dU3 dU2 dU3 dU12 dU3
d@HD d@HD d@HD d@HD d@HD d@HD d@HD d@HD d@HD d@PD
2 U1 + 4 U2 + 2 U1
2
1 d@HD d@HD d@HD d@HD d@HD d@HD d@HD d@HD
8I M
- 3 -2 - 2 U2 +2 +
d@HD 3 dU3 dU22 dU1 dU3 dU2 dU3 dU22 dU1 dU3 dU3 dU1 dU3
dU3
2 2
2 U2 - U1 -8 +
dU2 dU3 dU1 dU3 dU2 2 dU1 dU3 dU3 dU2 dU3 dU1 dU2
2 U2 - 4 U2 - 8 U2 -
dU32
dU22 dU1 dU2 dU3 dU22 dU3 dU1 dU2 dU3 2 dU1 dU3 dU1 dU2
2
3 U1 + U1 - 2 U1 - 8 U4
dU2 dU3 dU1 dU3 dU1 dU2 dU3 2 dU1 dU2 dU3 dU1 dU2 dU1 dU2 dU3
2
+ - + +
dU3 dU2 dU32 dU1 dU3 dU3 2 dU1 dU2 dU3 2 dU2 dU3 dU1 dU3 dU3 dU1 dU2 dU3
2
4 + 4 U2 + 2 U1 +
dU3 dU1 dU2 2 dU3 dU2 dU3 dU1 dU22 dU3 dU1 dU3 dU1 dU22
2
3 U3 + - +2 -
dU2 dU3 dU1 dU3 dU2 dU3 dU3 2 dU1 dU2 dU3 dU1 dU2 dU3
2 + - -
dU3 dU2 dU3 dU1 dU3
2
dU2 dU3 2 dU1 dU2 dU32
dU1 dU22
2
2 + 6 U2 + 2 U1 -
dU3 dU32 dU12 dU3 2 dU2 dU3 dU12 dU2 dU3 dU12
4 U2 + U1 - 2 U1 +
dU2 2
4 U2 + 2 U1
dU3 dU3 2
dU1 dU2 2 dU3 dU2 dU3 dU1 dU2 2 dU4
3 J3 N dU4
dU2 dU3 dU1 dU3
+2 -2
4I M
d@HD 2
dU3
3 J3 N dU4
dU2 dU3 dU1 dU3
+2 -2
4I M
d@HD 2
dU3
1 d@HD d@HD d@HD d@HD d@HD
I dU3 M
- 3 8 U4 - + +
d@HD 3 dU3 dU33 dU1 dU3 dU3 2
dU1 dU32
8
2
4 + 3 3 U3 - + -
dU3 dU1 dU2 dU3 dU3 2 dU2 dU3 dU1 dU3 dU3 2 dU1 dU2
2 2
2 U2 - + - +
dU2 dU3 dU1 dU3 dU3 2 dU2 dU3 dU1 dU2 dU3 2 dU1 dU3 dU32 dU12
2
U1 + -2 + -
dU3 2 dU1 dU3 dU1 dU2 dU2 dU3 dU1 dU3 dU32 dU12
2 2 + 2 U2 + U1 -
dU3 dU3 2 dU1 dU2 dU2 dU3 2 dU1 dU2 dU1 dU3 2 dU1 dU2
U1 + 3 U3 + - +
dU1 dU3 dU1 dU2 dU3 dU2 dU3 2 dU1 dU3 dU3 3 dU1 dU2 dU3 2 dU1 dU2 dU3
2 U2 + U1 - 2 U2 +
3 2 2 2 2
3 - 3 U3 - 2 U2 - U1
dU2 dU3 dU1 dU3 dU1 dU3 2 dU1 dU2 dU3 2
dU1 dU3 dU4
3 J5 N
-2
4I M
d@HD 2
dU3
3 J5 N
-2
4I M
d@HD 2
dU3
0
I dU3 M
-
d@HD 3
8
2 2
3 -10 - 2 U2 + 2 U2 - 4 U2 -
dU3 2
2 2
8 U4 + - + +
dU2 2
2
4 + 4 U2 +
dU3 dU2 3 dU3 dU2 dU3 dU23
3
3 U3 - +2 - + +
dU2 dU3 dU3 2 dU2 dU3 dU2 2 dU3 dU2 dU3 2
dU2 2
dU32 dU23
2
10 + 2 U2 - 8 U2 -
dU3 dU2 dU3 dU1 dU3 dU2 dU3 dU1 dU3 dU3 2
dU2 2 dU1 dU3
3 U1 + 4 U2 + 2 U1 +
dU2 dU3 dU2 2 dU1 dU3 dU3 dU2 dU3 dU1 dU3 2 dU3 dU2 3 dU1 dU3
2
6 U2 + U1 - 4 U2 +
dU3 2 dU2 dU3 dU1 dU2 dU2 dU3 dU1 dU2 dU3 dU2 dU32 dU1 dU2
2
2 U1 - 2 U1 -4 -
dU3 2
2 dU3 dU1 dU2 dU3
4 U2 - 2 U1 +
dU3 dU2 dU3 dU1 dU2 dU3 dU3 dU22 dU1 dU2 dU3
4 U2 + 2 U1
dU3 dU3 2
dU1 dU2 2 dU3 dU2 dU3 dU1 dU2 2 dU4
3 J3 I dU2 dU3 M + 2 N dU4

d@HD 2 d@HD d@HD d@HD d@HD d@PD
-2 dU3 dU22 dU3
dU32 dU22
4I M
d@HD 2
dU3
1 d@HD d@HD d@HD d@HD d@HD
I dU3 M
- 3 8 U4 - + +
d@HD 3 dU3 dU3 3 dU2 dU3 dU3 2
dU2 dU32
8
2 2
4 - + 3 3 U3 - + -
dU3 2
dU2 dU3 dU1 dU3 2
3 2
2 U2 + - + - +
dU2 dU3 dU3 2 dU2 dU3 dU2 2 dU1 dU3 dU3 2 dU1 dU2
2
U1 -2 + + -
dU2 2 dU1 dU3 dU3 2 dU2 dU3 dU1 dU2
2
2 3 + 3 U3 + 2 U2 - 2 U2 -
dU3 dU2 dU3 dU3 3
dU2 2
dU2 dU3 2
dU2 2
dU2 dU3 2 dU1 dU3
U1 + U1 + 2 U2 +
dU2 dU3 dU1 dU3
2
dU2 2
dU1 dU3 2
dU3 3 dU1 dU2
U1 - 2 U2 - + U1 +
dU2 dU3 2 dU1 dU2 dU2 dU3 dU2 dU3 2
dU1 dU3 2 dU1 dU2 dU3
2 - 3 U3 -5 - 2 U2
dU3 2
dU2 2
dU2 dU3 2 dU1 dU3 dU1 dU2 dU3 dU4
3 J5 N
-2
4I M
d@HD 2
dU3
3 J5 N
-2
4I M
d@HD 2
dU3
3 K5 J N -2 O dU4
d@HD 2 d@HD d@HD d@PD
dU32 dU3 dU33
4I M
d@HD 2
dU3
0
3 K5 J N -2 O dU4
d@HD 2 d@HD d@HD d@PD
dU32 dU3 dU33
4I M
d@HD 2
dU3
Solve@Simplify@D@p33, u4D - D@p34, u3DD ã 0, D@h, u3, u3, u3DD;
Derivative@0, 0, 3D@HD@u1, u2, u3D =

Part@Part@Part@Solve@Simplify@D@p33, u4D - D@p34, u3DD ã 0, D@h, u3, u3, u3DD, 1D, 1D, 2D;
Simplify@D@p23, u4D - D@p34, u2DD;










h111 = Simplify@D@h, u1, u1, u1DD
3 I dU1 dU3 M + 2
d@HD 2 d@HD d@HD
dU32 dU12
d@HD
2 dU3
d@HD d@HD d@HD d@HD

3 dU2 dU3 dU1 dU3
+2
dU32 dU1 dU2
d@HD
2 dU3
d@HD d@HD
5
dU32 dU1 dU3
d@HD
2 dU3
3 I dU2 dU3 M + 2
d@HD 2 d@HD d@HD
dU32 dU22
d@HD
2 dU3
d@HD d@HD
5
dU32 dU2 dU3
d@HD
2 dU3
5J N
d@HD 2
dU32
d@HD
2 dU3
References 168
References
[1] S.J. Alber, Associated integrable systems, J. Math. Phys. 32 (1991) 916-922.
[2] M.J.Ablowitz and H.Segur, Solitons and the Inverse scattering transform. Siam
(1981).
[3] D.J. Benney, Some properties of long nonlinear waves, Stud. Appl. Math. 52
(1973) 45-50.
[4] D.J. Benney, Conservation laws associated with long surface waves, Journal of
Computational and Applied Mathematics 190 (2006) 136-141.
[5] M. Blaszak, Classical R-matrices on Poisson algebras and related dispersionless

systems. Phys. Lett. A 297, no. 3-4 (2002) 191-195.
[6] V.M. Bukhshtaber, D.V. Leikin and M.V. Pavlov, Egorov hydrodynamic chains,
the Chazy equation, and the group SL(2, C), Funktsional. Anal. i Prilozhen.
37, no. 4 (2003) 13-26.
[7] J. Chazy, Sur les équations différentielles dont l’intégrale générale poss ède un
coupure essentielle mobile, C.R. Acad. Sci. Paris 150 (1910) 456-458.
[8] B.A. Dubrovin and S.P. Novikov, Hydrodynamics of weakly deformed soliton
lattices: differential geometry and Hamiltonian theory, Russian Math. Surveys,
44 , no. 6 (1989) 35-124.
[9] B.A. Dubrovin, A.T.Fomenko and S.P. Novikov, Geometry-Methods and Ap-
plications, Part I: The Geometry of Surfaces of Transformation Groups, and
Fields. Springer (1984)
[10] B.A. Dubrovin, Geometry of 2D topological field theories, Lect. Notes in Math.
1620, Springer-Verlag (1996) 120-348.
References 169
[11] B.A. Dubrovin, S. Liu and Y. Zhang, On Hamiltonian perturbations of hyper-

bolic systems of conservation laws. I. Quasi-triviality of bi-Hamiltonian pertur-
bations, Comm. Pure Appl. Math. 59, no. 4 (2006) 559-615.
[12] E.V. Ferapontov, Nonlocal Hamiltonian operators of hydrodynamic type: dif-

ferential geometry and applications, Amer. Math. Soc. Transl. (2), 170 (1995)
33-58.
[13] E.V. Ferapontov and K.R. Khusnutdinova, On integrability of (2+1)-

dimensional quasilinear systems, Comm. Math. Phys. 248 (2004) 187-206;
arXiv:nlin.SI/0305044.
[14] E.V. Ferapontov and K.R. Khusnutdinova, The characterization of two-

component (2+1)-dimensional integrable systems of hydrodynamic type, J.
Phys. A: Math. Gen. 37, N8 (2004) 2949 - 2963; arXiv:nlin.SI/0310021.
[15] E.V. Ferapontov and K.R. Khusnutdinova, Hydrodynamic reductions of multi-

dimensional dispersionless PDEs: the test for integrability, J. Math. Phys. 45,
N6 (2004) 2365-2377; arXiv:nlin.SI/0312015.
[16] E.V. Ferapontov and K.R. Khusnutdinova, Double waves in multi-dimensional

systems of hydrodynamic type: the necessary condition for integrability, Proc.
R. Soc. A 462 (2006) 1197-1219; arXiv:nlin.SI/0412064.
[17] E.V. Ferapontov, K.R. Khusnutdinova and M.V. Pavlov, Classification of in-
tegrable (2+1)-dimensional quasilinear hierarchies, Theor. Math. Phys. 144
(2005) 35-43.
[18] E.V. Ferapontov, K.R. Khusnutdinova, D.G. Marshall and M.V. Pavlov, Clas-
sification of integrable Hamiltonian hydrodynamic chains associated with Ku-
pershmidt’s brackets, J. Math. Phys. 47, no.1 (2006).
References 170
[19] E.V. Ferapontov, K.R. Khusnutdinova and S.P. Tsarev, On a class of three-
dimensional integrable Lagrangians, Comm. Math. Phys. 261, N1 (2006) 225-
243.
[20] E.V. Ferapontov and D.G. Marshall, Differential-geometric approach to

the integrability of hydrodynamic chains: the Haantjes tensor, (2005);
arXiv:nlin.SI/0505013.
[21] C. S. Gardner, C. S. Greene, M. D.Kruskal, and R. M. Miura, Method for

Solving the Korteweg-de Vries Equation, Phys. Rev. Lett. 19, (1967) 1095-
1097,
[22] J. Gibbons, Collisionless Boltzmann equations and integrable moment equa-

tions, Physica 3D (1981) 503-511.
[23] J. Gibbons and Y. Kodama, A method for solving the dispersionless KP hier-
archy and its exact solutions. II, Phys. Lett. A 135, no. 3 (1989) 167-170.
[24] J. Gibbons and S. P. Tsarev, Reductions of the Benney equations, Phys. Lett.
A 211 (1996) 19-24.
[25] J. Gibbons and S. P. Tsarev, Conformal maps and reductions of the Benney
equations, Phys. Lett. A 258 (1999) 263-271.
[26] J. Gibbons and A. Raimondo, Differential Geometry of Hydrodynamic Vlasov

Equations, arXiv:nlin.SI/0612022.
[27] J. Haantjes, On Xm -forming sets of eigenvectors, Indagationes Mathematicae

17 (1955) 158-162.
[28] K.R. Khusnutdinova and S.P. Tsarev, private communication.

References 171
[29] B.G. Konopelchenko and L. Martinez Alonso, Dispersionless scalar integrable

hierarchies, Whitham hierarchy, and the quasiclassical ∂-dressing method, J.
Math. Phys. 43, no. 7 (2002) 3807-3823.
[30] B.G. Konopelchenko, L. Martinez Alonso and O. Ragnisco, The ∂-approach to

the dispersionless KP hierarchy, J. Phys. A 34, no. 47 (2001) 10209-10217.
[31] I.M. Krichever, The τ -function of the universal Whitham hierarchy, matrix
models and topological field theories, Comm. Pure Appl. Math. 47, no. 4 (1994)
437-475.
[32] B.A. Kupershmidt and Yu.I. Manin, Long wave equations with a free surface.
I. Conservation laws and solutions. (Russian) Funktsional. Anal. i Prilozhen.
11, no. 3 (1977) 31-42.
[33] B.A. Kupershmidt and Yu.I. Manin, Long wave equations with a free surface.
II. The Hamiltonian structure and the higher equations. (Russian) Funktsional.
Anal. i Prilozhen. 12, no. 1 (1978) 25-37.
[34] B.A. Kupershmidt, Deformations of integrable systems. Proc. Roy. Irish Acad.
Sect. A 83, no. 1 (1983) 45-74.
[35] B.A. Kupershmidt, The quasiclassical limit of the modified KP hierarchy, J.

Phys. A: Math. Gen 23 (1990) 871-886.
[36] B.A. Kupershmidt, Normal and universal forms in integrable hydrodynamical

systems, Proceedings of the Berkeley-Ames conference on nonlinear problems in
control and fluid dynamics (Berkeley, Calif., 1983), in Lie Groups: Hist., Fron-
tiers and Appl. Ser. B: Systems Inform. Control, II, Math Sci Press, Brookline,
MA, (1984) 357-378.
[37] B.A. Kupershmidt, Extended equations of long waves, preprint, 2005.

References 172
[38] N.H. Ibragimov et al, Lie group analysis of differential equations.
[39] D.R. Lebedev and Yu.I. Manin, Conservation laws and Lax representation for
Benney’s long wave equations, Phys. lett. 74A (1979) 154-156.
[40] M. Mañas, On the rth dispersionless Toda hierarchy: factorization problem,

additional symmetries and some solutions, J. Phys. A: Math. Gen. 37 (2004)
9195-9224.
[41] M. Mañas, S-functions, reductions and hodograph solutions of the rth disper-
sionless modified KP and Dym hierarchies, J. Phys. A: Math. Gen. 37 (2004)
11191-11221.
[42] M. Mañas, L. Martinez Alonso and E. Medina, Reductions and hodograph

solutions of the dispersionless KP hierarchy, J. Phys. A: Math. Gen. 35 (2002)
401-417.
[43] Mathematica and Wolfram Mathematica are trademarks of Wolfram Research,

Inc.: www.wolfram.com
[44] V.G. Mikhalev, Hamiltonian formalism of Korteweg-de Vries-type hierarchies.

(Russian) Funktsional. Anal. i Prilozhen. 26, no. 2 (1992) 79-82; translation in
Funct. Anal. Appl. 26, no. 2 (1992) 140-142.
[45] R.M. Miura, Conservation laws for fully nonlinear long wave equations, Stud.
Appl. Math. 53 (1974) 45-56.
[46] M.V. Pavlov, Integrable hydrodynamic chains, J. Math. Phys. 44, no. 9 (2003)
4134-4156.
[47] M.V. Pavlov, Classification of integrable Egorov hydrodynamic chains, Teoret.

Mat. Fiz. 138, no. 1 (2004) 55-70.
References 173
[48] M.V. Pavlov, S.I. Svinolupov and R.A. Sharipov, An invariant criterion for
hydrodynamic integrability, Funktsional. Anal. i Prilozhen. 30 (1996) 18-29;
translation in Funct. Anal. Appl. 30 (1996) 15-22.
[49] M.V. Pavlov, private communication, 2005.
[50] M.V. Pavlov, Modified dispersionless Veselov–Novikov equations and corre-

sponding hydrodynamic chains, arXiv:nlin.SI/0611022.
[51] M.V. Pavlov, The Hamiltonian approach in classification and integrability of

hydrodynamic chains, arXiv:nlin.SI/0603057.
[52] M.V. Pavlov, The Hamiltonian approach in the classification of integrable hy-
drodynamic chains, nlin.SI/0603057.
[53] M.V. Pavlov, Algebro-geometric approach in the theory of integrable hydrody-

namic type systems, nlin.SI/0603054.
[54] M.V. Pavlov, Hydrodynamic chains and a classification of their Poisson brack-
ets, nlin.SI/0603056.
[55] Z. Peradzyński, Nonlinear plane k-waves and Riemann invariants, Bull. Acad.
Polon. Sci. Sr. Sci. Tech. 19 (1971) 625-632.
[56] B. Sévennec, Géométrie des systèmes hyperboliques de lois de conserva-

tion, Mémoire (nouvelle série) N56, Supplément au Bulletin de la Société
Mathématique de France 122 (1994) 1-125.
[57] L. Martinez Alonso and A.B. Shabat, Towards a theory of differential con-
straints of a hydrodynamic hierarchy, J. Nonlinear Math. Phys. 10, no. 2 (2003)
229-242.
References 174
[58] L. Martinez Alonso and A.B. Shabat, Energy-dependent potentials revisited:

A universal hierarchy of hydrodynamic type, Phys. lett. A 299, no. 4 (2002)
359-365; Phys. Lett. A 300, no. 1 (2002) 58-64.
[59] V.M. Teshukov, On the hyperbolicity of long wave equations, Dokl. Akad. Nauk
SSSR 284, no. 3 (1985) 555-559.
[60] V.M. Teshukov, Characteristics, conservation laws and symmetries of the ki-
netic equations of motion of bubbles in a fluid, Prikl. Mekh. Tekhn. Fiz. 40,
no. 2 (1999) 86-100; translation in J. Appl. Mech. Tech. Phys. 40, no. 2 (1999)
263-275.
[61] S.P. Tsarev, Geometry of Hamiltonian systems of hydrodynamic type. General-

ized hodograph method, Izvestija AN USSR Math. 54, no. 5 (1990) 1048-1068.
[62] V.E. Zakharov, Benney equations and quasiclassical approximation in the in-
verse problem method, Funktsional. Anal. i Prilozhen. 14, no. 2 (1980) 15-24.
[63] V.E. Zakharov, On the Benney equations, Physica 3D, no. 1,2 (1981) 193-202.

A Classification of Integrable Hydrodynamic Type Chains Using The Haantjes Tensor

Transféré par

Informations du document

Description originale:

Titre original

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

A Classification of Integrable Hydrodynamic Type Chains Using The Haantjes Tensor

Transféré par

Droits d'auteur :

Formats disponibles

Classification of integrable hydrodynamic

chains using the Haantjes tensor.

Department of Mathematical Sciences

The integrability of an m-component system of hydrodynamic type, ut = v(u)ux ,

Prof. E. V. Ferapontov and Dr K. R. Khusnutdinova, who gave me such superb and

All the of staff of the Mathematical Sciences Department at Loughborough Univer-

Loughborough University, particularly the Mathematical Sciences Department, for

All my friends for their moral support and understanding.

2 Hydrodynamic type systems in 1+1 dimensions 15

5.3 Conservation laws . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96

6 Manin-Kupershmidt’s bracket 103

7 Conclusions and further work 125

This thesis is a study in the area of integrable systems. It is devoted to a classification

where u = (u1 , u2 , ..., un )t is an n-component column vector and v(u) is an n × n

useful because calculating the eigenvalues and eigenvectors can be troublesome in

and introduce the Haantjes tensor [27]

Theorem 1 [27] A hydrodynamic type system with mutually distinct characteris-

where V (u) is a square matrix of infinite dimension, while u = (u1 , u2 , . . . , )t is

In Sect. 3 we propose that a hydrodynamic chain is said to be diagonalizable if all

example let us consider the chain

Definition 2 A hydrodynamic chain from the class C is said to be integrable if,

The method of hydrodynamic reductions is illustrated in Sect. 3.4 by considering

In Sect. 3.4 the following theorem is proved:

Theorem 2 The vanishing of the Haantjes tensor H is a necessary condition for

Theorem 3 provides an easy-to-verify necessary condition for testing the integrabil-

Conjecture 1 The vanishing of both tensors H and P is necessary and sufficient

In Sect. 4 the Haantjes tensor criterion is used to classify conservative chains.

u1t = u2x , u2t = g(u1 , u2 , u3 )x , u3t = h(u1 , u2 , u3 , u4 )x , ... . (4)

u1t = f (u1 , u2 )x , u2t = g(u1 , u2 , u3 )x , u3t = h(u1 , u2 , u3 , u4 )x , ... . (5)

are of the form

where u = (u1 , u2 , u3 , ...)t is an infinite-component column vector of the dependent

• Kupershmidt’s bracket: B ij = (α(i − 1) + β)ui+j−1 , h = h(u1 , u2 ), see [34];

• Manin-Kupershmidt’s bracket: B ij = (i − 1)ui+j−2 , h = h(u1 , u2 , u3 ), see [33];

so that the corresponding Hamiltonian system is integrable. The Kupershmidt’s

• p(u1 , u2 , u3 )t = q(u1 , u2 , u3 , u4 )x for Hamiltonian systems with Kupershmidt’s

• p(u1 , u2 , u3 , u4 )t = q(u1 , u2 , u3 , u4 , u5 )x for Hamiltonian systems with Manin-

The main results of this thesis were published in [18], [20].

2 Hydrodynamic type systems in 1+1 dimensions

An m-component system of hydrodynamic type has the form,

where v(u) is a m×m matrix and u is a column vector with components u1 , u2 , u3 , . . . ,

2.1 Riemann invariants

u1 , . . . , un → R1 (u), . . . , Rn (u) (8)

so that (7) goes to

To find Riemann invariants of a 2x2 system, proceed as follows:

An example follows to further highlight the steps to calculate Riemann Invariants.

Example Calculating Riemann invariants for the equations of gas dynamics

ρt + ux ρ + uρx = 0, ut + uux + γργ−2 ρx = 0. (10)

Write (10) in matrix form

Comparing (11) to (9) we see that

In step one we find the eigenvalues of vji ,

λ1,2 = u ± (γργ−1 )1/2 . (12)

In step two we find the Riemann invariants by requiring,

We obtain R2 by the same method, ultimately,

So we have equation (9) for i = 1, 2, thus

Rt1 + λ1 Rx1 = 0, Rt2 + λ2 Rx2 = 0. (14)

Note at this point that from (13) we have

Substitute (15) into (12) to obtain

If (16) is substituted into (14) then we have

Thus, R1 and R2 are Riemann invariants of (10).

uix + ai (u)t = 0, i = 1, · · · , n. (18)

First, we calculate the eigenvalues of matrix A where

uix + Aij ujt = 0, (20)