It cannot happen in a continuous time 2D system, simply due to uniqueness of ODE property. At least (2+1)-D is needed, i.e. this phenomenon can be seen in 2D maps derived from taking time-T sections of a time-dependent 2D ODE system.
A common example is forced Duffing system.
A good resource is this document by Wiggins: http://docenti.lett.unisi.it/files/4/2/13/1/Wiggins_CHAOS.pdf
Also take a look at Wiggin's "Global bifurcations and chaos".
Other keywords to search for : "Melnikov method", "conley-moser conditions".