mathematics//dynamical systems//nonlinear dynamics//bifurcation
A bifurcation is a change in the kind of behaviour a dynamical system shows when one of its parameters crosses a threshold: an equilibrium appears or vanishes, changes stability, or gives way to a sustained oscillation. For an engineer it is the mathematics of operating limits. Below the threshold the machine runs smoothly and small changes of the parameter produce small changes of behaviour; at the threshold that continuity breaks, and no amount of fine tuning on the safe side warns of what lies beyond.
A bifurcation is a change in the kind of behaviour a dynamical system shows when one of its parameters crosses a threshold: an equilibrium appears or vanishes, changes stability, or gives way to a sustained oscillation. For an engineer it is the mathematics of operating limits. Below the threshold the machine runs smoothly and small changes of the parameter produce small changes of behaviour; at the threshold that continuity breaks, and no amount of fine tuning on the safe side warns of what lies beyond.
The smallest example shows a resting state disappearing. Take
x˙=r+x2.\dot x=r+x^2 .x˙=r+x2.
For r<0r<0r<0 there are two equilibria at x=±−rx=\pm\sqrt{-r}x=±−r, the lower one stable and the upper one unstable. As rrr rises they approach each other, meet at r=0r=0r=0 and annihilate; for r>0r>0r>0 there is no equilibrium left and xxx runs away. That is a saddle-node bifurcation, and its industrial face is compressor surge: throttle a centrifugal compressor below a certain flow and stable operation ceases to exist, replaced by a violent oscillation of flow and pressure that damages seals, bearings and blades if it persists. Compressor maps draw that boundary as the surge line, and anti-surge control opens a recycle valve to keep the operating point a margin away from it.
The other common kind is the birth of an oscillation. When a complex pair of eigenvalues crosses the imaginary axis as a parameter moves (a loop gain raised, a damping lowered), a decaying spiral turns into a growing one and, in a nonlinear system, usually settles onto a limit cycle: a Hopf bifurcation, visible in the trace-determinant plane as the trace crossing zero.
A bifurcation announces itself in the linearization. At the threshold an eigenvalue of the Jacobian reaches zero (an equilibrium about to vanish) or a pair reaches the imaginary axis (an oscillation about to start), so the restoring force fades and the system recovers more and more slowly from small disturbances. That slowing down is used as an early warning in practice, with care: noise and drift can mimic it.
Bringing the parameter back does not always undo the jump. After surge or a saddle-node jump the system often stays on the other branch until the parameter goes well past the original threshold, a hysteresis that operators learn the hard way (nonlinear dynamics).
The linear analysis says where the threshold is and nothing about what lies beyond it; that part needs the nonlinear model or a simulation. In collective systems the same abrupt change of order with a parameter is called a phase transition (emergence).