<html><head>
<meta http-equiv="Content-Type" content="text/html; charset=utf-8">
</head>
<body>
<tt><font size="+1">Dear Antoine -<br>
<br>
You do not want to change the CFl factor, but you want to
control the source term actions better. In the case of
catastrophic cooling, the local cooling time becomes shorter
than the dynamical timescale. This is conceptually identical to
the nuclear burning/heating timescale issue.<br>
<br>
If there is no timestep limiter due to cooling, implement one
following the approach used in Burn_computeDt (or whatever the
module name is). You should be able to stabilize calculations by
limiting the source term contribution to about 10% of the
internal energy.<br>
<br>
It would be nice to hear if that helped.<br>
<br>
(BTW, I do not recommend using artificial viscosity. This
approach is largely obsolete these days and helpful only in very
particular situations.)<br>
<br>
Tomek<br>
--</font></tt><br>
<div class="moz-cite-prefix">On 2020-10-22 10:50, Antoine Gintrand
wrote:<br>
</div>
<blockquote type="cite" cite="mid:CAO_-DzROTMPdj9tOG352FREH_83YX_QpKejLOOP-5t9i-V-arA@mail.gmail.com">
<div dir="ltr">
<div>Dear flash users,</div>
<div><br>
</div>
<div>I have an issue that has been troubling me for a long time.
<br>
</div>
<div>I do the simulation of a strong shock in the interstellar
medium that cools <br>
</div>
<div>with a source term according to <span class="gmail-aCOpRe"><span>the
cooling tables of <em>Sutherland</em> & <em>Dopita</em>
(1993).</span></span></div>
<div><span class="gmail-aCOpRe"><span><br>
</span></span></div>
<div><span class="gmail-aCOpRe"><span>the problem comes when the
cooling becomes strong and I receive the following message
:</span></span></div>
<div><span class="gmail-aCOpRe"><span><br>
</span></span></div>
<div><span class="gmail-aCOpRe"><span><span class="gmail-aCOpRe"><span>
n t dt
( x, y, z) |
dt_hydro dt_Cool CFL </span></span></span></span></div>
<div><span class="gmail-aCOpRe"><span> <br>
</span></span></div>
<div><span class="gmail-aCOpRe"><span> 253369 8.7315E+09
2.1942E+03 ( 1.773E+16, -8.164E+15, 0.00 ) |
2.194E+03 4.696E+04 0.1000000<br>
253370 8.7315E+09 2.1958E+03 ( 1.773E+16, -8.164E+15,
0.00 ) | 2.196E+03 4.687E+04 0.1000000<br>
253371 8.7315E+09 3.0850E+01 ( 1.864E+16, -7.383E+15,
0.00 ) | 3.085E+01 4.678E+04 0.1000000<br>
253372 8.7315E+09 2.4375E-01 ( 1.864E+16, -7.383E+15,
0.00 ) | 2.438E-01 4.678E+04 0.1000000<br>
253373 8.7315E+09 2.4375E-01 ( 1.864E+16, -7.383E+15,
0.00 ) | 2.438E-01 4.678E+04 0.1000000<br>
253374 8.7315E+09 2.4375E-01 ( 1.864E+16, -7.383E+15,
0.00 ) | 2.438E-01 4.678E+04 0.1000000<br>
253375 8.7315E+09 1.0000E-10 ( 1.864E+16, -7.383E+15,
0.00 ) | 1.745E-16 2.569E+02 0.1000000<br>
253376 8.7315E+09 1.0000E-10 ( 1.864E+16, -7.383E+15,
0.00 ) | 1.289E-60 2.569E+02 0.1000000<br>
dtCheck= 0.0000000000000000 <br>
DRIVER_ABORT: [Hydro]: Computed dt is not positive!
Aborting!</span></span></div>
<div><span class="gmail-aCOpRe"><span><br>
</span></span></div>
<div><span class="gmail-aCOpRe"><span>and I can see that the
pressure drops dramatically so the movement of matter
accelerates and the compression increases drastically.</span></span></div>
<div><span class="gmail-aCOpRe"><span> I believe that this is
maybe due to the catastrophic instability of Falle (1981).</span></span></div>
<div><span class="gmail-aCOpRe"><span>Indeed, the cooling time
becomes smaller than the time needed for the sound wave to
travel the cooling region <br>
</span></span></div>
<div><span class="gmail-aCOpRe"><span>and the pressure could not
be stable anymore so it begins to fall inside this region.
<br>
</span></span></div>
<div><span class="gmail-aCOpRe"><span>The problem is that during
the process, the region becomes poorly resolved and the
gradient of pressure density and temperature are very
high.<br>
</span></span></div>
<div><span class="gmail-aCOpRe"><span>Also, the problem is that
when I increase the resolution, the code fails at earlier
times because the cooling and the compression are even
more efficient. <br>
</span></span></div>
<div><span class="gmail-aCOpRe"><span>When I chose a lower cfl
coefficient it would just delay the fail message.<br>
</span></span></div>
<div><span class="gmail-aCOpRe"><span><br>
</span></span></div>
<div><span class="gmail-aCOpRe"><span>One idea that seems to
work would be to use some artificial viscosity but I do
not know if this would be consistent and the physical
meaning of this.<br>
</span></span></div>
<div><span class="gmail-aCOpRe"><span>Indeed when I increase the
coefficient of artificial viscosity (cvisc), at some
value of the parameter (cvisc=1.5), the simulation will
continue to run but the time step drops a lot.<br>
</span></span></div>
<div><span class="gmail-aCOpRe"><span><br>
</span></span></div>
<div><span class="gmail-aCOpRe"><span>Is anyone have some ideas
to solve this issue,</span></span></div>
<div><span class="gmail-aCOpRe"><span><br>
</span></span></div>
<div><span class="gmail-aCOpRe"><span>Thank you for your help,</span></span></div>
<div><span class="gmail-aCOpRe"><span><br>
</span></span></div>
<div><span class="gmail-aCOpRe"><span>Best regards,</span></span></div>
<div><span class="gmail-aCOpRe"><span><br>
</span></span></div>
<div><span class="gmail-aCOpRe"><span>Antoine Gintrand.<br>
</span></span></div>
</div>
</blockquote>
<br>
</body>
</html>