Finding Liouvillian First Integrals of Rational ODEs of Any Order in Finite Terms
It is known, due to Mordukhai-Boltovski, Ritt, Prelle, Singer, Christopher and others, that if a given rational ODE has a Liouvillian first integral then the corresponding integrating factor of the ODE must be of a very special form of a product of powers and exponents of irreducible polynomials. Th...
Збережено в:
| Опубліковано в: : | Symmetry, Integrability and Geometry: Methods and Applications |
|---|---|
| Дата: | 2006 |
| Автор: | |
| Формат: | Стаття |
| Мова: | Англійська |
| Опубліковано: |
Інститут математики НАН України
2006
|
| Онлайн доступ: | https://nasplib.isofts.kiev.ua/handle/123456789/146113 |
| Теги: |
Додати тег
Немає тегів, Будьте першим, хто поставить тег для цього запису!
|
| Назва журналу: | Digital Library of Periodicals of National Academy of Sciences of Ukraine |
| Цитувати: | Finding Liouvillian First Integrals of Rational ODEs of Any Order in Finite Terms / Y.N. Kosovtsov // Symmetry, Integrability and Geometry: Methods and Applications. — 2006. — Т. 2. — Бібліогр.: 22 назв. — англ. |
Репозитарії
Digital Library of Periodicals of National Academy of Sciences of Ukraine| _version_ | 1859737816156602368 |
|---|---|
| author | Kosovtsov, Y.N. |
| author_facet | Kosovtsov, Y.N. |
| citation_txt | Finding Liouvillian First Integrals of Rational ODEs of Any Order in Finite Terms / Y.N. Kosovtsov // Symmetry, Integrability and Geometry: Methods and Applications. — 2006. — Т. 2. — Бібліогр.: 22 назв. — англ. |
| collection | DSpace DC |
| container_title | Symmetry, Integrability and Geometry: Methods and Applications |
| description | It is known, due to Mordukhai-Boltovski, Ritt, Prelle, Singer, Christopher and others, that if a given rational ODE has a Liouvillian first integral then the corresponding integrating factor of the ODE must be of a very special form of a product of powers and exponents of irreducible polynomials. These results lead to a partial algorithm for finding Liouvillian first integrals. However, there are two main complications on the way to obtaining polynomials in the integrating factor form. First of all, one has to find an upper bound for the degrees of the polynomials in the product above, an unsolved problem, and then the set of coefficients for each of the polynomials by the computationally-intensive method of undetermined parameters. As a result, this approach was implemented in CAS only for first and relatively simple second order ODEs. We propose an algebraic method for finding polynomials of the integrating factors for rational ODEs of any order, based on examination of the resultants of the polynomials in the numerator and the denominator of the right-hand side of such equation. If both the numerator and the denominator of the right-hand side of such ODE are not constants, the method can determine in finite terms an explicit expression of an integrating factor if the ODE permits integrating factors of the above mentioned form and then the Liouvillian first integral. The tests of this procedure based on the proposed method, implemented in Maple in the case of rational integrating factors, confirm the consistence and efficiency of the method.
|
| first_indexed | 2025-12-01T15:53:09Z |
| format | Article |
| fulltext |
Symmetry, Integrability and Geometry: Methods and Applications Vol. 2 (2006), Paper 059, 8 pages
Finding Liouvillian First Integrals
of Rational ODEs of Any Order in Finite Terms
Yuri N. KOSOVTSOV
Lviv Radio Engineering Research Institute, 7 Naukova Str., Lviv, 79060 Ukraine
E-mail: kosovtsov@escort.lviv.net
Received August 31, 2005, in final form May 12, 2006; Published online June 08, 2006
Original article is available at http://www.emis.de/journals/SIGMA/2006/Paper059/
Abstract. It is known, due to Mordukhai-Boltovski, Ritt, Prelle, Singer, Christopher and
others, that if a given rational ODE has a Liouvillian first integral then the corresponding
integrating factor of the ODE must be of a very special form of a product of powers and
exponents of irreducible polynomials. These results lead to a partial algorithm for finding
Liouvillian first integrals. However, there are two main complications on the way to ob-
taining polynomials in the integrating factor form. First of all, one has to find an upper
bound for the degrees of the polynomials in the product above, an unsolved problem, and
then the set of coefficients for each of the polynomials by the computationally-intensive
method of undetermined parameters. As a result, this approach was implemented in CAS
only for first and relatively simple second order ODEs. We propose an algebraic method
for finding polynomials of the integrating factors for rational ODEs of any order, based on
examination of the resultants of the polynomials in the numerator and the denominator of
the right-hand side of such equation. If both the numerator and the denominator of the
right-hand side of such ODE are not constants, the method can determine in finite terms
an explicit expression of an integrating factor if the ODE permits integrating factors of the
above mentioned form and then the Liouvillian first integral. The tests of this procedure
based on the proposed method, implemented in Maple in the case of rational integrating
factors, confirm the consistence and efficiency of the method.
Key words: differential equations; exact solution; first integral; integrating factor
2000 Mathematics Subject Classification: 34A05; 34A34; 34A35
1 The Prelle–Singer method
µ is an integrating factor for an nth order ODE in solved form
dny0
dxn
− f(x, y0, y1, . . . , yn−1) = 0, (1)
where y0 = y(x), yj = djy(x)
dxj , by standard definition [1], if µ(yn−f) is a total derivative of some
function ζ(x, y0, y1, . . . , yn−1), that is,
µ
(
dny0
dxn
− f
)
=
∂ζ
∂x
+
n−1∑
j=0
yj+1
∂ζ
∂yj
. (2)
Let
µ =
∂ζ
∂yn−1
. (3)
mailto:kosovtsov@escort.lviv.net
http://www.emis.de/journals/SIGMA/2006/Paper059/
2 Yu.N. Kosovtsov
Then (2) becomes
D(ζ) + f
∂ζ
∂yn−1
= 0, (4)
where
D =
∂
∂x
+
n−2∑
j=0
yj+1
∂
∂yj
.
We conclude that the integrating factor of ODE (1) is µ = ∂ζ
∂yn−1
, where the function ζ is a first
integral, a solution of the linear first-order PDE (4).
If we eliminate the function ζ from the system (3), (4), we obtain a PDE system for the
integrating factor. One of the equations of the PDE system obtained by double differentiation
of (4) with respect to yn−1 is as follows (n ≥ 2):
D
(
∂µ
∂yn−1
)
+
∂2(fµ)
∂y2
n−1
+ 2
∂µ
∂yn−2
= 0. (5)
Assuming that an integrating factor is known, it is well-known that we are able to obtain cor-
responding first integral by quadratures. If we know n independent first integrals for ODE (1),
we can then find its general solution, at least in implicit form. As it is seen from (3), the alge-
braic structure of the integrating factor is simpler than the structure of the first integral, so in
some cases finding the integrating factors is an easier problem.
The most beautiful way of finding integrating factors is the Darboux method and its refine-
ments. If f in (1) is a rational function with respect to all variables x, y0, y1, . . . , yn−1, we will
call such an ODE a rational one. In 1878, Darboux [2] proved that if a rational ODE has at
least m(m + 1)/2, where m = max(order(numerator(f), order(denominator(f)) invariant algeb-
raic curves Pi, now known as Darboux polynomials, then it has a first integral or an integrating
factor of the form
∏
i P
ai
i for suitable complex constants ai. So, in principle, the problem of
finding an integrating factor is reduced here to the problem of finding Darboux polynomials.
In 1983, Prelle and Singer [3] proved the essential theorem that all first order ODEs, which
possess elementary first integrals, have an integrating factor of the above mentioned form
∏
i P
ai
i .
Their result unifies and generalizes a number of results originally due to Mordukhai-Boltovski [4],
Ritt [5] and others. Singer in [6] refines this result for ODEs, which possess Liouvillian first
integral.
The following theorem is a corollary of the Proposition 2.2 [6] and of the definition of the
integrating factor (3) (see also [7]):
Theorem 1. If the n-th order ODE
dny0
dxn
=
A(x, y0, y1, . . . , yn−1)
B(x, y0, y1, . . . , yn−1)
, (6)
where A,B ∈ K[x, y0, y1, . . . , yn−1] are polynomials, has a local Liouvillian first integral, then
there exists an integrating factor of ODE (6) of the form
µ =
∏
i
P ai
i exp
b0
∏
j
Q
bj
j
, (7)
where Pi’s, Qj’s ∈ K[x, y0, y1, . . . , yn−1] are irreducible polynomials, ai’s, b0 are constants and
bj’s are integers.
Finding Liouvillian First Integrals of Rational ODEs of Any Order in Finite Terms 3
So, to find an integrating factor to ODE (6) in the conditions of Theorem 1, we first should
obtain the sets of polynomials Pi and Qj , and if we succeed in this task then we can obtain the
constants ai’s and bj ’s from the full PDE system for the integrating factor.
To consider the ways to find the sets of polynomials Pi and Qj , let us substitute the integrating
factor in the form (7) into (5). After some rearrangement, we arrive at
L1B = 2A
(
∂B
∂yn−1
)2∏
i
P 2
i
∏
j
Q2
j ,
where L1 is a polynomial. As A and B are relatively prime polynomials, we conclude that
B divides one of the polynomials Pi or Qj . Suppose that
A = A1(yn−1)A2(x, y0, y1, . . . , yn−1),
B = B1(yn−1)B2(x, y0, y1, . . . , yn−2)B3(x, y0, y1, . . . , yn−1)
and
µ = Aε0
1 Bε1
1 Bε2
2 Bε3
3
∏
i
P ai
i exp
b0A
η0
1 Bη1
1 Bη2
2 Bη3
3
∏
j
Q
bj
j
.
Substituting these expressions for A, B and µ into (5), we can conclude, by considering divis-
ibility of polynomials, that it is necessary that ε3 = 1 and η3 = 0. That is, if ∂Pi
∂yn−1
≡ 0 or
D(Pi) ≡ 0 ( ∂Qj
∂yn−1
≡ 0 or D(Qj) ≡ 0) then such polynomials are factors of A or B, and they can
be easily selected and included into the set of candidates for the integrating factor structure.
Without loss of generality, we will consider an integrating factor of the form
µ = B
∏
i
P ai
i exp
b0
∏
j
Q
bj
j
. (8)
and will focus our attention on the cases when ∂Pi
∂yn−1
6= 0 and D(Pi) 6= 0 ( ∂Qj
∂yn−1
6= 0 and
D(Qj) 6= 0).
For the case of ak 6= 1, substitution of (8) into (5) leads to
L2Pk = ak(ak − 1)
∂Pk
∂yn−1
∏
i
i6=k
P 2
i
∏
j
Q2
j
[
BD(Pk) + A
∂Pk
∂yn−1
]
(9)
and similarly for the case of bk < 0
L3Qk = b2
k
∂Qk
∂yn−1
∏
i
P 2
i
∏
j
j 6=k
Q
2(bj+1)
j
[
BD(Qk) + A
∂Qk
∂yn−1
]
, (10)
where L2 and L3 are polynomials.
All components involved in (9) and (10) are polynomials. We assume that Pi and Qj are
irreducible, relatively prime polynomials, so gcd
(
Pi,
∂Pi
∂yn−1
)
and gcd
(
Qi,
∂Qi
∂yn−1
)
are constants,
where greatest common divisor is denoted by gcd. We can conclude that (f | g means that f
divides g)
Pi | BD(Pi) + A
∂Pi
∂yn−1
(11)
4 Yu.N. Kosovtsov
and
Qj | BD(Qj) + A
∂Qj
∂yn−1
. (12)
The basic significance of relations (11) and (12) is that we can investigate each of the poly-
nomials Pi or Qj independently. Properties (11) and (12) constitute the basis of the so-called
Prelle–Singer method, which originates from Darboux [2], to obtain the integrating factors and
first integrals of ODEs of type (6). Supposing that
Pk =
N∑
i=0
N∑
j0=0
· · ·
N∑
jn−1=0
ckij0···jn−1x
iyj0
0 · · · yjn−1
n−1 (13)
and substituting Pk into (11), we can in principle obtain ckij0···jn−1 ’s and as a result Pk (or
similarly Qk) in explicit form.
To do so, we first have to establish an upper bound N for the degrees of the polynomial Pi
(or Qk). It is known that N exists, but there is no constructive method to evaluate it so far. It
is the principal weakness of the Prelle–Singer method. In the existing implementations of the
Prelle–Singer method, an upper bound N is predefined by a user of the method.
With an increase of equation order the number of undetermined variables ckij0···jn−1 and,
correspondingly, the dimension of the system of algebraic equations for these constants increases.
As a result, solving of this system becomes problematical. This approach was implemented in
CAS only for first and relatively simple second order ODEs [8, 9, 10, 11, 12, 13, 14, 15], and as
a rule the method can not be used for N > 4. In contrast, the modified method proposed in
the next section does not suffer from these restrictions.
We must also mention that a rational ODE does not always have a local Liouvillian first
integral. While generalizations of the Darboux integrability theory allow in some (“integrable”)
cases to find non-Liouvillian first integrals (see, e.g. [16, 17, 18, 19, 20]), we consider only
Liouvillian first integral theory.
2 The method of resultants
Let us investigate closely the relation (11) or equivalently (12). Let Rz(f, g) be the resultant of
polynomials f and g with respect to the indeterminate z. In the sequel, we will consider only
the case when both A and B are not constants and there are such z ∈ (x, y0, y1, . . . , yn−1) that
Rz(A,B) is not a constant.
The following Theorem 2 and Corollary are our main results:
Theorem 2. If a polynomial Pi satisfies the condition (11) then for any indeterminate z ∈
(x, y0, y1, . . . , yn−1) the resultant Rz(Pi, B) must divide the resultant Rz(A,B).
Proof. We may write Pi L = BD(Pi) + A ∂Pi
∂yn−1
for some polynomial L. Taking resultants of
both sides of this equation with B and with respect to an indeterminate z ∈ (x, y0, y1, . . . , yn−1),
we get the following polynomial equation
Rz(Pi, B)Rz(L,B) = Rz(A,B)Rz
(
∂Pi
∂yn−1
, B
)
.
As Pi and ∂Pi
∂yn−1
have no common roots, Rz(Pi, B) does not divide Rz
(
∂Pi
∂yn−1
, B
)
. Therefore,
Rz(Pi, B) | Rz(A,B) (14)
and it is obvious that we can choose Rz(Pi, B) as a divisor of Rz(A,B) up to a constant fac-
tor. �
Finding Liouvillian First Integrals of Rational ODEs of Any Order in Finite Terms 5
How can we recover Pi if we know the resultant Rz(Pi, B)? First of all, we note that the
case when A/B depends on only one indeterminate is trivial. As we have mentioned above, if Pi
depends only on one indeterminate, then such polynomials are factors of A or B and we have
already included them into the set of candidates. So, we assume that Pi depends at least on two
indeterminates and there exists such z that Rz(Pi, B) is a polynomial. We do not insist that
Rz(Pi, B) is irreducible.
It is known from elementary facts about resultants that there exist such polynomials αi, βi ∈
K[x, y0, y1, . . . , yn−1] that βi Pi = αi B + Rz(Pi, B).
As a corollary of Theorem 2, we have the following result. Assuming that Pi’s (and Qj ’s) are
irreducible polynomials we conclude that
Corollary. If polynomials Pi (and respectively Qj) satisfy (14) then the following holds
Pi hyp = αi B + ci [one of nonconstant factors of Rz(A,B)], (15)
where Pi hyp = βi Pi, βi is some polynomial, αi ∈ (−1, 0, 1) and ci are constants.
Proof. If αi 6= 0 identically, then for any z ∈ (x, y0, y1, . . . , yn−1) from
Rz(βiPi − αiB,Pi) = γRz(αi, Pi)Rz(B,Pi)
= ciRz([one of nonconstant factors of Rz(A,B)], B) = Rz(B,Pi),
where γ ∈ (−1, 1), we conclude that Rz(αi, Pi) = 1/γ. So as Pi depends at least on two
indeterminates then αi must be a constant and αi ∈ (−1, 0, 1).
Now we are able to modify the Prelle–Singer method by replacing the hypothesis (13)
with (15) and finding constants αi and ci from the following equation
Rz
(
BD(Pi hyp) + A
∂Pi hyp
∂yn−1
, Pi hyp
)
= 0.
Then Pi is a irreducible factor of Pi hyp which satisfies (11).
Thus, we can use algebraic, not differential, relations for obtaining polynomials as in the
original Darboux (and Prelle–Singer) approach. First of all, we bypass the principal weakness
of the Prelle–Singer method as we do not need to establish an upper bound N for the degrees
of polynomial Pi here. Second, we have the advantage that finding only two undetermined
constants αi and ci in the Pi hyp structure is much easier than finding the set of ckij0···jn−1 ’s. �
Example. Let us consider a relatively simple second order ODE
y′′ =
y′3 + y′(x− 2)− y
y′2 + (2y′ − 1)(x + y)− x
.
For this ODE, Rx(A,B) = −(2y2
1 + y1 − 2)(y2
1 − 2y1 − y0) and we can form the following
hypothesis Phyp = α(y2
1 + (2y1 − 1)(x + y0)− x) + c(y2
1 − 2y1 − y0). Calculation of the following
resultant leads to Ry1(BD(Phyp)+A
∂Phyp
∂y1
, Phyp) = α2(α−c)L = 0, where L is some polynomial.
So, there are two cases, when α = 0 then P1 hyp = c(y2
1 − 2y1 − y0), and when α = c then
P2 hyp = 2c(y1 − 1)(y1 + y0 + x).
By substitution of the irreducible factors of P1 hyp and P2 hyp to (11), we select the following
candidates for the integrating factor structure: P1 = (y2
1 − 2y1 − y0) and P2 = (y1 + y0 + x). So
the hypothesis of µ is
µhyp = (y2
1 + (2y1 − 1)(x + y0)− x)(y1 + y0 + x)X1(y2
1 − 2y1 − y0)X2.
6 Yu.N. Kosovtsov
After substitution of this µhyp into the PDE system for the integrating factor, we obtain that
X2 = −(X1 + 2) so an integrating factor of given ODE is
µ = (y′2 + (2y′ − 1)(x + y)− x)(y′ + y + x)X1(y′2 − 2y′ − y)−(X1+2),
where X1 is an arbitrary constant.
Examinations of the cases when ak = 1 and bk > 0, are more difficult. We are not able to
give here the full analysis of the problem which necessitates consideration of many subcases.
For brevity, we will only demonstrate that the method of resultants in principle enables us to
obtain all the needed polynomials in a finite number of steps. The main tool here is usage of
repeated resultants.
For compactness, we will use the following notation for the repeated resultants
Rz1,z2(f, (g, h)) = Rz2(Rz1(f, h), Rz1(g, h)).
For the case ak = 1, we obtain the following equation similar to (9):
L4Pk =
∏
i
i6=k
P̃i
∏
j
Q̃j(M1A + M2B)
+
∏
i
i6=k
P̃ 2
i
∏
j
Q̃2
j
[[
2
∂A
∂yn−1
+ D(B)
]
∂Pk
∂yn−1
+ D(Pk)
∂B
∂yn−1
]
,
where L4, M1, M2 are some polynomials, and P̃i (i 6= k), Q̃j are only such polynomials from
the integrating factor structure which do not satisfy conditions (11) and (12). So for this case
we can obtain the following properties ( ∂Pk
∂yn−1
6= 0):
Rz1,z2(Pk, (A,B)) | Rz1,z2
(∏
i
i6=k
P̃ 2
i
∏
j
Q̃2
j
[[
2
∂A
∂yn−1
+ D(B)
]
∂Pk
∂yn−1
+ D(Pk)
∂B
∂yn−1
]
, (A,B)
)
(16)
and further
Rz1,z2,z3
(
Pk,
(
A,B,
∂B
∂yn−1
))
| Rz1,z2,z3
(∏
i
i6=k
P̃ 2
i
∏
j
Q̃2
j
∂Pk
∂yn−1
[
2
∂A
∂yn−1
+ D(B)
]
,
(
A,B,
∂B
∂yn−1
))
. (17)
From (16) we can obtain the hypothesis for Pk and then with the help of (17) all its unde-
termined constants.
If there is a polynomial Ps in the integrating factor structure which satisfy (11), then
finding Pk can be simplified. Since
Rz(Pk, Ps) | Rz
(
A
∏
i
i6=k
P̃ 2
i
∏
j
Q̃2
j
∂Ps
∂yn−1
[
A
∂Pk
∂yn−1
+ BD(Pk)
]
, Ps
)
, (18)
Finding Liouvillian First Integrals of Rational ODEs of Any Order in Finite Terms 7
then
Rz1,z2(Pk, (Ps, B)) | Rz1,z2
(
A2
∏
i
i6=k
P̃ 2
i
∏
j
Q̃2
j
∂Ps
∂yn−1
∂Pk
∂yn−1
, (Ps, B)
)
. (19)
Among other things we can observe from (16) and (18) is that it is very likely that the Pk’s
satisfy the relation (11) even if ak = 1. Also, there are relations for Qk with bk > 0, similar to
(16)–(19).
The main elements of the method described above were implemented in demonstration pro-
cedures in Maple [21, 22] for the case of rational integrating factors.
We have tested routines in the following area: for a given ODE order we assigned the first
integral ζ of the following type
ζ =
∏
i N
αi
i∏
i M
βi
i
+
∑
i
(ηi lnΦi + θi arctanΨi),
where Ni, Mi, Φi, Ψi are some polynomials, αi > 0, βi > 0, ηi, θi are some constants. Such
algebraic first integral leads to an ODE of type (1) with f = −D(ζ)/ ∂ζk
∂yn−1
, which is potentially
tractable by the proposed procedure as this ODE has a rational integrating factor µ = ∂ζk
∂yn−1
=∏
i P
ai
i .
The first stage – finding candidates and hypothesis of the structure of an integrating factor – is
very fast. The second stage is relatively resource consuming – finding the unknown parameters ai
of the structure of an integrating factor.
We can conclude that our prototypes are workable for rational ODEs even with symbolic
constant parameters of orders from n = 1 to n = 4 − 5. Although in principle the method
provides ability to find an integrating factor for almost any of the above mentioned type of
ODEs, in practice the procedure heavily relies on some basic CAS functions such as factor,
simplify and so on, and in some cases fails. In addition, different outputs are observed in
different Maple versions for the same ODE.
As a rule, Maple cannot find any integrating factor for ODEs from the considered test
area while proposed procedure is usually successful. Often our procedure produces more than
one integrating factor. For example, the output may contain a set of µ’s or µ with arbitrary
parameters, and sometimes they lead to independent first integrals.
3 Conclusion
We have proposed an algebraic method for finding polynomials Pi and Qj of the integrating fac-
tor form (7) of rational ODE (6) for any order using resultants of the numerator and denominator
of the ODE’s right-hand-side.
The main advantage of the method of resultants is ability to obtain the polynomials Pi and Qj
without an a priori assumption of the upper bound for their respective degrees. Our method
bypasses the principal weakness of the Prelle–Singer (Darboux) method.
Moreover, the number of undetermined variables involved in calculations by our method is
radically smaller than in the Prelle–Singer (Darboux) method. This property allows us to obtain
polynomials Pi and Qj of arbitrary degrees without any complications whereas the Prelle–Singer
(Darboux) method can not be useful as a rule for N > 4.
Acknowledgements
I would like to thank the referees for extensive comments and suggestions regarding of earlier
versions of this paper and Reece Heineke for a careful reading of the paper.
8 Yu.N. Kosovtsov
[1] Olver P.J., Applications of Lie groups to differential equations, New York, Springer-Verlag, 1993.
[2] Darboux G., Mémoire sur les équations différentielles algébriques du premier ordre et du premier degré
(Mélanges), Bull. Sci. Math., 1878, V.2, 60–96, 12–144, 151–200.
[3] Prelle M., Singer M., Elementary first integral of differential equations, Trans. Amer. Math. Soc., 1983,
V.279, 215–229.
[4] Mordukhai-Boltovski D., Researches on the integration in finite terms of differential equations of the first
order, Communicatios de la Societe Mathematique de Kharkov, 1906–1909, V.X, 34–64, 231–269 (in Russian)
(English transl.: pages 34–64 by B. Korenblum and M.J. Prelle, SIGSAM Bulletin, 1981, V.15, N 2, 20–32).
[5] Ritt R.H., Integration in finite terms. Liouville’s theory of elementary functions, New York, Columbia Univ.
Press, 1948.
[6] Singer M., Liouvillian first integrals of differential equations, Trans. Amer. Math. Soc., 1992, V.333, 673–
688.
[7] Christopher C., Liouvillian first integrals of second order polynomial differential equations, Electron. J.
Differential Equations, 1999, N 49, 7 pages.
[8] Man Y.K., MacCallum M.A.H., A rational approach to the Prelle–Singer algorithm, J. Symbolic Comput.,
1997, V.24, 31–43.
[9] Man Y.K., First integrals of autonomous systems of differential equations and Prelle–Singer procedure,
J. Phys. A: Math. Gen., 1994, V.27, L329–L332.
[10] Duarte L.G.S., Duarte S.E.S., da Mota L.A.C.P., Analyzing the structure of the integrating factor for first
order differential equations with Liouvillian functions in the solutions, J. Phys. A: Math. Gen., 2002, V.35,
1001–1006.
[11] Duarte L.G.S., Duarte S.E.S., da Mota L.A.C.P., A method to tackle first-order ordinary differential equa-
tions with Liouvillian functions in the solution, J. Phys. A: Math. Gen., 2002, V.35, 3899–3910, math-
ph/0107007.
[12] Shtokhamer R., Glinos N., Caviness B.F., Computing elementary first integrals of differential equations, in
Proccedings of the Conference on Computers and Mathematics (July 30 – August 1, Stanford, California),
1986, V.1, 1–9.
[13] Man Y.K., Computing closed form solutions of first order ODEs using the Prelle–Singer procedure, J. Sym-
bolic Comput., 1993, V.16, 423–443.
[14] Duarte L.G.S., Duarte S.E.S., da Mota L.A.C.P., Skea J.E., An extension of the Prelle–Singer method and
a MAPLE implementation, Comput. Phys. Comm., 2002, V.144, 46–62.
[15] Duarte L.G.S., da Mota L.A., Skea J.E.F., Solving second order equations by extending the PS method,
math-ph/0001004.
[16] Kosovtsov Yu.N., The structure of general solutions and integrability conditions for rational first-order
ODE’s, math-ph/0207032.
[17] Kosovtsov Yu.N., The rational generalized integrating factors for first-order ODEs, math-ph/0211069.
[18] Garcia I.A., Gine J., Generalized cofactors and nonlinear superposition principles, Appl. Math. Lett., 2003,
V.16, 1137–1141.
[19] Garcia I.A., Giacomini H., Gine J., Generalized nonlinear superposition principles for polynomial planar
vector fields, J. Lie Theory, 2005, V.15, 89–104.
[20] Giacomini H., Gine J., An algorithmic method to determine integrability for polynomial planar vector fields,
European J. Appl. Math., to appear.
[21] Kosovtsov Yu.N., 2 Procedures for finding integrating factors of some ODEs of any order, Maple Application
Center, 2004, see here.
[22] Kosovtsov Yu.N., Particular solutions and integrating factors of some ODEs of any order, Maple Application
Center, 2005, see here.
http://www.maplesoft.com/applications/app_center_view.aspx?AID=1309
http://www.maplesoft.com/applications/app_center_view.aspx?AID=1652
1 The Prelle-Singer method
2 The method of resultants
3 Conclusion
|
| id | nasplib_isofts_kiev_ua-123456789-146113 |
| institution | Digital Library of Periodicals of National Academy of Sciences of Ukraine |
| issn | 1815-0659 |
| language | English |
| last_indexed | 2025-12-01T15:53:09Z |
| publishDate | 2006 |
| publisher | Інститут математики НАН України |
| record_format | dspace |
| spelling | Kosovtsov, Y.N. 2019-02-07T13:44:52Z 2019-02-07T13:44:52Z 2006 Finding Liouvillian First Integrals of Rational ODEs of Any Order in Finite Terms / Y.N. Kosovtsov // Symmetry, Integrability and Geometry: Methods and Applications. — 2006. — Т. 2. — Бібліогр.: 22 назв. — англ. 1815-0659 2000 Mathematics Subject Classification: 34A05; 34A34; 34A35 https://nasplib.isofts.kiev.ua/handle/123456789/146113 It is known, due to Mordukhai-Boltovski, Ritt, Prelle, Singer, Christopher and others, that if a given rational ODE has a Liouvillian first integral then the corresponding integrating factor of the ODE must be of a very special form of a product of powers and exponents of irreducible polynomials. These results lead to a partial algorithm for finding Liouvillian first integrals. However, there are two main complications on the way to obtaining polynomials in the integrating factor form. First of all, one has to find an upper bound for the degrees of the polynomials in the product above, an unsolved problem, and then the set of coefficients for each of the polynomials by the computationally-intensive method of undetermined parameters. As a result, this approach was implemented in CAS only for first and relatively simple second order ODEs. We propose an algebraic method for finding polynomials of the integrating factors for rational ODEs of any order, based on examination of the resultants of the polynomials in the numerator and the denominator of the right-hand side of such equation. If both the numerator and the denominator of the right-hand side of such ODE are not constants, the method can determine in finite terms an explicit expression of an integrating factor if the ODE permits integrating factors of the above mentioned form and then the Liouvillian first integral. The tests of this procedure based on the proposed method, implemented in Maple in the case of rational integrating factors, confirm the consistence and efficiency of the method. I would like to thank the referees for extensive comments and suggestions regarding of earlier versions of this paper and Reece Heineke for a careful reading of the paper. en Інститут математики НАН України Symmetry, Integrability and Geometry: Methods and Applications Finding Liouvillian First Integrals of Rational ODEs of Any Order in Finite Terms Article published earlier |
| spellingShingle | Finding Liouvillian First Integrals of Rational ODEs of Any Order in Finite Terms Kosovtsov, Y.N. |
| title | Finding Liouvillian First Integrals of Rational ODEs of Any Order in Finite Terms |
| title_full | Finding Liouvillian First Integrals of Rational ODEs of Any Order in Finite Terms |
| title_fullStr | Finding Liouvillian First Integrals of Rational ODEs of Any Order in Finite Terms |
| title_full_unstemmed | Finding Liouvillian First Integrals of Rational ODEs of Any Order in Finite Terms |
| title_short | Finding Liouvillian First Integrals of Rational ODEs of Any Order in Finite Terms |
| title_sort | finding liouvillian first integrals of rational odes of any order in finite terms |
| url | https://nasplib.isofts.kiev.ua/handle/123456789/146113 |
| work_keys_str_mv | AT kosovtsovyn findingliouvillianfirstintegralsofrationalodesofanyorderinfiniteterms |