Divergence detected in AMG solver in VOF-DPM simulation
 

I am simulating flow boiling in microchannel with nanofluids considering nanoparticles as separate phases by using the Discrete Phase model. The simulation converges for both constant and variable properties of water when DPM runs without activating the forces such as Thermophoretic Force, Saffman Lift Force, Virtual Mass Force, and Pressure Gradient Force, however, it is theoretically incorrect not to include them in my model. But when including them, divergence was detected in the AMG solver with divergences in all x-, y-, and z-momentum equations for both constant and variable properties of water, more frequently for variable properties.
I have tried
1. reducing the initial time step in Adaptive time settings to 1e-06, and
2. residuals of velocity to 1e-09, but the error occurs before completing the given time steps.
I would be trying
1. Changing momentum discretization from second-order upwind to QUICK
2. Reducing under-relaxation factors
3. Activating Non-iterative time advancement (NITA)
4. Warped Face Gradient Correction
5. High-order term Relaxation
6. If everything fails, improve the mesh.
If anyone has experience on this topic, please suggest me something helpful.