[mesa-users] Network types and other details

[Frank Timmes]

p.s. - happy to correct any typos you find or clarify any mysterious steps.

fxt



> On Feb 5, 2017, at 12:33 PM, Frank Timmes <fxt44 at mac.com> wrote:
> 
>> The benefit is that you can get away with far fewer isotopes 
> 
> the disadvantage, unseen by most, is the network evolution equations
> and their jacobian get significantly more complicated. no free lunch ;)
> 
>>> Is there an additional resource I can refer to on the approx networks and compound reactions?
> 
> attached is an unpublished pdf that reveals the sorcery behind equilibrium reactions.
> these techniques are used in the hardwired reaction networks aprox* and friends.
> 
> fxt
> 
> 
> 
> 
>> On Feb 5, 2017, at 12:01 PM, Robert Farmer <rjfarmer at asu.edu> wrote:
>> 
>>> So, assuming I had a model that started with a specific abundance profile, but I used a soft network that didn't include the same isotopes, the missing isotopes would not be treated. Would they retain their original absolute mass, or mass fraction?
>> 
>> They will be changed according to the rules mesa has for changing the network, see section 6.2.2 in the mesa3 paper, also see the star/private/adjust_xyz.f90 set_new_abundances function, isotopes not in the new net will be removed, but other similar isotopes will be re-scaled to compensate. Where similar is splits things up into neutron, hydrogen, helium, carbon, nitrogen, oxygen and others. Isotopes not in the old but in the new network get added based on solar abundances.
>> 
>>> Is there an additional resource I can refer to on the approx networks and compound reactions?
>> Maybe try and read through whats happening in net/private/net_approx21_procs.inc which is where the approx network logic is, though it can be hard to follow.
>> 
>> Rob
>> 
>> 
>> 
>> On Sun, Feb 5, 2017 at 11:39 AM, amber lauer <alauer1 at lsu.edu> wrote:
>> Thanks. I did check the logic of the code since I know it is preferred that we dig into it ourselves, unfortunately I still wasn't sure, although I was pretty sure.
>> 
>> So, assuming I had a model that started with a specific abundance profile, but I used a soft network that didn't include the same isotopes, the missing isotopes would not be treated. Would they retain their original absolute mass, or mass fraction?
>> 
>> Is there an additional resource I can refer to on the approx networks and compound reactions?
>> 
>> 
>> On Feb 5, 2017 1:20 PM, "Robert Farmer" <rjfarmer at asu.edu> wrote:
>> Hi
>> 
>> 1) Yes, though you can also specify the reactions if you want to do it manually but just setting the isotopes is easier. There other soft nets than the mesa_, the better way to think of it anything that is not an approx net is a soft network
>> 
>> 2) The basic idea is you go h->he->c12 then go up by adding alpha's following the alpha chain reactions. There are then special reactions for some equilibrium reactions to avoid evolving the intermediate isotope. The benefit is that you can get away with far fewer isotopes and reactions to evolve a star with than a corresponding soft net, less isotopes makes mesa faster. There is also special handling for the final fe56-> something so you don't have to follow all the heavy isotopes.
>> 
>> 3) Yes
>> 
>> 4) No, approx nets are handled separately to soft nets, the approx nets are literally hard coded (each reaction is individually written out) into mesa while the soft nets are flexible with the reactions.
>> 
>> 5) auto_extend_net only changes the net if you are currently using one of the nets specified in h_he_net or co_net. You can check the logic in star/job/run_star_support.f90 function extend_net
>> 
>> 6) auto_extend will only switch between those three networks, while the adaptive network may change all the isotopes, effectively making its own new network at each step
>> 
>> Rob
>> 
>> On Sun, Feb 5, 2017 at 10:01 AM, amber lauer <alauer1 at lsu.edu> wrote:
>> I'm trying to understand some details of fixed networks, please tell me if my assumptions are correct;
>> 
>> 1) Soft nets only specify a fixed list of isotopes, the code supplies any reactions that link them. Any isotope not listed in the net file is not treated. These nets are indicated by the name mesa_*.net where * is the number of species.
>> 
>> 2) Approx nets utilize compound reactions, which assume an equilibrium of intermediate isotopes not explicitly included in the net, and therefore can treat reactions involving these even if they are not explicitly included.  These are indicated by the name approx_*.net. If this is correct, how is this different from a network that explicitly lists the intermediate species?
>> 
>> 3) All others are standard fixed networks that treat only the isotopes and reactions explicitly listed. 
>> 
>> 4) Network files can be created that are a combination of other networks, listing them by name. Can these be a combination of types 1), 2),  and 3)?
>> 
>> 5) The auto_extend_net feature, which is set true by default, will extend beyond basic.net to cno_ burn.net and then to approx_21. What if the  network is changed to something else, will it still look to the co_net and adv_net networks if the specified net is insufficient?
>> 
>> 6) Auto_extend_net does not apply to adaptive networks (if this is false, what does it do?).
> 
> 
> 
> 
> 
> [The attachment effective_reaction_rates.pdf has been manually removed]