On the inverse problem of calculus of variations

arXiv:0809.4845v1 [nlin.SI] 28 Sep 2008

M.C. Nucci and A.M. Arthurs∗

Dipartimento di Matematica e Informatica, Università di Perugia, 06123 Perugia, Italy,

e-mail: [email protected]

Abstract
We show that given an ordinary differential equation of order four, it may be possible to
determine a Lagrangian if the third derivative is absent (or eliminated) from the equation.
This represents a subcase of Fels’conditions [M. E. Fels, The inverse problem of the calculus
of variations for scalar fourth-order ordinary differential equations, Trans. Amer. Math.
Soc. 348 (1996) 5007-5029] which ensure the existence and uniqueness of the Lagrangian
in the case of a fourth-order equation. The key is the Jacobi last multiplier as in the case
of a second-order equation. Two equations from a Number Theory paper by Hall, one of
second and one of fourth order, will be used to exemplify the method. The known link
between Jacobi last multiplier and Lie symmetries is also exploited. Finally the Lagrangian
of two fourth-order equations drawn from Physics are determined with the same method.

1 Introduction
It is well-known that a Lagrangian always exists for any second-order ordinary differential
equation [30]. What seems less known is that the key is the Jacobi last multiplier [11], [12],
[13], [30], which has many interesting properties, a list of which can be found in [26].
It is a matter for historians alike to find out why Darboux [3], Helmholtz [10], Koenigs-
berger himself1 and many other successive authors, e.g. Douglas [4] and Havas2 [7], never
acknowledged the use of the Jacobi last multiplier in order to find Lagrangians of a second-
order equation.
∗
Permanent address: Department of Mathematics, University of York, Heslington, York YO10 5DD, U.K.,
e-mail: [email protected]
1
In 1902-1903 Koenigsberger wrote Helmholtz’s biography [17] – which in 1906 was (abridged) translated into
English with a Preface by Lord Kelvin [19] – after he wrote his 1901-book on Mechanics [16]. Neither books cite
the connection between Jacobi last multiplier and Lagrangians. In 1904 Koenigsberger wrote Jacobi’s biography
[18] where the Jacobi Last Multiplier is extensively described.
2
Havas even cites the book by Whittaker [30] but only in connection with the formulation of Lagrangian
equations.

1
 The Norwegian Sophus Lie, who carefully studied Jacobi’s work [9], found a connection
between his groups of transformations and the Jacobi last multiplier [20], [21]. The Italian
Bianchi presented Lie’s work in his lectures on finite continuous groups of transformations,
and described quite clearly the Jacobi Last multiplier and its properties [1]. Neither of them
cited the connection with Lagrangians.
A deluge of papers have been dedicated to the solution of the inverse problem of calculus
of variation, namely finding a Lagrangian of differential equations. Most of them did not
acknowledge the seminal work of Jacobi. Some did (see references in [26], and [28]).
In [5] Fels derived the necessary and sufficient conditions under which a fourth-order equa-
tion, i.e.
u(iv) = F (t, u, u′ , u′′ , u′′′ ). (1)
admits a unique Lagrangian, namely:

∂3F
=0 (2)
∂(u′′′ )3
1 d2 1 ∂F 3
       
∂F ∂F d ∂F 3 ∂F d ∂F 1 ∂F ∂F
+ − − + + = 0. (3)
∂u′ 2 dt2 ∂u′′′ dt ∂u′′ 4 ∂u′′′ dt ∂u′′′ 2 ∂u′′ ∂u′′′ 8 ∂u′′′

Here we propose to extend the use of the Jacobi last multiplier in order to find the Lagrangian
for ordinary differential equations of order four satisfying a subset of Fels’ conditions (2)-(3).
The paper is organized in the following way. In the next section we present the properties of
the Jacobi last multiplier and its connection to Lie symmetries; then we show the connection
of the Jacobi last multiplier with Lagrangians for any second-order equation, and how and
when can this connection be extended to fourth-order equations. In section 3 two equations
from a Number Theory paper by Hall [6], one of second and one of fourth order, are used to
exemplify the method. The link between Jacobi last multiplier and Lie symmetries [20], [21]
is also exploited. In section 4 we consider two fourth-order equations drawn from Physics and
use the method of the Jacobi last multiplier to find their respective Lagrangians, and lastly
we provide some of the many examples where the method does not work. The last section
contains some final remarks.

2 Jacobi Last Multiplier

The method of the Jacobi Last Multiplier [11]-[13]) provides a means to determine all the
solutions of the partial differential equation
n
X ∂f
Af = ai (x1 , . . . , xn ) =0 (4)
∂xi
i=1

2
or its equivalent associated Lagrange’s system
dx1 dx2 dxn
= = ... = . (5)
a1 a2 an
In fact, if one knows the Jacobi Last Multiplier and all but one of the solutions, then the last
solution can be obtained by a quadrature. The Jacobi Last Multiplier M is given by

∂(f, ω1 , ω2 , . . . , ωn−1 )
= M Af, (6)
∂(x1 , x2 , . . . , xn )

where
∂f ∂f
 
···
 ∂x1 ∂xn 
∂ω1 ∂ω1
 
 
∂(f, ω1 , ω2 , . . . , ωn−1 ) 
∂x1 ∂xn

= det  =0 (7)
∂(x1 , x2 , . . . , xn )  .. .. 

 . . 

 ∂ωn−1 ∂ωn−1 
···
∂x1 ∂xn
and ω1 , . . . , ωn−1 are n − 1 solutions of (4) or, equivalently, first integrals of (5) independent
of each other. This means that M is a function of the variables (x1 , . . . , xn ) and depends on
the chosen n − 1 solutions, in the sense that it varies as they vary. The essential properties of
the Jacobi Last Multiplier are:

(a) If one selects a different set of n − 1 independent solutions η1 , . . . , ηn−1 of equation (4),
then the corresponding last multiplier N is linked to M by the relationship:

∂(η1 , . . . , ηn−1 )
N =M .
∂(ω1 , . . . , ωn−1 )

(b) Given a non-singular transformation of variables

τ: (x1 , x2 , . . . , xn ) −→ (x′1 , x′2 , . . . , x′n ),

then the last multiplier M ′ of A′ F = 0 is given by:

∂(x1 , x2 , . . . , xn )
M′ = M ,
∂(x′1 , x′2 , . . . , x′n )

where M obviously comes from the n − 1 solutions of AF = 0 which correspond to those

chosen for A′ F = 0 through the inverse transformation τ −1 .

3
 (c) One can prove that each multiplier M is a solution of the following linear partial differ-
ential equation:
n
X ∂(M ai )
= 0; (8)
∂xi
i=1

viceversa every solution M of this equation is a Jacobi Last Multiplier.

(d) If one knows two Jacobi Last Multipliers M1 and M2 of equation (4), then their ratio is a
solution ω of (4), or, equivalently, a first integral of (5). Naturally the ratio may be quite
trivial, namely a constant. Viceversa the product of a multiplier M1 times any solution
ω yields another last multiplier M2 = M1 ω.

Since the existence of a solution/first integral is consequent upon the existence of symmetry, an
alternate formulation in terms of symmetries was provided by Lie [20], [21]. A clear treatment
of the formulation in terms of solutions/first integrals and symmetries is given by Bianchi [1].
If we know n − 1 symmetries of (4)/(5), say
n
X
Γi = ξij (x1 , . . . , xn )∂xj , i = 1, n − 1, (9)
j=1

Jacobi’s last multiplier is given by M = ∆−1 , provided that ∆ 6= 0, where

 
a1 ··· an
 ξ1,1 ξ 1,n

∆ = det  .. .. . (10)
 
 . . 
ξn−1,1 · · · ξn−1,n

There is an obvious corollary to the results of Jacobi mentioned above. In the case that there
exists a constant multiplier, the determinant is a first integral. This result is potentially very
useful in the search for first integrals of systems of ordinary differential equations. In particular,
if each component of the vector field of the equation of motion is missing the variable associated
with that component, i.e., ∂ai /∂xi = 0, the last multiplier is a constant, and any other Jacobi
Last Multiplier is a first integral.
Another property of the Jacobi Last Multiplier is its (almost forgotten) relationship with
the Lagrangian, L = L(t, u, u′ ), for any second-order equation

u′′ = F (t, u, u′ ) (11)

namely [30]:
∂2L
M= (12)
∂u′2

4
where M = M (t, u, u′ ) satisfies the following equation
dM ∂F
+ M ′ = 0. (13)
dt ∂u
Then equation (11) becomes the Euler-Lagrange equation:
 
d ∂L ∂L
− + = 0. (14)
dt ∂u ′ ∂u
The proof is given by taking the derivative of (14) by u′ and showing that this yields (13). If
one knows a Jacobi last multiplier, then L can be obtained by a double integration, i.e.:
Z Z 
L= ′
M du du′ + f1 (t, u)u′ + f2 (t, u), (15)

where f1 and f2 are functions of t and u which have to satisfy a single partial differential
equation related to (11) [27]. As it was shown in [27], f1 , f2 are related to the gauge function
g = g(t, u, u′ ). In fact, we may assume
∂g
f1 =
∂u
∂g
f2 = + f3 (t, u) (16)
∂t
where f3 has to satisfy the mentioned partial differential equation and g is obviously arbitrary.
The importance of the gauge function should be stressed. In order to apply Noether’s theorem
correctly, one should not assume g ≡ const, otherwise some first integrals may not be found
(see [27] and the second-order equation in the next section).
We now consider a fourth-order equation (1). In this case the Jacobi last multiplier satisfies
the following equation
dM ∂F
+ M ′′′ = 0. (17)
dt ∂u
It is trivial to show from Fels’conditions (2)-(3) that if a Lagrangian L = L(t, u, u′ , u′′ ) is taken
such that
∂2L
M= (18)
∂u′′2
along with the constraints3
∂F
= 0, (19)
 ∂u 
′′′
∂F d ∂F
− =0 (20)
∂u′ dt ∂u′′
3
We note that (20) resembles the Euler-Lagrange equation (14) with F, u′ , u′′ replacing L, u, u′ respectively.

5
then equation (1) becomes the Euler-Lagrange equation:

d2
   
∂L d ∂L ∂L
− + = 0. (21)
dt2 ∂u′′ dt ∂u′ ∂u

We emphasize that because of the assumption (19), then a Jacobi last multiplier M is trivial
to find from equation (17), namely
M = const. (22)
Then L can be obtained by a double integration, i.e.:
M ′′2
L= u + f1 (t, u, u′ )u′′ + f2 (t, u, u′ ), (23)
2
where f1 and f2 are functions of t, u, u′ which have to satisfy some partial differential equations
related to (1). We can relate f1 , f2 to the gauge function g = g(t, u, u′ ). In fact, we may assume

∂g
f1 =
∂u′
∂g ′ ∂g
f2 = u + + f3 (t, u, u′ ) (24)
∂u ∂t
where f3 has to satisfy the mentioned partial differential equations and g is obviously arbitrary.
Again we stress the importance of the gauge function. In order to apply Noether’s theorem
correctly, one should not assume g ≡ const, otherwise some first integrals may not be found
as in the examples of the next section.

3 Two examples from Number Theory

In this section we show in details the application of the Jacobi Last multiplier along with its
connection to Lie symmetries. We also solve the inverse problem of calculus of variations as
given in [6].

3.1 A second-order equation

In [6] the following functional was introduced:
Z π
y ′4 + 6νy 2 y ′2 d x, (25)
0

where y = y(x) ∈ C 2 [0, π], y(0) = y(π) = 0 and ν ≥ 0. The corresponding Euler-Lagrange
equation is
y ′2 y ′′ + νy 2 y ′′ + νyy ′2 = 0 . (26)

6
If we apply Lie group analysis to this equation, we find that it admits a two-dimensional abelian
transitive Lie symmetry algebra (Type I) generated by the following two operators:

Γ1 = ∂x , Γ2 = y∂y . (27)

Then we can integrate equation (26). First, we introduce a basis of differential invariants of
Γ1 , i.e.:
y′
u = , v = x. (28)
y
Then equation (26) reduces to the following first-order equation:
du −νuv
= 2 , (29)
dv νv + u2
which admits the operator Γ2 in the space of variables u, v, i.e.

Γ2 = v∂v + u∂u . (30)

Then its general solution is implicitly given by:

√
1/4
u 2νv 2 + u2 = const (31)

and in the original variables4 :

1/4
y ′ 2νy 2 + y ′2
p
= const (32)

viz r
p 
νy 2 a1 + ν 2 y 4 a21 + 1 y 2 a1
y′ =  p  (33)
ya1 νy 2 a1 + ν 2 y 4 a21 + 1

with a1 an arbitrary constant Finally the general solution of (26) is given implicitly by:
 p 
Z ya1 νy 2 a1 + ν 2 y 4 a2 + 1
1
r  d y = x + a2 (34)
p
2 2 4 2
νy a1 + ν y a1 + 1 y a1 2

We note that if y is positive and a1 = 1 the integral on the left-hand side could be integrated
in terms of an hypergeometric function H, namely:
1√
    
2 1 1 1 1 1 1
2νy H − , , − , , ,− 4 2 (35)
2 2 4 4 2 2 y ν
4
The same first integral can be obtained by using Noether’s theorem (see below)

7
Let us try to find a Lagrangian for equation (26) by using the Jacobi last multiplier, namely
through (12). The two Lie point symmetries (27) yield a Jacobi last multiplier. In fact the
following matrix [20], [21]:  
νyy ′2
′ 2 ′2
 1 y − ν y +y 

 1 0 0

 (36)
0 y y ′

has determinant different from zero and its inverse is a Jacobi last multiplier, i.e.
2νy 2 + y ′2
M1 = − . (37)
y ′2 (νy 2 + y ′2 )
The corresponding Lagrangian is
√
  ′  
1 ′ y 2
L1 = − 2νy arctan √ + log(2νy + y )νy + 2 log(y )νy + f1 (x, y)y ′ + f2 (x, y),
′2 ′
4νy 2νy
(38)
where f1 , f2 are solutions of
∂f1 ∂f2
− = 0. (39)
∂x ∂y
If we impose the link between f1 , f2 with the gauge function g(x, y), namely (16), then f3 (x, y)
becomes just f3 (x), an arbitrary function of the independent variable x. The Lagrangian (38)
may appear ugly. Nevertheless the corresponding variational problem admits two Noether’s
symmetries, namely both Lie symmetries given in (27) are Noether’s symmetries. Consequently
the following two first integrals of equation (26) can be found by applying Noether’s theorem5 :
 Z 
2 ′2 ′2 x f3 (x)
Γ1 ⇒ I1 = (2νy + y )y , g=e dx + a2 (40)
ex
√
  ′  
1 ′ y ′
Γ2 ⇒ I2 = − 2νy arctan √ + 2νy − 4νxy , [g = s(x)y + x]. (41)
4νy ′ 2νy
with s(x) an arbitrary function of x. We note that the first integral I1 in (40) was already
derived in (32).
At this point one would like to know if it is possible to obtain the original Lagrangian given
in (25), i.e.
LH = y ′4 + 6νy 2 y ′2 (42)
A property of the Jacobi last multiplier is that if one knows a Jacobi last multiplier and a first
integral then their product gives another multiplier [26]. If we take the product of the first
5
Also the corresponding gauge function g is given. It is important to remark that in the case of the first
integral (41) the gauge function g cannot be constant, while it can be constant in the case of the first integral
(40). Naturally, we have left out any inessential additive constants.

8
integral I1 (40) and the multiplier M1 (37), then we obtain another Jacobi last multiplier of
equation (26), i.e.:
M2 = −y ′2 − νy 2 (43)
which can be integrated twice with respect to y ′ in order to yield the following Lagrangian6 :
1 ′4
L2 = − (y + 6νy 2 y ′2 ) + f1 y ′ + f2 , (44)
12
where f1 , f2 are solutions of (39). It is interesting to emphasize that this Lagrangian (namely
Hall’s Lagrangian) is such that the Lie operator Γ2 in (27) does not generate a Noether’s
symmetry for the corresponding variational problem. In fact only Γ1 is the generator of a
Noether’s symmetry for Hall’s Lagrangian.

If instead of (25) we consider the functional [6]

Z π
y ′4 + 6νy 2 y ′2 − 3λ(ν)y 4 d x. (45)
0

and apply Lie group analysis to its corresponding Euler-Lagrange equation, viz:

y ′2 y ′′ + νy 2 y ′′ + νyy ′2 + λy 3 = 0 , (46)

then we obtain the same Lie symmetry algebra generated by (27) which means that (46) can
be integrated by quadrature. In fact if we introduce the same variables as in (28) then equation
(46) reduces to the following first-order equation:

−v λv 2 + νu2


du
= , (47)
dv u (νv 2 + u2 )
which can be integrated to give

λv 4 + 2νu2 v 2 + u4 = const (48)

and in the original variables

λy 4 + 2νy ′2 y 2 + y ′4 = const (49)
a first integral of equation (46).
Also Lie group analysis implies that if λ = ν 2 then equation (46) admits an eight-dimensional
Lie symmetry algebra generated by the following operators:
√ √ √

Γ1 = −y − cos( νx)∂x + y ν sin( νx)∂y
√ √ √

Γ2 = −y sin( νx)∂x + y ν cos( νx)∂y
6
Note the inessential multiplicative constant.

9
 √ √ √
Γ3 = cos(2 νx)∂x − y ν sin(2 νx)∂y
√ √ √
Γ4 = − sin(2 νx)∂x − y ν cos(2 νx)∂y
Γ5 = ∂x (50)
Γ6 = y∂y
√
Γ7 = cos( νx)∂y
√
Γ8 = − sin( νx)∂y .
which means that equation (46) is linearizable or indeed linear. Indeed in this case equation
(46) is just
y ′′ = −νy. (51)
In order to find Lagrangians and first integrals for equation (45) we have to repeat mutatis
mutandis the same analysis given above. We may underline that in the case of λ = ν 2 , namely
equation (51), we can generate 14 different Lagrangians as it was shown in [27].

3.2 A fourth-order equation

Another functional in [6] is the following:
Z π
y ′4 + µy 2 y ′′2 d x . (52)
0
The corresponding Euler-Lagrange equation is
4µyy ′ y ′′′ + µy 2 y iv + 2µy ′2 y ′′ + 3µyy ′′2 − 6y ′2 y ′′ = 0 . (53)
If we apply Lie group analysis to this equation, we find that it admits a three-dimensional Lie
symmetry algebra generated by the following three operators:
X1 = ∂x , X2 = y∂y , X3 = x∂x , (54)
which means that we can reduce equation (53) to a first-order equation, i.e.:
dũ −7µũx̃ − µũ − 6µx̃3 − 4µx̃2 + 6x̃
= (55)
dx̃ µũ
with
y ′′′ y 2 y ′′2 y 2 y ′′ y y ′′ y
ũ = − 2 + , x̃ = . (56)
y ′3 y ′4 y ′2 y ′2
If µ = 3 then equation (53) admits an eight-dimensional Lie symmetry algebra L generated by
the following eight operators:
3 x3
Λ1 = x2 ∂x + xy∂y , Λ2 = x∂x , Λ3 = ∂x , Λ4 = y∂y , Λ5 = ∂y ,
2 y
x2 x 1
Λ6 = ∂y , Λ7 = ∂y , Λ8 = ∂y (57)
y y y

10
This means that equation (53), i.e.:

4y ′ y ′′′ + yy iv + 3y ′′2 = 0 (58)

is linearizable by means of a point transformation [21]. In order to find the linearizable transfor-
mation we have to find an abelian intransitive two-dimensional subalgebra of L and, following
Lie’s classification of two-dimensional algebras in the real plane [21], we have to transform it
into the canonical form
∂u , t∂u (59)
with u and t the new dependent and independent variables, respectively. We found that one
such subalgebra is that generated by Λ7 and Λ8 . Then we have to solve the following four
linear partial differential equations of first order:

Λ7 (t) = 0, Λ8 (t) = 0, Λ7 (u) = t, Λ8 (u) = 1. (60)

It is readily shown that the linearizable transformation is

t = x, u = y2 , (61)

and equation (58) becomes

uiv = 0. (62)
Finally, the general solution of (58) is
p
y = a1 + a2 x + a3 x2 + a4 x3 (63)

with ai (i = 1, 4) arbitrary constants.

We note that if we apply the transformation (61) to equation (53) in the case of any µ then
the following equation is obtained:

(µ − 3)(2uu′′ − u′2 )u′2

uiv = − (64)
4µu3

which does not contain u′′′ , namely the third derivative of u by x. Therefore a constant, say
1, is a Jacobi last multiplier of equation (64), and we can obtain a Lagrangian from (23), i.e.:
1
L1 = u′′2 + f1 (x, u, u′ )u′′ + f2 (x, u, u′ ) (65)
2
where f1 and f2 satisfy (24) and f3 has to have the following expression:

(−µ + 3)u′4
f3 (x, u, u′ ) = + h1 (x, u)u′ + h2 (x, u) (66)
24µu2

11
with h1 , h2 arbitrary functions of x, u.
We remark that a Jacobi last multiplier of equation (53) can be derived from equation (17),
i.e.:
1 dM y′
− 4 = 0 =⇒ M = y 4 , (67)
M dx y
although this Jacobi last multiplier is useless in order to find a Lagrangian of equation (53).
Instead, if we apply transformation (61) to (65) in order to go back to the original function
y(x), then adding also the particular assumptions f1 = −u′2 /(2u), and f2 = u′4 (µ + 1)/(16u2 )
yield the following Lagrangian for equation (53) :

Lm1 = (µ − 1)y ′4 + 2y 2 y ′′2 (68)

which, apart from an inessential multiplicative constant, is Hall’s Lagrangian in (52).

Let us apply Noether’s theorem. If we consider the Lagrangian in (65), we find that the
following two first integrals of equation (64) can be obtained:
3 1
X3 + X1 ⇒ I1 = (12µu3 u′′′ − 4µu2 u′ u′′ − 8µu2 u′ u′′′ x + 4µu2 u′′2 x
2 8µu2
+2µuu′3 − µu′4 x − 6uu′3 + 3u′4 x),
1
X2 ⇒ I2 = (−8µu2 u′ u′′′ + 4µu2 u′′2 − µu′4 + 3u′4 ). (69)
8µu2

A similar result is obtained if one uses Hall’s Lagrangian and equation (53). Moreover, if we
apply Noether’s theorem to the linearizable equation (58) with Hall’s Lagrangian, i.e.:

LH = y ′4 + 3y 2 y ′′2 , (70)

we obtain the following seven first integrals7 :

Λ1 ⇒ Im1 = −3y 3 y ′′ + 3y 3 y ′′′ x + y 2 y ′2 + 7y 2 y ′ y ′′ x

−2y 2 y ′ y ′′′ x2 + y 2 y ′′2 x2 − 2yy ′3 x − 4yy ′2 y ′′ x2 + y ′4 x2
3
Λ4 + Λ2 ⇒ Im2 = 9y 3 y ′′′ + 21y 2 y ′ y ′′ − 12y 2 y ′ y ′′′ x + 6y 2 y ′′2 x − 6yy ′3 − 24yy ′2 y ′′ x + 6y ′4 x
4
Λ3 ⇒ Im3 = −2y 2 y ′ y ′′′ + y 2 y ′′2 − 4yy ′2 y ′′ + y ′4
Λ5 ⇒ Im5 = −3y 2 + 6yy ′ x − 3yy ′′ x2 + yy ′′′ x3 − 3y ′2 x2 + 3y ′ y ′′ x3
Λ6 ⇒ Im6 = 2yy ′ − 2yy ′′ x + yy ′′′ x2 − 2y ′2 x + 3y ′ y ′′ x2
Λ7 ⇒ Im7 = −yy ′′ + yy ′′′ x − y ′2 + 3y ′ y ′′ x
Λ8 ⇒ Im8 = yy ′′′ + 3y ′ y ′′ (71)
7
Naturally, they are not all independent from each other.

12
 Although we do not write down the corresponding expressions of the gauge function, we un-
derline that it cannot always be set equal to a constant otherwise none of Im1 , Im5 , Im6 , Im7 ,
and Im8 could be obtained.
All seven first integrals (and even more) may be obtained without Noether’s Theorem.
In fact we just need to find the Jacobi last multipliers of equation (58) that are obtained
by inverting the nonzero determinants of the possible 70 matrices made out of the eight Lie
symmetries (57). Then the ratio of any two multipliers is a first integral of equation (58). For
example:
4y ′ y ′′′ + 3y ′′2
 
1 y ′ y ′′ y ′′′ −

 y 

 x2 3 yx 3 y − 1 xy ′ y ′ − 5 xy ′′ − 3 y ′′ − 9 xy ′′′ 
2 2 2 2 2 2
C1234 =  (72)
 

 x
 0 −y ′ −2y ′′ −3y ′′′ 

 1 0 0 0 0 
0 y y ′ y ′′ y ′′′

is the matrix obtained by considering the symmetries generated by operators Λ1 , Λ2 , Λ3 , Λ4 in

(57); its determinant is
3 9
∆1234 = 9y ′′ yy ′ y ′′′ − 2y ′3 y ′′′ − y ′2 y ′′2 + 6y ′′3 y + y ′′′2 y 2 , (73)
2 2
and the corresponding Jacobi last multiplier is:
1 1
M1234 = = . (74)
∆1234 18y ′′ yy ′ y ′′′ − 4y ′3 y ′′′ − 3y ′2 y ′′2 + 12y ′′3 y + 9y ′′′2 y 2
Similarly, the matrix C5678 yields the determinant ∆5678 = 12/y 4 , i.e. the Jacobi last multi-
plier8 :
M5678 = y 4 (75)
which we have already found in (67) as an obvious solution of (17). Note that
M5678
= y 4 (18y ′′ yy ′ y ′′′ − 4y ′3 y ′′′ − 3y ′2 y ′′2 + 12y ′′3 y + 9y ′′′2 y 2 ) (76)
M1234
is “another” first integral of equation (58).

4 More fourth-order equations

A similarity reduction of the fifth-order Korteweg-De Vries equation is the following fourth-
order equation [2], [22]
uiv = 20uu′′ + 10u′2 − 40u3 + αu + t. (77)
8
The multiplicative constant is inessential.

13
This equation satisfies Fels’ conditions (2)-(3) as well as (19)-(20). Therefore M = 1 is a Jacobi
last multiplier, and formula (18) yields the following Lagrangian of (77):

1 d
L = (u′′2 + t2 u′ ) + αtuu′ − 40tu3 u′ + 10uu′2 + g(t, u, u′ ) (78)
2 dt
where g = g(t, u, u′ ) is the gauge function.
In a paper on wave caustics [15], Kitaev derived the following equation:
1
uiv = (16uu′ u′′′ − 12αu5 + 8βu3 − 20u7 + 40u4 u′′ + 8u4 t + 20u3 u′2 − 8uu′
8u2
+12uu′′2 − 8uu′′ t + uy 2 − 16u′2 u′′ + 8u′2 t) (79)

which satisfies Fels’ conditions (2)-(3), and admits M = u−2 as a Jacobi last multiplier. The
simple transformation u = w2 transforms equation (79) into the following equation in w = w(t)
1
wiv = (−12αw8 + 8βw4 − 20w1 2 + 80w7 w′′ + 160w6 w′2 + 8w6 t + 160ww′2 w′′
16w3
−16ww′ − 16ww′′ t − 80w′4 + 16w′2 t + t2 ) (80)

This equation satisfies Fels’ conditions (2)-(3) as well as (19)-(20). Therefore M = 1 is a Jacobi
last multiplier, and formula (18) yields the following Lagrangian for equation (80):

1 ′′2 w′
L = w + (−36αw8 t + 24βw4 t − 60w12 t + 120w7 w′ + 12w6 t2 + 40ww′3
2 48w3
d
−24ww′ t + t3 ) + g(t, w, w′ ) (81)
dt
√
where g = g(t, w, w′ ) is the gauge function. Replacing w = u into this Lagrangian yields the
Lagrangian for equation (79), i.e.:
1
L = (12u2 u′′2 − 12uu′2 u′′ − 36αu5 u′ t + 24βu3 u′ t − 60u7 u′ t + 60u4 u′2 + 12u4 u′ t2
96u3
d
−12uu′2 t + uu′ t3 + 8u′4 ) + g(t, u, u′ ) (82)
dt
where g = g(t, u, u′ ) is the gauge function.
The method we proposed here is not the ultimate method. We can list many equations for
which it does not work. In [23] seven fourth-order equations were derived as similarity reduc-
tions of a mathematical model for thin liquid films [29] and its corresponding heir equations
[24]. None of the seven equations passes condition (3), although they all satisfy condition (2).

14
5 Final remarks
When one deals with a second-order differential equation the following remarks should be
enlightened and kept in mind:

• The most efficient Lagrangian, namely that which allows the most number of Noether’s
symmetries, may not be the Lagrangian with the simplest form.

• Lie symmetries are the key tool for finding Jacobi last multipliers and therefore La-
grangians.

In [5] the necessary and sufficient conditions under which a fourth-order equation (1) admits
a unique Lagrangian were determined. In this paper we suggest that the application of the
Jacobi last multiplier may yield that unique Lagrangian in some instances.
If one rewrites a fourth-order equation as either a suitable system of two second-order
equations [4], or a system of four first-order equations [8] the challenge of solving an inverse
problem of calculus of variations may still be open.

Acknowledgements
AMA is indebted to: Professor Gianluca Vinti for arranging his visits to the Dipartimento di
Matematica e Informatica, Università di Perugia; Professor Bryce McLeod for initiating the
connection in the first place; and finally to them and their colleagues, Professors Anna Rita
Sambucini and Carlo Bardaro, for excellent hospitality during his visits. The support of the
Erasmus Exchange scheme is gratefully acknowledged.

References
[1] L. Bianchi, Lezioni sulla teoria dei gruppi continui finiti di trasformazioni, Enrico Spoerri,
Pisa, 1918.

[2] C.M. Cosgrove, Higher-order Painlevé equations in the polynomial class I Bureau symbol
P2, Stud. Appl. Math. 104 (2000), 1-65.

[3] G. Darboux, Leçon sur la théorie générale des surfaces, Gauthier-Villars, Paris, 1894.

[4] J. Douglas, Solution of the inverse problem of calculus of variations, Trans. Amer. Math.
Soc. 50 (1941), 71-128.

[5] M. E. Fels, The inverse problem of the calculus of variations for scalar fourth-order ordi-
nary differential equations, Trans. Amer. Math. Soc. 348 (1996), 5007-5029.

15
 [6] R. R. Hall, A Wirtinger type inequality and the spacing of the zeros of the Riemann-Zeta
function, J. Number Theory 93 (2002), 235-245.

[7] P. Havas, The range of application of the Lagrange formalism - I, Nuovo Cimento Suppl.
Ser. X 5 (1957) 363-388.

[8] P. Havas, The connection between conservation laws and invariance groups: folklore,
fiction, and fact, Acta Physica Austriaca 38 (1973) 145-167.

[9] T. Hawkins, Jacobi and the birth of Lie’s theory of groups, Arch. Hist. Exact Sciences 42
(1991), 187-278.

[10] H. Helmholtz, Ueber die physikalische bedutung des princips der kleinsten wirkung, J.
Reine Angew. Math. 100 (1887), 137-166.

[11] C.G.J. Jacobi, Theoria novi multiplicatoris systemati æquationum differentia-

lium vulgarium applicandi, J. Reine Angew. Math. 27 (1844), 199-268.

[12] C.G.J. Jacobi, Theoria novi multiplicatoris systemati æquationum differentia-

lium vulgarium applicandi, J. Reine Angew. Math. 29 (1845), 213-279 and 333-376.

[13] C.G.J. Jacobi, Vorlesungen über Dynamik. Nebst fünf hinterlassenen Abhandlungen des-
selben herausgegeben von A. Clebsch, Druck und Verlag von Georg Reimer, Berlin, 1886.

[14] E.H. Kerner, Gibbs Ensemble: Biological Ensemble, Gordon and Breach, New York, 1971.

[15] A.V. Kitaev, Caustics in 1 + 1 integrable systems, J. Math. Phys. 35 (1994), 2934-2954

[16] L. Koenigsberger, Die Principien der Mechanik, Teubner, Leipzig, 1901.

[17] L. Koenigsberger, Hermann von Helmholtz, Continuum International Publishing Group,

London, 2001.

[18] L. Koenigsberger, Carl Gustav Jacob Jacobi, Teubner, Leipzig, 1904.

[19] L. Koenigsberger, Hermann von Helmholtz, translated by F.A. Welby, Clarendon Press,
Oxford, 1906.

[20] S. Lie, Veralgemeinerung und neue Verwerthung der Jacobischen Multiplicator-Theorie,

Christ. Forh. (1874), 255-274.

[21] S. Lie, Vorlesungen über Differentialgleichungen mit bekannten infinitesimalen Transfor-

mationen, Teubner, Leipzig, 1912.

[22] R. C. Littlewood, Hyperelliptic asymptotics of Painlevé-type equations, Nonlinearity 12

(1999), 1629-1641.

16
[23] S. Martini, N. Ciccoli and M.C. Nucci, Group analysis and heir-equations of a mathemat-
ical model for thin liquid films, J. Nonlinear Math. Phys. (2008) to appear

[24] M.C. Nucci, Iterating the nonclassical symmetries method, Physica D 78 (1994), 124-134.

[25] M. C. Nucci, Interactive REDUCE programs for calculating Lie point, non-classical, Lie-
Bäcklund, and approximate symmetries of differential equations: manual and floppy disk,
in CRC Handbook of Lie Group Analysis of Differential Equations. Vol. 3: New Trends in
Theoretical Developments and Computational Methods, ed. N. H. Ibragimov, CRC Press,
Boca Raton, 1996, pp. 415-481.

[26] M.C. Nucci, Jacobi last multiplier and Lie symmetries: a novel application of an old
relationship, J. Nonlinear Math. Phys. 12 (2005), 284-304.

[27] M.C. Nucci and P.G.L. Leach, Lagrangians galore, J. Math. Phys. 48 (2007) 123510 (16
pages).

[28] M.C. Nucci and P.G.L. Leach, Jacobi last multiplier and Lagrangians for multidimensional
linear systems, J. Math. Phys. 49 (2008) 073517 (8 pages).

[29] E. Ruckenstein and R.K. Jain, Spontaneous rupture of thin liquid films, Chem. Soc.
Faraday Trans. 70 (1973), 132-147.

[30] E.T. Whittaker, A Treatise on the Analytical Dynamics of Particles and Rigid Bodies,
Cambridge University Press, Cambridge, 1999.

17