scieee Science in your language
[en] (orig)

Bifurcation set for a disregarded Bogdanov-Takens unfolding: Application to 3D cubic memristor oscillators

Abstract

Motivated by the dynamical analysis of certain memristor-based oscillators, in this paper we derive the bifurcation set for a three-parametric Bogdanov-Takens unfolding that has not been previously considered in the literature (the saddle-focus-saddle case). By using several first-order Melnikov functions, we obtain for the first time analytical approximations for the bifurcation curves corresponding to homoclinic and heteroclinic connections, which along with the curves associated to local bifurcations organize the parametric regions with different qualitative phase planes. Our interest is the study of a family of 3D memristor oscillators, whose memristor characteristic function is a cubic polynomial. We show that these systems have a first integral; thus, after reducing the problem in one dimension, we can take advantage of the bifurcation set previously obtained. For a certain parameter region, the existence of closed surfaces completely foliated by periodic orbits in the original three-dimensional setting is shown. Additionally, we clarify some misconceptions that arise from the numerical simulations of these systems, emphasizing the important role played by the invariant manifolds associated to the involved first integral.

Read accessible full text

Bifurcation set for a disregarded Bogdanov-Takens unfolding: Application to 3D cubic memristor oscillators

Author: Amador, Andrés; Freire Macías, Emilio; Ponce Núñez, Enrique
Publisher: Springer Science and Business Media B.V.
Year: 2021
DOI: 10.1007/s11071-021-06352-z
Source: https://idus.us.es/bitstreams/324b274d-c868-462a-9b19-99ca30b934c0/download
a Xi :1811.04862 1 [ma h.DS] 12 No 2018
1
BIFURCATION SET FOR A DISREGARDED
BOGDANOV-TAKENS UNFOLDING. APPLICATION TO 3D
CUBIC MEMRISTOR OSCILLATORS.
ANDR´
ES AMADOR1AND EMILIO FREIRE2AND ENRIQUE PONCE3
Abs ac . We de i e he bi u ca ion se o a no p e iously conside ed h ee-
pa ame ic Bogdano -Takens un olding, showing ha i is possible exp ess i s
ec o ield as wo di e en pe u bed cubic Hamil onians. By using se e al i s -
o de Melniko unc ions, we ob ain o he i s ime analy ical app oxima ions
o he bi u ca ion cu es co esponding o homoclinic and he e oclinic connec-
ions, which along wi h he cu es associa ed o local bi u ca ions o ganize he
pa ame ic egions wi h di e en s uc u es o pe iodic o bi s.
As an applica ion o hese esul s, we s udy a amily o 3D mem is o oscilla-
o s, o which he cha ac e is ic unc ion o he mem is o is a cubic polynomial.
We show ha hese sys ems ha e an in ini y numbe o in a ian mani olds, and
by adding one pa ame e ha s a i ies he 3D dynamics o he amily, i is shown
ha he dynamics in each s a um is opologically equi alen o a ep esen an o
he abo e un olding. Also, based upon he bi u ca ion se ob ained, we show he
exis ence o closed su aces in he 3D s a e space which a e olia ed by pe iodic
o bi s. Finally, we cla i y some misconcep ions ha a ise om he nume ical sim-
ula ions o hese sys ems, emphasizing he impo an ole played by he exis ence
o in a ian mani olds.
Bi u ca ion se , Bogdano -Takens, homoclinic o bi , he e oclinic connec ion, Mel-
niko unc ion, Mem is o oscilla o s
1. In oduc ion
In plana sys ems, he exis ence o some local bi u ca ions may e eal he p esence
o o he bi u ca ions o global cha ac e [13, 14] and he cu es ha de e mine hese
global phenomena a e di icul o de e mine. This is, o ins ance, he case ega ding
he appea ance o homoclinic o he e oclinic connec ions.
A homoclinic connec ion is an o bi o he sys em ha joins a saddle equilib ium
poin o i sel , and gene ally c ea es o des oys pe iodic o bi s (see, o ins ance
[35]). A he e oclinic connec ion joins wo di e en equilib ium poin s o a sys em
and he exis ence o his connec ion can de e mine changes in he basin o a ac ion
o a posi i ely in a ian se .
Following [18], he echniques o s udy homoclinic o bi s in plana ec o ields
we e well de eloped du ing he 1920s in he wo ks o Dulac. The undamen al idea
is ha he ecu en beha io nea a connec ing o bi should be s udied in a ashion
simila o ha used in s udying pe iodic o bi s ia a Poinca ´e e u n map. Bu
he e a e some addi ional complica ions in he s udy o homoclinic o bi s compa ed
o ha o pe iodic o bi s which signi ican ly complica e he analysis.
1Facul ad de Ingenie ´ıa y Ciencias, Depa amen o de Ciencias Na u ales y Ma em´a icas, Pon i-
icia Uni e sidad Ja e iana-Cali, Cali, Colombia.
2,3Depa amen o de Ma em´a ica Aplicada, Escuela T´ecnica Supe io de Ingenie ´ıa, A da. de los
Descub imien os, 41092 Se illa, Spain.
1a amado @ja e ianacali.edu.co, 2e [email protected] 3eponce[email p o ec ed]
1
2 A. AMADOR AND E. FREIRE AND E. PONCE
Usually, he bi u ca ion cu es o homoclinic and he e oclinic connec ions a e
s udied by nume ical con inua ion echniques [17, 18, 9, 36]. On he o he hand,
when a plana sys em can be w i en as a pe u bed Hamil onian sys em, we can
calcula e some Melniko unc ions, in oduced by Melniko in [28], and unde ce ain
hypo heses, he ze os o he associa ed Melniko unc ion de e mine he exis ence
o pe iodic o bi s, homoclinic loops o he e oclinic connec ions, see o ins ance
[8, 20, 3].
As show la e , we will eso o such Melniko unc ions o se ing in o ma ion
on such global bi u ca ions cu es in a wo-pa ame ic plane o a speci ic amily o
di e en ial sys ems.
The no mal o m o he Bogdano -Takens (see [35]) bi u ca ion is gi en by
˙x=y, ˙y=µ1+µ2x+x2±xy.
Following he classi ica ion p oposed in [14], he de o ma ion o codimension h ee
o he p e ious no mal o m is gi en by he un olding
˙x=y,
˙y=µ1+µ2x+αx3+y(µ3+µ4x±x2),
(1)
The p esence o his ype o sys ems has been epo ed in di e en applica ions,
see [16, 2, 24]. On he s udy o bi u ca ion phenomena in hese sys ems many
con ibu ions ha e been made. In [15], he au ho s s udied he global bi u ca ion
diag am o he h ee-pa ame e amily
˙x=y,
˙y=µ1+µ2x−x3+y(µ3−3x2),
and ixing µ3>0, hey ob ained analy ical app oxima ions o he bi u ca ion cu es
o he homoclinic o bi s, by using Melniko unc ions. La e on, he abo e wo k was
quo ed in [23], whe e a nume ical analysis o he same model was pe o med. In
[10, 12, 11], i is conside ed he sys em
˙x=y,
˙y=µ1+µ2x−x3+y(µ3+µ4x−x2),
(2)
and he au ho s showed ha i can be w i en as a pe u bed Hamil onian sys em,
epo ing he maximum numbe o limi cycles. La e , by aking he pa ame e
µ4= 0 in (2), he au ho s in [4, 5, 7, 6] analyzed he sys em as a Li´ena d sys em, i s
local bi u ca ions we e cha ac e ized, and a nume ical s udy o he global bi u ca ions
was done.
While all he abo e e e ences deal wi h he ocus case, in his wo k we s udy
he saddle case
˙x=y,
˙y=µ1+µ2x+x3+y(µ3−3x2),
(3)
which up o he bes o ou knowledge, seems o be a dis ega ded case wi h a he
in e es ing dynamic beha io .
In ac , ou mo i a ion comes om he analysis o ce ain 3D mem is o oscilla o s
[1, 30, 22], whe e unde speci ic hypo heses on he mem is o cha ac e is ics, such
sys em appea s in a na u al way a e a dimensional educ ion achie ed hanks o
he exis ence o a i s in eg al.
BIFURCATION SET FOR A DISREGARDED BOGDANOV-TAKENS UNFOLDING 3
The pape is o ganized in he ollowing way. Fi s , in sec ion 2 we e iew he
in o ma ion ha can be gained by means o a local analysis o he sys em. Ou
main esul s appea in sec ion 3, whe e we apply Melniko heo y o app oxima e
he homoclinic and he e oclinic cu es on a con enien pa ame e plane. We ob-
ain he global bi u ca ion se o sys em (3), by spli ing such a plane in egions
wi h di e en quali a i e dynamical beha io . Nex , in sec ion 4, We show how he
abo e analysis is use ul o de i ing all he possible esponses o ce ain 3D canonical
mem is o oscilla o s, when he lux-cha ge cha ac e is ics unc ion is a speci ic cubic
polynomial, gene alizing some esul s gi en in [1]. As one o he possible dynamical
beha io s, we ocus o a en ion in showing, hanks o he p e ious analysis, he
exis ence o a opological sphe e in he 3D phase-space comple ely olia ed by pe i-
odic o bi s. Thus, we con i m p e ious nume ical esul s epo ed in [29, 25]. The
necessi y o inco po a ing igo ous echniques in he analysis o mem is o oscilla o s
is emphasized wi h he ma e ial o sec ion 5, whe e ollowing a simila p ocedu e
o he dimensional educ ion o sec ion 4, we can e u e se e al ecen ly published
s udies ha epo he exis ence o an in ini e numbe o hidden a ac o s in a
h ee-dimensional mem is o -based au onomous Du ing oscilla o . Some echnical
esul s a e elega ed o he appendix.
2. Local bi u ca ions
In his sec ion, we s udy he local bi u ca ions ha occu in sys em (3). Fi s , we
no e ha he sys em is in a ian unde he ans o ma ion
(x, y, µ1, µ2, µ3)→(−x, −y, −µ1, µ2, µ3).
The e o e, i is su icien o s udy he bi u ca ion diag am o µ1>0.The equilib-
ium poin s o he sys em a e o he o m (x, y) = (˜x, 0), being ˜xa solu ion o he
cubic µ1+µ2x+x3= 0,and i s Jacobian ma ix o is gi en by
(4) J(x, y) = 0 1
µ2+ 3x2−6yx µ3−3x2.
Rema k 1. No e ha o µ3≤0 he di e gence o sys em (3) does no change sign,
hus om Bendixson’s c i e ion [21], he sys em does no ha e pe iodic solu ions.
Fi s , we p o ide a echnical esul ha p o ides a s udy o he numbe o equi-
lib ia in sys em (3) and hei opological na u e.
Lemma 2. Conside sys em (3), he ollowing s a emen s hold.
(a) I µ2≥0o we ha e µ2<0wi h 27µ2
1+ 4µ3
2>0, hen he sys em has only
one equilib ium poin .
(b) I µ2<0and 27µ2
1+ 4µ3
2= 0 we ha e wo equilib ium poin s.
(c) I µ2<0and 27µ2
1+ 4µ3
2<0, hen he sys em has h ee equilib ium poin s
xi= (si,0) wi h i∈ {L, C, R}such ha
(5) sL<−(−µ2/3)1/2< sC<(−µ2/3)1/2< sR,
and sL+sC+sR= 0.Fu he mo e, xLand xRa e saddles while xCis an
an isaddle (node o ocus).
P oo . We s udy he oo s o he polynomial p(x) = µ1+µ2x+x3.Since p′(x) =
µ2+ 3x2,i µ2≥0 we ob ain p′(x)≥0 and so he polynomial has only one oo .
4 A. AMADOR AND E. FREIRE AND E. PONCE
In he es o he p oo we assume µ2<0. The de i a i e p′(x) anishes a he
poin s x±=±(−µ2/3)1/2being a maximum and minimum local espec i ely, also a
di ec compu a ion gi es p(x±) = µ1∓2(−µ2/3)3/2. When p(x−)<0 o p(x+)>0
he g aph o p(x) only c osses once he x-axis and so hese inequali ies p o ides he
condi ion 27µ2
1+ 4µ3
2>0,and he s a emen (a) ollows. Assuming p(x−) = 0 o
p(x+) = 0, he s a emen (b) ollows.
Finally, i p(x+)<0< p(x−) hen we ha e h ee oo s as indica ed in (5).
Mo eo e , using he ela ion be ween oo s and coe icien s o polynomials, we ge
µ1=−sLsCsR, µ2=sCsL+sCsR+sLsR, sL+sC+sR= 0.
Fo he Jacobian ma ix gi en in (4), we ge J(si,0) = −(µ2+ 3s2
i) = −p′(si).As
we know ha p′(sL)>0, p′(sC)<0 and p′(sR)>0, he conclusion ollows and he
p oo is comple e. 
In he nex esul , we gi e a cha ac e iza ion o he local bi u ca ions o sys em
(3) on he pa ame ic plane (µ2, µ1),assuming a ixed alue o he pa ame e µ3.
No e ha wee ha e pu he µ1axis in he plane (µ2, µ1) e ically, being he µ2axis
he ho izon al one.
P oposi ion 3. The ollowing s a emen s hold o sys em (3).
(a) Gi en µ3∈R he pa ame e alues in he se
(6) ϕsn ={(µ2, µ1) : 27µ2
1+ 4µ3
2= 0},
co espond wi h saddle-node bi u ca ion poin s o equilib ia. In pa icula ,
he sys em has a cusp bi u ca ion o equilib ia a µ2=µ1= 0.
(b) Gi en µ3>0, he pa ame e alues in he se
(7) ϕH={(µ2, µ1) : µ1=±(µ3/3)3/2∓(µ3/3)1/2µ2, µ2<−µ3},
ep esen And ono -Hop bi u ca ion poin s o codimension one o he cen al
equilib ium poin xC,see Lemma 2(c).
(c) The se de ined in (7) de e mines a symme ic pai o s aigh hal lines
emana ing om wo poin s co esponding o Bogdano -Takens bi u ca ion
poin s, namely
(8) BT±≡−µ3,±2 (µ3/3)3/2.
P oo . S a emen (a) is a di ec consequence o he equa ions
x3+µ2x+µ1= 0, µ2+ 3x2= 0,
o be ul illed o any non-hype bolic equilib ium (x, 0) a a saddle-node bi u ca ion.
Le (˜x, 0) an equilib ium poin o sys em (3). Conside ing he Jacobian ma ix J
gi en in (4), hen J(˜x, 0) has wo pu ely imagina y eigen alues when aking µ3>0,
he alue ˜xsa is ies ˜x=±pµ3/3 wi h µ2<−µ3<0,because hen µ2+ 3˜x2<0.
The las inequali y is ul illed only o he equilib ium poin xC,see Lemma 2(c).
Since (˜x, 0) is an equilib ium poin we ha e
µ1+µ2±pµ3/3+±pµ3/33= 0,
and s a emen (b) ollows. To show s a emen (c) is su icien o conside he equa-
ions ace (J(˜x, 0)) = de (J(˜x, 0)) = 0.
BIFURCATION SET FOR A DISREGARDED BOGDANOV-TAKENS UNFOLDING 5
3. Global bi u ca ions
In his sec ion we will comple e he bi u ca ion analysis o sys em (3). We will
w i e he sys em as a pe u bed Hamil onian o which he Melniko heo y can be
applied. This can be done in di e en ways, as indica ed in he nex esul . The
possibili y o eso ing o one o he wo nex epa ame iza ion o ms will be help ul
la e .
P oposi ion 4. Sys em (3) can be w i en as wo di e en pe u bed Hamil onian
sys ems, as ollows.
(a) Taking
(9) µ1=ε4ν1, µ2=−ε2ν2, µ3=ε2ν3,
he sys em can be ew i en as
˙x=y,
˙y=−ν2x+x3+εν1+ν3y−3x2y,
(10)
which o ε= 0 co esponds o he Hamil onian
(11) H1(x, y) = y2
2+ν2
x2
2−x4
4.
(b) Taking
(12) µ1=ε3ν1, µ2=−ε2ν2, µ3=ε2ν3,
he sys em can be ew i en as
˙x=y,
˙y=ν1−ν2x+x3+ε(ν3y−3x2y),
(13)
which o ε= 0 co esponds o he Hamil onian
(14) H2(x, y) = y2
2−ν1x+ν2
x2
2−x4
4.
P oo . The blow-up ans o ma ion x1= (1/ε)x, y1= (1/ε2)y, and ˜
=ε , allows
o ew i e sys em (3) as
x′
1=y1, y′
1=x3
1+µ2
ε2x1+µ1
ε3+µ3
εy1−3εx2
1y1,
whe e he p ime deno es de i a i es wi h espec o he new ime ˜
. Now, using (9)
and (12), a e some elemen a y algeb a we ob ain sys ems (10) and (13), espec-
i ely. 
The phase po ai o he unpe u bed Hamil onian sys ems (10) and (14) a e
shown in Figu e 1. No e ha when ν1= 0 we ob ain H1(x, y) = H2(x, y),and so in
ha case i is su icien o s udy he p ope ies o he Hamil onian H1.
Now, we will conside he he e oclinic connec ions o unpe u bed Hamil on-
ian sys em (11). The Hamil onian has a pai o he e oclinic connec ions Γ±( ) =
(x( ),±y( )),pa ame e ized by
x( ) = √ν2 anh pν2/2 ,
y( ) = ν2
√2sech2pν2/2 ,
(15)

6 A. AMADOR AND E. FREIRE AND E. PONCE
whe e −∞ < < ∞and ν2>0.In he nex esul , we compu e he Melniko unc-
ion along he he e oclinic connec ion Γ+ o he unpe u bed Hamil onian sys em
(10), and by using (9), we ob ain he app oxima e bi u ca ion cu es o he e oclinic
connec ions o sys em (3).
Figu e 1. (a) Phase po ai o unpe u bed Hamil onian sys em
(10) wi h ν2= 0.2. We show in g een he wo he e oclinic o bi s, while
he non-closing s able and uns able mani olds o he saddle poin s a e
shown in ed. (b) Phase po ai o unpe u bed Hamil onian sys em
(13) wi h ν1= 0.3 and ν2= 1. We d aw in g een he homoclinic o bi ,
he s able and uns able mani olds o he saddle poin s a e shown in
ed.
P oposi ion 5. I we conside pe u bed Hamil onian sys em (10) and ν= (ν1, ν2, ν3)
wi h ν2>0and ν2
2/4< ν3
2/27 (see Lemma 2(c)) hen he Melniko unc ion along
o he he e oclinic connec ion Γ±is gi en by
(16) Mh (ν) = 2
15√ν215ν1+ 5√2ν2ν3−3√2ν2
2.
P oo . The sys em can be w i en as
( ˙x, ˙y)T= (x, y) + εg(x, y),
whe e (x, y) = (y, −ν2x+x3)Tand g(x, y) = (0, ν1+ν3y−3x2y)T.Thus, we ha e
∧g=y(ν1+ν3y−3x2y). Acco dingly, he Melniko unc ion is de ined by
Mh (ν) = Z∞
−∞
(x( ),±y( )) ∧g(x( ),±y( ))d =
=Z∞
−∞ ±y( )ν1±(ν3−3x2( ))y( )d ,
whe e x( ) and y( ) a e de ined as in (15). A e a di ec compu a ion we ob ain
(16). 
BIFURCATION SET FOR A DISREGARDED BOGDANOV-TAKENS UNFOLDING 7
By using he Melniko heo y and by ixing one pa ame e o sys em (3), we
can gi e an app oxima ion o he he e oclinic connec ion cu es in he emaining
pa ame e s plane.
P oposi ion 6. Conside sys em (3) wi h µ3>0su icien ly small and he pa ame -
ic plane (µ2, µ1). Then he sys em has a unique hype bolic he e oclinic connec ion
in a neighbo hood o he cu e
(17) ϕh ={(µ2, µ1)∈R2:µ1=±√2
15 µ2(3µ2+ 5µ3), µ26=−5µ3/3}.
P oo . Fixing ν3= 1 in he Melniko unc ion gi en in (16), and imposing he
condi ion Mh (ν1, ν2) = 0,we ob ain
ν1=√2
15 ν2(3ν2−5) .
F om (9) we ge ε=√µ3,µ1=µ2
3ν1,and µ2=−µ3ν2,so ha
µ1=µ2
3ν1=−µ2
3
√2
15
µ2
µ33−µ2
µ3−5,
and he conclusion ollows. 
When µ1= 0 and µ3>0 on he pa ame e plane (µ2, µ1),we ob ain he poin o
double he e oclinic connec ions
(18) DHT ≡(−5µ3/3,0) .
We ecall ha Schec e ’s poin s a e co-dimension wo poin s de ined by he in-
e sec ion o a saddle-node cu e and a homoclinic o he e oclinic cu e, o mo e
de ails see [32]. Taking he in e sec ion poin s o he saddle-node bi u ca ion cu e
and he he e oclinic cu es gi en in (6) and (17) espec i ely, we ob ain a i s -o de
app oxima ion o Schec e ’s poin s o he sys em. Since he sys em is symme ic
wi h espec o he pa ame e µ1, he sys em has ou Schec e ’s poin s (see Figu e
3), hese poin s a e
S±
1≡ρ1(5/27),∓(5√10/729) p18µ3+ 5 + √5,
S±
2≡ρ2(5/27),±(5√10/729) p18µ3+ 5 −√5,
(19)
whe e
ρ1=9µ3+ 5 −√5p18µ3+ 5, ρ2=√5p18µ3+ 5 −9µ3−5.
Now, by using he homoclinic connec ion o Hamil onian sys em (14), we compu e
he associa ed Melniko unc ion o sys em (3) when ν3= 1.
P oposi ion 7. I we conside sys em (13) and ν= (ν1, ν2, ν3)wi h ν1>0,ν2>0
and ν3= 1, hen he Melniko unc ion associa ed o he homoclinic o bi wi h
connec ion poin (0, sR), i is gi en by
(20) M(ν) = √2cosh2(θ)
cosh2(θ) + 2 (F1(θ) + ν2F2(θ)) ,
8 A. AMADOR AND E. FREIRE AND E. PONCE
whe e
F1(θ) =720θ−320 sinh θ+ 240θcosh3θ−320 cosh2θsinh θ−
−80 cosh4θsinh θ+ 480θcosh θ,
F2(θ) =1440θcosh θ−768 sinh θ−cosh3θ−1344 cosh2θsinh θ−
−48 cosh4θsinh θ
(21)
and 0< θ < ∞, wi h
cosh θ=2s
ω, ω2= 2(ν2−s2
R)>0, ν1=ν2sR−s3
R,
being sR he bigges posi i e oo o he equa ion ν1−ν2x+x3= 0, see Figu e 2.
P oo . We conside he unpe u bed Hamil onian sys em gi en in (13) wi h ν2>0.
F om Lemma 2(c), he sys em has 3 equilib ium poin s xi= (si,0), whe e xLand
xRa e saddle poin s and xCis a ocus o node and
sL< sC< sR, sL+sC+sR= 0, sLsCsR=−ν1.
We s udy only he case ν1>0, o he case ν1<0 is analogous.
Sys em (13) can w i en as
( ˙x, ˙y)T= (x, y) + εg(x, y).
Now, assuming ν1>0,by G een’s Theo em, he homoclinic Melniko unc ion o
he sys em can ew i en as
Mh(ν) = Z ZD(ν1,ν2)−∂g(x, y)
∂y dA,
whe e Dis he egion bounded by he homoclinic o bi which joins he equilib ium
poin (sR,0) o i sel . By ixing ν3= 1 ( ha is µ3>0), and aking p(x) =
ν1−ν2x+x3,we ge p(sR) = ν1−ν2sR+s3
R= 0, ha is
(22) ν1=sR(ν2−s2
R),
and so ν2−s2
R>0.Taking he auxilia y unc ion
q(x) = Zx
0
p(x)dx =ν1x−ν2
x2
2+x4
4,
and using (14), he homoclinic loop is gi en by he poin s (x, y±
s(x)) whe e x≤x≤
sR,
y±
s(x) = ±√2pq(x)−q(sR),
and y±
s(x) = y±
s(sR) = 0,see Figu e 2. Now, he Melniko unc ion is hanks o he
symme y o he loop
Mh(ν) = 2 ZsR
x
(3x2−1)dx Zy+
s(x)
0
dy =
=√2Zs
x
(3x2−1)(sR−x)q(x+sR)2−2(ν2−s2
R)dx =
=√2Zs
x
(3x2−1)(sR−x)p(x+sR)2−ω2dx,
BIFURCATION SET FOR A DISREGARDED BOGDANOV-TAKENS UNFOLDING 9
whe e om (22) ω2= 2(ν2−s2
R),and we ha e used ha
q(x)−q(sR) = 1
4(x−sR)2(x+sR)2−ω2.
Taking he change o a iable
x+sR=ωcosh θ,
and no ing ha q(sR)−q(x) = 0 we see ha x+sR=ω, which co esponds o θ= 0,
while o x=sR he co esponding alues o θ=θsRsa is y cosh θsR= 2sR/ω, o
also ω2cosh2θsR= 4s2
R, ha is (s2
R+ν2) cosh2θsR= 2s2
R,and so we ge
s2
R=cosh2θsR
2 + cosh2θsR
ν2.
Now we a i ed o
Mh(ν) = √2ω2ZθsR
0
(1 −3(ωcosh θ−sR)2)(2sR−ωcosh θ) sinh2θdθ,
and a e some compu a ions, we ob ain (20) and (21), whe e θsRhas been simpli ied
o θ.
0
x
0
y
y+
s(x)
y−
s(x)
sC
sLxsR
Figu e 2. Homoclinic o bi which joins he saddle equilib ium poin
(sR,0) o i sel .
As a di ec consequence o he abo e esul , we gi e an analy ical app oxima ion
o he bi u ca ion cu es o homoclinic connec ions o sys em (3).
P oposi ion 8. Conside sys em (3) wi h µ3>0su icien ly small and he pa ame -
ic plane (µ2, µ1).Then he sys em has a unique homoclinic o bi in a neighbo hood
o he cu e
(23) ϕh={(µ2, µ1)∈R2:µ2=−µ3ν2(θ), µ1=±µ3/2
3ν1(θ),0< θ < ∞},
whe e
ν2(θ) = 10(cosh 2θ+ 5)(9 sinh θ+ sinh 3θ−12θcosh θ)
3(370 sinh θ+ 115 sinh 3θ+ sinh 5θ−60θ(11 cosh θ+ cosh 3θ)),
ν1(θ) = ν2(θ)s−s3, s2=cosh2θ
2 + cosh2θν2(θ).
(24)
16 A. AMADOR AND E. FREIRE AND E. PONCE
sys em (46). This esul gua an ees he exis ence o a opological sphe e in he 3D
phase-space comple ely olia ed by pe iodic o bi s.
P oposi ion 11. Conside sys em (30) wi h β, ξ > 0,, he unc ion qde ined as in
(31),a2−3b < 0and
(37) a2−3b+ 3β > 0
su icien ly small. Addi ionally, suppose ha he ollowing inequali ies hold
0<3b−a2<3ξ/β,
β(3b−a2)−3ξ < (5/2) 3b−a2−3ββ < 0.
(38)
Then o all h∈Rwi h
−A
27 < h < B
27,
whe e
A=4a2β+ 3β2−12bβ + 9ξpa2−3b+ 3β+ 9aξ + 2a3β−9abβ,
B=4a2β+ 3β2−12bβ + 9ξpa2−3b+ 3β−9aξ −2a3β+ 9abβ,
(39)
he sys em has a s able pe iodic o bi . Mo eo e , he e exis a opological sphe e Ω
(see Figu e 10) olia ed by such pe iodic o bi s.
P oo . F om Rema k 10, and a e subs i u ing he alues o µ1, µ2and µ3gi en
in (35), we ob ain he inequali ies (37)-(38). Now, om (36) we ob ain |µ1|<
(1/3) (µ3/3)1/2(µ3−3µ2),so ha µ3−3µ2>0,since om hypo heses we ha e
µ2<−(5/2)µ3<0.Now a e some algeb a we ob ain
|27h+ 9aξ + 2a3β−9abβ|<4a2β+ 3β2−12bβ + 9ξpa2−3b+ 3β.
Taking in o accoun he absolu e alue, and g ouping e ms, we ob ain he alues o
Aand Bde ined in (39). Finally, om Rema k 10 sys em (30) has a s able pe iodic
o bi on each Shde ined in (32), so a ying he pa ame e h, we ob ain a sphe e
olia ed by such pe iodic o bi s. 
5. False Hidden A ac o s in Mem is o -Based Au onomous
Du ing Oscilla o s
An a ac o is called a hidden a ac o i i s basin o a ac ion does no in e sec
any neighbo hood o equilib ia; o he wise, i is called a sel -exci ed a ac o , o
mo e de ails see [26, 27]. Recen ly in [33, 34, 19] i was epo ed he exis ence o
an in ini e numbe o hidden a ac o s in a mem is o -based au onomous Du ing
oscilla o s, whose mem is ance unc ion is a cubic polynomial. He e, by using a
simila app oach o he ollowed in he p e ious sec ion, we will show ha such
hidden a ac o s a e no possible, so ha he nume ical simula ions included in
[33, 34, 19] a e misleading.
The quo ed mem is o based au onomous Du ing oscilla o is modeled by he
dynamical sys em
˙x=y,
˙y=z,
˙z=−αz −M(x)y,
(40)

BIFURCATION SET FOR A DISREGARDED BOGDANOV-TAKENS UNFOLDING 17
Figu e 10. Using (62) on each in a ian mani old Shde ined in (32),
some slices o he su ace Ω gi en by P oposi ion 11 o sys em (30)
wi h pa ame e s a= 1, b = 4.8, β = 5 and ξ= 80 a e shown. Fo
his se o pa ame e s we ge µ3= 0.106 >0, µ2=−2.3<−(5/2)µ3,
A= 1180.1 and B= 152.2.
whe e he mem is ance unc ion M(possibly discon inuous) is de ined as
(41) M(x) = dφ(x)
dx
and φis a con inuous unc ion. Sys em (40) has a con inuum o equilib ia, since any
poin o he x-axis is an equilib ium poin . In he nex esul , we show ha e en
sys em (40) does no belong o he amily (46) o he appendix, he sys em also has
he p ope y o possessing an in ini e numbe o in a ian mani olds.
P oposi ion 12. Conside sys em (40) wi h he unc ion Mde ined as in (41). Fo
any h∈R he se
(42) Sh={(x, y, z)∈R3:H(x, y, z) = h}
is an in a ian mani old o he sys em, whe e we ha e in oduced he con inuous
unc ion
(43) H(x, y, z) = φ(x) + αy +z.
The e o e, he sys em has an in ini e numbe o in a ian mani olds olia ing he
whole R3, and so he dynamics is essen ially wo-dimensional.
P oo . Taking Has in (43), de ine o any solu ion (x(τ), y(τ), z(τ)) o (40) he
auxilia y con inuous unc ion
h(τ) = H(x(τ), y(τ), z(τ))
Now, a di ec compu a ion gi es, excep ing he poin s o possible non-di e en iabili y,
h′(τ) = dφ(x)
dx ˙x+α˙y+ ˙z=M(x)y+αz −αz −M(x)y= 0.
Then his piecewise cons an along he o bi s o (40), bu as his con inuous by
de ini ion, i should be globally cons an . In sho , he le el se s o Ha e in a ian
o he low. 
18 A. AMADOR AND E. FREIRE AND E. PONCE
Now, by using he abo e esul , we educe he s udy o he dynamical beha io
o he sys em, o he s udy o a plana sys em.
P oposi ion 13. Conside sys em (40) wi h he unc ion Mde ined as in (41).
Then on each in a ian se Shde ined in (42) he sys em is opologically equi alen
o he plana sys em
˙x=y,
˙y=−φ(x)−αy +h.
(44)
Mo eo e , (x(τ), y (τ)) ∈R2is a solu ion o he abo e sys em i and only i Eh(x(τ), y (τ))
is a solu ion o sys em (40), whe e
(45) Eh(X(τ), Y (τ)) = 
x(τ)
y(τ)
h−φ(x(τ)) −αy(τ)


P oo . F om P oposi ion 12 we can sol e o zin he equa ion H(x, y, z) = h, and
w i e
z=h−φ(x)−αy.
Replacing his exp ession in o he i s and second equa ion o (40) we ob ain sys em
(44). Suppose ha (x(τ), y (τ)) ∈R2is a solu ion o sys em (44). Taking
z(τ) = h−αy(τ)−φ(x(τ))
we ob ain
˙z(τ) = −α˙y(τ)−dφ(x(τ))
dx ˙x(τ) = −α(h−φ(x(τ)) −αy(τ)) −M(x(τ))y(τ) =
=−α(z(τ)) −M(x(τ))y(τ).
and he p oposi ion ollows. 
In he ollowing esul , we show ha o α6= 0, he sys em does no ha e pe iodic
solu ions.
P oposi ion 14. Conside sys em (44). The ollowing s a emen s hold.
(a) Fo α= 0 he sys em is Hamil onian.
(b) Fo α6= 0 he sys em does no ha e pe iodic solu ions.
P oo . The di e gence o he sys em is ∆ = −α. Then, when α= 0 he sys em
co esponds o he Hamil onian
H(x, y) = y2
2+φ′(x).
Fo α6= 0 he di e gence o sys em (44) does no change sign, hus om Bendixson’s
c i e ion [21], sys em (40) does no ha e pe iodic solu ions. 
Rema k 15. No e ha as a consequence o p oposi ion 13 and 14, he 3Dsys em
(40) sys em canno ha e pe iodic o bi s o any con inuous unc ion φand α6= 0.
Howe e , when α= 0 he sys em could ha e an in ini e numbe o pe iodic o bi s on
each in a ian se Shde ined in (42).
BIFURCATION SET FOR A DISREGARDED BOGDANOV-TAKENS UNFOLDING 19
Rega ding [33, 34, 19] au ho s conside sys em (40) wi h he unc ion φ(x) =
ωx +βx3and he se o pa ame e s α= 0.0001, ω = 0.35, β = 0.85.In bo h quo ed
e e ences, au ho s epo ed he exis ence o an in ini e numbe o s able pe iodic
o bi s coexis ing wi h an in ini e numbe o s able equilib ia, by aking in o accoun
se e al nume ical simula ions, so concluding he exis ence o hidden a ac o s.
F om P oposi ions 12 and 13, we ob ain he in a ian mani olds
Sh={(x, y, z)∈R3:ωx +βx3+αy +z=h},
and he plana sys em uling he dynamics on each Shgi en by
˙x=y, ˙y=−ωx −βx3−αy +h.
F om Rema k 15 we no e ha , he sys em canno ha e pe iodic o bi s, and so,
he s a emen made in he quo ed pape s is clea ly w ong, p obably a e gi ing
oo much c edi o nume ical simula ions. This emphasizes he ele ance o he
app oach ollowed in his wo k which allows o a oid misconcep ions coming jus
om nume ical simula ions.
6. Conclusions
Mo i a ed by he dynamical analysis o 3D mem is o oscilla o s whose nonlin-
ea cha ac e is ics is a cubic polynomial, and a e showing ha hei dynamics
is essen ially wo-dimensional, he need o conside a dis ega ded un olding o he
Bogdano -Takens singula i y na u ally a ose. The co esponding bi u ca ion se ,
including bo h local and global bi u ca ions has been desc ibed. While local bi u -
ca ions can be easily de ec ed, he cha ac e iza ion o global bi u ca ions pa ame e s
cu es is much mo e in ol ed; only by eso ing o Melniko ’s heo y i was possible
o ob ain such cu es p o iding a comple e desc ip ion o he bi u ca ion se .
Rega ding he conside ed 3D mem is o oscilla o s, and by wo king wi hin some
pa ame e s egions o he abo e bi u ca ion se , i has been possible o show igo -
ously he exis ence o mul iple pe iodic o bi s leading o a opological sphe e.
When he same app oach is applied o a di e en amily o 3D mem is o oscilla-
o s, i has been shown ha he oscilla ions a e no possible, con a ily o wha had
been ecen ly claimed.
Acknowledgemen s
The i s au ho is suppo ed by Pon i icia Uni e sidad Ja e iana Cali-Colombia.
E. F ei e and E. Ponce a e pa ially suppo ed by MINECO/FEDER g an MTM2015-
65608-P and by he Conseje ´ıa de Econom´ıa, Inno aci´on, Ciencia y Empleo de la
Jun a de Andaluc´ıa unde g an P12-FQM-1658.
Re e ences
[1] A. Amado , E. F ei e, E. Ponce, and J. Ros. On discon inuous piecewise linea models o
mem is o oscilla o s. In e na ional Jou nal o Bi u ca ion and Chaos, 27(06):1730022, jun
2017.
[2] H.J. De Blank, Y. Kuzne so , M.J. Pekk´e , and D.W.M. Veldman. Degene a e Bogdano -
Takens bi u ca ions in a one-dimensional anspo model o a usion plasma. Physica D:
Nonlinea Phenomena, 331:13 – 26, 2016.
[3] T.R. Blows and L.M. Pe ko. Bi u ca ion o limi cycles om cen e s and sepa a ix cycles o
plana analy ic sys ems. SIAM Re iew, 36(3):341376, 1994.
20 A. AMADOR AND E. FREIRE AND E. PONCE
[4] H. Chen and X. Chen. Dynamical analysis o a cubic li´ena d sys em wi h global pa ame e s.
Nonlinea i y, 28(10):35353562, Ap 2015.
[5] H. Chen and X. Chen. Dynamical analysis o a cubic li´ena d sys em wi h global pa ame e s
(ii). Nonlinea i y, 29(6):1798, 2016.
[6] H. Chen and X. Chen. Global phase po ai o a degene a e bogdano - akens sys em wi h
symme y II. Disc e e & Con inuous Dynamical Sys ems-B, 23(10):4141–4170, 2018.
[7] H. Chen, X. Chen, and J. Xie. Global phase po ai o a degene a e bogdano - akens sys em
wi h symme y. Disc e e & Con inuous Dynamical Sys ems-B, 22(4):1273–1293, 2017.
[8] G. Dangelmay and J. Guckenheime . On a ou pa ame e amily o plana ec o ields.
A chi e o Ra ional Mechanics and Analysis, 97(4):321352, 1987.
[9] A. Dhooge, W. Go ae s, Y. Kuzne so , H.G.E. Meije , and B. Sau ois. New ea u es o he
so wa e ma con o bi u ca ion analysis o dynamical sys ems. Ma hema ical and Compu e
Modelling o Dynamical Sys ems, 14(2):147175, 2008.
[10] F. Dumo ie and C. Li. Pe u ba ions om an ellip ic hamil onian o deg ee ou . Jou nal o
Di e en ial Equa ions, 175(2):209243, 2001.
[11] F. Dumo ie and C. Li. Pe u ba ion om an ellip ic Hamil onian o deg ee ou -III global
cen e. Jou nal o Di e en ial Equa ions, 188(2):473511, 2003.
[12] F. Dumo ie and C. Li. Pe u ba ion om an ellip ic Hamil onian o deg ee ou IV global
cen e. Jou nal o Di e en ial Equa ions, 188(2):512566, 2003.
[13] F. Dumo ie , R. Roussa ie, and J. So omayo . Gene ic 3-pa ame e amilies o ec o ields
on he plane, un olding a singula i y wi h nilpo en linea pa . he cusp case o codimension
3. E godic Theo y and Dynamical Sys ems, 7(03), 1987.
[14] F eddy Dumo ie . Bi u ca ions o plana ec o ields: nilpo en singula i ies and Abelian
in eg als. Sp inge -Ve lag, Be lin New Yo k, 1991.
[15] F. Fe n´andez, E. F ei e, L. Piza o, and A.J. Rod ´ıguez-Luis. P oc NDES 96: 4 h wo kshop on
nonlinea dynamics o elec onic sys ems (se ille, spain). ed J.L. Hue as and A.J. Rod ´ıgues-
Luis, page 3216, 1996.
[16] E. F ei e, L. Piza o, A. J. Rod ´ıguez-Luis, and F. Fe n´andez S´anchez. Mul ipa ame ic bi-
u ca ions in an enzyme-ca alyzed eac ion model. In e na ional Jou nal o Bi u ca ion and
Chaos, 15(3):905947, 2005.
[17] E. F ei e, L. Piza o, and A.J. Rod ´ıguez-Luis. Examples o non-degene a e and degene a e
cuspidal loops in plana sys ems. Dynamics and S abili y o Sys ems, 14(2):129161, 1999.
[18] E. F ei e, L. Piza o, and A.J. Rod ´ıguez-Luis. Nume ical con inua ion o homoclinic o bi s
o non-hype bolic equilib ia in plana sys ems. Nonlinea Dynamics, 23(4):353375, 2000.
[19] P. Goda a, D. Dudkowski, A. P asad, and T. Kapi aniak. New opological ool o mul is able
dynamical sys ems. Chaos: An In e disciplina y Jou nal o Nonlinea Science, 28(11):111101,
no 2018.
[20] J. Guckenheime , R. Rand, and D. Schlomiuk. Degene a e homoclinic cycles in pe u ba ions
o quad a ic hamil onian sys ems. Nonlinea i y, 2(3):405, 1989.
[21] Jack Hale. Dynamics and Bi u ca ions. Sp inge New Yo k, New Yo k, NY, 1991.
[22] M. I oh and L.O. Chua. Mem is o oscilla o s. In e na ional Jou nal o Bi u ca ion and Chaos,
18(11):3183–3206, 2008.
[23] A.I. Khibnik, B. K auskop , and C. Rousseau. Global s udy o a amily o cubic liena d
equa ions. Nonlinea i y, 11(6):1505, 1998.
[24] L. Kong and C. Zhu. Bogdano - akens bi u ca ions o codimensions 2 and 3 in a Leslie-Gowe
p eda o -p ey model wi h michaelis-men en- ype p ey ha es ing. Ma hema ical Me hods in
he Applied Sciences, 40(18):67156731, Ma 2017.
[25] I.A. Ko nee and V.V. Semeno . And ono - Hop bi u ca ion wi h and wi hou pa ame e in
a cubic mem is o oscilla o wi h a line o equilib ia. Chaos: An In e disciplina y Jou nal o
Nonlinea Science, 27(8):081104, aug 2017.
[26] N.V. Kuzne so , G.A. Leono , and S.M. Seledzhi. Hidden oscilla ions in nonlinea con ol
sys ems. IFAC P oceedings Volumes, 44(1):2506–2510, jan 2011.
[27] G.A. Leono , N.V. Kuzne so , and V.I. Vagai se . Localiza ion o hidden Chua’s a ac o s.
Physics Le e s A, 375(23):2230–2233, jun 2011.
[28] V.K. Melniko . On he s abili y o a cen e o ime-pe iodic pe u ba ions. T ans. Moscow
Ma h, 12:1–57, 1963.
BIFURCATION SET FOR A DISREGARDED BOGDANOV-TAKENS UNFOLDING 21
[29] M. Messias, C. Nespoli, and V.A. Bo a. Hop bi u ca ion om lines o equilib ia wi hou pa-
ame e s in mem is o oscilla o s. In e na ional Jou nal o Bi u ca ion and Chaos, 20(02):437–
450, eb 2010.
[30] E. Ponce, A. Amado , and J. Ros. Un a elling he dynamical ichness o 3D canonical mem-
is o oscilla o s. Mic oelec onic Enginee ing, 182:15–24, oc 2017.
[31] A.J. Rod ´ıguez-Luis, E. F ei e, and E. Ponce. A me hod o homoclinic and he e oclinic con-
inua ion in wo and h ee dimensions. Con inua ion and Bi u ca ions: Nume ical Techniques
and Applica ions, Sp inge Ne he lands, pages 197–210, 1990.
[32] S. Schec e . The saddle-node sepa a ix-loop bi u ca ion. SIAM Jou nal on Ma hema ical
Analysis, 18(4):11421156, 1987.
[33] V. Va shney, S. Saba a hinam, A. P asad, and K. Thamilma an. In ini e numbe o hidden
a ac o s in mem is o -based au onomous du ing oscilla o . In e na ional Jou nal o Bi u -
ca ion and Chaos, 28(01):1850013, 2018.
[34] V. Va shney, S. Saba a hinam, K. Thamilma an, M.D. Sh imali, and A. P asad. Exis ence
and con ol o hidden oscilla ions in a mem is i e au onomous du ing oscilla o . Nonlinea
Dynamical Sys ems wi h Sel -Exci ed and Hidden A ac o s S udies in Sys ems, Decision and
Con ol, page 327344, 2018.
[35] S ephen Wiggins. In oduc ion o applied nonlinea dynamical sys ems and chaos. Sp inge ,
2003.
[36] V. Wi e, W. Go ae s, Y. Kuzne so , and M. F iedman. In e ac i e ini ializa ion and con in-
ua ion o homoclinic and he e oclinic o bi s in MATLAB. ACM T ansac ions on Ma hema ical
So wa e, 38(3):134, Jan 2012.
Appendix: Dimensional educ ion in 3D mem is o oscilla o s
We conside a amily o h ee-dimensional sys ems, which is gene al enough o
cap u e all he ma hema ical models o mem is o oscilla o s gi en in (30). Such
amily was been s udied in [1] and [30], whe e he au ho s showed ha he dynamics
o such a amily o h ee-dimensional sys ems is essen ially uled by a one pa ame e
se o wo-dimensional sys ems. We conside he sys em
(46)
˙x=a11W(z)x+a12y,
˙y=a21x+a22y,
˙z=x,
whe e he cons an s a11, a12, a21, a22 ∈Rand he unc ion Wallows o de ine a
con inuous unc ion
(47) q(z) = Zz
0
W(s)ds.
The nex esul gua an ees ha he dynamics o sys em (46) is essen ially wo-
dimensional, see [1] o a p oo .
P oposi ion 16. Conside sys em (46) whe e he unc ions Wand qa e ela ed as
in (47). Fo any h∈R, he se
(48) Sh={(x, y, z)∈R3:−a22x+a12y−a12a21z+a11a22q(z) = h}
is an in a ian mani old o he sys em. The e o e, he sys em has an in ini e amily
o in a ian mani olds olia ing he whole R3, and so he dynamics is essen ially
wo-dimensional.
In he ollowing esul we show ha on each in a ian se Shgi en in (48), and o
any con inuous unc ion qde ined as in (47), he dynamics is opologically equi alen
o a Li´ena d sys em. Fu he mo e, we gi e o any solu ion o he Li´ena d sys em
wi h a gi en alue o h, he co esponding solu ion o he 3D canonical model (46).

22 A. AMADOR AND E. FREIRE AND E. PONCE
This esul is a gene aliza ion o Theo em 3 gi en in [1], whe e he unc ion qwas
conside ed o be a con inuous piecewise linea unc ion.
P oposi ion 17. Conside sys em (46) wi h he unc ion qde ined as in (47). I
a12 6= 0, hen on each in a ian se Shgi en in (48), he dynamics is opologically
equi alen o he Li´ena d sys em
(49) ˙
X=Y−F(X),˙
Y=−g(X) + h,
whe e Fand ga e gi en by
(50) F(X) = −a11q(X)−a22X, g(X) = a11a22q(X)−a12a21X
Mo eo e , (X(τ), Y (τ)) ∈R2is a solu ion o he Li´ena d sys em (49) o a gi en
h∈R, i and only i Eh(X(τ), Y (τ)) ∈R3is a solu ion o sys em (46) on Sh,
whe e
(51) Eh(X(τ), Y (τ)) = 

Y(τ)−F(X(τ))
1
a12 [(a2
22 +a12a21)Y(τ)−a22Y(τ) + h]
X(τ)

.
P oo . Fi s , wi h a12 6= 0 he change o a iables
(52) x=x, y =a22x−a12y, z =z
ans o ms sys em (46) in o he sys em
˙
x= 1(z)x−y,(53)
˙
y= 2(z)x,
˙
z=x,
whe e he unc ions 1and 2a e de ined as
(54) 1(z) = a11W(z) + a22, 2(z) = a22a11W(z)−a12a21.
F om P oposi ion 16, he in a ian mani olds (48) o sys em (53)-(54) can be w i en
in he new a iables as
(55) e
Sh={(x, y, z)∈R3:−y+g(z) = h}.
Now, eplacing he condi ion gi en in (55) in he i s equa ion o (53) and emo ing
he unnecessa y second equa ion, we ob ain he sys em
(56) ˙
x= 1(z)x−g(z) + h,
˙
z=x.
whe e he unc ion gis de ined by
(57) g(u) = a11a22q(u)−a12a21u.
A e he change o a iables
(58) X=z,
Y=−˜
F(z) + x,
whe e Fis
(59) ˜
F(z) = a11q(z) + a22z,
we ob ain ˙
X=˙
z=x=Y+˜
F(X) = Y−(−˜
F(X)),
BIFURCATION SET FOR A DISREGARDED BOGDANOV-TAKENS UNFOLDING 23
so ha
˙
Y=−˜
F′(z)˙
z+˙
x=−(a11q′(z) + a22) + ( 1(z)x−g(z) + h) =
=− 1(z)x+ 1(z)x−g(z) + h=−g(z) + h,
and aking F(X) = −˜
F(X) we ob ain sys em (49)-(50).
I (X(τ), Y (τ)) ∈R2is a solu ion o sys em (49)-(50) o a gi en h∈R, we ha e
om (58) ha x(τ)
z(τ)=Y(τ)−F(z(τ))
X(τ)
is a solu ion o sys em (56). F om (55), we ob ain on e
Sh ha y=g(z)−h, wi h g
as in (57). Thus, 
x(τ)
y(τ)
z(τ)

=
Y(τ)−F(X(τ))
g(X(τ)) −h
X(τ)

,
is a solu ion o sys em (53) on Sh. Finally, om (52) we ob ain o sys em (46) he
solu ion x(τ) = x(τ),
y(τ) = 1
a12
[a22x(τ)−y(τ)] = 1
a12
[a22Y(τ)−a22F(X(τ)) −g(X(τ)) + h]
=1
a12 ha22Y(τ) + a22 ˜
F(X(τ)) −g(X(τ)) + hi,
and z(τ) = z(τ). The conclusion ollows om he ac ha o all Xwe ha e
a22 ˜
F(X)−g(X) = (a2
22 +a12a21)X.

In o de o apply he analysis pe o med o sys em (3), in wha ollows we conside
he unc ion qde ined by a cubic polynomial, ha is, we assume
(60) W(z) = 3cz2+ 2az +b, q(z) = cz3+az2+bz,
wi h c6= 0. As a di ec consequence o P oposi ions 16 and 17, we ob ain he nex
esul .
Co olla y 18. Conside sys em (46) wi h he unc ions qand Wde ined as in (60).
I a12 6= 0, hen on each in a ian se Shgi en by
Sh={(x, y, z)∈R3:−a22x+a12y+a11a22cz3+aa11a22z2+(ba11a22 −a12a21)z=h}
he dynamics is opologically equi alen o he Li´ena d sys em
˙x=y+ca11x3+aa11x2+ (ba11 +a22)x,
˙y=−a11a22cx3−a11a22ax2+ (a12a21 −a11a22b)x+h.
(61)
Mo eo e , (x(τ), y (τ)) ∈R2is a solu ion o he Li´ena d sys em (61) o a gi en
h∈R, i and only i Eh(x(τ), y (τ)) ∈R3is a solu ion o sys em (46) on Sh,whe e
(62) Eh(x(τ), y (τ)) = 

y(τ) + ca3
11x(τ)3+aa2
11x(τ)2+ (ba11 +a22)x(τ)2
1
a12 [(a2
22 +a12a21)y(τ)−a22y(τ) + h]
x(τ)

.
In he nex P oposi ion, we show ha sys em (61) can be w i en in o he o m
(1).
24 A. AMADOR AND E. FREIRE AND E. PONCE
P oposi ion 19. The ollowing s a emen s hold o sys em (61).
(a) I a22 6= 0 and a11a22 <0 hen he sys em can be w i en in o he o m
(63) ˙x=y, ˙y=µ1+µ2x+cx3+µ3y+ 3ca11x2y.
whe e he new pa ame e s µ1, µ2and µ3a e gi en by
µ1=27ch +a11a22a(9cb −2a2)−9caa12a21
27c2(−a11a22)5/2,
µ2=a11a22(a2−3cb) + 3ca12a21
3c(a11a22)2, µ3=a11(a2−3cb)−3ca22
3ca11a22
.
(64)
(b) I a22 = 0 hen he sys em can be w i en in o he o m
(65) ˙x=y, ˙y=µ1+µ2x+µ3y+ 3ca11x2y,
whe e he new pa ame e s µ1, µ2and µ3a e de ined by
(66) µ1=h−aa12a21
3c, µ2=a12a21, µ3=ba11 −a2a11
3c.
P oo . Fi s , he change o a iables
u=x+a
3c, =y+2
27
a3
c2a11 −1
3
a
ca22 −1
3ab
ca11,
ans o ms sys em (61) in o
˙u= +ca11u3+λ1u,
˙ =−ca11a22u3+λ2u+λ3,
(67)
whe e he new pa ame e s a e
λ1=a22 +ba11 −1
3
a2
ca11, λ2=a12a21 −ba11a22 +1
3
a2
ca11a22,
λ3=h+1
3ab
ca11a22 −1
3
a
ca12a21 −2
27
a3
c2a11a22.
(68)
I a11a22 <0, he change o a iable
x=1
(−a11a22)1/2u, y = , τ =1
−a11a22
,
ans o ms sys em (67)-(68) in o
˙x=1
(−a11a22)3/2y+ca11x3−λ1
a11a22
x,
˙y=λ3
−a11a22
+λ2
(−a11a22)1/2x+c(−a11a22)3/2x3
and aking in o accoun ha
¨x=1
(−a11a22)3/2˙y+ 3ca11x2˙x−λ1
a11a22
˙x,
and a e some algeb a, s a emen (a) ollows.
I a22 = 0, hen om sys em (67), we ob ain s a emen (b) a e a di ec compu a ion.
