On time steps and gravity force calculation

From: Aldo Alberto Batta Marquez <abatta_at_astro.unam.mx>
Date: Mon, 18 Nov 2013 21:03:55 -0600

Hi everyone,

I've been working with the collapse of rotating polytropic envelopes onto a
sink particle (BH) originally set at the centre of the spherical
distribution. I'm studying the effect of cooling and self gravity of the
accretion disk in the dynamics of the system (gas and BH).

I've noticed that BH dynamic's (whose mass is comparable to the gas mass)
is calculated using a huge time step compared to the integration time used
in the system (determined mainly by high density SPH particles lying close
to the BH), which translates on sudden kicks on the BH that mess with
momentum conservation. The large time step is due to the small
gravitational accelerations acting on the BH (because of the disk's
symmetry). Up to now I've chosen a gravitational softening for the BH
smaller than it's accretion radius and I've implemented a Paczynski-Wiita
potential for the BH.

I've tried to fix this issue by choosing a smaller softening length for the
BH but that can result in rather slow simulations when the resulting time
step is much smaller than the minimum time step for SPH particles.
Apparently this solution is not enough because linear momentum is still not
being conserved as it should. Only simulations with really thick disks
(with highly unefficient cooling), conserve momentum in a proper way. I've
noticed that increasing the accuracy on the gravitational force, by
choosing smaller node opening parameters, improves linear momentum
conservation but only up to some point. At this point I'm not sure if this
is a problem with time integration (time step) or with the gravity force
calculation.


Do you think an adaptive gravitational softening would improve things?
Has anyone implemented it in Gadget2?
Any ideas are welcomed.

I also tried to set the BH time step as the minimum time step obtained for
all particles, but apparently I was missing something and couldn't get that
to work properly.

I'm looking forward to any feedback.

Thanks

Aldo Batta


-- 
Aldo Alberto Batta Márquez
Instituto de Astronomia
Universidad Nacional Autonoma de Mexico (UNAM)
http://www.astroscu.unam.mx
Received on 2013-11-19 08:02:35

This archive was generated by hypermail 2.3.0 : 2023-01-10 10:01:32 CET