---
title: "(k1,k2)-Mode Turing-Hopf Bifurcation"
url: https://www.emergentmind.com/topics/k1-k2-mode-turing-hopf-bifurcation
type: topic
---

# (k1,k2)-Mode Turing-Hopf Bifurcation

A \((k_1,k_2)\)-mode Turing-Hopf bifurcation is a codimension-two Hopf-steady-state singularity in which a simple zero eigenvalue is associated with one spatial mode \(k_1\), while a simple conjugate pair \(\pm i\omega_0\) is associated with another spatial mode \(k_2\). In the formulation that explicitly uses this terminology, the bifurcation is called a \((k_1,k_2)\)-mode Turing-Hopf bifurcation when the steady branch is spatially inhomogeneous, i.e. \(k_1\neq 0\) [1802.10286]. The resulting local dynamics organize the interaction of diffusion-driven pattern formation and temporal oscillation, and may generate patterned steady states, homogeneous or inhomogeneous periodic orbits, and mixed spatiotemporal states. A terminological complication is that some continuation literature uses “Turing-Hopf” or “wave” for a Hopf crossing at a single nonzero spatial mode, whereas the \((k_1,k_2)\)-mode usage refers to simultaneous steady and oscillatory criticality on possibly different spatial modes [1604.01600].

## 1. Definition and terminological scope

In the explicit normal-form framework for delayed reaction-diffusion systems, the underlying codimension-two hypothesis is that for mode \(k_1\) the characteristic equation has a simple real eigenvalue \(\gamma(\alpha)\) with \(\gamma(0)=0\), while for mode \(k_2\) it has a simple pair \(\nu(\alpha)\pm i\omega(\alpha)\) with \(\nu(0)=0\) and \(\omega(0)=\omega_0>0\); all other spectral branches stay off the imaginary axis [1802.10286]. Under homogeneous Neumann boundary conditions, \(k=0\) denotes the homogeneous spatial mode, while \(k\neq 0\) denotes a spatially inhomogeneous Laplacian mode. In that language, \(k_1\) is the steady or Turing mode, and \(k_2\) is the oscillatory or Hopf mode [1802.10286].

The phrase is not used uniformly across the literature. In the pde2path study of the extended Brusselator, a Hopf-type crossing at nonzero spatial mode is called “Turing-Hopf (aka wave),” so the relevant object is a single oscillatory finite-\(k\) instability rather than a Hopf-zero codimension-two interaction [1604.01600]. This distinction matters because the \((k_1,k_2)\)-mode formulation is specifically a two-mode codimension-two concept.

A second recurrent misconception is that the Hopf component must be spatially homogeneous. The general \((k_1,k_2)\)-mode framework does not require that: \(k_2\) may be \(0\) or nonzero, and the corresponding critical eigenfunctions may both be spatially inhomogeneous [1802.10286]. The same point appears in the non-instantaneous Kerr cavity, where the Hopf instability may occur at \(k_H=0\) or at \(k_H\neq 0\), while the Turing instability is selected by a finite critical wave number \(k_T\) [1611.04123].

## 2. Spectral formulation and modal decomposition

The general delayed reaction-diffusion formulation used for \((k_1,k_2)\)-mode analysis is
\[
\dot{u}(t)=D(\alpha)\Delta u(t)+L(\alpha)u_t+G(u_t,\alpha),
\]
posed on a bounded domain with homogeneous Neumann or Dirichlet boundary conditions [1802.10286]. After separation in Laplacian eigenfunctions \(\{\beta_k\}\), the characteristic equation becomes modewise:
\[
\Delta_k(\lambda)=\lambda I-\mu_kD_0-L_0(e^{\lambda\cdot}I),\qquad \det\Delta_k(\lambda)=0.
\]
The codimension-two point is therefore located by solving one steady condition at mode \(k_1\) and one Hopf condition at mode \(k_2\) [1802.10286].

In one-dimensional Neumann problems, the spatial basis is typically \(\beta_k(x)=\cos(k\pi x)\), so the mode number has a direct wavelength interpretation. In the delayed Schnakenberg system, the \(k\)-th modal characteristic equation is
\[
D_k(\lambda,\tau,\varepsilon)=\lambda^2+p_k\lambda+r_k+(s_k\lambda+q_k)e^{-\lambda\tau}=0,
\]
and a \(k\)-mode Turing bifurcation means that \(D_k\) has a simple zero root, whereas a \(k\)-mode Hopf bifurcation means that \(D_k\) has a simple pair of purely imaginary roots [1803.00164]. The paper then states that the system undergoes a \((k_1,k_2)\)-mode Turing-Hopf bifurcation at
\[
\varepsilon=\varepsilon_*(k_1,d),\qquad \tau=\tau_{k_2},
\]
with \(k_1\) the Turing index and \(k_2\) the Hopf index [1803.00164].

A closely related linear-algebraic formulation appears in the diffusive epidemic model. There the Neumann decomposition on \([0,\ell\pi]\) yields modal Jacobians
\[
J_k=J_0-\frac{k^2}{\ell^2}\begin{bmatrix}r_1&0\\0&r_2\end{bmatrix},
\]
with modal characteristic polynomial
\[
P_k(\lambda)=\lambda^2-\mathcal T_k\lambda+\mathcal D_k.
\]
The Turing condition is \(\mathcal D_{k_1}=0\), and the Hopf condition is \(\mathcal T_{k_2}=0\) with \(\mathcal D_{k_2}>0\). The paper’s theorem then locates a \((k_1,k_2)\)-mode Turing-Hopf point at
\[
(r_1,\beta)=\bigl(r_1^{(k_1)},\beta_H^{(k_2)}\bigr),
\]
subject to \(0\le k_2<k_1\le\bar k\) and condition \((\mathbb H)\) [2509.09000].

## 3. Center-manifold reduction and normal forms

At a \((k_1,k_2)\)-mode Turing-Hopf point, the center space is three-dimensional over \(\mathbb R\): one real amplitude for the steady mode and one complex amplitude for the Hopf mode. In the general PFDE theory this yields
\[
v(t)=\phi_1 z_1(t)\beta_{k_1}+(\phi_2 z_2(t)+\bar\phi_2\bar z_2(t))\beta_{k_2}+y(t),
\]
and the reduced normal form
\[
\dot z = Bz+\frac12 g_2^1(z,0,\alpha)+\frac1{3!}g_3^1(z,0,\alpha)+\text{h.o.t.},
\qquad
B=\operatorname{diag}(0,i\omega_0,-i\omega_0),
\]
or, more explicitly,
\[
\dot z_1=a_1(\alpha)z_1+\cdots,\qquad
\dot z_2=i\omega_0 z_2+b_2(\alpha)z_2+\cdots
\]
with quadratic and cubic coupling terms such as \(z_2\bar z_2\), \(z_1z_2\), \(z_1z_2\bar z_2\), and \(z_2^2\bar z_2\) [1802.10286].

The mode dependence is not merely bookkeeping. The coefficients depend on Fréchet derivatives of the original PDE or PFDE and on overlap integrals such as
\[
\langle \beta_{k_2}^2,\beta_{k_1}\rangle,\qquad
\langle \beta_{k_1}\beta_{k_2},\beta_{k_2}\rangle,
\]
which determine whether quadratic couplings vanish. The theory distinguishes five mode configurations, including \(k_2=0,\ k_1\neq 0\) and \(k_1\neq k_2,\ k_1,k_2\neq 0\), and therefore accommodates both homogeneous and inhomogeneous Hopf components [1802.10286].

Two reduced amplitude systems recur. In the Hopf-transcritical case, the planar reduction is written in \((r,z)\)-coordinates with \(r\) the Hopf amplitude and \(z\) the steady amplitude [1802.10286]. In the Hopf-pitchfork case, which is especially common when the Turing branch is symmetry-related, the reduction takes the form
\[
\dot r=r(\varepsilon_1(\alpha)+r^2+b_0 z^2),\qquad
\dot z=z(\varepsilon_2(\alpha)+c_0 r^2+d_0 z^2),
\]
and its equilibria correspond to the base equilibrium, a pure Hopf branch, a pure Turing branch, and a mixed branch with both amplitudes nonzero [1802.10286]. In the delayed Schnakenberg system the same planar form is obtained for \(k_1\neq 0,\ k_2=0\), and explicit examples are worked out for \((k_1,k_2)=(1,0)\) and \((3,0)\) [1803.00164].

The attractor interpretation is standard across these normal forms. Pure \(z\)-branches correspond to spatially inhomogeneous steady states; pure \(r\)-branches correspond to periodic solutions; mixed equilibria correspond to spatially inhomogeneous periodic orbits. In the delayed reaction-diffusion framework of the two-component Turing-Hopf analysis, periodic orbits of the planar amplitude system can further correspond to spatially inhomogeneous quasi-periodic solutions of the original PDE [1710.10411].

## 4. Pattern classes: steady, periodic, localized, and front-pinned states

The local codimension-two reduction organizes several distinct pattern classes. In the non-instantaneous Kerr cavity, the linear problem yields a finite-wavelength Turing instability together with a Hopf instability, and the numerics show a transition from a stationary Turing branch to a stable mixed-mode Turing-Hopf branch. The same model also supports localized mixed-mode solutions: stationary localized structures begin to pulse in time, producing time-periodic states that remain localized in space [1611.04123].

Two-dimensional effects introduce additional structure that is not captured by one-dimensional front interaction alone. In the study of forced oscillatory media near a Hopf-Turing point, the authors derive a two-dimensional amplitude description with one Hopf mode and two Turing modes \(T_\parallel\) and \(T_\perp\), associated with wave vectors along \(x\) and \(y\), respectively. This produces comb-like localized Turing states embedded in oscillatory Hopf backgrounds, including localized stripes inside \(\pi\)-shifted Hopf oscillations and in spiral cores [1702.08556]. The key mechanism is a local dynamics pinning of Hopf fronts, which allows the Turing mode perpendicular to the front to become locally selected and pinned; the resulting states occur outside the one-dimensional flip-flop or homoclinic-snaking picture and extend frequency locking beyond the classical \(2{:}1\) resonance region [1702.08556].

The same literature therefore shows that the \((k_1,k_2)\)-mode viewpoint is not restricted to spatially uniform oscillatory backgrounds or to one-dimensional pattern branches. A plausible implication is that the codimension-two mechanism is best understood as a mode-selection problem with geometry-dependent coupling coefficients rather than as a single universal front scenario.

## 5. Symmetry, geometry, and multimode generalizations

On a disk, the scalar wave number \(k\) is replaced by a pair of mode labels \((n,m)\), where \(n\) is the azimuthal index and \(m\) the radial index. Nonradial modes with \(n\ge 1\) are symmetry-degenerate, so a Turing-Hopf point may involve equivariant Turing or equivariant Hopf branches rather than simple scalar modes. The disk analysis distinguishes three types: ET-H, T-EH, and ET-EH. Their center manifolds have dimensions \(4\), \(5\), and \(6\), respectively, and the associated patterns include breathing, standing wave-like, and rotating wave-like states [2309.06133]. A resonance condition
\[
n_{T_3}=2n_{H_3}
\]
changes the normal form by introducing additional phase-coupling terms, so the disk problem provides a symmetry-based analogue of wave-number resonance in standard mode-interaction theory [2309.06133].

A different extension appears in the codimension-three Hopf-Turing-Turing framework for two-component reaction-diffusion systems on an interval. There a \(0\)-mode Hopf instability interacts simultaneously with two adjacent Turing modes \(m\) and \(m+1\), giving a \(0:m:m+1\) organizing center. For the explicitly derived \(0:1:2\) case the reduced normal form contains resonance terms of the form \(\tilde B z_1 z_2\) and \(\tilde C z_1^2\); for \(m\ge 2\) these resonance terms drop out in the stated \(0:m:m+1\) normal form [2311.07045]. The dynamics of this extension include mixed spatiotemporal oscillations, invariant \(2\)-tori, possible invariant \(3\)-tori, heteroclinic structures, and numerically suggested chaos [2311.07045]. Although this is not a codimension-two \((k_1,k_2)\)-mode Turing-Hopf bifurcation in the strict sense, it shows how nearby Turing modes can enrich the standard two-mode picture.

## 6. Computation, continuation, and representative models

The computational side of \((k_1,k_2)\)-mode problems is developed most explicitly in pde2path. After spatial discretization of
\[
\partial_t u=-G(u,\lambda),\qquad M\dot u(t)=-G(u(t),\lambda),
\]
Hopf points are detected from the generalized eigenvalue problem
\[
\mu M\phi=\partial_u G\,\phi,
\]
localized by bisection, and continued as periodic orbits with phase and continuation conditions; Floquet multipliers are computed either by direct monodromy products or by periodic Schur decomposition [1604.01600]. In the two-dimensional Brusselator example on
\[
\Omega=(-l_x,l_x)\times(-l_y,l_y),\qquad l_x=\pi/2,\quad l_y=\pi/

Source: https://www.emergentmind.com/topics/k1-k2-mode-turing-hopf-bifurcation