Spurious energy balance increase at the end of dynamic explicit analysis

Hello fellow engineers!

This is my first time writing here, although I regularly consult some useful forum threads.
I face a quite strange issue with Abaqus Explicit. I am running a linear dynamic plane-strain analysis of a soil domain using CPE4R elements (1mx1m) and CINPE4 (infinite) elements to model the boundary conditions as you can see in the attached picture. A real earthquake record in terms of horizontal force is used as uniform input excitation at the base nodes (PGA=0.24g).



I use no material damping so far and I have used no fixed boundary conditions. The reason I'm running this analysis is to check the efficiency of the infinite elements in absorbing the seismic wave energy.
The problem arises when I plot the energy histories. As you can see, the dissipated energy accounts for almost the entire external work done by the seismic forces. However, the last three seconds consist of unexpected high frequency oscillations with increasing amplitude of the kinetic and strain energy, and also the artificial energy becomes noticeable (~10% of strain energy). As a result, the total energy balance starts to deviate from zero and becomes an increasing function of time. This obviously violates the conservation of energy law.




I've tried enhanced hourglass control to solve the issue but with no results.
Do you have any idea about what may cause this issue?
I would be grateful for any insight from more experienced people in Abaqus Explicit .