Professional Documents
Culture Documents
Bifurcation
Bifurcation
0). The cal-
culation of a, allows the location of this transition to be
predicted, and the normal-form analysis yields a detailed
lnderstanding of the dynamics near this critical region.
For simplicity we assume the two stable waves have
equal damping rates T and equal amplitudes a, a3.
‘The dynannical variables are then (a,,a;,4), where a, is
the amplitude of the unstable wave and "4, —4)~—¢3 is
the phase difference between the waves. Following Wers-
inger et al. (1980), we introduce the coordinates
(x,y,2)=(a,c084,<;8ing,a3), so that the wave interac-
tion is described by
aPl ft -@ 2 Blo
S\ylela 1 oo {pla} |, on
2} lo o -arj[z) [=x
where 1=0,—o,—0 measures the detuning from the
resonance @;=0;+a;. Both parameters and Tare
non-negative. For additional background on the plasma,
physics ancestry of Eq, (9.1) see Wersinger et al. (1980).
‘The chaotic dynamics of the model in the regime of large
damping (P+ and 0/1 fixed) has been analyzed by
Hughes and Proctor (1990). The analysis of the Hopf bi-
furcation in these equations follows Crawford (1983)
Let V(x,y,2) denote the two-parameter family of vec~
tor fields defined by the model (9.1), leaving the depen-
Rev. Mod. Py, Vol 62, No.4, tober 1991
John David Crawford: Introduction to bifurcation theory
dence on (0,1) implicit. The divergence of this family is
divV=21-P) 2)
For I°<1 the flow expands volumes in R” and there are
‘no stable bounded solutions; for I'> 1 the flow contracts,
volumes (Verhulst, 1990). Since the equations are un-
changed by the shift (,y)+(~0,—y), we may assume
{to be non-negative,
1. Linear analysis
V has two fixed points. There is a trivial fixed point at
(ey,2)=(0,0,0) corresponding to no waves; this solution,
is unstable, since the high-frequency wave is unstable.
There is a nontrivial fixed point at
or a
r iit D
23—I | ar-1? Hl
03)
(=o.70:20)= |
whose stability depends on and I’, If we shift the o:
gin 10 Konyoroh X=X"+xX0, Y=Y"tYg, 2=2'-+Z9 and
drop the primes, then
af | 1 tu 1) fe} fa
ap -p =n ol |yf42}ay |,
2) (wtp 0 ojlz} [xe
oa)
where w=2P and p=G/qu—1). The eigenvalues A of
the linearization at (x,y,2)™(0,0,0) satisiy
I+ (p—BWE+ [LAL 2p TA la D+ p?)=0 -
(9.5)
For 422, all coeficients are non-negative and the con-
FIG. 19. Surfaces of constant ¥ in the (9,P) parameter space.
‘The Hopf bifurcation surface is 0; for 7 <0 the fixed point
(9.3) is stable.John David Crawford: Introduction to bifurcation theory
stant term is positive, This implies that any real root
‘must be negative; in particular, A==0 cannot occur in this
region of parameter space. If cigenvalues with ReA=0
secur, they must form a conjugate pair tie. ‘Thus, in
the regions of parameter space where the stability of the
fixed point changes, there will be a negative real eigenval-
tue and a conjugate pair. From the characteristic polyno-
rial, a complex root, 7 +io, satisfies
37a + 2-207 H+ A+ I=O, °o
eet ee
a P2127 2017)
Now given parameters (y,j1) we ean determine p and
hence @ from O=piu—1). The (0,41) parameter space
for 020, 4=0 corresponds to 750.3 and #20, a8
shown in Fig. 19. The curve =0 determines the Hopf
bifurcation surface where a complex-conjugate pair of e
‘geavalues reaches the imaginary axis.
The center-manifold reduction for this bifurcation re-
quires that we determine the two-dimensional center sub-
space. For the eigenvalues (2;,Ay/43) we have eigenvee-
1085 (04,0250)
dp +n)
=A0a,)
[pull +pAV1+e)
where A(20=2?—A+y(1+p"). For the real eigenvalue
dy the eigenvector is real; for the conjugate pair %y=Ky
we have
123 5 9)
@.10
aa(1—27) abo
BG | |-ale*+ 40) aun] abia—7)
aal2y—D) ~aBo
with
a=ullte, B=pitp) »
detS = aB'al ACA) Aly +0" + (24-71-27)
‘Next we implement the linear change of variables
e148)
in Eq, (9.4), to express the vector field in the standard
form of Eq. (7.2):
-[2.%
Fev. Med. Phy. Vol 63, No $, October 1991
a
a
aa @.19)
y" 0)
Ryoxty'2")
+
Ray’
= 2y =I + 4yly— D421 74? By +5)]
1023
130") +(p— 2077-02)
F[1+ e142 ly tule D+p9=0. 0.7)
Although (0,0) are the physical parameters, and jt
fare more convenient, as y directly measures the distance
in parameter space from criticality for a Hopf bifurca-
tion, ie. 7=0. We may express the dependence of 2 on
(ym) by solving Eq, (9-6) for 0, eliminating w* from Eq.
(977), and solving for p*:
os)
———
where
~rp\l+p)
aan |,
allt p+ w)
—pll+n)
y=0| 1-20
°
span E° at y=0. The linear transformation
Sel, 4 9) (9.1)
puts the linear problem in block-diagonal form;
yoo
-0o y 0
0 Om
=s-us, (o.12)
where L is the matrix in Eq, (9.4) and St is given by
of A +AL=27)}
RAPA) Dyw?
ole? AQ”) —Y1-27)]
(9.13)
eS
ast
Go mAs + Ry") 0.16)
where
Ryx',y',2")
Ry(x',y',2") |=25 + 5 17)
Ry(x'y's2")
with 9,2) expressed in terms of (x',y',2") using Eq
(6.14), For convenience in our discussion ‘below, the re-
sult of fully expanding the right-hand side of Ea. 9.17)
will be denoted
Rx y ZY REE R TERY
FRyx'2 +R"? +R? 8)1024 John David Crawford: Introduction to bifurcation theory
for each component i=1,2,3. The coefficients Ry are
readily worked out, but we shall not require the detailed
expressions, which tend to be unwieldy, e.g.,
= 2080 11 24 0? AGP
Ry det (27M Aly)
+B yo Aly)
+BY AAD +AI—27]- 9.19)
2. Approximating the center manifold
Near (x',»’,2')=(0,0,0) we represent the center mani-
fold by a function h:i—K describing the 2' coordinate
of the manifold, ic. z’=A(x',y"). This function satisfies
(ef. Eq. (7.7)
ah
Lyx!boy' +R yx'y'sh)]
+ Bente by]
s(x! y ERA yh) (9.20)
with
ah
a
‘An asymptotic solution for A(x',y") near (x',y')=(0,0)
hhas the form
ety"
room, S000 200=0.
1x ay hx ty + 21)
where terms in (x',y") of third degree or higher have
been dropped. A straightforward evaluation of the quad-
ratic coefficients yields
2wlRy~ Ry )+(2y~ A Ros
(oF+Qy—mP cae
ons tRy
ee 23)
cohytRy
a 02
yy
Given h(x',y'), the two-dimensional vector field on the
‘center manifold follows directly from Eqs. (9.15), (9.18),
and (9.21),
br
E:
RuxtRisy"
Fax thay they") |p
Yo
-o7
[Rix',y'sh)
+ feiecran|
RyxP+ Ruy? AR yx'y!
e
+
“25
24x" + Rasy" |
(925)
Fev. Mod. Pry, Vol 63, No 4, Oetober 1991
where terms of fourth degree and higher have been omit
ted.
3. Determining the normal form
‘The quadratic terms in Eq, (9.25) may be removed by
rear-identity normal-form transformation to new vari-
ables xy)=(x',y+6%x',9") with inverse
Oy) =(x,9)- 8x, +O) [ef Bq. (83)). From
Eqs. (9.25) and (8.8), the equation for 6 is
Ry xP+Ryy?+R yxy
Ryx? + Ray + Rasy
where L(g) is defined in Eq. (8.16). Following the dis-
cussion in VIL-A.2, we solve Eq. (9.26) by rewriting it rel-
ative to the basis ("defined in Eq. (8.19). Thus
(Ryx?+R yy? +R yxy
Lig?) 0.26)
eee:
Raix?+ Ray? + Rasy | }
20
(0.288)
(9.280)
=e", (9.286)
and
Hem SP RNeleh-g gag ; 9.29)
hence from Eqs. (8.22) and (9.26)
Ro
sears 9.30)
x
for !=0,1,2. This change of coordinates must now be
carried out in Eq. (9.25) to obtain the transformed vector
field up to terms of fourth degree:
fe] [7 o]{x]
I} [-e r} by)
[Rex +R yy
Hh thay thy) Reet Ray
vay, 9, RRR ay? +Rasay
PIM [Raxtt Ray Rory [TOM
eanJohn David Crawford: Introduction to bifurcation theory
Here we see the additional terms of third degree generat-
ed by the nonlinear coordinate change removing the
quadratic terms. The final task is to consider the terms
of third degree in Eq, (9.31) relative to the basis {£2"")
and determine the coeficient az of £2" fet. Eq, (8.251)
‘This calculation yields
ey=RQ?—6EREM ERD)
~2929R9 + 92R2), 9.32)
where Ris the component of the “origmal” cubic
terms in Eq, (9.25) along the basis vector £4),
ROP ALIA Ra thaRas+haR
FhyR sth Rao thy Rau]
+ E1h Ra haRu)thaRacthsRas
—hyRis—hyRy) - (9.33)
We now have the normal form for this bifurcation to
leading nonlinear order (ef. Eq. (8.26)]:
PHyrtRelayr + Or),
6=0-Imia,? +O")
(9.34a)
(9.340)
‘The dependence of a, on parameters is complicated, and
the behavior of the cubic coefficient Rea along the criti-
cal curve y=0 in parameter space is best examined nu-
‘merically. ‘The graph of a, =Rea, vs p for y=0 in Fig.
20 indicates a region of supercritical bifurcation a, <0
and a region of subcritical bifurcation a, >0, with the
transition a, =0 occurring at j1,~3.29. Thus for damp-
ing rates greater than y1, the instability will saturate at
‘ry in a small stable oscillation of the wave amplitudes
[cf Eq. (5.25)]. For <1, the analysis implies that there
FIG. 20. At criticaly
4, =Re(a,) and a,=Reay) in Ea, (9.34a) are plotted against j.
ling the function f(x) -2xgn(x)log(1.0+ Ix).
Fev. Mod Phys, Vol: 63, No, Otober 9
1025
is no stable solution in the neighborhood of r=0 when
is positive; in fact, numerical studies indicate that the
wave amplitudes gow without bound.
‘These conclusions indicate that the stable Hopf period-
ic orbit must be destroyed in a separate bifurcation in the
parameter neighborhood of (=H, Y=0), since there is
no periodic orbit in the neighborhood of the fixed point
for j=j1,,0
uolsx,r)= Alxyrie tec, 5 oan
in terms of a complex amplitude function; at higher order
the GL equation (9.39) for A arises as a “solvability”
condition, which must be satisfied to avoid secular behav-
ior. The basic equilibrium A=0 for GL corresponds,
therefore, to the spatially uniform state one observes if
j2<0; in addition, for :>0 there are spatially periodic
cquilibria (‘pure modes”)
Aglxid)=Vp—QPeitel™ 0.42)
that describe patterns with wave number q=q.+ Vi.
‘As j1 varies there are bifurcations from A =O and 4g
that can be studied using center-manifold theory; howey~
ef, this analysis is more subtle for two reasons. First, the
GL equation (9.39) is highly symmetric. ‘The group of
symmetries is generated by reflections and translations in
2There has also been interesting recent work on the necessity,
of allowing for finite end effects in order to descr
features of the experiments in long cylinders (Edwards, 1990).1028 John David Crawford: Introduction to bifurcation theory
x, complex conjugation, and phase shifts; these opera-
tions we denote by x, Ty, C, and R g, respectively:
(nA MEI AC=2) 5 (9.438)
(Ty-A I= Ala +a), (9.430)
(CAN AGE), (9.430)
(Ry AYx)= 08 A(x) (9.430)
‘Thus if A (x,t) is a solution then (7-4 )(x,0) is also a
solution for y'=K, Ty, C, Ro, oF any combination of these
operations. For bifurcation problems with symmetry
there exists a generalization of theory presented in Secs.
II-VI that incorporates @ variety of group-theoretic,
techniques. We do not require this generalization for this
example, but we will indicate how the symmetry (9.43)
affects the bifurcation analysis. Golubitsky, Stewart, and
Schaeffer (1988) provide a comprehensive introduction to
equivariant bifurcation theory, and there are also the
more concise reviews by Stewart (1988), Gaeta (1990),
and Crawford and Knobloch (1991). A second novelty
arises because Eq. (9.39) describes an infinite-dimensional
dynamical system; i.e, it defines a flow on an infinite-
dimensional phase space—the space of functions (x).
Center-manifold theory can be rigorously extended to
partial differential equations, but this generalization is
rather technical for the present discussion (see the recent,
review by Vanderbauwhede and Tooss, 1991). However,
if we assume there are center manifolds associated with
the bifurcations in Eq. (9.39), then the corresponding
reduction and bifurcation analysis can be carried through
Jjust as for an ordinary differential equation.
‘The assumption of a finite-dimensional center manifold
requires a consideration of boundary conditions for Eq.
(9.39). This necessity is clear if we analyze the linear sta~
bility of A =0. Linearizing Eq, (9.39) defines the opera
tor L,
2 ympst
SAcust SG asta, (9.44)
with eigenvectors and eigenvalues given by
ho=e@, A=p-g?, (94s)
where ~200. In addition there is a “mixed”-mode
solution given by pi=p.=p and p= }(n—Q").
From Eq. (9.60) the lincar stabilities within the center
‘manifold of each of these states may be esleulated. With
respect to perturbations in the amplitudes (p,,p;), the
pure modes are stable but the mixed mode is unstable; in
each case there are also two zero eigenvalues correspond:
ing to perturbations in the phases (9.60c). The phase por-
trait Fig. 24 for (p,,p,) summarizes this analysis.
In addition, for all of these solutions (pure and mixed
modes) there can be unstable directions transvers: to the
center manifold because of the unstable directiens for
‘A=0. The number of these unstable directions is equal
to the dimension of E* at criticality (see Table 1. We
shall see that these initially unstable pure modes Ag re-
gain their stability as js increases further above = Q*.
60)
° i
FIG. 24. Phase diagram for the low on the center manifold as:
sociated with bifureation from 0. The pure modes are the
stable fixed points on the p; axis and the p: axis. The unstable
‘mixed mode lies on the diagonal
Fe. Med. Phys, Vol 69, No, Ccaber 1901
TABLE 1. Dimension of the unstable subspace for A=0 as a
funetion of
Value of
#50
o
You might also like