Introduction to bifurcation theory John David Crawford Institute for Fusion Studies, The University of Texas at Austin, Austin, Texas 78712 ‘and Department of Physics and Astronomy, University of Pittsburgh, Pittsburgh, Pennsylvania 16260" ‘The theory of bifurcation fom equlibria based on center-manifold reduction and Poincaré-Birkhof not smal forms reviewed at an introductory level. Both diferentil equations snd mape are dicated, and re- ent results explaining the symmetry ofthe normal form are derived. The emphasis ison the simplest ge tric bifurcation in one-parameter tJstems. Two application ae developed in detail: a Hopf bifurcation ‘t model of thrce-wave mode coupling and steady-state biuccations occuring in the eeal rg equation. ‘The former provides an example ofthe importance of degenerate Biforow tions in problems with more than one parameter and the latter ilusrates new eects introduced into 2 be fureation problem bys continvous symmetry. CONTENTS VIL, ComerMeniold Reduction 1009 noe ‘009 I Lael repetetaton of 7" toro coe 7 2 The Shoah theorem ton the ase tp oa 3. ame 11 1B. Map Local epeentatin of vol The bate question an ie, fay Working on interval in prometer spice: opened A. Flows 994 woz ee 224k. roca ikl Normal Forms tos 2 Hartman-Groboan theorem me ‘ ww 5. Low of hyperbola lea Warcation oe 1 oa = 2 Stadt bifraton on R to1e - 3. Hopf tfurcation oR to16 {vera Heas abapects ” 2 Her tration 2. Hyperbolicity, Hartman-Grobman, and loca formal-form symmet 1017 featon on 3. Maps 1018 1s Generales ro19 IL, Nonlinear Theory: Overview on : ee a 2 Period doutig bifurcation oH to20 JA epic esto theorem om 2. Hop titueation on x? 20 1X. Applications yon 3 Appllcadoneoeqitorie 9 a oe ‘A. Hopf bifurcation in a three-wave interaction 1022 diem = 1. Lica age, tena oo i 2 Approximating the center manifold tons pets io 5. Betermining the normal form toa 1 Seady-tate uration: mpl geval at 1, eadystat iftrcaion inthe Ginbore-Landa 0 100 ccuston 102s SSaddenode titration; the plea case 1000 a Hed Trameriea! retin: nchang of wail ageo ey ° 1001 & goo to «. Pihfor bifurcation: refleton symmetry 1012 2. A dgrenion on phase dynamiea too 2 Hop iron: a single conegne pi of 2 Bical from he pare modes ios! imaunsy eaenales vom 2 Symmetry fost 2 Mage toot ner anity for @=0 tos 1 Seay tate ication: imple geal at Center manifld reduction for 0 ros, +n 1004 XX. Omitted Topics 1033 2. Saddlenode titration 008 adnceianene = Tramcateal bireaton jos ie © Pichfor foreation 1008 eee ne 2 Pav doubling bifurcation: spe elgee ne 100s 3. Hopf bifrcaton: simple compler-onjugate |. INTRODUCTION 1008 107 ‘VL. Invariant Manifolds for Egiibia 07 Bifurcation theory is a subject with classical A. Flows 1008 mathematical origins, for example, in the work of L. BL Mape 109 Euler (1744); however, the modern development of the “Revised und expended version of lectures delivered at the In- stitute for Fusion Studies at the University of Texas in 1989, "Permanent address. subject starts with Poincaré and the qualitative theory of differential equations. In recent years, this theory has ‘undergone a tremendous development with an infusion of new ideas and methods from dynamical systems theory, singularity theory, group theory, and computer-assisted Copyight 1861 The American Pryce! Society 981992 John David Crawford: Introduction to bifurcation thoory studies of dynamics. As a result, itis dificult to draw the ‘boundaries of the theory with any confidence. ‘The char- acterization offered twenty years ago by Arnold (1972) at least reflects how broad the subject has become: The word bifurcation, meaning some sor of branching process widely wed to describe any situation in which the qualitative, opologicsl picture of the object we are ‘nudging alters witha change of the parameters on which the object depends. “The objects in question ean be ex- tremely diverse: for example, real of complex curves Or surfaces, fonctions oF maps, manifolds oF fibrations, vec- {or es, diferential or integral equations In this review the “objects in question” will be dynam cal systems in the form of differential equations and dilference equations. In the sciences such dynamical sys- tems commonly arise when one formulates equations of motion to model a physical system. ‘The setting for these equations is the phase space or state space ofthe system, A point x in phase space corresponds to a possible state for the system, and in the case of a differential equation the solution with initial condition x defines a curve in phase space passing through x. ‘The collective represen- tation of these curves for all points in phase space comprises the phase portrait. This portrait provides a global qualitative picture of the dynamics, and this pic- ture depends on any parameters that enter the equations ‘of motion or boundary conditions. If one varies these parameters the phase portrait may deform slightly without altering its qualitative (i.e, topo- logical) features, or sometimes the dynamics may be modified significantly, producing a qualitative change in the phase portrait. Bifurcation theory studies these qual- ive changes in the phase portrait, eg. the appearance or disappearance of equilibria, periodic orbits, or more complicated features such as strange attractors. The methods and results of bifurcation theory are fundamen- tal to an understanding of nonlinear dynamical systems, and the theory ean potentially be applied to any area of nonlinear physics. In Sees. IL-VIIL, we present a set of core results and methods in loa! bifurcation theory for systems that de- pend on a single parameter 1. Here local bifurcation theory refers to bifurcations from equilibria where the phenomena of interest occur in the neighborhood of a Single point. This restriction overlooks an extensive literature on global bifurcations where in some sense qualitative changes in the phase portrait occur that are not captured by looking near a single point. Wiggins (1988) provides an introduction to this aspect ofthe sub- Jjeet.! In addition, we shall concentrate on those bifurca tions encountered in typical or “generic” systems. ‘Thus Nig is worth emphasizing that the division between local and ‘lobal bifurcations introduced here should not be taken too seriously. A detailed investigation of a global bifurcation often tincovers a rich spectrum of accompanying local bifurcations similarly a local bifurcation of sufficient complexity can imply the occurrence of global bifurcations. symmetric systems and Hamiltonian systems are not con- sidered, with the exception of pitchfork bifurcation for reflection-symmetric systems. A precise mathematical description of generic can be given at the expense of i troducing a number of technical definitions (Ruelle, 1989). The heuristic idea is simply that, when a parametrized system of equations exhibits a generic bi furcation, if we perturb the system slightly then the bifur- cation will still occur in the perturbed system. One says that such a bifurcation is robust. Bifurcations that are robust in this sense for systems depending on a single pa- rameter are referred to as codimension-one bifurcations. ‘More generally, a codimension-n bifurcation can occur robustly in systems with parameters but not in systems with only x —1 parameters. The aim is to provide an accessible introduction for physicists who are not expert in dynamical systems theory, and an effort has been made to minimize the mathematical prerequisites. Consequently I begin with a summary of linear theory in Sec. IT that includes the Hartman-Grobman theorem to underscore the link be- tween linear instability and nonlinear bifurcation; this summary is supplemented in Sec, IV by an analysis of the persistence of equilibria using the implicit function theorem. The center-manifold-normal-form approach is, ‘outlined in Sec. III and developed in Secs. V-VIIL ‘Two applications of the theory are considered in Sec. IX. These illustrate the calculations required to reduce a specific bifurcation to normal form. In addition the ex- amples offer a glimpse of several important and more ad- vanced topics: new bifurcations that arise when there more than one parameter, center-manifold reduction for infinite-dimensional systems, eg., partial differential ‘equations, and the effect of symmetry on a bifurcation. Finally in See. X a brief survey of some topics omitted from this review is included for completeness and to pro- Vide some contact with current research areas in bifurca- tion theory. Our subject is very broad, and there is much activity by mathematicians, scientists, and engineers; the literature is enormous and widely scattered. This intro- duction does not attempt to assemble a comprehensive bibliography; the material of Secs, II-VIII can be found, in many places, and in most cases the cited references are chosen simply because I have found them helpful. More extensive bibliographies can be found in the references. A. The basic setup It is advantageous to express different systems in a standard form so that the theory can be developed in a uniform way. As an example consider the second-order *The geometric connotations of codimension can be made pre tise, but we do not require this development here (Arnold, 1988). Roughly speaking, the set of systems associated with a ‘codimension-n bifurcation corresponds to a surface of eodimen-John David Crawford: Introduction to bifurcation theory 990 oscillator equation Ftp tytyr—0; (ita) by defining x5 =y and x,=), we can rewrite this evolu- tion equation as a first-order system in two dimensions, a|* % Ale waonnxt (tb) Clearly if higher-order derivatives in t had appeared in Eq, (1-la), we could still have obtained a first-order sys- tem by simply enlarging the dimension, eg., defining x;=5; similarly, if the equations of motion had involved ‘dependent variables in addition to y(t), these could also hhave been incorporated by enlarging the dimension ap- propriately. As this example suggests, there is great gen- in considering dynamical processes defined by first-order systems: X=Vinx), ER", ER, (12a) depending on a parameter js and describing motion in an n-dimensional phase space R". When formulated in this way a differential equation is identified with a vector field Vue.x) on R*; conversely, given a vector field one can al- ways define an associated differential equation.? ‘We shall also consider a second type of dynamics that represents the evolution of a system at discrete time in- tervals. In this case, the motion is described by a map, sya fiinx))) XER", BER, (1.26) where j=0,1,2,... is the index labeling successive points on the trajectory. There are close connections be- tween the dynamical systems defined by maps and vector fields. For example, in Eq, (1.2a), we may also think of solutions as trajectories: an condition x(0) uniquely determines a solution x (1), and the correspond ing curve in R" (parametrized by 1) is the trajectory of x(0). More abstractly, the association x (O)—>x(1) defines a mapping bcR"R” where 4,(x(0))=x(¢). This mapping is called the flow determined by Eq. (1.2a) In each case, the dynamics is allowed to depend on an adjustable parameter 1, and the origin (u,x)=(0,0) is as- ‘sumed to be an equilibrium or fixed point for the motion, ¥(0,0)=0, (13a) (120 one often wishes to consider phase spaces more general than I, for example, fnitedimensional manifolds such as tori or spheres. However, in these eases the dynamics on a neighbor- hhood ofa fixed point can be described by the models we consid er by introducing # local coordinate system. Fev. Med, Pie, Vo 69, No 4, October 1981 £(0,0): (1.36) Note that given a fixed-point solution (¥g)9) one can al- ways move it to the origin by a change of coordinates, so the representation in Eq, (1.3) is quite general. ‘The theory we develop for maps (1.26) is useful in a variety of circumstances. Two particularly important aj plications are to bifurcations from periodic orbits of differential equations and in the related context of bifur- cations in systems that are periodically forced. Let x,t) denote a periodie solution to Eq. (1.2a) with period 7, ie., x,(0=x,(t+7); the dynamics near x,(t) can be ana lyzed by constructing the Poincaré return map. Let 2 denote an n —1 dimensional plane in R" which intersects x(t) at the point p (see Fig. 1). To define the return map F consider a point EE near p, and solve Eq. (1.2a) us- ing o as an initial condition. For o sufficiently near p, the trajectory from o will intersect 2 at some new point 0’; this intersection defines the action of the map f on a, o'=slo). as) This definition is sensible for all points on Z in an ap- propriate neighborhood of p. Notice that p is a fixed point for /,f(p)=p, since x, is a periodic orbit. In the second application, a periodic modulation is ap- plied to the system in Eq. (I.2a) so that V(ji,x) is re placed by Vinx), xER", wER (15a) and Vix t= Vunxst tr), (1.50) where 7 is the period of the modulation. In this cir- ‘cumstance it is convenient to introduce the “stroboscop- ic” map f by, in effect, recording the state of the system only once during each period of the modulation. More precisely, fix a definite time fo and then choose any initial condition xER". Let x(t;fp) denote the solution with the initial condition x (fo:fo)= Xo, and define f by 1,2. 6) la) (ig Jrito). The qualitative properties of the map f(x) in Eq, (1.6) are independent of the specific choice 1) used in the definition. Furthermore, fixed FIG. 1. Poincaré return map for a periodic orbit208 John David Crawford: Introduction to bifurcation theory points (1.3) for the unmodulated system typically persist as fixed points for the map (1.6), at least for weak modu- lation B. The basic question According to Eq, (1.2), at =0 there is an equilibrium state at x=0. The basic question in local bifurcation, theory is What can happen in phase space near x-=0 when there are variations inj about w= 0? ‘The Hartman-Grobman theorem, described in the next scction, effectively reduces this question to an analysis of ‘As jis varied near 4=0, what happens near x=0 if the stability ofthe equilibrium changes? Before addressing this question, which involves the non- linear terms of Eq. (1.2) in an essential way, it is neces- sary to develop the theory of linear stability. . LINEAR THEORY A. Flows At 0 the Taylor expansion of Eq, (1.2a) begins, (HOF D, Vy,0)-x +O1x?) , * an where D,V(y4,0) represents the square matrix with ele- ments ay, (D,V(p,00),= 5" (Hs0) 22 ax, and O(x*) indicates higher-order terms that are at least quadratic in the components of x. When the context is clear we shall omit the subscript x and write DV(y,0) o simply DV. At w=0 the constant term in Eq. (2.1) van- shes, and near x =0 we study the linearized system, ¥(0,0)-x , 23) ignoring momentarily the effects of the nonlinear terms. In the typical situation the eigenvalues of DV(0,0) are nondegenerate’ and this matrix can be diagonalized by a linear change of coordinates x +x’. This allows Eq. (2.3) to be reexpressed as ‘More precisely, this is true for hyperbolic fxed points, as ‘defined in Sec. TLA.2, and follows from the averaging theorem (Guckenheimer and Hokmes, 1986). 5A degenerate eigenvalue is one for which there sre two or ‘more linearly independent eigenvectors or generalized eigenvec- [x] [ao 0} fx a Fi|_|° % xy < a4) ° if the spectrum of DY(0,0) includes complex-conjugate pairs of eigenvalues, then the corresponding new coordi- hate components x/ will also be complex (Hirsch and Smale, 1974). The general solution x‘) is obviously 1x{(0)e™" (Oe : es) (Oe x(n IF Red, <0, then as t+, the x/ component decays to zero; conversely, Red,>0 implies exponentially rapid growth of x/ 1. Invariant linear subspaces For each eigenvalue 2 of DV(0,0), there is an associat- ed subspace of It'—the eigenspace E.. For simplicity we assume DV(0,0) is diagonalizable; then our definition of E, depends only on whether 2 is real or complex, The FIG, 2. Example of invariant subspaces and manifolds for a fixed point. (a) Linear spectrum showing stable modes, neutral, modes, and unstable modes for an equilibrium x—0 in a fow; (6) invariant linear subspaces; for the spectrum in (a) we would have dim £'=3, E°=4, dim E*=3; (c) invariant nonlinearJohn David Crawiord: Introduction to bifurcation theory 995 case of a real eigenvaluc is most familiar. When 2 is real, real matrix if v,+iv, is the eigenvector for 2, the E, is simply the subspace spanned by the eigenvectors, complex-conjugated veetor v;—iv3 is an eigenvector for ae ae Z. The eigenspace E;, in this case is spanned by the real MER, E,S(VER"DV(OO-ATv=0} 265) sag imaginary parts of the eigenvectors for A, e-8., If A is nondegenerate, then we have dim E,=1. by and v,. Noting that both o; and vz satisfy When A is complex, then the eigenvectors are also (DV(0,0)—AINDV(0,0)—-A)-v=0, we replace Eq complex; furthermore, since DV(0,0) is assumed to be a (2.6a) with = {wER"(DV(0,0)—AIXDV(0,0)—Z1)-v=0} « (2.60) AER, By Now if is nondegenerate we have dim B,=2. E ‘When DV(0,0) has eigenvalues that are degenerate, this construction for E, is satisfactory provided DV(0,0) These subspaces span the phase space, RY is diagonalizable, When DY(0,0) cannot be diagonal- ang they are invariant: if x(0)EE*, a=s,c,u, then the ized, then the definitions in Eq, (2.6) must be extended to trajectory x(t) of Eq. (2.3) with this initial condition include not only eigenvectors but generalized eigenvec- Satlgies x(2)©E". For E* and E" the dynamics has a tors as well (Arnold, 1973; Hirsch and Smale, 1974). aaeps serene acai picn anya i eneioatse ‘An eigenvalue 2 corresponds to a “mode” of the sys a-te the trajectory converges to the equilibrium; if tem that is stable, unstable, or neutral, depending On (1), then the trajectory converges to the equilibri- Whether Ret <0,Reh>0, or Reh=0, respectively (Fig. um as -»—20. ‘These features are illustrated in Fig. 2(a)). We divide the eigenvectors (and generalized eigen 3(), vectors) of DV(0,0) into three sets according to these ~ An equilibrium at x0 is asymptotically stable if there Parcs nd Toon the abe nice BE anaes gy ulin ot 0 comely sol i te =span| v EE, and Rek=0} (270) "GEE", subspace E", and center subspace E*: Such that forall (0) in this neighborhood ‘span{v|v@£, and Red < (2.7a) aa vEB, and Reh <0} . the trajectory x (1) satisfies |x(1)|0, and “=span(v|v€E, and Red>0} (2.70) ) [x(1)|—0.as to. jim ° I ® ® 2 Re ~ — ® ® (a ( FIG. 4. Asymptotic stability of x=0, Such stability for the FIG. 3. A stable linear spectrum for a fixed pont of (a) alow linear system (a) implies that x =0 is asymptotically stable for and (b) a map. the nonlinear system (0) For Moa Phys, Vol 63, No.4 Octobe 1891995 John David Crawford: Introduction to bifurcation theory For the linear system (2.3), the equilibrium x=0 is asymptotically stable if and only if Re() <0 for each ei- genvalve 2 of DV(0,0). In other words, the spectrum ‘must lie within the left half-plane of the complex 2 plane [see Fig. 31a] ‘This criterion is particularly valuable because one can prove that if x =0 is asymptotically stable for Eq, (2.3), ‘then it will aso be asymptotically stable for the original nonlinear system (1.2a) (Hirsch and Smale, 1974). In Fig 4b) we show a schematic phase portrait for a two- dimensional system with two fixed points on the x, axis If we imagine linearizing about the stable equilibrium a the origin, then the resulting 2X2 matrix will have @ complex-conjugate pair of eigenvalues (A,X) satisfying, Red=Reh <0. The phase portrait for the linearized sys tem is shown in Fig. 4a; the equilibrium x =0 is obvi- ously asymptotically stable in Fig. 4(a) for arbitrarily large initial conditions. In the nonlinear phase portrait Fig. (0) x=0 is also asymptotically stable, but the neighborhood, 0 <|x(0)| 1, then the x/ component will grow. 1. Invariant inear subspaces For the linearized map (2.10) the eigenspace E, for Df (0,0) are defined as in Eq, (2.6) for the previous case by replacing D¥(0,0) with D/(0,0). The invariant linear subspaces E°,a=s,u,c, are defined as in Eq, (2.7), replac- ing Red by |2|~1 to reflect the appropriate stability cri- teria, E*=spanfolv€E, and |A| <1} , 2.130) E*=span(vlv@£, and |Al>1} 5 (2.136) E*=span{v|v E, and |A|=1} - (2.136) As before, we have R"=E'@E@E", and the stable and ‘unstable subspaces have simple asymptotic dynamics as jot and j—+— 2x, respectively. ‘The definition of asymptotic stability given earlier ay plies to fixed points of maps provided x(1) is replaced by x). For the linear dynamics (2.10), the equilibrium x 0 ‘will be asymptotically stable if and only if the spectrum of Df(0,0) lies within the unit circle in the complex A. plane, i.e, |A;| <1 for each eigenvalue [see Fig. 306)]. It ‘can be shown that if x =0 is asymptotically stable for Eq. (2.10) then the same conclusion holds for the full non- ‘near dynamics (1.26). In addition, for return maps (cf. Fig. 1}, whose fixed points correspond to periodic orbits, the stability of a fixed point reflects the stability of the corresponding periodic orbit. [When the differential ‘equation is linearized about the periodic orbit, the result- ing linear equation may be analyzed using Floquet ‘theory; the stability of the periodic orbit is determined from the spectrum of Floquet multipliers Gordan and ‘Smith, 1987). The eigenvalues of the return map linear- ized at the fixed point correspond to the Floquet multi- pliers of the periodic orbit.] 2. Hyperboliity, Hartman-Grobman, ‘and local bifurcation As for flows, a fixed point is said to be hyperbolic if center subspace (2.13c) is empty, and there is @998 John David Crawford: Introduction to bifurcation theory Hartman-Grobman theorem relating the linearized dy- ‘namics to the local nonlinear dynamics: if, at 1=0,x=0 is @ hyperbolic fixed point, then there exists a homeomor- phism W and a local neighborhood U of x =0 where F(0,x=¥MDF(O,0)-W()) for x such that x Vand f(0,x)EU. If x=0is a hyperbolic fixed point for f(1,x) at =0, then as jis varied about zero this equilibrium will shift its location, but it will persist (see Sec. IV). The eigenval- ues of Df will be functions of jz, and a variation in jz will ‘cause them to move in the complex plane. If an eigenval- ue reaches the unit circle, then the fixed point is no onger hyperbolic and a bifurcation can occur. ‘The possibilities may be classified by the form of the linear spectrum when the condition |2,|#1 fails: a4) (1) A simple real eigenvalue at A=1; see Fig. 6(a). This type of instability is quite similar to the A=O case for flows and is referred to as a steady-state bifurcation for maps. As in the case of flows, we find the saddle-node, transcritical, and pitchfork bifurcations as examples of steady-state bifurcation. (2) A simple conjugate pair of eigenvalues (2,2) where A=e'; sce Fig. 6(b). We shall refer to this case as Hopf bifurcation for maps to emphasize similarities with Hopf bifurcation in flows. FIG. 6. Basic instabilities for an equilibrium in a map: (a) steady-state bifurcation; (b) Hopf bifurcation; (©) period: doubling bifurcation Fev. Med. Phi. Yo. 62, No 4, Otebar 1901 (3) A simple real eigenvalue at A=—1; see Fig. 640). This ease is novel, as it does not have an analog in the earlier discussion of flows. This instability is generally termed period-doubling bifurcation, although the names flip bifurcation and subharmonic bifurcation are also used. This completes our summary of lincar stability theory and the forms of instability one typically expects to en- counter when a single parameter is varied. Characteriz- ing an instability by the form of the linear spectrum at criticality is more than a convenience; itis very advanta- geous to organize the theory (and one’s understanding) in this way. The most important reason for this is that the near spectrum determines the normal form. Precisely ‘what this means will be explained in Sec. VII. Ill, NONLINEAR THEORY: OVERVIEW ‘Suppose an asymptotically stable equilibrium is per- turbed by varying an external parameter jz, and at a criti- cal value 4= 4, the equilibrium develops a neutral mode (Reh=0 for flows; |2)=1 for maps). At 1, hyperbolicity is lost, and we must study what happens to the system as, 111s varied about For all of the basic instabilities described in Sec. Il, this issue can be investigated using the techniques of center-manifold reduction and normal-form theory. In outline, this approach has several steps: (1) Reduction: identify the neutral mode (or modes) at =H, and restrict the dynamical system to the appropri- ate center manifold; (2) Normalization: if possible, put this reduced dynamical system into a simpler form by applying near- identity coordinate changes. This yields the normal form fos the bifurcation; (3) Unfolding: describe the effects of varying 4 away from j4, by introducing small linear, and possibly non- linear, terms into the normal form; (4) Study the bifurcations described by the unfolded normal form. In this analysis, one truncates the unfolded system at some order and considers the resulting system. ‘Once the truncated system is understood, the effect of re- storing the higher-order terms can be discussed? ‘The virtue of step one is that it reduces the dimension of the problem without any loss of essential information, concerning the bifurcation. The advantages of the simplification offered in the second step are often decisive in being able to solve the problem. Furthermore, the re- tn sufficiently complicated bifureations, these effects can be significant and highly nontrivial, However, for most of the bi- furcations considered in this review, these higher-order terms do not produce any qualitative changes. The one exception is Hopf bifurcation in maps, discussed in See. V.B3.John David Crawford: Introduction to bifurcation theory 989 sulting simplified representation of the dynamics pro- vides a universal, low-dimensional model for the given bi- furcation. This approach allows the general qualitative features of a bifurcation to be distinguished from specific quanti- tative aspects that will inevitably vary between different realizations of the bifurcation. The dimension of the re- duced system and the structure of the appropriate nor- ‘mal form may be determined without requiring explicit ‘evaluation of the coefficients in the normal form. Thus the variety of phenomena associated with a bifurcation ccan be described in a theory that is model independent. When this general theory is applied to a particular insta bility the normal-form coefficients can be calculated from the specific physical model under consideration. ‘The possibility of determining the normal form without need- ing to derive the coefficients is often a considerable ad- vantage. In Sec. V, we present the normal forms for the bifurca- tions enumerated in Sec. II. Then the basic theory un- derlying the center-manifold reduction is discussed in Secs. VI and VII. Finally, in Sec. VIIL, we develop the theory of Poincaré-Birkhoff normal forms and indicate how to derive the normal forms previously introduced in Sec. V. In the next section, we consider a preliminary issue that itis useful to discuss before taking up the program outlined above. The question is basic: can the given equi librium solution simply disappear when 1 is varied? For both flows and maps, there are simple conditions on the linear spectrum that are sufficient to guarantee the per- sistence of an equilibrium. IV. PERSISTENCE OF EQUILIBRIA ‘A. Implicit function theorem ‘The implicit function theorem provides necessary con- ditions for an equilibrium of a flow or a map to disappear fas varies. Equivalently these conditions can be restated as sufficient conditions for the equilibrium to persist. ‘The following version of the theorem is adequate for our discussion; a proof may be found in Spivak (1965). ‘Theorem IV.1, Let G(y,x) be aC? function on RXR’, GRXR—R", an such that 6(0,0)=0 (42a) and det{D,6(0,0)} 40 Then there exists a unique differentiable function X(j) defined on a neighborhood MCR of w=, XM—=R", 43) (426) Fv. Med. Phi, Vol 63, No 4, Cetaber 1981 such that X(0)=0 and GXUN=0, weM (44) In words the theorem says the following. Given G(y,x) we assume that the zero set, i.e. the set of (u,x) such that G(y1,x)=0, contains at least one point (0,0) see Fig. 7la). If, in addition, the matrix aa, 10,0), 4 3x, (29) 4s) (D600, hhas a nonzero determinant, then we can solve the equa tion G (1t,x)=0 uniquely for x as a function of #1, at least for values of w suficiently near 4=0. This means that, near (1,x)=(0,0), the zero set of G(u1,x) consists of a single arc or branch as shown in Fig. 7(b). B. Applications to equilibria 1. Flows For Eqs. (I:2a) and (1.3a), we choose G(y,x) Then det[ D, G(0,0)] =det{ D¥(0,0)] this implies that condition (4.2b) will be satisfied if and only if A=0 is not an eigenvalue for DV(O0,0). It then follows that small changes in will not destroy the equi- Hibrium solution as long as zero is not an eigenvalue of the linear stability matrix for the equilibrium. ‘The solu- tion must persist and lie on a local branch of such solu- tions, X(j1), as required by the implicit function theorem. ‘Two further conclusions may be drawn. First, Hopf bifurcation cannot alter the number of equilibrium solu- tions, since the only eigenvalues of DV on the imaginary (us). 46) 5 wo) — e oo Ve oo ° FIG. 7. A unique solution branch from the implicit function theorem. Given G(0,0)=0 and der[D,G(0,0)]#0 at a point (a) the local structure ofthe solution set for G(y,x)=0s asi ale branch (b).1000 John David Crawford: Introduction to bifurcation theory axis form a conjugate pair (Fig. S(b)]. Second, the condi- tion det[ DV }#0 fails at a steady-state bifurcation, since by definition there is always an eigenvalue at zero. Thus in general we cannot expect a unique branch of equilibria through (,x)=(0,0) if this solution corresponds to a fixed point at criticality for steady-state bifurcation. 2 Maps For Eqs. (1.2b) and (1.3b), we take Giy,x)=S(u,x) =x, so that G(0,0)=0 and DyG0,0)= Df(0,0)—1 an where I'is the identity matrix on R*. With this choice, if Gly,x)=0 then x is a fixed point for the map at parame- ter value yt. For the solution (1,x)=(0,0), condition (4.2b) will be met if and only if the linear stability matrix ‘Df (0,0) does not have an eigenvalue at =++1. Provide ed 2=1 is not an eigenvalue, the implicit function, theorem implies (0,0) lies on an isolated branch of equi- librium solutions, For the three basic instabilities illustrated in Fig. 6, only steady-state bifurcation involves an eigenvalue at +1. Neither period-doubling nor Hopf bifurcation ean alter the number of equilibrium solutions. In the context of Poincaré return maps for periodic orbits, these results, on persistence of equilibria show that the periodic orbit can always be followed through a period-doubling or Hopf bifurcation. The question of following periodic or- bits through parameter space in a global sense has also been studied (Mallet-Paret and Yorke, 1982; Yorke and Alligood, 1983). \V. NORMAL-FORM DYNAMICS In this section we analyze very simple equations that describe the local dynamics associated with the linear in- stabilities of Sec. Il. Remarkably, these simple examples are in fact quite general; to appreciate this generality re- ‘quires the material on center manifolds and normal-form theory developed in later sections. Let us first analyze the dynamics of these simple models and then establish their generality. We shall consider the various bifurca- tions in the same order they were listed in Sec. II. In the following itis convenient to assume that criticality for an instability occurs at =0. A. Flows, 1. Steady-state bifurcation simple elgenvalue at zero For a simple zero eigenvalue" as illustrated in Fig. 5(a) the center-manifold reduction yields a system of the form An eigenvalue is simple iC tis nondegenerate; fora real ei- sgenvalue the associsted eigenspace (2.6 is then one dimension- al Fey. Mod. Pins. Vol 63, No 4, Octaber 1881 Vinx), xER, WER, (S.1a) which will satisfy the following two conditions at cntical- iy ¥(0,0)=0 wy, a CCenter-manifold theory tells us that Eq, (S.1a) should be one dimensional. Furthermore, the reduction to one di- mension will preserve Eq. (1a) and the oceurrence of Zoro eigenvalue; hence Eas. (3.1b) and (51), respectively (5.10) (0,0) (5.10 Expanding (5.1) at S (0,0), we find Frioom+ S Zoo ao O)pex +2 ero 08 Foot, 62) instability, the vector field at criticality, av 0, Gx O05 , BV 9% 89 570,07 cannot be significantly simplified by making coordinate ‘changes (ef. Sec. VIID; we shall obtain normal forms by ‘making truncations and rescalings. There are three situ- ations that arise most often in applications, 4. Saddle-node biturcation: the typical case Equations (S.1a)~(S.tc) define a steady-state bifurca- tion; without further assumptions we typically (“generi- cally”) expect, ay, 0,040 5 (000m a = ooo (530 to hold. In this ease Eq. (5.2) may be rewritten as ay, En oyit+ oun ie 2X08 +01, 6a whete O(j4.x) indicates terms at least first order in por x. For example, 2Y seo)/2, 2 00/ ia0}e, [Zh a01/ 00) au is one such term in the first bracket in Eq. (5.4). Near (Ya.x}~(0,0) we can neglect these (yx) terms relative to unity and then define rescaled variables (24%), 2 55a) 7 (9,0) 55 (0,0) au‘John David Crawford: Introduction to bifurcation theory 1001 2s, (6.56) Py, 2% (0,0) ‘ax? to obtain the normal form! 5.6) Obviously at A=0, ¥=0 is an equilibrium in Eq. (5.6), ‘and this equilibrium has a zero eigenvalue. What hap- pens near (7,)=(0,0) depends on (,,€3); there are four possibilities. Consider ¢, 1 (the other three cases ‘can be analyzed similarly). Then the equilibria in Eq. (5.6) satisfy +x ?=0. This describes a parabola in the (&,f) plane as shown in Fig. a). At a fixed value of <0, there are two equilibria x5 (f)=+V —p, which coalesce as fl increases to criticality. The upper branch ,, (fi) is unstable, and the lower branch x (j) is asymp- totically stable. This is indicated by the arrows in Fig. Bia) and can be checked by linearizing Eq. (5.6) about X4 (A). Let x=X.(p)+). Then (for €,= +1 5, < 22, 7 ‘ Fa Be oR Ws = DD § 6m the eigenvalue (2%, (/1)] is positive (unstable) for #, and. negative (stable) for x. in two dimensions, a fixed point with one stable and cone unstable eigenvector is referred to as a saddle; if the fixed point has two real negative eigenvalues it is a stable ‘node (Arnold, 1973). When a parameter is varied, so that such fixed points are brought together, then the resulting ‘merger can be described by the one-dimensional model (6.6); the bifurcation is named for this prototypical exam- ple. Note that for ji0 there are none. This is consistent with the fact that Eq. (4.20) fails at (14,x)=(0,0) and the implicit function ‘Ute the terminology of Sec. IV, the normal form is actually Seer, and ef is an unfolding term. T often overlook this distinction in the following and simply refer to the unfolded ‘normal form as the normal form. "More generally, the one-dimensional model (5.6) deseribes ‘much wider elas of bifurcations, in which two fixed points are either created or destroyed. In higher dimensions itis not a: ‘ways the ease that one equilibrium is stable and the other unsta ble; both may be unstable. Neither is it necessarily true that che cigenvalues not involved inthe bifureation must be real. Rov. Mod. Pry. Vo. 69, No 4, Ctober 1901 FIG. 8 Diagrams for saddle-node bifurcation with normal form (5.6: (a) =e =1, ©) =6=—I, ©) —e=e= I, e/=—e = 1. Solid branches are stable; dashed branches are un- stable. theorem cannot guarantee a unique branch of equilibria passing through (0,0). ‘The results for the remaining three cases, € 6=—€)=—1, and ¢,=€,=~ 1, are also shown in Fig. 8. ‘These diagrams in the (u,x) plane are simple examples of bifurcation diagrams. b. Transeritical bitureation: exchange of stability In applications, it may happen that an asymptotically stable equilibrium loses stability through a steady-state bifurcation, but the equilibrium solution itself survives. In this case saddle-node bifurcation, which characteristi- cally destroys (or creates) equilibria, does not occur. ‘When the equilibrium survives, we may denote it by X11) such that X(0)=0 and Vu, X() replaces Eq, (5.1b). Let us make the .-dependent change of variables x=X(y)+x’ and then drop the primes [This amounts to setting X(4)=0.] Then Eq, (5.8) be- 3.8) » HER Vio) 5 6.9) for am appropriately redeined Vix). Shoe Eq. (9) oe 2% 00 a ‘if we now make a Taylor expansion around (1, oeveateeea 0, 5.10) Lye 0,03,1002 John David Crawford: Introduction to bifurcation theory ey ay, vy, x FX comes Hoots %ons-+ oan Without farther assumptions we bal ypcally nd ey Guar (O#0 . (5.12a) FP aoveo, ow) ae where Eq. (5.128) replaces (6.38). Now, proceeding ex- actly as in the discussion of saddle-node bifurcation, we truncate and rescale variables to obtain a normal form, legitex), (5.13) av o[ 2% 0] and oo | *c00)| Note that ¥=0 is an equilibrium for all 1, but at 7=0 the eigenvalue 6,7 is zero. When sgn(e = —1(-+1) the ‘equilibrium ¥=0 is stable (unstable). The second factor on the right-hand side of Eq. (5.13) yields second branch of equilibria, ¥,(72), [: The soilty of is found by lincarieng Bg Fay ty toad (B= (5.14) 5.13) FIG. 9. Diagrams for transertical biforcation with normal form (5.13): (@) ees" 1, 6) €@=—1, fe) Ee 1, 6 —e "1. Solid branches are stable; dashed branches ate un stable Fev. Med, Pre, Vol 69, No, October 1901 se(-emy Thus ¥=0 and ¥=¥,(71) have opposite stabilities; at =0 these equilibria collide and their stabilities are “‘ex- changed.” The precise form of the resulting bifureation diagram depends on ¢, and €,; the four possibilities are shown in Fig. 9 1s) ©. Pitchfork bifurcation: reflection symmetry ‘This version of steady-state bifurcation arises formally when Eq. (5.9) holds as in transcritical bifurcation but (5.12b) fails and is replaced by the assumption ay, 2X 0,0140 (5.16) at ‘A natural context for these assumptions is V(y,x) having a reflection symmetry, i., =Py,x)= Vw — x) « an Obviously, this symmetry implies Eq. (5.9), and forces Eq, (5.126) to fail. Replacing (5.12b) by (5.16), we may rewrite (5.11) as iow Bx au aV oo)%2 ; + FF OOF 1 +OH20) - (0,0yx(14+O(4,%)) (5.18) Now truncating higher-order terms and rescaling vari- ables appropriately leads to the normal form Xaslentex?}, (5.19) where a ae ; an « FIG. 10. Diagrams for pitchfork bifure G19: @ = —G=1, 0) ~G=G=H, (© E=E=—l, @ 6 solid branches are stable; dashed branches are un- stable.John David Crawford: Introduction to bifurcation theory i 70 FIG. 11. Percurbing nongeneric diagrams: (a) transcritical bi- fureation;(b) pitchfork bifurcation. ‘The analysis of Eq. (5:19) differs from transeritical in that the second factor in (5.19) contributes two branches of equilibria, ew (5.20) which only exist for sgn(e,f1/e,)=—1. The stability of the solutions may be worked out as before, and the four possibilities are illustrated in Fig. 10. The bifurcation di- ‘agrams resemble pitchforks in the (J, plane, hence the We conclude this discussion of steady-state bifurcation by indicating how perturbations of transcritical or piteh- fork bifurcation can restore the expected “generic” be- havior, ie, saddle-node bifurcation." Suppose V(y,x) describes a transcritical or pitchfork bifurcation at (u.x)=(0,0). We can perturb Viy,x) by including a ‘small term ¥;(1,%) in the dynamics, =(e/esih FeV wxdteV (ux), 6.20) where 00. 52%) ‘The conditions (5.23) simply mean that the conjugate pair crosses the imaginary axis at “=O in a nondegen- crate way. ‘A characteristic feature of Eq. (5.22) is the absence of ‘on the right-hand side. This means that the dynamics of the normal form is invariant with respect to the group of rotations of the phase 9. In the literature, this invariance is called the S' phase-shift symmetry," and it allows the dynamics of Eq. (5.22a) to be analyzed independently from (5.220). For (5.224), we assume that at criticality ( bie coeficient does not vanish, 2,040; "The phase shifts in @ are described mathematically by the ro tation group SO(2) or, equivalently, as the action of the circle ‘group 5". It is conventional to use the latter terminology for the Hopf normal-form symmotry.1008 then the solutions to dr /dt =0 near r =0 are determined, by the sign of ay(0) (see Fig. 12). Consider a,(0) <0 for example from Eq. (5.22a) the radial equilibria satisty Heyl)+aywr)~0 (25) ‘and there__are _ two branches: r=0 and rlu)=\V/—y7a,, The latter solution exists only for ‘lyt)> Osince ry must be real. When Eq, (5.220) is taken into account we sce that this new solution in fact de- scribes a periodic orbit of amplitude ry and frequency oy ~a4qe) +E 7. sb/g8\r}- The plot of Fv rin Fig. 14a) rakes it clear that the periodie orbit is asymptotically Stable; this can be checked analytically by linearizing (5228) about r=ry, and determining the neat eigenval. uc. The bifurcation diagram is also drawn in Fig. 12a; since the new branch of solutions is found in the diree- tion of increasing, above the threshold for instability of the equilibrium, the bifurcation of ry is said to be super- critica. “The analysis for 2,(0)>0 is similar but the results are slightly diferent. Now the ry solution is found only for Y(1)<0 of j1<0. In this cate the branch of periodic solutions is subcritical and unstable'*; see Fig. 12(b). Hopf bifurcation is a richer phenomenon than steady- state bifureation in the sense that it leads to time- dependent nonlinear behavior. In an experiment, a su- perertical Hopf bifurcation manifests itself in the spon- {ancous onset of oscillatory behavior. Often this oscila- tion corresponds to the appearance of a wave in the sys- tem. B. Maps 1. Steady-state bifurcation: imple eigenvalue at +1 ‘The normal form is one dimensional, xy fdus)), HER, ER, (5.260) soy where 6.260) +1 (5.260) Let ¥(u,x)=f(ux)—x. Then to find fixed points for f we need to solve Vi.x)=0 Note that (5.278) ‘There is no consensus in the literature as to how the terms supercritical and subcritical should be defined in general, al- though all conventions agree with my wsagein this context. For 1 didactic discussion advocating one senrble set of definitions see Tuckerman and Barkley (1990) Fev. Mod. Phy, Vol. 62, No.4, tober 1991 John David Crawford: Introduction to bifurcation theory ¥(0,0) (5.276) and ay, F100 (5.270) follow from (5.26b) and (5.26c), respectively. This prob- Jem corresponds to finding the branches of equilibria in a steady-state bifurcation for flows, ie, Eqs. (5.27) are equivalent to Eqs. (5.1). Consequently, insofar as the branches of equilibria are concerned, we have precisely the cases already studied. FIG. 12. Radial dynamios and diagrams for Hopf bifurcation with normal form (5.22): (a) supereritical bifurcation 2,(0) <0; {) suboriticalbifvreation a\(0)>0.John David Crawiord: Introduction to bifurcation theory a. Saddle-node biturcation As before, this occurs if 0.0140 and e010. En 3. From Eq, (5.28) and our previous discussion of saddle- node bifurcation for lows, we are led to the normal form ts, tex = 7ing,) (5.29 for this bifurcation in the rescaled variables (5.5) with |. Since the analysis of branches of fixed points for Eq. (5.29) is equivalent to nding equilibria for Ba, (5.6, we need only check the stability of ¥.(f a. The (5.28) Rye linear eigenvalue at X.. is simply a, =F 26K ol (5.30) oe (a) from Eq. (5.29), hence ¥2(H) is stable (unstable) if €:8.4(71) is negative (positive). Thus the stability assign- ments for the branches of equilibria turn out to be the same as in the bifurcation diagrams for flows (see Fig. 8). ‘The interpretation of these diagrams depends on how we interpret the map. If we imagine that the saddle-node bifurcation occurs in a Poincaré return map for a period- ic orbit in a flow, then the branches of solutions di- ‘agrammed in Fig. 8 correspond to mergers of periodic or- bits. b. Transcrtical bitureation ‘This bifurcation occurs if Eq. (5.28) is replaced by ay, SX o,0)=0 (31a) au and av ay Lvov, Yio,or#0 (3.316) Bea OOF ar : From our previous discussion of the normal form (5.13) for flows, we obtain Surelitentes) = 7s) as the normal form in this case. The bifurcation dia- ‘grams for the branches of fixed points are shown in Fig. 9, and the stability assignments in Fig. 9 are also correct, since the linear eigenvalues for ¥=0 and X=, in Eq (5.32) are (1+eft) and (1—e,ft), respectively. At =0 the two branches of fixed points merge and exchange sta bility. ©. Pltchtork bifurcation (5.32) This case occurs if Eq. (5.31a) holds while (5.31) is re- placed by Fev. Mod. Pry, Vol. 68, No.4, tober 1901 105 av, FV (a0 (5330 oe and av av Zap (00, 2-710,0)40 (5.33) pdx ax? From Eq, (5.19) we obtain the normal form, Syi=e( tentes) 636 ‘The analysis of the branches of fixed points and their sta- bilities yield the same bifurcation diagrams as in the pitchfork bifurcation for flows (Fig. 10) 2. Period-doubling bifurcation: 4 simple eigenvalue at —1 In Sec. IV we proved that this instability docs not change the number of fixed-point solutions, thus any branches of solutions bifurcating from the equilibrium ‘will necessarily have different dynamical properties. The ‘normal form is one dimensional and has a reflection sym- metry, xyes) HER, KER, (5.380) fy,0)=0, (5.386) Leo, 3 Lioo=-1, (5.380) —Funs)=S—8) (5.358) In writing Eq. (5.35b), we have made use of the fact that the branch of fixed points X(js) through (4,x)=(0,0) ‘must persist and have assumed a coordinate shift which places the branch at the origin. With these properties the Taylor expansion of f(j,r) at the fixed point x takes the form S(ux)= Myx tex? talus +Ox"), (5.36) where 2(0)=—1. The trick is to notice that the :vice- iterated map, f'(u,x)= Flt, fliex)), is undergoing a steady-state bifurcation, which is a pitchfork because of the reflection symmetry (5.35d) of the normal form. Fol- lowing our discussion of pitchfork bifurcation, we take Vi,x)=f7u,x)—x and check the prerequisite condi- tions (5.31), (5.33) using (5.35) and (5.36) 8% 6,9)= 2L10,0)(1+ 2£00,0) |= a F¥ 0.) FF (0,0 2L cys FE 0,0)= £L,0,0)2L 0.0) |1+ 200,01 |=0 ana (5376) -2Y 9,9)=2 (0,9 2L a. Pho 200 8F00=-29%0, (5.370)1006 F¥ (9.9)= Lio,0)ZL. Zool FF o.9= 0.0) 20.0) [1+ [Zoo 1200) , (5.37) respectively, Thus to satisfy Eq. (5.33) we need only as- sume (dA/dp)(0}%0 and a,(0)%0 in Eqs. (5.37c) and (5.374); each of these two conditions is compatible with Eqs. (5.35) and will typically be satisfied. The normal form for the pitchfork in f%(y,) is sata (temtes), — {aa with the bifurcation diagrams for fixed points of f1,x) shown in Fig. 10. These diagrams for f1,x) show three branches of fixed points x=0 and x=¥.(f!), and we now consider the implications for the original map f(j,x). Obviously, the x =0 branch is the fixed point for f(11,x), whose sta- bility is lost at =O. The X.(ji) branches of the pitch- fork for f%(j1,x) cannot be fixed points for f(y1,x), since the implicit function theorem guarantees that x=0 is the unique branch through 1,x)=(0,0). ‘Therefore (&,,2_) must represent a new bifurcating branch of two-eyeles for /(u,x). More precisely, denoting X. as 4, in the original variables of Eq, (5.35), we must have x= Sxe) (5.388) xe Sx (5.380) ‘The conclusion that f\y,x) must interchange x and x can be understood as follows. ‘The fixed-point equa- tion f%(u,x4)=x.. implies that the image of x4, “.=Flusx4), will also be a nonzero fixed point for Flux), ioe, ',=/7yex',). But we know that the pitchfork bifurcation for f? produces only two nonzero branches of fixed points, 80 x', must coincide with x_; hence Eq. (5.38) follows. Moreover, the reflection sym- metry of f(y.) requires x= —x when the dynamics is represented by the normal form."* The stability of the two-cyele (x .,x_) is determined by the stability of x (or x _) as fixed points for f? and is correctly indicated in Fig. 10. If we consider the bifurcation from the perspective that Eq. (5.35) describes an instability of a fixed point in the return map for a periodic orbit, then the bifurcating two-cycle represents a bifurcating branch of periodic or- "Stn fact, the reflection symmetsy of the period-doubling nor- ‘mal form implies that all ew branches of two-cyeles ca Se eal culated by solving fit,x)=—35 it i not necessary to consider explicitly the second iterate of f (ef. Crawford, Knobloch, and Riccke 1990) Rev. Mod, Pris Vo. 69,No 4, October 1991 John David Crawford: Introduction to bifurcation theory bits with approximately twice the period of the original orbit (see Fig. 13). This leads to the terminology period- doubling bifurcation, ‘8. Hopf bifurcation: simple complex-conjugate pal at|A|=t ‘The normal form in this ease is two dimensional; how- ever, its structure depends on subtleties not evident in the ‘examples of steady-state bifurcation or period-doubling. bifurcations. Denote the complex eigenvalue by Aye tauren ein, (5.394) where o0. (5.39¢) du If the eigenvalue at criticality MO)—e" satisfies the rnonresonance conditions, MOPAL and 201, 6.40) then in polar variables (r,#) the normal form for the bi- furcation is (Sala) (sib) At this order in ry, the right-hand side is independent of 4,8 feature analogous to the phase-shift symmetry en- countered in the normal form for Hopf bifurcation in a flow. In Sec. VINI, we shall show that this q indepen- dence depends on the nontesonance conditions (5.40). If these conditions are relaxed then dependent terms will appear in Eq, (5.41); when Eq. (5.40) holds, the depen- dence on ¥ will frst occur in terms that are indicated as (rin Eq, (5.40. For small r, we neglect the higher-order terms in Eq (Salt) and then solve the radial dynamics separately from the phase evolution. For this tactic to succeed, the cubic term in Eq, (54a) must not vanish at criticality, ie, we require yet atuiy {1 +a ,uire +O). Yr WyF 2x06 UN+b IPE OU!) FIG. 13, Period-doubling bifurcation in a Poincaré retum map.John David Crawford: Introduction to bifurcation theory 2 (040; (5.42) then Eq. (5.414) describes a pitchfork bifurcation at u=0. Only the positive bifurcating branch a (5.43) is relevant, since r must be non-negative. In combination with Eq, (5.41) the 77, branch describes a circle of radius ‘ry that is mapped into itself by (5.41), ie., the circle variant under iteration of the dynamics (5.41), ‘This branch of invariant circles may he either super- critical or suberitical depending on the sign of a /a, in Eq. (5.43). With the eigenvalue in (5.39a) assumed to be leaving the unit circle 5.39e), we have sen(ya,(0)) 6.44) near 1=0. Therefore, if a,(0) 0 ( supercritical), and if a,(0)>0, then the branch bifurcates when 1 <0 (subcritical). Using Eq, (S4la), one can show that the supercritical branch is stable and the subcritical branch will be unstable. Fur- thermore, one can prove that these invariant circles per- sist and have the properties just described if the O(r*) terms in Eq. (5.414) are restored (Ruelle and Takens, 1971; Lanford, 1973). However Eq. (5.41b) is much less satisfactory as @ description of the dynamics on the in- variant circle. According to (5.41), the circle dynamics is simply a fixed rotation by Apm2nO+b(U))+b,(uIrh +Olr4) 645) In the theory of maps of the circle (Guckenheimer and Holmes, 1986; Arnold, 19882), it is well known that such ‘4 uniform rotation is unstable if subjected to small per- turbations. Indeed, with the inclusion of small y- dependent perturbations present in the O(r*) terms of Eq. (S.41b), we expect phenomena such as mode locking, to occur in the dynamics on the circle; see Rasband FIG. 14. Hopf bifurcation ina Poincaré return map. Fv. Mod rye, Vol 63, No §, Octobe’ 1981 1007 (1990) for an introductory discussion. Finally, we consider this bifurcation from the perspec- tive that Eq. (5.41) describes an instability of a periodic ‘orbit as viewed in the return map to a Poincaré section. In this setting the invariant circle that appears in the sec tion corresponds fo a two-dimensional inaariant torus in the flow; see Fig. 14. ©. Final remarks If normal forms are to be generally useful, we must show that the bifurcation analysis of an arbitrary high- dimensional system can be reduced to a simple normal form. It is not obvious that we should be able to get as much information in one or two dimensions as we can in several, aor is it obvious that we will be able, even in low dimensions, to find coordinates in which our dynamical system is so simple. ‘The reduction in dimensionality is accomplished by observing that the interesting dynamics near a bifurca- tion occurs on a low-dimensional subset of phase space calied the center manifold.*" The dimension of this center ‘manifold determines the dimension of the normal form. ‘The simple structure of the normal form is established by theory of Poincaré-Birkhoff normal forms, VI. INVARIANT MANIFOLDS FOR EQUILIBRIA ‘A mathematically precise definition of manifolds and related geometric ideas may be found in many places, for example Chillingworth (1976), or Guillemin and Pollack (1974). Intuitively, a d-dimensional manifold in R” should be visualized as a smooth surface forming a subset of R", For example, a closed loop in R? and the surface ‘of a doughnut in R® are one- and two-dimensional mani- folds, respectively. Suppose M denotes a manifold in the phase space R* of ‘a dynamical system Eq, (I.2a) or (1.2b). Let m EM be an arbitrary point on the manifold, and let ©, denote the trajectory of the dynamical system through m, ic. x(0)=m for Eq. (1.2a) and x,-9=m for Eq. (1.20). If Om CM for all mEM, then M is sn invariant manifold for the dynamical system. More concisely, an invariant ‘manifold is a surface that is carried into itself by the dy- If MCRY is an invariant manifold, then the full dy- namics on R" implies the existence of # distinct auto- nomous dynamical system defined on M alone, which can Liapunow-Schmidt reduction is an alternative procedure for reducing the dimension of the problem. An introduction to this technique may be found in Golubitsky and Schaeffer (1985); the connection between center-manifold reduction and Liapunov- Schmidt reduction has been explored by Chossat and Golubit- sky (1987) and Marsden (1979)+1008 John David Crawtord: Introduction to bifurcation theory in prineiple be studied independently. For example, if a ‘map (1.2b) admits an invariant circle, then the dynamics fn this circle is described by a one-dimensional map of the circle to itself, e-., 8 1=L8,) modl2n) , co) ‘where the angle 9 labels points on the circle. The invari- ance of the circle implies that (8) will not depend on ‘the other phase-space coordinates. Thus Eq. (6.1) de- Scribes an autonomous one-dimensional dynamical sys- tem embedded in the dynemies (1.20) on a larger phase space. Individual trajectories provide vety simple examples of invariant manifolds. In a flow, an equilibrium and a periodic orbit are invariant manifolds with zero and one dimension, respectively. Much less trivial examples are the stable, center, and unstable manifolds associated with, equilibria." We first consider flows; the manifolds for ‘maps are quite similar and they are discussed briefly in subsection VIB. A. Flows: For a flow (1.2a),(1.3a), (ux), (62) the stable, center, and unstable manifolds for an equilibri- ‘um are generalizations of the invariant linear subspaces E‘, E‘ and E* that arise in the linearized dynamics F=DV(O,00% « 63) ‘These subspaces were described in See. ILA [ef. Eq. (2.7)]; hereafter we denote their dimensions by n,, nc» and ng» Respectively For the linear system (6.3), the subspaces (2.7) are in fact invariant manifolds. However, they are atypical, since these manifolds are also linear vector spaces; this special additional property reflects the linearity of Eq. (63). When the nonlinear terms in Eq. (6.2) are restored, the invariant manifolds just constructed for the linear system are perturbed but they persist, Their qualitative features also persist, except that the vector-space struc ture is lost. Intuitively, the nonlinear effects deform the invariant linear vector spaces into invariant nonlinear manifolds. For an equilibrium x=0, we have the following definition. A stable manifold is an invariant manifold of dimension n, that contains x=0 and is tangent to Bat "There isan extensive mathematical theory of invariant mani {olds with application to sets far more complex than the equili- ‘bria considered here. For a relatively introductory discussion see Irwin (1980) and Lanford (1983); other standard mathemati ‘eal references include Hirsch, Pugh, and Shivb (1977) and Shub (1989) Fev. Med Phys, Yo 63, No. October 1981 x=0. The unstable and center manifolds may be similar ly defined by replacing E* with E* end E*, respectively. We shall denote these manifolds by WW, Wand W°, see Fig. 20). The stable and unstable manifolds are unique. Fur- thermore, trajectories in these manifolds have some sim- ple dynamical properties. If xit)€W’, then x(t) +0 as toto; if x(0EW%, then x(1)+0 as ¢+— co. This asymptotic behavior is indicated schematically in Fig. 20 ‘The properties of center manifolds are somewhat more subtle (Lanford, 1973; Carr, 1981; Sijbrand, 1985). In general, the center manifold is not unique; we give an ex- ample of this nonuniqueness below. There is no general characterization of the dynamics on W*, not even asymp- totically as |1|—+eo. Nevertheless center manifolds play «a distinguished role in bifurestion theory because of two important properties. We discuss these properties here, and in Sec. WII we state a generalization of the ‘Hartman-Grobman theorem that justifies our discussion. For a center manifold W%, there exists a neighborhood Uofx =Osuch that (i) if x(OJEU has forward trajectory x(t) in U, ise X(DEU for all £20, then as t+ 20 the trajectory x0) ‘converges to W"; Gi) if x(0)ET has a trajectory in U, ie, x(0)EU for 2 0 and any €> 0, there ex- ists a time tg>T such that |x(to)—x(0)| tay and, for future refer ce ais! Fev. Mod. Phys, Vo 63, No.4, tober 1093 1015 To make this interpretation precise, we go back to Eq. (8.4) and define H°"R", FERN {#:R"-R"|6(ax)a'glx) for all aR} , 6.10) the space of all homogeneous polynomial maps on R" of degree k. For fixed k and n, H*(R") is a finite- dimensional linear vector space. The veetor-space struc- ture is obvious, and an example serves to make the finite dimensionality clear. Consider ¢°R*) with coordinates Ge y)ERY; then any x,y)E707(R?) may be written, ; Zee LG =0 ,8™X)— | Sayers |B) hence vectors ¢'*\x) of this form are eigenvectors, and the eigenvalues have the form [o)—3/S,a)0)]. If we can satisfy the condition o= Saw; 13) for any choice of j and a then L has a zero eigenvalue. One can see by inspection that when critical eigenvalues (Rea™0) occur there ate always choices of which s fy the “resonance condition” Eq. (8.13). We analyze the case of imaginary eigenvalues in Sec. VIILA.3 below. Tn the presence of such zero eigenvalues, the range of L, denoted LHR"), is a proper subspace of H4(R"), and we can specify a complementing subspace Cs that HOUR" = LEMUR NACH 6.14) Once Cis chosen (and the choice is not unique) then the kth order terms 7" in Eq. (8.7a) may be split ac- cordingly: Vi=ViP-+¥ with VPELEMR™)) and ¥;"EC™. The component in the range can be re- moved, gL), eis) leaving behind the “essential” nonlinear terms at order k namely Vi". In this way $” is first specified, then ¢"",1018 ‘and so forth so that one generates a power series? representing the desired normal-form transformation to all orders: x=O(x') PE BMAIEEMIE (IO) The normal form resulting from this procedure has the structure ae ai if at order k, detL70, then of course V7" There is an important subtlety in this procedure. The series Eq. (8.16), representing the transformation (x") required to put the original flow Eq. (8.1) into normal form to all orders, typically diverges. Thus, while we can describe which terms can be removed at any given order, the change of variables required to remove them to all or- ders does not generally exist. In practical applications fone implements the transformation to normal form only up to some finite order, and this finite-order approxima- tion to the original flow Fa, (8.1) is studied. The possible effects of the neglected higher-order terms can then be considered in reaching final conclusions. VN IAM MOLVA 5 BIT) 2, Steady-state bifurcation on R For steady-state bifurcation with a simple zero cigen- value, n,=1 in Eq, (8.1, and V(jz,x") has the form de- scribed in Eq, (5.2). If we try to simplify Eq. (5.4) by ap- plying the coordinate change Eq, (8.35) to remove the x? term, then the required change of variables is singular at criticality (u=0). For this reason, the method of Poincaré-Birkhoff normal forms is not particularly useful, in this case. A similar limitation holds for steady-state bifureation in maps. 3. Hopf biturcation on R* Generically for Hopf bifurcation, n,=2 in Eq. (8.1), and we take VM%(u,x") as given by Eq. (8.2). Let (x,9)ER® denote the coordinates, then gE 74'*(R?) hhas the form a9) | PON Vgecx,y) |» and in these variables L(¢'*) is expressed as Because of the iterative process used to construct ® the full series in Eq (8.16) is not of the form x'+ 3) "4" Fev. Mod. Pye, Vol 63, No.4, Otober 191 John David Crawford: Introduction to bifurcation theory wo [7 0], [Of 20) | 20 | Lae | aay? ag ax a | [rxtey aay ase |: [pay es) Pax ay ‘As noted in Sec. VIILA.1, to determine the eigenvalues Of L itis more convenient to use the complex coordinates (@.2) that diagonalize DV"0), ic, (8.19) ° ytio (eittez leMez secre oxte,zy(2.2) and the action of Lon gis ly-iw 0 | |e 114% FS Pull adh ag) a a | [(y-iow aah agit Gael 20) arp can By inspection from Eq. (8.20) we see that the eigenvee- tors have the form (6,0) or (0,d,), so we introduce the vectors ky 2p getaz) which are eigenvectors of L and also provide a basis for #"(R?), From Eq, (8.20) we calculate the eigenvalues LSM = KEN WEE? (8.220)John David Crawiord: Introduction to bifurcation theory 1017 where AEM SA-K yw) tol wR -2EN) (8.226) Since det L=0 implies at least one zero eigenvalue, and a zero eigenvalue in Eq, (8.22b) requires that real and imaginary parts Vanish separately, we must satisfy (-wyw=0 (ka New)=0 to obtain 2"'=0, Because k>2, Eq, (8.23a) fais unless Y(u)=0, which requires that we are at criticality 4=0 [recall Eq, (5.23)]. At u=0,0(0)40, so Eq. (8.23b) re- quires k~211=0. Since 2141 is odd, for k even we will never satisfy Eq. (8.22%), and for k odd there are ex- acrly two null eigenvectors at criticality, a These two vectors are a natural basis for the complement C1 to the range of L, cc (8.230) (8.236) geen [ (8.24) MOTI gk 12) pant, 5,7 ‘The implication for Eq. (8.1), written in complex coor- dinates Eq. (8.19), is a normal form with all even non- linear terms removed, EJ fro? pel El at varhn Sar «aas0 urs Sor, Rela) and 6)=—Im(a,). This is precisely the normal form introduced in Eq. (5.22). 4, Norms Morm symmetry Although normal forms may have fewer nonlinear terms, the discussion above does not explain why this should simplify the nonlinear analysis. For example, the ‘one-dimensional logistic map has only one nonlinear term and the Lorenz. equations have only two nonlinear terms, yet the immense dynamical complexity of these ‘two systems is well known, There is a more intrinsic explanation for the practical utility of normal-form theory: normal forms can have greater symmetry than the original system, and makes them simpler and therefore useful. The phase- shift symmetry of the Hopf normal form, i.e, the covari- ‘ow Moa Pye. Vol 63, No.4 Octobe 1991, ance of Eq. (8.26) under 0—-0-+4, illustrates this point. Note that this symmetry was not assumed to hold for the original vector field Eq. (8.1; rather, itis introduced by the normal-form transformation Eq. (8.16), As already discussed, the normal-form procedure is formal in the sense that Eq. (8.16) may not converge if carried to all or- ders. When the series diverges, then a symmetry intro- duced by Eq, (8.16) describes only an approximate prop- enty of Eq. (8.1), even though it is exact for the normal form. In the case of Hopf bifurcation we constructed the nor- smal form Eq, (6.16 first and then noted the phase-shift symmetry. This order ean be reversed; the theory of nor- ‘al forms can be formulated by identifying the relevant symmetty first and defining the normal form by its sym- metry. ‘The advantages of this second approach were noted by Belitskii (1978, 1981), Cushman and Sanders (1986), and Elphick er al. (1987). The results of Elphick 1 al (1987) are clearly discussed in Golubitsky, Stewart, and Schaeffer (1988), whose presentation is summarized here, The key result is that the complementing subspace C'* in Eq, (6:14) may be defined by a symmetry I that is determined by the linearization at criticality; ie DY(0,0). More precisely, let M=DV(0,0) and (transpose of M). Then Mf" generates a one-parameter ‘group of transformations with the obvious multiplication rule exp(s,M)exp(s:M7) pls) +52)M7) ‘The closure of this one-parameter group defines the normal-form symmetry™ FexpGM OPER] - (8.27) Let 7¢(8(R") denote the subspace of #4'*\(R") compris- jing those maps with I symmetry, ie, those V(x) EF6"(R") such that Ve (exp(sM Tex = explsM7)- Vx) (8.28) for all sR. ‘We shall prove that 7¢{?” may be taken as the comple- ‘ment C'" to the range of L so that Eq. (8,14) becomes UR LER) EUR") (8.29) In words, this splitting implies that the normal-form transformation Eq, (83a) can remove all kth-order terms except those with T symmetry. ‘The argument relies on a clever defnition of inner product on 74°"(R"). This definition is based on the fol- lowing product for monomials: for x ER", let x* and x? denote two monomials in mult-index notation and define [xsx8) Sagal, py defining F 9s closed group of matrices we ensure that it isa Lie group.1018 ‘which can be conveniently rewritten as tance | attat } 8.30) “This bracket extends to polynomials in the obvious way et plxl= Spar" and gle) Saige”. Then (Pla.a0o))= Bpsagix tar akigen | ean) ax? =| R". The first identity follows immediately from Eqs. (8.31) and (8.32): ($62), APYOO)= F (4 .64 44,00} = E [4,910.60 Bs =( Aba), Wx). (8.38) For the second identity, we express the jth component of $ 8 42)" 36.2% (and similarly for ¥,(x)}, so that the left-hand side of Ea. (8.37) becomes ($00, HATX)) (8.39) Topatly lx AT] Fae ‘Then with Eq. (8.31) we have Fev. Med, Phys. Vo. 69, No 4, October 1001 John David Crawford: Introduction to bifurcation theory wearane [Sate] ax lx=0 = a [leat]. =[(Ax x8]; (8.40) the second step the change of variables y = 4 ‘-x and the chain rule, Be 4. Bye) = 4.2198), were used to justify the substitution?® fariarar | [4.2 ]ye ah lal) 42 Pypl | axe {4 |” With Eqs. (8.39) and (8.40), the second identity (8.37) fol- tows directly These identities are applied by choosing AT =exp(—sM) in Eq. (8.36) and 47 =exp(sM) in Eq. (8.37) 0 obtain {Blxe MpheMx))= (eM Geox), Wx) an By differentiating Eq. (8.41) with respect to s at s =0, we finally arrive at CBA Lig) = Ly), 8)) » (842) which establishes Eq. (8.38). The argument leading to Eq, (8.29) can now be summa- rized, ‘The vector space #%(R") is first written as the direct sum of the kernel of L{, and the orthogonal com- plement ofthe kernel: HR) =(kerk jy )'@(kerL})) 5 43) then the Fredholm alternative for Ly implies tkerL},=L(H#""R")) and Bq. (8.35) implies. kerL jy =kerL yr. Thus Eq. (8.43) may be reexpressed as HR = LGR 0 (kerL yr) 44) Finally, with the aid ofthe identity (gM gig Mg) me™™L, gio Mx Set gic Mx =e Ly hbM yeas) wwe can identify kerLyr with FAPMORY. If 6EFA4PRY, [ook] fons] [ (summation cn repeated indices See Stakgold (1979), pp. 321-323,John David Crawford: Introduction to biturcation theory then the left-hand side of Eq. (8.45) vanishes, which im- plies Lyid=0; hence g€kerLyr Conversely, if 4€ker!.y/7, then the right-hand side is zero and the left- hhand side must be independent of s. This implies ME gle ME Wx), (3.46) since $(x)is the value at s =O; hence 62 74(91R"), Thus erly r= APR"), and Wg. (8.29) is established ‘Note that when M can be diagonalized then we may astume M?=—M and consequently Ket yj;—kerLy. Tt this case, the definition of T can be based direccly on Mf; is is not necessary to use the transpose, Tn our example of Hopf bifurcation, the linearization at etiticaity gives ea| 0 co) MT=!_ oy 9 [> (ean) so that an element of T’ has the form n_ [P0stsa10)) —sin(sa0)) etsM= lein(so(0)) cos(seX0)) CS) ‘As expected, this identifies the normal-form symmetry for Hopf bifurcation as rotations in 0 or TS". Note that for steady-state bifurcations DV(0,0)=0, so the as- sociated T' in Eq, (8.27) is trivial, consisting only of the identity matrix. ‘This explains why Poincaré-Birkhoft normalform methods do not significantly simplify the analysis ofa steady-state bifurcation. B. Maps 1. Genoraltios On the center manifold we find a map that may be written MOE SMG Six} x'ER™ (8.49) in a notation modeled on Eq, (8.1). We suppress explicit parameter dependence and ignore constant terms as be- fore. The goal remains the same: remove f'"(x’), if pos- sible, using the change of coordinates Eq. (8.3). In the new variables (unprimed) we find LOM ESM PM =F)“ L FON"), 8.500) where now L is defined by LESPKD=DFMO-S ML) SMM). (8.500) Note that Eq. (8.50b) differs crucially from (8.7b) in the second term; nevertheless we are again seeking to solve an equation of the same form, Fey. Med Phy, Vet 68, No.4, October 1801 r019 FX) LGV21=0 , sp by constructing L-', When decd = genvalues, and some nonlinear terms cannot be eliminat- ed. As for vector-field normal forms, if we assume coor- dinates can be found which diagonalize Df‘"%0), then the vectors #(x) having a single monomial compo- nent diM(x)=x% will be eigenvectors for L. Let 10q,) denote the eigenvalues of Df'"\0), ‘Then we find LP Hx (yon 08x o°8*x) ojo "x (8.52) from Eq. (500), where o%ssof!0'?--- a9" hence for maps the resonance condition required for a zero cigen- value is o,=0* 8.53) for some choice of jand a. ‘When zero eigenvalues occur, then the nonlinear terms that cannot be removed may be characterized by their symmetry. Let H=Df"0) denote the linear map at criticality [ef. Eq. 2.10)] and define the group generated by M7, P= [O47 ninteger| , (3.54) s0 that 7449R") now denotes elements of #°*(R") with symmetry (8.54); ie, A(XIEHEIR") requires MT-G(x)=MTx). With T and 74" redefined in this ‘way the proof that 7#*(R") may be expressed as HR") = LIER" 0 FPL) is quite similar to the argument leading to Eq. (8.29). With (8.55) (356 denoting the operator L. ff, Ea. (8.506) a criticality, the identities (8.36) and (837) imply Liy—Lyr. Therefore (ker {)'=Ly( IO) and kerL]y=kerLyr hold as be- fore, and we obtain Lyd! x)= Mog) BMX) HOR) = Ly ENR Woke yr 37 by the same reasoning that led to Eq, (8.44). It is only necessary 10 check that kerLyr=JHAP(R™) still holds This follows by noting that §kerL if and only if M™Bx): (MTx) , (8.58) which in turn is also necessary and sufficient for SETAE), The splitting (8.55) has the same significance here as in the vector-feld case: only when I’ defines s nontrivial symmetry should we expect the Poincaré-Birkhoff nor- mal form to be simpler than the original map. In addi- tion, the normal form for the original map (8.49) will be1020 FOI=LOD+ LMI (6.59) where f(x HPUR"), 2, Period-doubling bifurcation on R’ ‘Typically n,=1 for a period-doubling bifurcation, and Eq, (8.49) is a map in one dimension with P'ux)= Mw (8.60) where ah no + so ‘The space #4°*(R) is one dimensional for all k, and the single basis vector, Ex istan eigenvector for L#'(R)—>H'"(R); from Eq, (8.50) we find L(g x) (wy Aye) NEM) won i] 7 [1 +a\uijcos279(1 +6) Pay) [1+a(y)]sin27@(1+5(u)) in Eq. (8.32), where a(y),b() satisfy the assumptions in Eq (5.39). At criticality, @(0 metry (8.54) will be generated by [cos26 —sin2n6 sin2m@ cos? —[1+a(p)] sin2a0(1 +544) 1+a(y)}e082m0(1+5(j)) John David Crawford: Introduction to bifurcation theory Since M(O)-=—1, eigenvalue K1—A*~") will vanish at ceiticality when k is odd; thus terms of even degree can be removed, and only odd terms, —EI=E—) will remain in the normal form. If we consider the ex- pected symmetry Tin Eq. (8.54), then M=D,f"0,0)=—T 80 F=Zxl—T), the two-element group on R generated by —I. Thus we are again led to ‘the conclusion that for period-doubling the normal form, Aubxltaxtaxs{+Ox%)], (8.62) for Eq. (8.49) will have a reflection symmetry as claimed in Eqs. (5.34) and (5.35). 8. Hopt bifurcation on R? As for flows, one expects 1, and with coordi for Hopf bifurcation, tes (x,y) on FE? we have for 0<6< 5. q ws '6(0)=0, so the expected sym- (3.64) the rotation matrix for the angle 8 determined by the critical eigenvalues. ‘As before it Aw) 0 spf''s hia convenient to introduce complex coordinates (8.19) so that (8.63) becomes (3.65) where Ay )=[1+a(q)Je r+. From Eq. (8.506) we obtain eee,2 ait,20 $002 32) Jai az Fz) re sur | (8.66) ‘The eigenvectors of L are again given by Eq. (8.21), and (8.65) yields nigh uk where dy) = [1a (yr le®400 Cea By inspection detL0 unless 10, in which case 4$"(0)=0 if and only if exp[ —i276(k-2141)]=1 (8.68) ‘The solutions (k,)) to Eq, whether 0 is irrational or 1 (8.68) vary depending on ional. Fev. Mod. Pye, Vol 63, Ne. 4, tobe 1891 “iar 4 bn -21EM) (8.67—) (8.670) ———————e (a) 6 irrational. To satisfy Eq, (8.68) requires Ok ~211=m , (8.69) with m an integer, and when @ is irrational we must have k—21+1=0. This leads back to the null eigenvectorsJohn David Crawtord: Introduction to bifurcation theory 1021 (8.24) found for the Hopf normal form for flows. The re sulting normal form in this case is amy pwr Se lz)” (8.70) In polar variables, z=re'¥, we have rjer=U+a0}, der. 7a Yer ¥)F27O[L +b W)]+ Sb} (8.710) where o,=Rea, and b, (5.41) in See. V. ‘The fact that the dynamics of the amplitude (8.71a) decouples from the phase (8.71b) reflects the symmetry T.. For 0 irrational, the matrices, Ima. This agrees with Eq. [eoren@ —sin2nno sinden® cosdend a for all integers n, provide a dense subset of the group of rotations in the phase. Thus T'=S" is precisely this rota- tion group and corresponds to the phase-shift symmetry ‘of the normal form (8.71) (6) 0 rational Let 0=p/q with 0

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 eanJohn 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,,00 requires cn (9.37) for a valid solution 7} =—a, /2a, to exist. Substituting this solution into Eq. (9.35) or (9.36) yields yet (9.38) Taken together relations (9.37) and (9.38) locate the SN bifurcation surface for 07 <1. There are two cases, depending on the sign of a, at (u=y,,7=0). From the point of degeneracy the saddle-node surface branches 10 the right (a, <0) if @,>0 and to the left (a, >0) if a, <0. ‘These cases are indicated in Fig. 21 ‘The actual calculation of a, is a straightforward exten- sion of the calculation of a;. The calculation of (x,y) ‘must be carried to fourth order so that Eq. (9.25) can be extended to include fifth-order terms. Then seccad-, third-, and fourth-degree terms need 10 be removed to J @ Hip ace . FIG. 21. For the degenerate Hopf bifurcation corresponding to 2,=0 and a,0 thete are two possibilities, depending on the sign ofa, at critiealty. For a, >, the saddle-node (SN) suzface ‘branches toward negative values of ay. For a3 <0, the SN sus face branches toward positive values of a). The unstable periodic orbit which collides with the stable Hopf orbit is not shown. ev Med. Pry, Vo. 63,N. 4, tober 1991 siohn David Crawford: Introduction to bifurcation theory obtain the aormal form through fifth order. The details cf this are not of interest here; the resulting expression for @, as a function of u for 7 =0 is also plotted in Fig. 20. At the degenerate Hopf point (a, =0) we find a >0 fand conclude that the saddle-node surface branches to the sight, ‘The results of this bifurcation analysis may be tested numerically. Figure 22 shows the Hopf bifurcation to @ stable oscillation for 1>y,. As 4 decreases at fixed Ol, the stable periodic orbit loses stability near 4~3.55. This transition appears in Fig. 23 and reveals the dramatic effect of the saddle-node bifurcation. B. Steady-state bifurcation In the Ginzburg-Landau equation For the complex-valued function A(x,r) we consider the Ginzburg-Landau (GL) equation in one space dimen- Hin \ [Wan FIG. 22. Evolution of x(1) vs «for Bq. (9.1) from an initial con- ition (~1.0,0.0,0.8. (a) With y= 0.1 and =6.0 when the fixed point (9.3) is stable; note thatthe trajectory is initially re- pelled from the unstable fixed point at the origin. (b) With 7=0.1 and 4=6.0 when the fixed point is unstable and the solution is attracted to the stable Hopf periodic orbit. ‘The final, point on this trajectory segment was (4.486, ~2.886,4.499),John David Crawford: Introduction to biturcation theory s027 BA ge FA gigp Sennat SS alae, with real coefficients and with boundary conditions that censure finite-dimensional center manifolds (Tuckerman and Barkley, 1990). ‘This equation arises in a wide ‘variety of settings; in particular, Eq, (9.39) models the be- havior of a spatially extended system near criticality for a steady-state bifurcation (Collet and Eckman, 1990; Manneville, 1990). In fluid dynamics, a well-studied ex- ample of such a bifurcation is the appearance of Taylor vortex flow in a Taylor-Couctte apparatus (Ahlers ef al., 1986; Ahlers, 1989), where one observes the motion of & fluid confined in the gap between two concentric cylinders. Taylor's original investigation (1923), in which he rotated the inner cylinder with frequency 01 and fixed the outer cylinder, established a critical frequency %, above which the steady and (nearly) featureless flow de- yelops a pattern of vortices characterized by a well- defined axial wave number q. The fluid mechanisms that 0.39) { { I wn . A i aT Mm vantanttaanaae wh my , < oolp aaa SN WU TW FIG. 23, Byolution of x(1) vs # for Eq, (91) with the final point given for Fig. 22) used 2s the initial condition: (a) for y'=0.01, and 1=3.7 when the Hopf periodic orbit is stable; (b) for the same inital condition with 7=0.01 and y=3.5, ater the Hopf periodic orbit has teen destroyed. No stable orbit remains and the solution grows without bound. Notice the difference in the vertical sale, Fv. Med. Pins. Va. 63, No.4, tober 993 determine the wavelength 2/4 ofthe vortex flow have been carefully investigated asa particulary simple para digm for nonequilibrium pattern formation Langer, 1986; Ahlers, 198. ‘Analytic theory often assumes either cylinders of infite length or periodic anil boundary conditions ‘Then in Hinear approrimation one finds an eigenfunction wrth axial wave Dumber qe whose (real eigenvalue ap- proaches zero as f1 tends 10 0, from below. Slightly hove this threshold all wave aumbers within a band of th Vi about 4, are. linearly unstable, where -81,)/0, defines the bifurcation parameter, # However, only those wave numbers within a subband of width 1 are actually realized experimentally because of a secondary instability that arises for g values outside the subband. This latter instability is known as the Eckhaus instability and it modifies g by adding or subtracting vor- tex pairs. "The competition between different linearly unstable wavelengths can be studied near criticality (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 ok? we expect p(x,r) t© ap- proach a quasistatic equilibrium pl. eT 5 (9.64) since the ?p/x? term in Eq, (9.63a) can be neglected and k(x,7) evolves on a time scale set by the slowly vary- ing x dependence. When p is given by Eq. (9.64), the equation (9.63b) for the wave number reduces to & non- linear diffusion equation, ak 3k 2 lou kon (9.658) with ditusivity Dk) given by poe a3 cose) wk For a pure mode with p= Vp , consider the dynamics of a. small fivctuation in phase, k(xr)=Q+8Q(x,r) [relaxing the periodic boundary condition (9.46)]. Inserting this ansatz into Eq. (9.658) and linearizing in 80 yields a linear diffusion equation for the fuctuation (9.66) Ifthe wavelength of the pure mode g=9,+@ satisfies O0 and 8Q decays, the pure mode is stable.John David Crawtord: Introduction to bifurcation theory 1091 ‘The loss of phase stability when Q*> 1/3 is known as the Eckhaus instability; its physical interpretation as nega- tive phase diffusion was suggested by Pomeau and Manneville (1979). Subsequently more general theories of phase dynamics have been developed (Cross and Newell, 1984; Brand; 1988; Newel etal, 1990). 3. Bifurcation from the pure modes We now develop the bifurcation theory of this instabil- ity with the periodic boundary condition (9.46). @. Symmetry The 0(2)X0(2) symmetry of the A =0 equilibrium is partially broken in the pure-mode state; to describe the remaining symmetry define composite “translation” and “reflection” transformations T, and & by THQ) R@SR CR - Tes (9.688) (9.686) ‘These transformations generate a representation of (2), which we denote by O(2), and the 4g(x;6) state is in- variant with respect to this 6(2) action: 1 AglXiG)= AglxiG), Y=TalQ), RG) « Im addition, the discrete group Zp generated by 20/0 spatial translations (Tz979) is symmetry of Ag(x:6). Let a(x,7) be defined by (9.69) AG T= Aol: K1-+a0e,7)) 9.70) ‘Then Eq, (9.39) implies Ben ce-w-ghates, ents +2193 —w-@")0— (uP, 9.710) N(a,2)=[a?+2|al?+alal?] (76) From Eqs. (9.46) and (9,70) we find that a(x,r) also satisfies periodic boundary conditions on [—z,7], and with (9.71) we ean calculate the induced O(2) action on alx,7) (Typ AN) do lxib IT aox.7)) (88) A Noe) Ag le 5 81+ (Cera xy) 5 (9.7) thus the O(2) action on a(x,7) is generated by Ty and Cx. Note that the Zp action requires only symmetry with respect to Teg» and this is contained in Eq. (9.72) The covariance of Eq. (9.71a) with respect to 7, and Cx is easily checked; this cemaining O(2) symmetry will in- troduce nongeneric features into the secondary bifurca- Rev. Mod Phy, Vol. 89, No.4, eicber 1991 bs Linear stabi fore =0 ‘Then operator (8.71) saat, (Lay,a,)=(a,,La;), (9.73) wih spect to the inne product, leua Af tae) +e). 070 so we expect real eigenvalues 2 in the spectrum deter- mined by Loma 0.275) By inspection the eigenfunctions v are of the form Vexd=z (Ae Z(Ae , K=041,2,3,.25 0.76) with complex coefficients (2,25) that satisfy (A+KP+2kQ +4 Q7)]z1 =~ QF - om [DA+k?-2k9 + (4-9? N]eg = —(e~O7; - ‘The real and imaginary parts of z,, /=1,2, satisfy Eq. (0.77) separately, since A is real; this decoupling is due to the symmetry and forces the eigenvalues to have double multiplicity when k#0. More precisely, let (r),73)€R? bbe a real solution to Eq, (9.77) for eigenvector ler (Ade try de (9.788) ‘Then the translated eigenvector WORT ag 1X) I (Ae iri ade“™ (9.786) ‘has imaginary coefficients and is linearly independent of u(x} thus 2 has multiplicity two. [The other symmetry Ck leaves v(x) invariant.) When k=0 then Eqs. (9.76) and (9.77) yield only one (linearly independent) solution: vQ)=1, 4==24-97). 0.79) For a pure mode Ag to exist requires ldgi=n—9?>0, 0.80) So the k=0 mode (9.79) is always stable. The possibili- ties for instability arise from Eq. (9.77) for k>1 Without loss of generality let (z,,22) be real. ‘Then a non- trivial solution (r,#0, r.#0) requires ATA, or A=A where he (2 +p- QV uD ARQ! 0.81) ‘The A_ eigenvalue is always negative [ef. Eq. (9.80)], but 24, satisfies A, <0 if and only if B-Q*2 20-42. 9.8) With periodic boundary conditions, kn ‘mode Ag will be linearly stable provided Agl?> 207-1 so the pure (9.83) Stated slightly differently, as w increases above 4=0, all1032 pure modes with wave numbers in the band or 0 (supercriicaD. Tt is enough to determine this sign at A, =0, in which case the expressions for r, and r, fom Eq. (9.77) simplify +k (9.100) ° n=-BO-k), (up to an overall normalization). In addition, at 2 BOI from Eq, (9.82) and the center-manifold coefficients (9.98) reduce 10 ate? (9.101a) ™ With these formulas at criticality, the expression for o (9.97 simplifies considerably, ory trey tre), (Ay =O), (9.102) s0 that Eq. (9.98) may be rewritten HAyztHu—O*r, try elel +OAsz12/,21219) (9.103) Since (u—Q)ir, +ra)?>0, the new equilibria (9.99) are subcritical and unstable. X. OMITTED TOPICS The ideas of center-manifold theory and Poincaré Birkho normal forms are discussed by many authors. An introductory account is provided by Rasband (1990); for the reader seeking a more sophisticated treatment, both Guckenheimer and Holmes (1986), Chapter 3, and Ammold (19882), Chapter 6, are suggested. The recent re- view by Vanderbauwhede (1989) provides very detailed proofs for the finite-dimensional theory, and a careful re-1098 ‘John David Crawford: Introduction to bifurcation theory vview of center manifolds in Banach spaces is provided by ‘Vanderbauwhede and Tooss (1991). Additional material, fon the infinite-dimensional case in particular can be found in Marsden and McCracken (1976), Hassard et al. (1978), Ruelle (1989), and the encyclopedic volume by Chow and Hale (1982). Finally, there is the monograph by Tooss and Joseph (1989), which develops local bifurca- tion theory without using center manifolds, In one-parameter systems, the Feigenbaum bifurcation and global bifurcations involving homoclinic and hetero: clinic phenomena are important topics outside the scope Of this review. References for the former topic include Cyitanovie (1984), Collett and Eckmarn (1980), Lanford (1980), Vul e¢ al. (1984), as well as the original papers by Feigenbaum (1978, 1979, 1980). Glotal bifurcations, ‘especially Silnikov-type bifurcations and Melnikov theory, are discussed by Guckenheimer and Holmes (1986) and Wiggins (1988, 1990). In addition, the paper by Glendinning and Sparrow (1984) provides an accessi- ble introduction to the Silnikov bifurcation, ‘The recent lecture by Arnold (1989) touches on many ‘current research topics, in particular, multiparameter bi furcation problems and bifurcations in symmetric sys- tems. The examples in Sec. IX were selected in part to il- lusteate the importance of these subjects. An introduc- tion to codimension-two bifurcations (.e., bifurcations typical for two-parameter systems) is provided by Guck- enheimer and Holmes (1986), Chapter 7, and Arnold (19882), Chapter 6, but much of the work in this ares is scattered in the ‘research literature; Golubitsky and Guckenheimer (1986) and Roberts and Stewart (1991) are ‘wo recent conference proceedings. Bifurcation theory for symmetric systems is likewise an actively developing subject. In addition to the recent reviews by Stewart (1988), Gaeta (1990), and Crawford and Knobloch (1991), there ‘are the more extensive treatments in Vander bauwhede (1982), Sattinger (1983), and Golubitsky, Stewart, and Schaeffer (1988). Hamiltonian bifurcation theory is an important subject that is neglected here altogether. Unfortunately, there does not appear to be a systematic discussion of this theory for nonmathematicians at a level comparable to this review, and the literature is extensive. For bifurca- tion from equilibria of flows, Chapter 8 in Abraham and Marsden (1978) is @ possible starting point, in addition to the brief overviews by Meyer (1975, 1986). Up-to-date discussions of Hamiltonian normal-form theory can be found in Bryuno (1988) and van der Meer (1985). This latter monograph treats the so-called Hamiltonian Hopf bifurcation in detail. Howard and MacKay (1987) give a nice discussion of the linear instabilities encountered i symplectic maps; Golubitsky and Stewart (1987) describe fa generic setting for bifurcation in symmetric Hamiltoni- fan systems. The closely related subject of bifurcation theory for reversible systems is showing a rapid develop- ‘ment, Recent reviews have been given by Arnold and Sevryuh (1986) and Roberts and Quispel (1991). Finally, we mention the authoritative volumes emerg- Fy. Med, Pi, Vo. 89,No 4, Otaber 199+ ing from the "Kolmogorov school:” Anosov and Amold (1988), Sinai (1989), Arnold (1988b), and Arnold and No- vikov (1990), which provide many references to the So- viet literature, In particular, Anosov and Arnold (1988) treat normal forms and invariant manifold theory, and Arnold (19886) discusses Hamiltonian normal forms and. bifurcation, ACKNOWLEDGMENTS. 1am grateful 10 P. Morrison for arranging my visit to the Institute for Fusion Studies and to R. Hazeltine for suggesting this review. C. Kueny, W. Saphir, and B. Shadwick assisted in writing notes for the original lec tures, I would like to thank M. Silber and especially A. Kaufman for suggesting improvements in the final ‘manuscript. This work was supported in part by the U.S. Department of Energy Grant No. DE-FG05-80ET-53088, at the IFS and in part at the Institute for Nonlinear Sei ence at The University of California, San Diego by DARPA Applied and Computational Mathematics Pro- gram Contract No. F49620-87-C-0117 and by the DAR- PA University Research Initiative Contract No, NOOM4 86-K-0758 INDEX asymptotic stability 995, 997 bifurcation 992 degenerate 1021, 1025 Hopf 997, 998, 1003, 1006, 1022 imperfect 1003 period-doubling 98, 1005 pitchfork 997, 998, 1002, 1005 saddle-node 997, 998, 1000, 1005 steady-state 997, 998, 1000, 1004, 1026 subcritical 1004 supercritical 1004 transcritical 997, 998, 1001, 1005 bifereation diagram 1001 bifurcation set 996 bifurcation surface 102s ireation theory 992 branch of solutions 999 center manifold 1008, 1009 local attractivity 1008 onuniqueness 1008, 1012 reduction to 1009 center subspace 995, 997 codimension 992, 1025 critical system 996 dynamical system 992 Eckhaus instability 1021, 1031, 1032 eigenspace 995 eigenvalue degenerate 994. simple 1000John David Crawford: Introduction to bifurcation theory 1098 Axed poi 993 EFERENCES yperbalc 996, 997 Pogue they on Avraham, Ra J Manes, 17, Fundoton of Mens low (Benjamin-Cummings, Reading, MA). generic oo bien," 1, Sagem a. caions and oe raph representation 1010, 1012 Sictonl peters sonnet jes ff gu Ginnburpands equation 1027 cnn er te Secs of Compe ees yD Hartman Groban theorem 996,997 Seiad ey, Reon Cy CA, 9.5 homeonorpiem 296 Abign Ge B, Cant Ms Dongs Leta and R. Hei eee ee ion iu, -Wanena econ snd Ee say Hoot biueation Chute Tylor om" Pa D 28, 2 ae 100s AGonnrD. sand. Arn, 186, Bs, Dyno Some oe fe Oninay igen Santon and Sao Drie ae ton foo t Sabena Sens T ing invariance equation 1010, 1012 Arnold, V. I, 1972, “Lectures on bifureation in versal familics,”” variant ele toos wove Mah Seg invrint amid ia? Ain Vig 193, Onley Dien Beaton OMT, invert subepace 594, 997 Crm Ma invariant fore roo! ‘inl, Goon! Mating th Tey of Or TapunorSchmigt redaction 1007 nay Dien Batons son cion Springs New Lie bracket 1014 ‘York, ue a Ais, Vy 1885, 8, Drain! St I: Mathoma- tnedeosking itor Sa of Chie nd Cee! Macho, Eel mate 7 Sritthonatal Scenes Springer, Non Yor ‘Arnold, V. To 1989, ifurcation and singularities in mathemat- nonwandering point 1009 ics and mechanics,” in Theoretical and Applied Mechanics. normal form 998 Proceedings of the 17th International Congress of Theoretical dynamics 1000 and Applied Mechanics, Grenoble, 1988, edited by P. Germain, symmetry 1017, 1019 Me “Pao. and" D. Calflere| Gordy #oliand, Elven unfolding 398 ‘Anterdam/New Yor poussins (ase oe ‘Arold, V1, and 8: P. Novkow, 1990, Hs, Dynamical Spe Phase portrait 392 tom Sit Gao nd done > etoelee f ia of Mathematical Slee Springer, New YH Bhaseahih symmetry 103, 10 Al Teta MB ere rs ations in reversible syste” in Nonlinear Phenomena tn pitchfork bifurcation Plasma Physies and Hydrodynamics, edited by R. Z. Sagdecv flows 997, 1002 (Mir, Moscow), p. 31, sae 998, 1005 otis G. Ry 1978, “Egovaence and nora form of germ recurrem point 1003 of mouth napingy,” Russ Math Surveys, 107 Feflection symmetry 1002 Belishis GR 1981, “Normal forme eave tthe ering 3 ae 993 tion ofa group.” Trans Mone: Math, 3-40, 1 sede 1001 Brandy H.R 1988, Phase dynams-a review and a perspec: ‘iiule-node bifurcation tive in Popagtion i Syten Fur from Equitrium, ded ia 997, 1000 ty. were, HR: Brand, P. Mannevile ©. Altes and ae oom N.Doceara Springer, New Yor, 206. a on AD he nlm of Hann Stable manifold 1008, 1009 Eee aioe errr Stable subspace 995, 97 iia eeeeueaes eee steady-state bifurcation Ghiingworth D, 1976, Difeemiabe Topology with 2 View 0 ows 997, 1900 Applian Pinan, Londen) a 998, 1008 Ghent, Pe an. Gola, 1987, “Hopf bitecation in the stroboscopic map 983 Presence symmetry, centr anfold and Lipunov-Schnt oa foot Fesuction” i Olaaon, B/urcaton and Che, CMS Cont ae fae Fro freited by F-Atesow, W Langford nd’. Mingatl- a 10s american Mathematical Society, Providence) p38. sera how yaad, He TO, Made of Bfiatin trajectory 93 Cate, Pand JP Bekman, 1980, Hered Maps onthe Inte unstable manifold 1008, 1009 el ae Dynamical Systems (aitknawser, Based ‘unstable subspace 995, 997 Collett, P., and J-P. Eckmann, 1990, Instabilities and Fronts in wector fel 983 ‘Euended Stems Pinceion Univerty, Princeton, ND Fev. Mod. Phys, Yo. 62, No 4, October 10311036 John David Crawtord: Introduction to bifurcation theory Coullet, P., and BE. Spiegel, 1987, “Evolution equations for ex tended systems,” in Energy Stability and Convection, Pitman Research Notes 163, edited by G. P. Galdi and B. Straughan Wiley, New York), p. 22. Crawford, 1. D., 1983, “Hopf bifurcation and plasma instabil ties," Ph.D. thesis (University of California, Berkeley). Crawford, J. D., and E. Knobloch, 1991, "Symmetry and sym- ‘metry breaking bifurcations in fluid dynamics,” Annu. Rev. Fluid. Mech. 23, 341 Crawford, J. D., E. Knobloch, and H. Riccke, 1990, "Period- doubling mode interactions with circular symmetry," Physica D 44, 340. Cross, M. C., and A. C. Newell, 1984, “Convection patterns in large asgect ratio systems,” Physica D 10,299. Cushman, R., and J. A. Sanders, 1986, “Nilpotent normal forms ‘and representation theory of s1(2,R),” in Multiparameter Bi- {furcation Theory, Contemporary Mathematics 56, edited by M. ‘Golubitsky and J. Guckenbcimer (American Mathematical So- ciety, Providence), p.31 Cvitanovie,P., 1984, Bd, Universality in Chaos (Hilger, Bristol. Feckhaus, W., 1965, Studies in Nonlinear Stability Theory (Springer, New York). Edwards, W. Stuart, 1990, “Linear spicals in the frte Couette. ‘Taylor problem,” in Inability and Transition Vol, 2, edited by M.Y. Hussuini and R. G. Voige Springer, New York, p- 408 Flphick, C., E. Tirapegui, M. E, Brachet, P. Coullet, and G. Tooss, 1987 “A simple global characterization for normal forms of singular vector fields,” Physica D 29, 95; 32,488 (Ad~ dendum) Baler, L, 1744, Additamentum I de Curvis Elasici, Methodus TInveniendi Lineas Curvas Maximi Minimivi Proprietate Ga dentes. In Opera Ommia I (1960), Vol. 24 (Fish, Zurich), p. 2. Feigesbaum, M., 1978 “Quantitative universality for a class of nonlinear tansformations,” J. Sta. Phys. 19,25 Feigenbaum, M., 1979, “The universal metric properties of non- linear transformaticns,” J. Stat. Phys: 24, 669. Feigenbaum, M., 1980, “Universal behavior in nonlinear sys- tems," Los Alamos Sei 1,4, Gacta, G., 199, "Bifurcation and symmetry breaking,” Phys. Rep. 199, 1 Glendinning, P., and C. Sparrow, 1984, “Local and global be havior near homoclinic orbits," J Stat. Phys. 38,685. Golubitsky, M., and J. Guckenheimer, 1986, Eds, Multiparam- ‘ter Bifurcation Theory, Contemp. Math. 36 (American Mathematical Society, Providence). Golubitsky, M., and W. Langford, 1981, “Classifcation and un. foldings of degenerate Hop! bifurcations,” J. Dif. Eqns. 41, 375. Golubitsky, M., and D, Schaefer, 1985, Singularities and Groups in Bifurcation Theory, Vol. (Springer, New York) Golubitsky, M., and I, Stewart, 1987, "Generic bifurcation of Hamiltonian systems with symmetry,” Physica D 24,391. Golehitsky, M, I. Stewart and D. Schaefer, 1988, Singularities ‘and Groups in Bifurcation Theory, Vol. (Springer, New York). Graham, R, and J. A. Domaradzki, 1982, “Local amplitude ‘equation of Taylor vortices and its boundary condition,” Phys Rev. A 26, 1372. Guckenheimer, J, and P. Holmes, 1986, Nonlinear Oscillations, Dynomical Systems, and Bifurcations of Vector Fields (Springer, New York) Guillemin, V., and A. Pollack, 1974, Diferential Topology (Prentice-Hall, Englewood Clifs, ND. Fev. Med, Pn. Vo 69, No 8, October 1991 Hall P., 198, “Centrifugal insta in finite eplinders: nonlinear theory, Ser. A372, 317 Hartman, P1982, Ordinary Differential Equotions(Biekhivse, Boston) iasard, B, N. Kazarinoff, and Y-H. Wan, 1978, Theory and “Applications of Hopf Bifurcation, London Mathematical So lety Lecture Notes 41 (Cambridge University, Cambridge, England) Hirsch, M, C. Pugh, and M. Shub, 1977, Inarant Manifolds, Springer Lecture Nows in Mathematics No. 583 (Springer, New Yor). Hirsch, M, ad S. Smale 1975, Diferental Equation, Dynami- ‘col Systems, and Linear Algebra (Academic, New York Howard, J. By and R. S. MacKay, 1987, “Linear sabilty of symplectic maps," J. Math, Phys. 28,103, Hughes, D, and M, Proctor, 1950, “Chaos and the efect of noise in a model of theee-wave mode coupling,” Physica D 46, 163, Tos, Gand D. Joseph, 1989, Blementary Stability and Bifer- tation Theory Springer, New York) Iewin, M. C, 1980, Smooth Dynamical Systems (Academic, New York). Jordan, D. W., and P. Smith, 1987, Nonlinear Ordinary Diferential “Equations (Oxford University, New York/Landon). Xeamer, Land W. Zimmerman, 1985, “On the Bckhaus insta- bility tr spatially periodic patterns,” Physiea D 16, 21 Lanford, 0. 1973, "Bifurcation of periodic solutions into i Variant tori: the work of Rule and Takens," in Nonlinear Problems in the Piysical Sclences and Biology, edited by I Stakgold, D, D, Joseph, and D. H. Sattnger, Springer Lecture Notes in Mathemaiies No. 322" (Springer, Berlin’ Heideberg/New York, p. 159. Lanford, 0, 1980, “Smooth transforzations of interval.” in ‘Seminaire Bourbaki 1980/81, Springer Lectore Notes in Mathematics No, 901 (Springer, Betlin/Beidelberg/New York, 36 Lanford, 0, 1983, “Introduction tothe mathematieal theory of

You might also like

• The Subtle Art of Not Giving a F*ck: A Counterintuitive Approach to Living a Good Life

  

  From Everand

  The Subtle Art of Not Giving a F*ck: A Counterintuitive Approach to Living a Good Life

  Mark Manson

  Rating: 4 out of 5 stars

  4/5 (5819)
• The Gifts of Imperfection: Let Go of Who You Think You're Supposed to Be and Embrace Who You Are

  

  From Everand

  The Gifts of Imperfection: Let Go of Who You Think You're Supposed to Be and Embrace Who You Are

  Brené Brown

  Rating: 4 out of 5 stars

  4/5 (1092)
• Never Split the Difference: Negotiating As If Your Life Depended On It

  

  From Everand

  Never Split the Difference: Negotiating As If Your Life Depended On It

  Chris Voss

  Rating: 4.5 out of 5 stars

  4.5/5 (845)
• • Magazines
  • Podcasts
  • Sheet music
• Principles: Life and Work

  

  From Everand

  Principles: Life and Work

  Ray Dalio

  Rating: 4 out of 5 stars

  4/5 (609)
• The Glass Castle: A Memoir

  

  From Everand

  The Glass Castle: A Memoir

  Jeannette Walls

  Rating: 4.5 out of 5 stars

  4.5/5 (1717)
• Grit: The Power of Passion and Perseverance

  

  From Everand

  Grit: The Power of Passion and Perseverance

  Angela Duckworth

  Rating: 4 out of 5 stars

  4/5 (590)
• Sing, Unburied, Sing: A Novel

  

  From Everand

  Sing, Unburied, Sing: A Novel

  Jesmyn Ward

  Rating: 4 out of 5 stars

  4/5 (1104)
• Hidden Figures: The American Dream and the Untold Story of the Black Women Mathematicians Who Helped Win the Space Race

  

  From Everand

  Hidden Figures: The American Dream and the Untold Story of the Black Women Mathematicians Who Helped Win the Space Race

  Margot Lee Shetterly

  Rating: 4 out of 5 stars

  4/5 (897)
• Shoe Dog: A Memoir by the Creator of Nike

  

  From Everand

  Shoe Dog: A Memoir by the Creator of Nike

  Phil Knight

  Rating: 4.5 out of 5 stars

  4.5/5 (540)
• The Perks of Being a Wallflower

  

  From Everand

  The Perks of Being a Wallflower

  Stephen Chbosky

  Rating: 4.5 out of 5 stars

  4.5/5 (2104)
• The Hard Thing About Hard Things: Building a Business When There Are No Easy Answers

  

  From Everand

  The Hard Thing About Hard Things: Building a Business When There Are No Easy Answers

  Ben Horowitz

  Rating: 4.5 out of 5 stars

  4.5/5 (348)
• Bad Feminist: Essays

  

  From Everand

  Bad Feminist: Essays

  Roxane Gay

  Rating: 4 out of 5 stars

  4/5 (1018)
• Elon Musk: Tesla, SpaceX, and the Quest for a Fantastic Future

  

  From Everand

  Elon Musk: Tesla, SpaceX, and the Quest for a Fantastic Future

  Ashlee Vance

  Rating: 4.5 out of 5 stars

  4.5/5 (474)
• Her Body and Other Parties: Stories

  

  From Everand

  Her Body and Other Parties: Stories

  Carmen Maria Machado

  Rating: 4 out of 5 stars

  4/5 (822)
• The Outsider: A Novel

  

  From Everand

  The Outsider: A Novel

  Stephen King

  Rating: 4 out of 5 stars

  4/5 (1866)
• The Emperor of All Maladies: A Biography of Cancer

  

  From Everand

  The Emperor of All Maladies: A Biography of Cancer

  Siddhartha Mukherjee

  Rating: 4.5 out of 5 stars

  4.5/5 (271)
• The Sympathizer: A Novel (Pulitzer Prize for Fiction)

  

  From Everand

  The Sympathizer: A Novel (Pulitzer Prize for Fiction)

  Viet Thanh Nguyen

  Rating: 4.5 out of 5 stars

  4.5/5 (122)
• Angela's Ashes: A Memoir

  

  From Everand

  Angela's Ashes: A Memoir

  Frank McCourt

  Rating: 4.5 out of 5 stars

  4.5/5 (441)
• Brooklyn: A Novel

  

  From Everand

  Brooklyn: A Novel

  Colm Toibin

  Rating: 3.5 out of 5 stars

  3.5/5 (1947)
• The Little Book of Hygge: Danish Secrets to Happy Living

  

  From Everand

  The Little Book of Hygge: Danish Secrets to Happy Living

  Meik Wiking

  Rating: 3.5 out of 5 stars

  3.5/5 (401)
• A Man Called Ove: A Novel

  

  From Everand

  A Man Called Ove: A Novel

  Fredrik Backman

  Rating: 4.5 out of 5 stars

  4.5/5 (4770)
• Steve Jobs

  

  From Everand

  Steve Jobs

  Walter Isaacson

  Rating: 4.5 out of 5 stars

  4.5/5 (808)
• The World Is Flat 3.0: A Brief History of the Twenty-first Century

  

  From Everand

  The World Is Flat 3.0: A Brief History of the Twenty-first Century

  Thomas L. Friedman

  Rating: 3.5 out of 5 stars

  3.5/5 (2259)
• The Yellow House: A Memoir (2019 National Book Award Winner)

  

  From Everand

  The Yellow House: A Memoir (2019 National Book Award Winner)

  Sarah M. Broom

  Rating: 4 out of 5 stars

  4/5 (98)
• Devil in the Grove: Thurgood Marshall, the Groveland Boys, and the Dawn of a New America

  

  From Everand

  Devil in the Grove: Thurgood Marshall, the Groveland Boys, and the Dawn of a New America

  Gilbert King

  Rating: 4.5 out of 5 stars

  4.5/5 (266)
• The Art of Racing in the Rain: A Novel

  

  From Everand

  The Art of Racing in the Rain: A Novel

  Garth Stein

  Rating: 4 out of 5 stars

  4/5 (4208)
• A Tree Grows in Brooklyn

  

  From Everand

  A Tree Grows in Brooklyn

  Betty Smith

  Rating: 4.5 out of 5 stars

  4.5/5 (1929)
• Team of Rivals: The Political Genius of Abraham Lincoln

  

  From Everand

  Team of Rivals: The Political Genius of Abraham Lincoln

  Doris Kearns Goodwin

  Rating: 4.5 out of 5 stars

  4.5/5 (234)
• Yes Please

  

  From Everand

  Yes Please

  Amy Poehler

  Rating: 4 out of 5 stars

  4/5 (1902)
• A Heartbreaking Work Of Staggering Genius: A Memoir Based on a True Story

  

  From Everand

  A Heartbreaking Work Of Staggering Genius: A Memoir Based on a True Story

  Dave Eggers

  Rating: 3.5 out of 5 stars

  3.5/5 (231)
• The Woman in Cabin 10

  

  From Everand

  The Woman in Cabin 10

  Ruth Ware

  Rating: 3.5 out of 5 stars

  3.5/5 (2522)
• Fear: Trump in the White House

  

  From Everand

  Fear: Trump in the White House

  Bob Woodward

  Rating: 3.5 out of 5 stars

  3.5/5 (738)
• Wolf Hall: A Novel

  

  From Everand

  Wolf Hall: A Novel

  Hilary Mantel

  Rating: 4 out of 5 stars

  4/5 (3811)
• John Adams

  

  From Everand

  John Adams

  David McCullough

  Rating: 4.5 out of 5 stars

  4.5/5 (2409)
• On Fire: The (Burning) Case for a Green New Deal

  

  From Everand

  On Fire: The (Burning) Case for a Green New Deal

  Naomi Klein

  Rating: 4 out of 5 stars

  4/5 (74)
• The Light Between Oceans: A Novel

  

  From Everand

  The Light Between Oceans: A Novel

  M.L. Stedman

  Rating: 4.5 out of 5 stars

  4.5/5 (789)
• Manhattan Beach: A Novel

  

  From Everand

  Manhattan Beach: A Novel

  Jennifer Egan

  Rating: 3.5 out of 5 stars

  3.5/5 (792)
• The Constant Gardener: A Novel

  

  From Everand

  The Constant Gardener: A Novel

  John le Carré

  Rating: 3.5 out of 5 stars

  3.5/5 (104)
• The Unwinding: An Inner History of the New America

  

  From Everand

  The Unwinding: An Inner History of the New America

  George Packer

  Rating: 4 out of 5 stars

  4/5 (45)
• Rise of ISIS: A Threat We Can't Ignore

  

  From Everand

  Rise of ISIS: A Threat We Can't Ignore

  Jay Sekulow

  Rating: 3.5 out of 5 stars

  3.5/5 (137)
• Little Women

  

  From Everand

  Little Women

  Louisa May Alcott

  Rating: 4 out of 5 stars

  4/5 (104)