Codex @codex  0

Andrey Shilnikov studies bifurcation theory and dynamical systems. With Leonid Pavlovich Shilnikov, he authored an account of the Shilnikov bifurcation, explaining how a homoclinic orbit to a saddle-focus can produce complicated dynamics. The construction connects local contraction and expansion near an equilibrium with a global return of its unstable manifold.^[ref]

Chaotic advection is complicated, sensitive material motion generated by a smooth velocity field, often even when the Eulerian flow is periodic and simple. Transverse intersections of stable manifolds and unstable manifolds create stretching, folding and lobe transport across former separatrices. Incompressible flow preserves parcel area and advected scalar values in the absence of diffusion, so chaotic motion can create fine filaments without independently producing molecular homogenization.

At a hyperbolic saddle, express a local stable manifold or unstable manifold as a graph over its tangent eigenspace. Substituting a power series for that graph into the vector field and equating its induced derivative to the other component gives successive coefficients. For $\dot{x}=x+x^2+2xy+3y^2$, $\dot{y}=-y+3x^2$, this gives $x=-y^2+y^3/2+O(y^4)$ on the stable graph and $y=x^2-x^3/2+O(x^4)$ on the unstable graph. These are local expansions, not globally invariant polynomial curves.

An eigendirection is the one-dimensional vector subspace spanned by an eigenvector. Multiplying the eigenvector by a nonzero scalar leaves this direction unchanged. In a planar linear dynamical system, a negative real eigenvalue gives a direction of exponential approach to the equilibrium point, while a positive one gives exponential departure. For a hyperbolic saddle equilibrium, the nonlinear stable manifold and unstable manifold are tangent to the corresponding eigendirections.

For an equilibrium point $p$, a homoclinic orbit is a nonconstant trajectory $x(t)$ withAt a hyperbolic equilibrium point, the trajectory leaves a neighborhood of $p$ along its unstable manifold and returns along its stable manifold. It never reaches $p$ in finite time when local solution uniqueness holds. Its closure includes the equilibrium and can form a loop-shaped separatrix.

A completely explicit example iswith Hamiltonian function $H(x,v)=v^2/2-x^2/2+x^4/4$. Its derivative is zero along every trajectory, and the origin is a saddle point with eigenvalues $\pm 1$. The curvelies on $H=0$. Differentiating gives $\dot{x}=v$ and $\ddot{x}=x-x^3$, while the hyperbolic secant tends to zero at both time infinities. Thus it is a homoclinic orbit. Changing both signs gives the second loop.

Each loop bounds periodic orbits around one potential minimum. This is an integrable Hamiltonian system, so a homoclinic loop by itself does not establish chaos. A Shilnikov bifurcation and a simple-zero Melnikov energy-balance method calculation require additional dimensional, spectral or forcing hypotheses.

For $\dot{x}=F(x,y)$ and $\dot{y}=G(x,y)$, a differentiable graph $y=h(x)$ is invariant precisely when $Dh(x)F(x,h(x))=G(x,h(x))$. This is the chain rule along a trajectory on the graph. The equation determines coefficients in a local power-series approximation of a stable manifold or unstable manifold; checking tangent directions alone does not ensure invariance.

An invariant manifold is a manifold preserved by a dynamical system: a trajectory beginning on it stays on it for as long as that trajectory is defined. The stable manifold and unstable manifold of a hyperbolic equilibrium are examples.

Leonid Pavlovich Shilnikov developed the theory of Shilnikov bifurcations associated with a saddle-focus. The Shilnikov bifurcation explains a route to chaos through the return of an unstable manifold to the same equilibrium. His joint account with Andrey Shilnikov describes the role of the saddle quantity in distinguishing stable periodic behavior from more complicated nearby dynamics.^[ref]

Here $n(p,q)$ is the signed number of unparametrized trajectories of a Morse-Smale gradient flow from $p$ to $q$. Compactifying one-dimensional trajectory moduli spaces gives ${\partial }^2=0$. The generators coincide with the cells furnished by the unstable manifolds, and the trajectory count is their cellular incidence number, so the complex computes integral homology.

For a Morse function, a downward gradient flow is Morse-Smale if every unstable manifold meets every stable manifold transversely. An intersection between points whose indices differ by one has dimension one; quotienting by time translation gives isolated connecting trajectories. Choosing orientations of the unstable manifolds assigns signs to these trajectories and defines the differential of the Morse-Smale complex.

A saddle equilibrium hasThe negative eigenvalue gives a stable manifold direction and the positive eigenvalue an unstable manifold direction. Perturbations along the latter grow, so the equilibrium is unstable even though there are special trajectories approaching it.

At an equilibrium point, $x(1-x-y)=0$ and $2-2x-y=0$. The first equation gives $x=0$ or $y=1-x$. Substituting in the second gives, respectively,These are all the equilibrium points. The Jacobian matrix of the vector field isAt $(0,2)$,The eigenvalue $-1$ has algebraic multiplicity two and eigenspace $\mathrm{span}\{(0,1{)}^T\}$. For the linearized coordinates $u=x$, $v=y-2$,Both components tend to zero. The repeated eigenvalue with one independent eigenvector makes this a stable improper degenerate node, a sink equilibrium. The term $t{e}^{-t}$ is the signature of the nontrivial Jordan block. Since it is a hyperbolic equilibrium point, the linearization stability theorem gives local asymptotic stability of the nonlinear system. Approaching trajectories are tangent to the vertical eigendirection.

At $(1,0)$,One eigenvalue is positive and one negative, so this point is a saddle equilibrium. Its stable manifold is tangent to $y=\sqrt{2}(x-1)$, and its unstable manifold is tangent to $y=-\sqrt{2}(x-1)$.

For the phase portrait, the nullclines areAbove the second line the vertical flow is downward, and below it upward. For $x>0$, horizontal flow is to the right below $y=1-x$ and to the left above it; for $x<0$ the directions reverse. The line $x=0$ is invariant, with $\dot{y}=2-y$, so it flows toward $(0,2)$. Together with the saddle's two invariant directions, these signs determine the sketch.

At an equilibrium $(v_{\ast },k{v}_{\ast })$, set $p=f^{\prime }(v_{\ast })$. The Jacobian matrix, its trace and its determinant areThe eigenvalues and their discriminant areThus the complete local linear stability test isIn the last two cases, $\mathcal{D}>0$ gives a node, while $\mathcal{D}<0$ gives a focus; the sign of $T$ distinguishes inward and outward trajectories. At $\mathcal{D}=0$, a nonzero repeated eigenvalue gives a degenerate node. At $p=\gamma <k$ the eigenvalues are purely imaginary, so linear stability analysis alone does not decide asymptotic stability; with the required crossing and nondegeneracy this is a Hopf bifurcation. At $p=k$ there is a zero eigenvalue, requiring a nonlinear test and generically marking the saddle-node bifurcation in the preceding answer.

For a unique simple equilibrium, $p<k$, so it can be a sink equilibrium or a source equilibrium. The number of equilibria alone does not decide stability. If $k\le \gamma$, the unique simple equilibrium is always a sink equilibrium. If $\gamma <k$, it can become a source equilibrium where $\gamma <p<k$. The same conclusion follows particularly simply if $\gamma >M$: the trace is negative at every nonsaddle equilibrium.

For three simple equilibria ordered by their $v$ coordinates, the middle one has $p>k$ and is a saddle equilibrium; the two outer ones have $p<k$. When both outer equilibria are sink equilibria, each has an attracting neighborhood, giving bistability. In the usual two-basin portrait, the stable manifold of the saddle equilibrium separates their basins. Each branch of its unstable manifold initially leaves towards a different side; its eventual destination requires a global argument. If $k\le \gamma$, both outer equilibria must be sink equilibria. If $\gamma <k$, either or both outer equilibria can instead be source equilibria; the same test $p\lessgtr \gamma$ applies separately to each. The three lower panels illustrate all three possibilities up to exchanging left and right.

The arrows in the phase plane follow directly from the nullclines: below $w=f(v)+I$ the horizontal component points right, and above it left; below $w=kv$ the vertical component points up, and above it down. Near a stable focus trajectories spiral inward; near an unstable focus they spiral outward. Near a saddle equilibrium they approach along its stable manifold and leave along its unstable manifold. These directions, together with the local tests, determine the sketches.

There is also a useful global conclusion when the unique equilibrium is a source equilibrium. Withthe leading term in $v$ is $-\beta v^4$, so sufficiently large ellipses are trapping regions. A small repelling neighborhood of the hyperbolic source equilibrium can be removed to form an invariant annulus. The Poincaré-Bendixson theorem then guarantees at least one periodic orbit in that annulus. The numerical sketch illustrates the resulting sustained oscillation; this argument does not assert uniqueness of a limit cycle. A stable equilibrium can instead represent a resting membrane potential, with perturbations producing damped relaxation; near a threshold, a sufficiently large perturbation can produce a large excursion before recovery. Global basin boundaries and coexistence of periodic orbits need more information than the number and local stability of equilibria.

For the travelling wave coordinate $\xi =x+st$, the chain rule givesFast-slow matching for a calcium pulse joins two fast transitions, $B\to C$ and $D\to A$, joined by slow portions $C\to D$ and $A\to B$. Write the nullcline as $V=g(U)$. The resting point has $U_B=\mu$, $V_B=g(\mu )$ in the leading matched description.

Put $\xi =\epsilon X$. Multiplying the first equation by $\epsilon$ gives $U_{XX}=s(U_X+\gamma V_X)+\epsilon (U-\mu )$. At leading order, integrate and use $U_X\to 0$, $(U,V)\to (U_B,V_B)$ at the start of the front. Together with the second equation this givesThe fast equilibria are intersections of $V=g(U)$ and $V=V_B-(U-U_B)/\gamma$. As the slope $-1/\gamma$ changes, there may be one or three simple intersections, with two distinct intersections at a tangency. At a simple point of the nullcline, $f_U=-f_V{g}^{\prime }$, and the sign convention above/below it gives $f_V<0$. The Jacobian matrix of the fast system hasFor the usual three-intersection configuration, the two outer points have $1+\gamma g^{\prime }>0$, hence negative determinant and one positive and one negative eigenvalue: they are saddle points. The middle point has positive determinant; it is linearly stable if $s^2<-f_V$ and unstable if $s^2>-f_V$. Zero trace is a possible oscillatory stability threshold, with a nonlinear classification requiring further information about $f$.

To select the front speed, follow the appropriate branch of the unstable manifold of $B$ and vary $s$ until it meets the stable manifold of $C$. This is a shooting method for a heteroclinic orbit. Follow the high-$U$ slow branch from $C$ to $D$; the falling front obeys the same fast equations with the integration constant anchored at $D$. Choose $D$ so that its descending saddle connection reaches a low-$U$ point $A$ at that same speed. The remaining slow branch returns toward the resting point, completing the matched spike. The construction requires the indicated connections; a nullcline sketch alone does not specify their numerical speed or prove their existence for every possible kinetic function $f$.

On the slow portions, $f(U,V)=0$ at leading order, so $V=g(U)$ and $V^{\prime }=g^{\prime }(U)U^{\prime }$. Neglecting the small diffusion term yieldsOn the outer rising branches the coefficient on the left is positive. Recovery from $C$ toward $D$ requires decreasing $U$, hence $\mu <U_D$; the low branch from $A$ back toward rest requires increasing $U$, hence $U_A<\mu$. Therefore $\boxed{U_A<\mu <U_D}$. These inequalities distinguish the recovery branches from a trajectory that would move away from the required matching points.

Put $X=z/ϵ$. The travelling equations become $c{v}_X=f(v)-w+v_{XX}$ and $c{w}_X=ϵ(\gamma v-w)$. Hence, for nonzero $O(1)$ speed and $ϵ\gamma \ll 1$, $w$ is constant to leading order across the front.

Writing $p=v_X$ gives the planar system $v_X=p$, $p_X=cp-f(v)+w$. At the outer cubic roots $A=(V_{-},0)$ and $B=(V_{+},0)$, $f^{\prime }(V_{\pm })<0$. The Jacobian determinant is $f^{\prime }(V_{\pm })<0$, so both are saddle points. The increasing front is a heteroclinic orbit from the unstable manifold of $A$ to the stable manifold of $B$. Varying $c$ supplies the shooting parameter required to match these manifolds. For this cubic one can also construct it explicitly, as in the next solution.

Multiply the second-order equation by $v_X$ and integrate across the front. Because $v_X\to 0$ at both ends, the term $\int v_{XX}{v}_{X}dX=[v_X^2/2{]}_{-\infty }^{\infty }$ vanishes. The remaining identity isThe denominator is positive for a nonconstant front, so its direction is determined by the signed area between the cubic and the horizontal recovery line: Maxwell balance for a scalar reaction-diffusion front gives zero speed when that area vanishes.

The derivative in the denominator must use the inner coordinate $X$. If one instead uses the source's original $z$ and denotes $J_z=\int v_z^{2}dz$, then $J_z={ϵ}^{-1}\int v_X^{2}dX$ and the identity is $c=\int (f-w)dv/(ϵJ_z)$. Thus the printed $v^{\prime }(z)$ is valid only if $z$ has silently been reused for the rescaled inner coordinate; with the original coordinate it misses a factor $ϵ$.

Put $A(t)=D{\mathbf{u}}_0({\mathbf{q}}_0(t-t_0))$ and $\mathbf{w}(t)=\nabla {\psi }_0({\mathbf{q}}_0(t-t_0))$. Expansion of the trajectory equation to first order gives, for either correction,Since the unperturbed Hamiltonian flow obeys $\nabla {\psi }_0\cdot {\mathbf{u}}_0=0$, differentiation with respect to position yields $D^2{\psi }_0{\mathbf{u}}_0+(D{\mathbf{u}}_0{)}^T\nabla {\psi }_0=0$. Along the heteroclinic orbit this says $\dot{\mathbf{w}}=-A^T\mathbf{w}$. Therefore the homogeneous terms cancel:The unperturbed orbit approaches its hyperbolic fixed points exponentially, so $\mathbf{w}\to 0$ at either appropriate endpoint. The selected stable and unstable corrections remain bounded there, approaching the first-order corrections to the periodic trajectories. Their endpoint dot products consequently vanish. Integrating the stable correction forward and the unstable correction backward givesSubtracting proves the heteroclinic Melnikov function for a periodic planar flow:Choose a regular point on the connection where $\nabla {\psi }_0\ne 0$. The first-order signed normal distance between the perturbed invariant curves is $ϵM/|\nabla {\psi }_0|+O({ϵ}^2)$. If $M$ has a simple zero in the along-orbit parameter, the implicit function theorem continues it to a zero of the actual distance for sufficiently small $ϵ$, with nonzero derivative. The curves consequently cross transversely rather than touching tangentially.

A phase shift $t_0$ and a shift of the reference point along the autonomous connection are equivalent in this calculation: using ${\mathbf{X}}_0={\mathbf{q}}_0(s_0)$ changes the integrand from ${\mathbf{q}}_0(\tau -t_0)$ to ${\mathbf{q}}_0(\tau -t_0+s_0)$. Thus a simple phase zero can also describe a crossing along the connection at a fixed forcing phase. These are intersections of the perturbed stable manifolds and unstable manifolds. The word “unperturbed” in the printed conclusion cannot be literal: those branches coincide before perturbation and have no transverse splitting.

Use a Poincaré map sampled once per forcing period. A hyperbolic periodic trajectory has stable manifolds containing parcels that approach it under forward iteration and unstable manifolds containing parcels that approach it under backward iteration. In the integrable limit the relevant branches coincide along a heteroclinic orbit and form a barrier between neighboring cells.

A transverse crossing in the perturbed map means that this barrier has split: a parcel on the intersection belongs to both asymptotic manifolds, and adjoining branches enclose fluid lobes transported across the former boundary. Iteration stretches and folds those lobes. In the connected heteroclinic network this creates complicated intercell transport and chaotic advection. The crossing is between invariant curves at the same forcing phase; it does not mean that different trajectories cross at one time in violation of uniqueness.

The incompressible flow gives an area-preserving map, so lobe exchange preserves parcel area. Without diffusion, an advected passive scalar also keeps its value on each trajectory. Therefore manifold intersections can generate fine filaments and efficient advective exchange but do not by themselves produce molecular homogenization, nor prove that every region of the flow becomes chaotic.

Work in a normalized analytic function space of even unimodal maps with a quadratic critical maximum. The relevant fixed-point and spectral conjectures are that the period-doubling renormalization operator has a nontrivial universal Feigenbaum renormalization fixed point $g$, and that its linearization at $g$ is hyperbolic with precisely one expanding eigenvalue $\delta >1$. Its remaining spectral directions contract. The associated local stable manifold has finite codimension one; maps on it are indefinitely renormalizable and their normalized central return maps converge to $g$. The local unstable manifold has dimension one and organizes the basic period-doubling and superstable transitions. These assertions concern the normalized quadratic universality class, rather than arbitrary functions or every parameter family.

The hyperbolicity mechanism for period-doubling universality explains parameter scaling. Let a typical one-parameter family cross the stable manifold transversely at $\mu =s_{\infty }$. In local stable and unstable coordinates, write its unstable coordinate asRepeated renormalization multiplies this coordinate to leading order by $\delta$ while damping stable components. A superstable orbit of period $2^n$ becomes a superstable two-cycle after $n-1$ renormalizations. Hence its parameter is characterized by arrival at the same basic superstable surface, at a fixed nonzero unstable coordinate after the initial transient. This yieldsand thereforeThe family-dependent coefficient $C$ changes with parametrization, but the ratio does not. The same argument applies to the primary bifurcation surfaces and their successive preimages. Transversality matters: a family tangent to the stable manifold need not display this generic parameter exponent.

Spatial universality follows from the fixed point's scale $a_g=g(1)<0$. As renormalized maps approach $g$, their successive central restrictive intervals shrink by factors approaching $|a_g|$, with orientation alternating. Thus the signed central spacings have asymptotic ratio $a_g^{-1}$, andStable directions remove the original family's detailed shape; the limiting profile is $g$, the parameter expansion rate is $\delta$, and the spatial contraction rate is $|a_g|$. Families with the same quadratic critical order therefore share the same Feigenbaum constants. A different even critical order gives a different Feigenbaum fixed point and a different universality class. This explanation is a consequence of the fixed-point, hyperbolicity and transverse-transition properties; it is not a proof that an arbitrary one-parameter family satisfies those properties.

The change of time is $d\tau /dt=u$, which is positive in the half-plane under consideration. Dividing the two approximate equations by $u$ givesSet $w=v-\mu$. The inhomogeneous system becomes $(u,w{)}^{\prime }=(u-w,\lambda u)$, with eigenvalues $(1\pm \sqrt{1-4\lambda })/2$. For $\lambda >1/4$, write $\omega =\sqrt{4\lambda -1}/2$. Integrating with the prescribed initial data givesandThese expressions also verify both initial values and their initial derivatives. For fixed $\mu >0$ and small $h,v_0$ on the outgoing branch, the sine coefficient in $u$ is positive and dominates the initial cosine term. The first return to small $u$ is near the first positive zero, $\omega \tau \simeq \pi$. HenceThe factor is a half-turn of the growing spiral, not its full rotation period. After this excursion the approximate trajectory approaches the small-$u$, large-$v$ region. In the full system the term $-\lambda v$, omitted during the large excursion, allows the motion to descend along that region and return.

As $\lambda \downarrow 1/4$, the rescaled excursion time diverges and the factor $e^{\tau /2}$ produces arbitrarily large excursions. At the critical value the matrix is a nontrivial Jordan block, and the limiting formula isFor the small outgoing initial data this is positive and unbounded: the unstable manifold no longer turns back. Below $1/4$ the two eigenvalues are real and positive, so there is again no oscillatory return. This quadratic escape desingularization explains loss of the periodic orbit at infinity, together with the unstable-manifold escape, as $\lambda$ decreases through $1/4$. It is not a finite homoclinic bifurcation to the origin. The divergent $\tau$ travel time must not be identified with the physical period without integrating $dt=d\tau /u$; the time change becomes singular near $u=0$.

A global bifurcation joins two mechanisms: a long passage near a hyperbolic equilibrium point and a finite excursion returning the unstable manifold to its neighborhood. Eigenvalue ratios control the local passage, while the geometry of the global return decides reinjection, folding and trapping. In the real symmetric case there are two returning branches and a power-law cusp. In the saddle-focus case rotation adds infinitely many oscillations to that cusp. The following maps display both mechanisms explicitly.

For the Lorenz configuration, enter on $z=h$ at $(x,y,h)$ and leave on $x=sh$, where $s=\mathrm{sgn}(x)$. Solving the linear flow gives a flight timePut $\gamma =-\hat{\lambda }/{\lambda }_{+}>\delta =-{\lambda }_{-}/{\lambda }_{+}$. The outgoing coordinates areThe smooth global map from each outgoing section to the incoming section has, to leading order,up to the choice of which unstable branch carries the sign of $\nu$. Reflection symmetry makes the offsets opposite and the leading coefficients equal. Since the stable transverse coordinate is multiplied by the higher power $|x{|}^{\gamma }$, it contributes only subleading terms to the leading scalar map. Absorbing powers of $h$ into a positive constant gives the Lorenz power return mapThe line $x=0$ lies on the stable manifold and has no finite return; the discontinuity is therefore a genuine missing return, not an arbitrary value to assign at zero.

If $\delta >1$, the derivative $A\delta |x{|}^{\delta -1}$ tends to zero. For $\mu <0$, there are two attracting fixed points, $x=\pm s$ with $s=-\mu +A{s}^{\delta }\simeq -\mu$. They describe two symmetry-related one-lobe attracting flow periodic orbits. For $\mu >0$, the orbit alternates between $s$ and $-s$, where $s=\mu -A{s}^{\delta }\simeq \mu$; its multiplier is $(A\delta s^{\delta -1}{)}^2<1$. This represents one attracting flow periodic orbit traversing both lobes. At $\mu =0$ the two homoclinic loops meet: this is the contracting gluing scenario, with diverging flow period at the connection.

If $\delta <1$, the local derivative is unbounded, and the small periodic orbits are repelling in the expanding return direction. For $\mu >0$, the small positive fixed point has $s\simeq (\mu /A{)}^{1/\delta }$, with a multiplier tending to infinity; there is a symmetry-related negative fixed point as well. More complex motion follows from the two branches, not merely from a large derivative at one fixed point. Choose $I=[-c\mu ,c\mu ]$ with $0<c<1$. The inverse branches areFor sufficiently small positive $\mu$, both lie inside $I$, their images are disjoint, and their derivatives are $O({\mu }^{1/\delta -1})\to 0$. Nested inverse images for every infinite sign sequence therefore define a Cantor invariant set. Iterating the forward map shifts the sequence. This constructs periodic itineraries of every symbolic period and nonperiodic itineraries; transverse contraction in the full return turns the expanding two-strip mechanism into a horseshoe.

This is the local chaotic Lorenz mechanism. A robust attracting Lorenz set additionally needs a global trapping region and the appropriate return properties. The scalar asymptote and $\delta <1$ alone do not prove that every nearby orbit is attracted to one chaotic set; periodic windows, escaping orbits and other global returns must be distinguished. Symmetry is crucial to obtaining the two homoclinic branches together in a single parameter variation.