Mehmet Emre Tiryaki
Back to Notes

Systems of ODEs and Phase Portraits

Linear Systems

For x′=Axx' = Ax, eigenpairs of AA give the solutions.

  • Distinct real eigenvalues rir_i with eigenvectors viv_i: x=∑icieritvix = \sum_i c_i e^{r_i t} v_i.
  • Complex pair r=λ±iμr = \lambda \pm i\mu with eigenvector v=a+ibv = a + ib: the real solutions are eλt(acos⁡(μt)−bsin⁡(μt))e^{\lambda t}\big(a\cos(\mu t) - b\sin(\mu t)\big) and eλt(asin⁡(μt)+bcos⁡(μt))e^{\lambda t}\big(a\sin(\mu t) + b\cos(\mu t)\big).
  • Repeated eigenvalue rr with one eigenvector vv: find a generalized eigenvector (A−rI)w=v(A - rI)w = v. Then x=c1ertv+c2(tertv+ertw)x = c_1 e^{rt} v + c_2 \big(t e^{rt} v + e^{rt} w\big).

See also eigenvalues and diagonalization.

Phase Portraits of 2-by-2 Systems

Get the eigenvalues and eigenvectors and draw the eigenvector lines. As t→∞t \to \infty, trajectories line up with the eigenvector of the larger eigenvalue.

  • λ1,λ2<0\lambda_1, \lambda_2 < 0 distinct: nodal sink, asymptotically stable.
  • λ1,λ2>0\lambda_1, \lambda_2 > 0 distinct: nodal source, unstable.
  • λ1λ2<0\lambda_1 \lambda_2 < 0: saddle point, unstable.
  • complex λ±iμ\lambda \pm i\mu: spiral point, asymptotically stable if λ<0\lambda < 0, unstable if λ>0\lambda > 0. The direction of rotation follows from evaluating x′x' at a point, e.g. on an axis.
  • purely imaginary ±iμ\pm i\mu: center, closed elliptical orbits around the origin, stable but not asymptotically stable.
  • repeated λ\lambda: a proper (star) node if A=λIA = \lambda I, an improper node otherwise. Asymptotically stable if λ<0\lambda < 0, unstable if λ>0\lambda > 0.

Trajectories

For x′=F(x,y)x' = F(x,y), y′=G(x,y)y' = G(x,y), the trajectories satisfy

dydx=G(x,y)F(x,y)\frac{dy}{dx} = \frac{G(x,y)}{F(x,y)}

and solving this, when possible, gives a conserved quantity H(x,y)=cH(x,y) = c whose level sets are the trajectories.

Periodic Solutions and Limit Cycles

In polar coordinates x=rcos⁡θx = r\cos\theta, y=rsin⁡θy = r\sin\theta, r2=x2+y2r^2 = x^2 + y^2,

rdrdt=xdxdt+ydydt,r2dθdt=xdydt−ydxdtr\frac{dr}{dt} = x\frac{dx}{dt} + y\frac{dy}{dt}, \qquad r^2 \frac{d\theta}{dt} = x\frac{dy}{dt} - y\frac{dx}{dt}

A limit cycle is an isolated closed trajectory. It is asymptotically stable if trajectories on both sides approach it, unstable if trajectories on both sides move away, and semistable if they approach on one side and move away on the other.

Bendixson's criterion If Fx+GyF_x + G_y has the same sign everywhere in a simply connected region DD, there is no closed trajectory lying entirely in DD.