Tracing seismic rays - possible intersection

[FermatPell]

This problem is about seismic wave propagation in a non-homogeneous layer over a halfspace. I'm not asking you to solve anything, I've already solved the problem both algebraically and in Matlab. However, the graph that I've gotten mildy surprises me. According to the graph, the seismic rays intersect after returning from the half space.

Briefly, I was asked to trace the seismic rays originating from a point on the surface.

The velocity in the layer is a function of depth v=v0+k1z where k1=0.025
The velocity in the halfspace is a function of depth v=v0+k1H+k2(z-H) k2=0.01

where H=20km is depth of the layer

NOTE: The y-axis is inverted, z increases with depth.

 

[mfb]

Different refractive indices can do this with light (it is the basic idea of optical lenses), I am not surprised to see a similar effect for seismic waves.

Related: Northridge Earthquake Damage Caused by Geologic Focusing of Seismic Waves