REBOUNDx not correctly applying some forces with TRACE

[tigerchenlu98]

@dtamayo and I were looking into this during the workshop. This is what happens if you run the Migration notebook example in REBOUNDx with TRACE:

Part of what's happening here is at the point of disagreement, TRACE starts flagging the inner particle as undergoing pericenter approach. There's currently a bug in integrator.c where additional forces are not applied during pericenter flags. We can fix this issue by adding in a check for r->ri_trace.mode == 3 here.

rebound/src/integrator.c

Line 237 in 68944aa

if (r->additional_forces && (r->integrator != REB_INTEGRATOR_MERCURIUS || r->ri_mercurius.mode==0) && (r->integrator != REB_INTEGRATOR_TRACE || r->ri_trace.mode==0)){

Unfortunately, that's not the whole issue -- this makes things much better, but we still see disagreement.

[hannorein]

Is the screenshot before or after the fix?

[tigerchenlu98]

This is before the fix. The PR I just put in (#834) adds in the extra check for the FULL PERI modes. This is what it looks like after:

Worth noting: this works perfectly fine for PARTIAL_BS and FULL_IAS15. The only mode it doesn't work for is FULL_BS the default.

[hannorein]

Hm. I don't have an immediate idea...