Mesocite CGMD with MS Martini 3 — does bead-type Radius affect non-bonded energy? Dynamics unstable

I'm running CG-MD in Mesocite (MS 2025 SP1) with MS Martini 3: SDS (1 head + 3 tail beads), Na, and W beads, 250 Å cell, ~130,000 beads. Bead types defined by hand; forcefield types assigned via Properties Explorer.

  1. Does the Bead Types Radius value affect the non-bonded energy in a CGMD calculation, or only packing and display? The MS Martini 2 bilayer tutorial specifies 2.35 Å; the MS Martini 3 tutorial specifies 4.7 Å. Which applies to MS Martini 3?

  2. Should Radius differ for tiny/small/regular Martini 3 bead sizes?

  3. Geometry optimisation gives vdW energy ~10¹⁸–10²¹ kcal/mol and forces ~10¹⁶–10²⁰ kcal/mol/Å at the initial structure; valence terms are normal. Dynamics then fails on energy deviation. Overlaps, wrong Radius, or missing forcefield terms?

  4. Must a system-specific forcefield always be generated with the MS Martini 3 Forcefield Generator and merged with the solvent forcefield, or can the stock MS Martini 3 be used directly for a bonded molecule?