Changes in r22.11.1

Backwards-incompatible changes


A large amount of internal clean up has occurred since the last release. This lists some of the most important changes, but the list is not exhaustive.

Module-level changes


The main controls for the selection of parameters and non-seismic constraints (which were dubbed “variables”) has changed. The defaults files document the new interface but the most important changes are repeated here.

Each non-seismic constraint is now given a name, target value, uncertainty and flag for whether to include it in the \(\chi^2\) calculation. The default work folder will either try for one of the custom options included for backward compatibility (e.g. Rcz) or fall back to computing the matching history column (e.g. for log_g). So whereas an effective temperature constraint would previously be included using, say

include_Teff_in_chi2_spectro = .true.
Teff_target = 6000
Teff_sigma = 100

you would now use

constraint_name(1) = 'Teff'
include_constraint_in_chi2_spectro(1) = .true.
constraint_target(1) = 6000
constraint_sigma(1) = 100

The maximum number of such constraints is currently 100 but can trivially be increased at compile time by modifying max_constraints in astero/public/astero_def.f90.

Similarly, each parameter now has a name, initial value, minimum, maximum and grid-spacing. So whereas the mixing-length parameter was previously controlled with something like

vary_alpha = .true.
first_alpha = 1.7
min_alpha = 1.5
max_alpha = 1.9
delta_alpha = 0.1

you would now use

param_name(1) = 'alpha'
vary_param(1) = .true.
first_param(1) = 1.7
min_param(1) = 1.5
max_param(1) = 1.9
delta_param(1) = 0.1

Again, the maximum number of parameters is 100 and can be increased at compile time by modifying max_parameters in astero/public/astero_def.f90.

The default run_star_extras.f90 defines the hooks set_constraint_value and set_param so that the old options remain available, though with a new syntax. Users can also use those routines to define their own parameters and constraints.

The output files contain information for constraints or parameters with names that are not ''. Thus, the column order now varies but the same information is present and now follows the same structure as histories and profiles.


For wind mass loss, schemes that scale with metallicity now employ Zbase rather than Z (as long as Zbase is set to a non-negative value, otherwise we fall back to Z). This reflects the fact that wind recipes primarily account for the opacity of iron-group elements, which have surface abundances that are unlikely to change during evolution. This change therefore avoids unphysical influence on winds by, e.g., evolution of surface CNO abundances.


All test cases have now had the inlist option, makefile variable, and shell script variable, MESA_DIR removed. This means that you no longer need to do anything to make a MESA test case work outside of the test suite. Test cases now infer their MESA_DIR variable entirely by the environment variable $MESA_DIR.

The history output option tri_alfa (and other quantities that relate to the triple-alpha nuclear reaction) have been renamed to tri_alpha for better consistency with other _alpha reactions.


The derived type net_info (conventional given the symbol n) is no longer a pointer. If you declare a local copy of the variable, you should also ensure to do n% g => g to make sure that net_info knows about the net_general_info derived type. g can be had from a call to get_net_ptr(handle, g, ierr).

The pointer array net_work and its size net_lwork have been removed from the net interface, thus these variables should be removed form any other_net_get and other_split_burn hooks. The following routines have also been removed as they are no longer needed net_work_size, get_net_rate_ptrs, net_1_zone_burn_work_size, get_burn_work_array_pointers, net_1_zone_burn_const_density_work_size, and get_burn_const_density_work_array_pointers

Previously you could pass arg_not_provided for either the temperature (density) or log(temperature) (log(density)). Now you must pass both explicity.


ADIPLS now has a USE_ADIPLS flag in utils/makefile_header to enable is build to be disabled.

Changes in r22.05.1

Backwards-incompatible changes


A large amount of internal clean up has occurred since the last release. This lists some of the most important changes, but the list is not exhaustive.

Module-level changes


&astero_search_controls now has an option astero_results_directory to specify a folder into which all of astero’s results are saved (like log_directory in star). The default is outputs, so if you can’t seem to find your astero output, have a look there.

&astero_search_controls also now has options

astero_results_dbl_format = '(1pes26.16)'
astero_results_int_format = '(i26)'
astero_results_txt_format = '(a26)'

by which the user can set the formats of floats, integers and strings in the astero results file, much like star_history_*_format does for history files.

The format of the astero results file has changed to match histories and profiles. The contents of the file are unchanged.


The 7Be(e-,nu)7Li has been switched from REACLIB rate to that of Simonucci et al 2013. This is due to the fact that the REACLIB rate does not take into account the neutral ion rate below 10^7 K.

The ability to set the rates preferences has been removed. This added alot of complexity to the rates code handling NACRE and REACLIB and made it difficult to reason about where a rate actually came from. From now on we excusivily use NACRE for any rate that cares about temperatures below 10^7K (for all temperatures), REACLIB for almost all other rates, and a small number of rates from CF88 (if they aren’t in REACLIB or NACRE).

Of note is that the default C12(a,g)O16 rate has thus changed from NACRE to that of REACLIB.

The options set_rates_preferences, new_rates_preference, and set_rate_c1212 have been removed without replacements.

The options set_rate_c12ag, set_rate_n14pg, and set_rate_3a have been removed. However, those rates can now be access thorugh a new rate selection mechanism. In $MESA_DIR/data/rates_data/rate_tables we now ship a number of MESA provided rate tables. These can be loaded, either by the normal mechanism of adding the filename to a rates_list file, or by using the new option filename_of_special_rate. This option sets the filename to load the rate from for the rate specified by reaction_for_special_factor.

Thus the options:

num_special_rate_factors = 1
reaction_for_special_factor(1) = 'r_c12_ag_o16'
special_rate_factor(1) = 1
filename_of_special_rate(1) = 'r_c12_ag_o16_kunz.txt'

replaces the previous:

set_rate_c12ag = 'Kunz'


As part of this new scheme we now ship a set of rates from NACREII Xu et al 2013. These rates do not, by default, override the default NACRE rates. You must explicitly ask for them with filename_of_special_rate.

There is now a new hook other_rate_get to provide a simple way to change an existing rate in a run_star_extras.f90. Note this hook only works on rates that are NOT currently in your rates_cache. It is recommended when using this option to set a custom rates_cache_dir otherwise the cache files in MESA_DIR will be over written.

The previous option:

use_rate_3a = 'Fl87'

has been replaced with:

use_3a_fl87 = .true.


There is a new hook other_net_derivs that allows for modifying the dydt term MESA computes for each zone inside net/. This allows adding changes in composition due to nuclear reactions that MESA could otherwise not handle or does not know about. This hook only works with soft networks (thus no approx nets). This hook requires many derivatives to be set, thus users should look at net_derivs.f90 for reference to what needs setting.

There is now a hook other_split_burn for replacing MESA’s split burn routine.


Diffusion coefficients for white dwarf interiors are now included based on Caplan et al. (2022). By default, these coefficients are used for strong plasma coupling \(\Gamma > 10\), but there is an inlist option to turn them off and revert to the previous default Stanton & Murillo (2016) coefficients if desired.

Fixed a combination of bugs whereby the atmosphere data written to pulsation file formats (e.g. FGONG) was incorrect or wouldn’t work if tau_factor or atm_T_tau_opacity differed from their defaults (1.0 and 'fixed', respectively).


Due to re-organization of the star_type derived type, all pgstar controls have been moved into a separate pgstar derived type. If you access a pgstar option XX in your run_star_extras.f90 then you need to replace s% XX with s% pg% XX.


In r21.12.1 an experimental RSP solver feature was turned on by default, leading to convergence issues in nonlinear model integration. This is now turned off by default. Users that continue to use RSP in r.21.12.1 should include RSP_do_check_omega = .true. in the &controls section of their inlists to get rid of this issue.

Changes in r21.12.1

Backwards-incompatible changes


A large amount of internal clean up has occurred since the last release. This lists some of the most important changes, but the list is not exhaustive.

Simplification of energy equation options

The desired form of the MESA energy equation is now selected via the control energy_eqn_option. The available options are 'dedt' (default) and 'eps_grav'. See the documentation at energy_eqn_option for more information about these forms.

The controls use_dedt_form_of_energy_eqn, always_use_dedt_form_of_energy_eqn, and use_eps_grav_form_of_energy_eqn were removed and replaced by the functionality of energy_eqn_option.

Simplifications to the energy equation code mean that this selection applies globally (i.e., to all cells in the model and at all timesteps).

  • The per-cell energy equation controls max_eta_for_dedt_form_of_energy_eqn and max_gamma_for_dedt_form_of_energy_eqn were removed.
  • The form-switching control steps_before_always_use_dedt_form_of_energy_eqn was removed.

Name changes

  • The star_job option saved_model_name has been replaced with load_model_filename everywhere.
  • The controls options power_c_burn_{lower,upper}_limit were replaced with the more generic power_z_burn_{lower,upper}_limit.
  • The controls option delta_lgL_phot_limit was renamed to delta_lgL_power_photo_limit (“phot” was easily confused with photosphere instead of photodisintegration).
  • The controls options surf_w_div_w_crit_limit and surf_w_div_w_crit_tol were renamed to surf_omega_div_omega_crit_limit and surf_omega_div_omega_crit_tol
  • The core/layer mass values c_core_*, c_rich_layer, and o_core_* have been renamed to co_core_*, co_rich_layer_*, and one_core_*. This better reflects the typical carbon/oxygen and oxygen/neon compositions of these regions. This affects the names of both the relevant controls and history columns.
  • The controls option use_d_eos_dxa was renamed to fix_d_eos_dxa_partials. This control originally had a broader function during the implementation of eos composition derivatives, but is now restricted to selecting whether we do a finite-difference-based fix up when on a component EOS that doesn’t provide composition derivatives.
  • The history and profile columns burn_* where replace with *_alpha.
  • History, profile, and binary history column files are now case insensitive.

Removed options

  • The time-smoothing scheme for mixing diffusion coefficients was removed. All associated options (e.g., new_D_smooth_flag and D_smooth_replacement_fraction) were removed.
  • Removed option semiconvection_upper_limit_center_h1. This can be implemented by setting s% alpha_semiconvection in run_star_extras.f90/extras_start_step.
  • Removed the option use_brunt_gradmuX_form. Alternative forms of the Brunt can be calculated using the other_brunt hook.

Removed history and profile columns

A major clean up of the history and profile columns was undertaken. Some of the removed values include:

  • Removed profile columns total_energy and total_energy_integral.

Relocation of eos hooks

The other_eos hooks have been removed from star. See the eos section for information about their replacements.

Hook interface changes

  • The Teff argument has been removed from the other_surface_PT hook. (Teff is instead available in the star_info pointer.)
  • other_mesh_delta_coeff_factor no longer takes eps_h, eps_he or eps_z as arguments.

Auto diff

We now make more extensive use of the new autodiff module for automatically differentiating variables. If you are using a hook in your run_star_extras.f90 then you will need to add use auto_diff to the top of your run_star_extras.f90 file.

If you see errors such as:

Error: Cannot convert REAL(8) to TYPE(auto_diff_real_star_order1) at (1)

Then this means you are missing the use auto_diff statement.

An example of using autodiff in a hook can be found at Example of auto_diff in run_star_extras

Module-level changes


Many of the one-dimensional arrays of mode data (e.g. l0_obs) have been consolidated into two-dimensional arrays (e.g. freq_target) in which the first index is the angular degree l. The following controls in &astero_search_controls have changed:

Old New

































The call signatures to the surface correction subroutines have also changed, generally from

subroutine get_some_freq_corr(...,
      nl0, l0_obs, l0_sigma, l0_freq, l0_freq_corr, l0_inertia,
      nl1, l1_obs, l1_sigma, l1_freq, l1_freq_corr, l1_inertia,
      nl2, l2_obs, l2_sigma, l2_freq, l2_freq_corr, l2_inertia,
      nl3, l3_obs, l3_sigma, l3_freq, l3_freq_corr, l3_inertia)


subroutine get_some_freq_corr(...,
      nl, obs, sigma, freq, freq_corr, inertia)


There are new hooks other_binary_photo_read and other_binary_photo_write. These allow the user to save/restore values in run_binary_extras.


There are new module-level eos hooks (see eos/other) that replace the star-level eos hooks (previously in star/other). Usage of these hooks is similar to hooks in star. However, the relevant procedure pointer is part of the EOS_General_Info structure and not the star_info structure. Therefore, in extras_controls, the procedure pointer statement should look like s% eos_rq % other_eos_results => my_other_eos_results. The boolean option use_other_eos_results controlling whether to use the hook is part of the eos namelist rather than controls. For the first required argument handle, pass s% eos_handle. This ensures that the routine uses the same configuration options as other calls from star to the eos module.

The hook other_eos_component allows the user to replace all or part of the MESA EOS by providing a new component EOS and to control the location of the blends between this and the other component EOSes. It is controlled by the option use_other_eos_component. The user-provided routine must return a complete set of EOS results. This EOS component has the highest priority in the blend. This hook should be used along with the hook other_eos_frac, which defines the region over to use other_eos_component.

The hook other_eos_results allows the user to modify the results returned by the EOS. The user-provided routine receives the results from the EOS right before they are returned, after all components have been evaluated. This allows the user make minor modifications to the results from the existing EOS without having to provide a full replacement.

Two alternative eos module entry points (eosDT_HELMEOS_get and eosDT_ideal_gas_get) and the star options that replaced the standard eosDT calls to be with these routines (use_eosDT_ideal_gas and use_eosDT_HELMEOS). This enables significant simplifications of eos_support. Restriction to a single component EOS can be achieved through the eos namelist options and replacement of the EOS should be performed through the other hook.

The HELM table was updated to a new, larger 100 points per decade version.

The HELM-related controls logT_ion_HELM, logT_neutral_HELM, and max_logRho_neutral_HELM were removed. These were used in an now-unsupported variant of HELM that blended the normal, fully-ionized HELM and a neutral version (which dropped the electron-positron terms).

The HELM-related controls always_skip_elec_pos and always_include_elec_pos were combined in the simplified control include_elec_pos which defaults to .true..

There is a new backstop EOS (ideal) which analytically models an ideal ion gas with radiation pressure. The purpose of this EOS is to provide coverage over the whole density-temperature plane for times when MESA needs to run to extreme densities or temperatures. No electrons are included in this EOS.


The call signatures of kap_get and the hook other_kap_get have changed. The set of arguments is now conceptually equivalent between the two subroutines. The inputs include the density, temperature, and full composition vector. The free electron/positron number and the electron degeneracy parameter (and their derivatives) are also required. The outputs include the opacity and its derivatives as well as information about the fractions of various opacity sources used in the blended opacity.

subroutine kap_get( &
   handle, species, chem_id, net_iso, xa, &
   logRho, logT, &
   lnfree_e, d_lnfree_e_dlnRho, d_lnfree_e_dlnT, &
   eta, d_eta_dlnRho, d_eta_dlnT , &
   kap_fracs, kap, dlnkap_dlnRho, dlnkap_dlnT, dlnkap_dxa, ierr)

   ! INPUT
   integer, intent(in) :: handle ! from alloc_kap_handle; in star, pass s% kap_handle
   integer, intent(in) :: species
   integer, pointer :: chem_id(:) ! maps species to chem id
   integer, pointer :: net_iso(:) ! maps chem id to species number
   real(dp), intent(in) :: xa(:) ! mass fractions
   real(dp), intent(in) :: logRho ! density
   real(dp), intent(in) :: logT ! temperature
   real(dp), intent(in) :: lnfree_e, d_lnfree_e_dlnRho, d_lnfree_e_dlnT
      ! free_e := total combined number per nucleon of free electrons and positrons
   real(dp), intent(in)  :: eta, d_eta_dlnRho, d_eta_dlnT
      ! eta := electron degeneracy parameter from eos

   real(dp), intent(out) :: kap_fracs(num_kap_fracs)
   real(dp), intent(out) :: kap ! opacity
   real(dp), intent(out) :: dlnkap_dlnRho ! partial derivative at constant T
   real(dp), intent(out) :: dlnkap_dlnT   ! partial derivative at constant Rho
   real(dp), intent(out) :: dlnkap_dxa(:) ! partial derivative w.r.t. species
   integer, intent(out) :: ierr ! 0 means AOK.

The Compton scattering opacity routine has been updated to use the prescription of Poutanen (2017).

The conductive opacity routine has been updated to include the corrections from Blouin et al. (2020) for H and He in the regime of moderate coupling and moderate degeneracy. These are on by default, controlled by the kap option use_blouin_conductive_opacities.

There are new module-level kap hooks (see kap/other) that allow individual components of the opacity module to be replaced with a user-specified routine given in run_star_extras. Usage of these hooks is similar to hooks in star. However, the relevant procedure pointer is part of the Kap_General_Info structure and not the star_info structure. Therefore, in extras_controls, the procedure pointer statement should look like s% kap_rq % other_elect_cond_opacity => my_routine. The boolean option use_other_elect_cond_opacity controlling whether to use the hook is part of the kap namelist rather than controls. For the first required argument handle, pass s% kap_handle. This ensures that the routine uses the same configuration options as other calls from star to the kap module.


The call signature of other_neu has changed. You no longer need to pass in z2bar

The value of the Weinberg angle was updated to be be consistent with CODATA 2018.


The screening mode classic_screening has been removed. Anyone using other_net_get needs to remove theta_e_for_graboske_et_al from its argument list.

The options reuse_rate_raw and reuse_rate_screened have been removed from other_net_get (and eval_net)


The format for custom weak rate tables (see e.g., data/rates_data/rate_tables/weak_rate_list.txt and test suite case custom_rates) no longer supports the (previously optional) Coulomb correction datasets delta_Q and Vs.

When this capability was first added, the energetics associated with the change in the composition were calculated in rates and included in eps_nuc. This meant the rates module needed to have access to information about the Coulomb-induced shifts in the electron and ion chemical potentials.

After the changes in the definition of eps_nuc and the energy equation described in MESA V, the energetics associated with the changing composition are self-consistently accounted for in the energy equation using information provided by the MESA EOS. Therefore, the ability to provide these unneeded and unused quantities has been removed.

Other changes

  • Analogous to kap_frac_Type2, information about the fractional contribution of the lowT tables, highT tables, and Compton opacities to the final result from the opacity module are now included in star_info arrays and profile columns with the names kap_frac_lowT, kap_frac_highT, kap_frac_Compton.
  • The control format_for_FGONG_data has been replaced by the integer fgong_ivers, which can be either 300 or 1300. This enforces adherence to the FGONG standard. In addition, users can now set the four-line plain-text header of FGONG output using the new controls fgong_header(1:4).
  • mixing_type now reports the mixing process that generates the largest D_mix, rather than prioritizing convection and thermohaline mixing over all others.
  • Added profile panel and history panel controls in pgstar to specify same yaxis range for both left and right axes (e.g., Profile_Panels1_same_yaxis_range(1) = .true.)
  • Experimental options have been moved into *_dev.defaults files and experimental test cases are now prefixed with dev_. These options and test cases are not ready for general use.
  • The ionization module has been removed. The eval_typical_charge routine has been moved into mod_typical_charge.f90 within the star module. The eval_ionization routine is no longer supported, as it was untested, undocumented, and unused.
  • A new module hdf5io for working with HDF5 files has been added.
  • The controls diffusion_gamma_full_{on,off} are no longer used by default. The EOS now returns phase information and by default that EOS phase will automatically turn off diffusion for crystallized material.
  • The issue with the value of free_e when using FreeEOS has been corrected. Thanks to Jason Wright for the report.
  • An other_screening hook was added.
  • All parts of test suite cases are now run by default. To skip running the optional inlists, set the environment variable MESA_SKIP_OPTIONAL (to any value). Previously, optional parts were skipped by default, and running all parts required setting MESA_RUN_OPTIONAL.
  • The headers for history and profile data now contain the value of Msun (grams), Rsun (cm), and Lsun (erg/s) used.
  • A bug has been identified and fixed in the Brown_Garaud_Stellmach_13 thermohaline mixing routine. The routine was meant to use Newton-Raphson relaxation to converge to a solution for the Nusselt number based on an initial guess from the asymptotic analysis in Appendix B of Brown, Garaud, & Stellmach (2013). However, a bug previously caused the routine to immediately return the asymptotic guess and skip the NR relaxation step. The asymptotic guess is usually fairly accurate, so this usually still produced a thermohaline result that was fairly close to the right answer, but the bug has been fixed now so that the NR relaxation is applied as well.

Changes in r15140

Backwards-incompatible changes

Addition of eos and kap namelists

The options associated with the eos and kap modules have been moved into their own namelists. (That is, there now exist &eos and &kap at the same level as &star_job and &controls.) User inlists will need to be updated. See Module-level changes for more specific information.

If you previously accessed the values of eos/kap related options from star_job or controls via run_star_extras, you will need to adjust your code to access the option values using the pointers to the EoS_General_Info and Kap_General_Info structures. These are exposed in star as s% eos_rq and s% kap_rq, respectively. So for example, the inlist value of Zbase is now accessible via s% kap_rq% Zbase (instead of s% Zbase).

Some file suffixes changed to .f90

Many source file names have been changed to have an .f90 suffix. For users, the most important changes are to the star and binary work directories.

In an existing star work directory (i.e., a copy of star/work or star test suite case), rename the files

  • src/run.fsrc/run.f90
  • src/run_star_extras.fsrc/run_star_extras.f90

In an existing binary work directory (i.e., a copy of binary/work or binary test suite case), rename the files

  • src/binary_run.fsrc/binary_run.f90
  • src/run_star_extras.fsrc/run_star_extras.f90
  • src/run_binary_extras.fsrc/run_binary_extras.f90

Changes to local makefiles that are not part of MESA might also need to be updated to reflect these changes.

Removal of backups

MESA no longer has the concept of a “backup”. (In a backup, after the failure of a retry, MESA would return to the previous model and evolve it with a smaller timestep.)

Models that previously relied on the use of backups in order to complete should instead use appropriate timestep controls such that retries alone are sufficient to enable the model to run.

All backup-related options and output quantities have been removed. Users migrating inlists or history_column.list files from previous MESA versions will need to remove these options, all of which contain the string “backup”.

Changes to solver reporting

MESA can report information about the progress of the iterative Newton–Raphson solution process that forms a key part of taking a timestep. The names of numerous options related to the solver have changed. These changes follow two main patterns.

First, the word “newton” was replaced with the word “solver”. For example, the history column that records the number of iterations changed from num_newton_iterations to num_solver_iterations. The controls option that defines a number iterations above which to reduce the timestep changed from newton_iterations_limit to solver_iters_timestep_limit and the terminal output correspondingly shows the message solver iters instead of newton iters. (The control newton_iterations_hard_limit was removed and not renamed.)

Second, the word “hydro” was removed or replaced with the word “solver” in the controls related to monitoring the solver internals. For example, the control report_hydro_solver_progress is now report_solver_progress and report_hydro_dt_info is now report_solver_dt_info. The use of these and other related controls is described in the developer documentation.

Changes to eps_grav and eps_mdot

A new method for handling the energetics associated with mass changes in MESA models was presented in MESA V, Section 3.2. The approach discussed therein, incorporated in a term named eps_mdot, has now become standard. As such, the option use_eps_mdot has been removed (because it is now effectively always true).

This eps_mdot approach supersedes the approach described in MESA III, Section 7, and so that implementation has been removed. This resulted in the removal of the &controls options

  • eps_grav_time_deriv_separation
  • zero_eps_grav_in_just_added_material
  • min_dxm_Eulerian_div_dxm_removed
  • min_dxm_Eulerian_div_dxm_added
  • min_cells_for_Eulerian_to_Lagrangian_transition
  • fix_eps_grav_transition_to_grid

the history columns

  • k_below_Eulerian_eps_grav
  • q_below_Eulerian_eps_grav
  • logxq_below_Eulerian_eps_grav
  • k_Lagrangian_eps_grav
  • q_Lagrangian_eps_grav
  • logxq_Lagrangian_eps_grav

and the profile columns

  • eps_grav_h_effective
  • eps_mdot_sub_eps_grav_h_effective
  • eps_mdot_rel_diff_eps_grav_h_effective
  • eps_grav_h
  • eps_mdot_sub_eps_grav_h
  • eps_mdot_rel_diff_eps_grav_h

Removal of lnPgas_flag

The option to use gas pressure instead of density as a structure variable has been removed. Users migrating inlists from previous MESA versions will need to remove these options, all of which contain the string “lnPgas_flag”.

Removal of logQ limits

As a consequence of the changes to eos, star no longer enforces limits on the quantity logQ (logQ = logRho - 2*logT + 12 in cgs). Therefore the controls options

  • logQ_limit
  • logQ_min_limit

and the pgstar option

  • show_TRho_Profile_logQ_limit

have been removed.

The removal of these controls does not indicate that the EOS is reliable at all values of logQ. Users should consult the description of the component EOSes and the regions in which they are applied to understand if MESA provides a suitable EOS for the conditions of interest.

Removal of GR factors

The control use_gr_factors and corresponding code has been removed. (This provided only a simple correction to the momentum equation and not a full GR treatment of the stellar structure equations.) Users wishing to include GR corrections to MESA’s Newtonian equations can achieve the same effect by using the other_cgrav or other_momentum hooks. For an example, see the neutron star test cases (ns_h, ns_he, and ns_c).

Change in STELLA file output

The options to create output files suitable for input to STELLA have been removed from MESA/star and the star_job namelist. These capabilities are now included as part of the ccsn_IIp test case (see inlist_stella and run_star_extras.f90). Users desiring STELLA-format output should re-use the code from that example.

This affects the options

  • save_stella_data_for_model_number
  • save_stella_data_when_terminate
  • save_stella_data_filename
  • stella_num_points
  • stella_nz_extra
  • stella_min_surf_logRho
  • stella_min_velocity
  • stella_skip_inner_dm
  • stella_skip_inner_v_limit
  • stella_mdot_years_for_wind
  • stella_mdot_for_wind
  • stella_v_wind
  • stella_show_headers

Removal of mesh adjustment parameters around convective boundaries

Controls matching the following patterns, which adjust the mesh resolution around convective boundaries, have been removed:

  • xtra_coef_czb_full_{on,off}
  • xtra_coef_{a,b}_{l,u}_{n,h,he,z}b_czb
  • xtra_dist_{a,b}_{l,u}_{n,h,he,z}b_czb
  • xtra_coef_scz_above_{n,h,he,z}b_cz

Convective boundaries can be resolved using a custom mesh-spacing function or mesh_delta coefficient. The simplex_solar_calibration test case has an example custom mesh-spacing function.

Change to mixing_type codes

The mixing_type codes (defined in const/public/const_def.f90) have changed. User code and/or analysis routines (e.g., scripts interpreting the mixing_type profile column) may need to be revised. We recommend that users use the mixing_type variables rather than the corresponding integers in their own code. e.g. rather than writing

if (mixing_type == 1) then


if (mixing_type == convective_mixing) then

assuming use const_def appears somewhere, as in the default run_star_extras.f90.

Limitations on use of varcontrol_target

A new variable min_allowed_varcontrol_target (default 1d-4) has been introduced to discourage the use of small values of varcontrol_target. MESA will exit with an error if the value is below this threshold.

The value of varcontrol is an unweighted average over all cells of the relative changes in the structure variables. For situations that need tighter timestep limits, there are many specific timestep controls that should be used instead of reducing the general target. The use of controls that specifically apply to the problem being studied will typically provide more effective and efficient timestep limiters. In addition, small values of varcontrol_target can lead to poor performance when it forces the size of the step-to-step corrections to become too small.

The option varcontrol_target is NOT the recommended way to push time resolution to convergence levels. To perform temporal convergence studies, use the new control time_delta_coeff, which acts as a multiplier for timestep limits (analogous to mesh_delta_coeff for spatial resolution).

One strategy for choosing effective timestep limits is to first set varcontrol_target = 1d-3. Then add some additional specific timestep limits relevant to the problem. Do a run, watching the reason for the timestep limits and the number of retries. Summary information about the conditions that limited the timestep can be printed at the end of run using the star_job option show_timestep_limit_counts_when_terminate. Repeat the runs, adding/removing or adjusting timestep limits until there are few retries and few places where the timestep is limited by varcontrol. Finally, repeat the calculation with a smaller value of time_delta_coeff (e.g., 0.5) and compare the results to gain confidence that they are numerically converged.

Module-level changes


Material previously present in star/astero and test cases using these capabilities have been promoted into their own module.

The csound_rms observational constraint has been removed.

The options for executing an arbitrary shell script (shell_script_num_string_char and shell_script_for_each_sample) have been removed. The usual use for these options—renaming output files at the end of each sample—can be replicated using the system tools available through utils_lib. For example, the following extras_after_evolve in run_star_extras.f90 moves the best profile and FGONG file to outputs/sample_#.{profile,fgong}.

subroutine extras_after_evolve(id, ierr)
   use astero_def
   use utils_lib, only: mv
   integer, intent(in) :: id
   integer, intent(out) :: ierr
   character (len=256) :: format_string, num_string, basename
   ierr = 0

   write(format_string,'( "(i",i2.2,".",i2.2,")" )') num_digits, num_digits
   write(num_string,format_string) sample_number+1 ! sample number hasn't been incremented yet
   basename = trim(sample_results_prefix) // trim(num_string)
   call mv(best_model_fgong_filename, trim(basename) // trim('.fgong'), skip_errors=.true.)
   call mv(best_model_profile_filename, trim(basename) // trim('.profile'), skip_errors=.true.)

end subroutine extras_after_evolve


This new module implements local theories of turbulence, including MLT, TDC, semiconvection, and thermohaline turbulence. These used to be a part of star. TDC (which stands for time-dependent convection) is now the recommended method for situations where the time dependence of convection must be taken into account. Other methods for time dependent convection present in the code have been removed, including the options min_T_for_acceleration_limited_conv_velocity and set_conv_vel_flag. TDC can be turned on with the option MLT_option = "TDC" in the controls section of an inlist.

Users will not generally need to interact with this module, but it can be used within run_star_extras by writing use turb.


This new module provides Fortran types that support algorithmic differentiation via operator overloading. Users will not generally need to interact with this module, but it can be used within run_star_extras to make derivatives easier to calculate (e.g. in the implicit hooks like other_surface).

Usage is by writing use auto_diff. This imports types such as auto_diff_real_4var_order1, which supports first-order derivatives with respect to up to four independent variables. A variable of this type could be declared via:

type(auto_diff_real_4var_order1) :: x

This variable then holds five fields: x%val stores the value of x. x%d1val1 stores the derivative of x with respect to the first independent variable. x%d1val2 is the same for the second independent variable, and so on. All d1val_ fields are initialized to zero when the variable is first set.

Once an auto_diff variable it initialized, all mathematical operations can be performed as they would be on a real(dp) variable. auto_diff variables also interoperate with real(dp) and integer types.

So for instance in the following f%d1val1 stores df/dx and f%d1val2 stores df/dy.

x = 3d0
x%d1val1 = 1d0

y = 2d0
y%d1val2 = 1d0

f = exp(x) * y + x + 4

Similar types are included supporting higher-order and mixed-partial derivatives. These derivatives are accessed via e.g. d2val1 (d²f/dx²), d1val1_d2val2 (d³f/dx dy²).


The const module has been updated to account for the revision of the SI and now uses CODATA 2018 values of the physical constants.

For astronomical constants, MESA follows IAU recommendations. MESA adopts nominal solar and planetary quantities from IAU 2015 Resolution B3 and now follows the recommended procedure of deriving nominal solar and planetary masses from the mass parameters \((GM)\) and the adopted value of \(G\).

As a result of these changes, most constants now have slightly different values than in previous MESA versions. For example, \({\rm M}_\odot\) changed from 1.9892e33 g to 1.9884e33 g.


EOS-related options have been moved into their own eos namelist. The module controls and their default values are contained in the file eos/defaults/eos.defaults.

The PTEH EOS has been removed. Tables from the FreeEOS project now provide coverage of similar conditions.

The region covered by the PC EOS has been increased. The boundaries of the region where PC is used no longer consider composition and so now include H/He-dominated material. The upper limit of the region where PC is used is now determined using the electron Coulomb coupling parameter and generally corresponds to higher temperatures than the previous approach.

For more information about the component EOSes and the regions in which they are applied, see the new overview of the EOS module.


GYRE has been upgraded to version 6.0. See the GYRE Documentation for information about this release.


Opacity-related options have been moved into their own kap namelist. The module controls and their default values are contained in the file kap/defaults/kap.defaults.

The OPAL Type 2 opacity tables are now on by default (use_Type2_opacities = .true.). These tables separately account for carbon and oxygen enhancements. Since this is especially important during core helium burning, the default transition from the OPAL Type 1 tables to the Type 2 tables occurs when material becomes nearly hydrogen free. As a result of this change, by default, users are required to specify the base metallicity of material using the kap namelist control Zbase. Usually, this physically corresponds to the initial metallicity of the star.

For more information about the opacity tables and how they are combined, see the new overview of the kap module.

rates & net

A number of rates have had their defaults switched to using JINA’s REACLIB.

When using a custom user rate (i.e from a rate table) the reverse rate is now computed in detailed balance from the user rate. Previously the reverse rate was computed using the default rate choice.

A bug with burning li7 at low temperatures rate has been fixed. Users stuck using previous versions of MESA and a soft network (something that is not an approx net) should add these lines to their nuclear network as a fix until they can update to a newer MESA:


With thanks to Ian Foley for the bug report.

We now define the forward reaction to always be the exothermic reaction, not the reaction as defined by REACLIB. This fixes an issue with exothermic photo-disintegrations which would generate wrong values when computed in detailed balance.

A lot of work has been done getting operator split burning (op_split_burn = .true.) to work. This option can provide a large speed up during advanced nuclear burning stages. See the various split_burn test cases for examples.

Other changes

  • Saved model files now contain a version_number field in their header. This indicates the version of MESA that was used to generate the model.
  • binary now automatically writes photo (restart) files at the end of the run.
  • If not provided with an argument, the binary ./re script will now restart from the most recent photo (determined by filesystem modification time). The star ./re script also has this behavior as of r12778.
  • The test case for building C/O white dwarf models has been overhauled to be more robust. See documentation for the new version in make_co_wd.
  • The builder for NS envelopes (test case neutron_star_envelope) has been replaced with a more general envelope builder (test case make_env). The test cases ns_{h,he,c} have been overhauled to start from these new models.
  • Added other_remove_surface. This routine is called at the start of a step and returns an integer k. All cells with j < k will be removed from the model at the start of the step, making cell k the new surface.
  • Installations are now blocked from using sudo. This is generally not what you want to use to fix installation issues. If you want to install MESA in a root location then you will need to edit out the check in the install file.
  • The install script now blocks attempts to use a MESA_DIR which contains spaces in it. This has never really worked as makefiles can not handle the spaces. To work round this either move MESA_DIR to a folder location with no spaces in its path or symlink your MESA_DIR to another location with no spaces in its path and set MESA_DIR to point at the symlink.
  • The option to create a pre main sequence model now relaxes the model until a radiative core forms. This is activated with the star_job option pre_ms_relax_to_start_radiative_core, which can be set to .false. to restore the old behavior.


Thanks to all who reported problems and asked or answered questions on mesa-users. Special thanks to Siemen Burssens, Mathias Michielsen, Joey Mombarg, Mathieu Renzo, and Samantha Wu for their assistance in testing pre-release versions.

Changes in r12778

This section describes changes that occurred since r12115.

SDK changes (Version 20.3.1 or later required)

To use the this MESA release, you must upgrade your SDK to 20.3.1.

In previous releases of MESA, we have included the CR-LIBM library to provide versions of standard math functions (exp, log, sin, etc) that guarantee correct rounding of floating-point numbers. In this new release, we made the decision to move CR-LIBM into the software development kit (SDK), where it properly belongs and can be maintained as one of the pre-requisites of MESA.

This means that to use this release (and subsequent releases) of MESA, you’ll need to upgrade to version 20.3.1 of the SDK or later. MESA checks the SDK version during compilation, and will stop with an error message if the SDK is too old.

Backwards-incompatible changes

Replacement of crlibm_lib with math_lib

MESA now talks to CR-LIBM via an intermediate module called math_lib. To make sure any code you add can properly access the CR-LIBM math routines, you’ll need to make sure that a use math_lib statement appears in the preamble of the file. At the same time, you should remove any use crlibm_lib statements, as they will no longer work (and are not needed). With math_lib, the names of the correctly rounded math functions are the same as the Fortran intrinsics (i.e., they no longer have a _cr suffix).

Existing run_star_extras, run_binary_extras, or other user-written code will need to be updated. To migrate, you should replace use crlibm_lib with use math_lib and remove the _cr suffix from any math functions (e.g., exp_crexp).

Removal of DT2 and ELM EOS options

The ELM and DT2 EOS options have been removed. (These options were underpinned by HELM and OPAL/SCVH data, but used bicubic spline interpolation in tables of lnPgas, lnS, and lnE as a way to get numerically accurate 1st and 2nd partial derivatives with respect to lnRho and lnT. A more detailed description can be found in previous release notes and Appendix A.1 of MESA V.) These options were introduced in r10398 and were turned on by default in r11532.

The numerical issues that ELM/DT2 were designed to address have been dealt with via another approach. MESA now separately treats quantities that appear in the equations (and happen to be partials) and the places where these theoretically equivalent, but numerically different quantities appear in the Jacobian (as partials of other quantities that appear in the equations). This is an implementation detail that should be transparent to users.

This change has two pleasant side effects. One, it lowers the memory demands of many MESA models, which should aid users of virtualized, containerized, or otherwise memory-constrained systems. Two, it removes small, interpolation-related wiggles that were present in partial derivative quantities such as \(\Gamma_1\).

These changes may require inlists that made use of DT2/ELM related options to be updated.

The following controls options have been deleted:

  • use_eosDT2
  • max_logT_for_max_logQ_eosDT2
  • max_logQ_for_use_eosDT2
  • use_eosELM
  • logT_max_for_ELM
  • logQ_min_for_ELM
  • check_elm_abar_zbar
  • check_elm_helm_agreement

The following star_job options have been renamed:

  • eosDT2PTEH_use_linear_interp_for_X to eosPTEH_use_linear_interp_for_X

The following controls options have been renamed/removed, as well as moved to star_job (see next entry):

  • logRho_max_for_all_PTEH_or_DT2 to logRho_max_for_all_PTEH
  • logRho_max_for_any_PTEH_or_DT2 to logRho_max_for_any_PTEH
  • logQ_max_for_low_Z_PTEH_or_DT2 (removed)
  • logQ_max_for_high_Z_PTEH_or_DT2 to logQ_max_for_PTEH

Change in location of PTEH EOS options

Options that modify the parameters associated with the PTEH EOS have be moved from controls to star_job. This brings PTEH in line with the behavior of the other component EOSes.

If you explicitly set any of following options in your inlist, you will need to move them from controls to star_job. Their meaning and default values remain unchanged.

  • use_eosPTEH_for_low_density
  • use_eosPTEH_for_high_Z
  • Z_for_all_PTEH
  • Z_for_any_PTEH
  • logRho_min_for_all_OPAL
  • logRho_min_for_any_OPAL
  • logRho_max_for_all_PTEH
  • logRho_max_for_any_PTEH

In addition, you must add the new option set_eosPTEH_parameters = .true. to star_job to indicate that these values should override the eos module-level defaults.

The removal of DT2 (see previous entry) has also resulted in the change that the controls option logQ_max_for_low_Z_PTEH_or_DT2 has been removed (as it applied primarily to DT2) and logQ_max_for_high_Z_PTEH_or_DT2 (which applied primarily to PTEH) has been renamed to logQ_max_for_PTEH and moved from controls to star_job.

New overshooting controls

The new controls for overshooting, briefly described in the release notes of version 12115, are now the default in MESA (and the old controls have been removed). All test_suite cases now use these new controls.

There are two schemes implemented in MESA to treat overshooting: a step overshoot scheme and an exponential scheme that follows Herwig (2000).

The old “double exponential overshoot scheme” is no longer accessible through simple controls. An example of how to implement such a scheme via the other_overshooting_scheme hook is contained in the other_physics_hooks test suite case.

The new overshooting controls are based on convection-zone and convection-boundary matching criteria. In the new set of controls, for each convective boundary it is possible to define an overshoot_zone_type, overshoot_zone_loc and an overshoot_bdy_loc, as well as values for the overshooting parameters.

The permitted values are the following:

  • overshoot_scheme = exponential, step
  • overshoot_zone_type = burn_H, burn_He, burn_Z, nonburn, any
  • overshoot_zone_loc = core, shell, any
  • overshoot_bdy_loc = bottom, top, any

The following controls assign values for the diffusive or step overshooting parameters:

  • overshoot_f
  • overshoot_f0
  • overshoot_D0
  • overshoot_Delta0

overshoot_f0 is defined so that the switch from convective mixing to overshooting happens at a distance overshoot_f0*Hp into the convection zone from the estimated location where grad_ad = grad_rad, where Hp is the pressure scale height at that location.

For exponential overshoot, D(dr) = D0*exp(-2*dr/(overshoot_f*Hp0) where D0 is the diffusion coefficient D at point r0, Hp0 is the scale height at r0.

For step overshoot: overshooting extends a distance overshoot_f*Hp0 from r0 with constant diffusion coefficient D = overshoot_D0 + overshoot_Delta0*D_ob where D_ob is the convective diffusivity at the bottom (top) of the step overshoot region for outward (inward) overshooting.

These “new” controls replace the following “old” controls:

  • overshoot_f_above_nonburn_core
  • overshoot_f0_above_nonburn_core
  • overshoot_f_above_nonburn_shell
  • overshoot_f0_above_nonburn_shell
  • overshoot_f_below_nonburn_shell
  • overshoot_f0_below_nonburn_shell
  • overshoot_f_above_burn_h_core
  • overshoot_f0_above_burn_h_core
  • overshoot_f_above_burn_h_shell
  • overshoot_f0_above_burn_h_shell
  • overshoot_f_below_burn_h_shell
  • overshoot_f0_below_burn_h_shell
  • overshoot_f_above_burn_he_core
  • overshoot_f0_above_burn_he_core
  • overshoot_f_above_burn_he_shell
  • overshoot_f0_above_burn_he_shell
  • overshoot_f_below_burn_he_shell
  • overshoot_f0_below_burn_he_shell
  • overshoot_f_above_burn_z_core
  • overshoot_f0_above_burn_z_core
  • overshoot_f_above_burn_z_shell
  • overshoot_f0_above_burn_z_shell
  • overshoot_f_below_burn_z_shell
  • overshoot_f0_below_burn_z_shell
  • step_overshoot_f_above_nonburn_core
  • step_overshoot_f_above_nonburn_shell
  • step_overshoot_f_below_nonburn_shell
  • step_overshoot_f_above_burn_h_core
  • step_overshoot_f_above_burn_h_shell
  • step_overshoot_f_below_burn_h_shell
  • step_overshoot_f_above_burn_he_core
  • step_overshoot_f_above_burn_he_shell
  • step_overshoot_f_below_burn_he_shell
  • step_overshoot_f_above_burn_z_core
  • step_overshoot_f_above_burn_z_shell
  • step_overshoot_f_below_burn_z_shell
  • step_overshoot_D
  • step_overshoot_D0_coeff

The “new” control overshoot_D_min replaces the “old” control D_mix_ov_limit.

The “new” control overshoot_brunt_B_max replaces the “old” control max_brunt_B_for_overshoot.

The “new” control overshoot_mass_full_on replaces the “old” control mass_for_overshoot_full_on.

The “new” control overshoot_mass_full_off replaces the “old” control mass_for_overshoot_full_off.

The following example will apply exponential overshoot, with f = 0.128 and f0 = 0.100, at the bottom of non-burning convective shells; and exponential overshoot, with f = 0.014 and f0 = 0.004, at all other convective boundaries.

overshoot_scheme(1) = 'exponential'
overshoot_zone_type(1) = 'nonburn'
overshoot_zone_loc(1) = 'shell'
overshoot_bdy_loc(1) = 'bottom'
overshoot_f(1) = 0.128
overshoot_f0(1) = 0.100

overshoot_scheme(2) = 'exponential'
overshoot_zone_type(2) = 'any'
overshoot_zone_loc(2) = 'any'
overshoot_bdy_loc(2) = 'any'
overshoot_f(2) = 0.014
overshoot_f0(2) = 0.004

Other examples are illustrated in the test_suite cases. Examples for exponential overshooting can be found in the following test_suite cases:

  • 1.4M_ms_op_mono
  • 25M_pre_ms_to_core_collapse
  • 5M_cepheid_blue_loop/inlist_cepheid_blue_loop
  • 7M_prems_to_AGB/inlist_7M_prems_to_AGB
  • accretion_with_diffusion
  • agb
  • axion_cooling
  • black_hole
  • c13_pocket
  • cburn_inward
  • envelope_inflation
  • example_ccsn_IIp
  • example_make_pre_ccsn
  • gyre_in_mesa_rsg
  • high_mass
  • high_z
  • hot_cool_wind
  • magnetic_braking
  • make_co_wd
  • make_metals
  • ppisn
  • pre_zahb
  • radiative_levitation

Examples for step overshooting can be found in the following test_suite cases:

  • high_rot_darkening
  • relax_composition_j_entropy

Version number

The MESA version_number is now represented as a string internally and in the headers of history/profile output. User scripts that assume this is an integer may need to be revised.

other_wind hook

The interface of the other_wind hook changed from

subroutine other_wind_interface(id, Lsurf, Msurf, Rsurf, Tsurf, w, ierr)
   use const_def, only: dp
   integer, intent(in) :: id
   real(dp), intent(in) :: Lsurf, Msurf, Rsurf, Tsurf ! surface values (cgs)
   real(dp), intent(out) :: w ! wind in units of Msun/year (value is >= 0)
   integer, intent(out) :: ierr
end subroutine other_wind_interface


subroutine other_wind_interface(id, Lsurf, Msurf, Rsurf, Tsurf, X, Y, Z, w, ierr)
   use const_def, only: dp
   integer, intent(in) :: id
   real(dp), intent(in) :: Lsurf, Msurf, Rsurf, Tsurf, X, Y, Z ! surface values (cgs)
   real(dp), intent(out) :: w ! wind in units of Msun/year (value is >= 0)
   integer, intent(out) :: ierr
end subroutine other_wind_interface

Existing user routines will need to be updated.

Removal of id_extra from run_star_extras.f

Most routines in run_star_extras.f had an argument id_extra. This argument generally did not do anything and so has been removed. Existing user routines will need to be updated.

A simple way to migrate from routines written for previous versions of MESA is to find and replace the string “, id_extra” with the empty string in run_star_extras.f.

Change of extras_startup from function to subroutine

The interface of extras_startup changed from integer function to subroutine. The current empty version of this routine is:

subroutine extras_startup(id, restart, ierr)
   integer, intent(in) :: id
   logical, intent(in) :: restart
   integer, intent(out) :: ierr
   type (star_info), pointer :: s
   ierr = 0
   call star_ptr(id, s, ierr)
   if (ierr /= 0) return
end subroutine extras_startup

Existing user routines will need to be updated to reflect this new interface.

Hooks for extra header items

The interface of the routines

  • how_many_extra_history_header_items
  • data_for_extra_history_header_items
  • how_many_extra_profile_header_items
  • data_for_extra_profile_header_items

has changed. If these routines are included in your run_star_extras.f (even if they have not been customized), you will need to update them. You should replace the old versions with:

integer function how_many_extra_history_header_items(id)
   integer, intent(in) :: id
   integer :: ierr
   type (star_info), pointer :: s
   ierr = 0
   call star_ptr(id, s, ierr)
   if (ierr /= 0) return
   how_many_extra_history_header_items = 0
end function how_many_extra_history_header_items

subroutine data_for_extra_history_header_items(id, n, names, vals, ierr)
   integer, intent(in) :: id, n
   character (len=maxlen_history_column_name) :: names(n)
   real(dp) :: vals(n)
   type(star_info), pointer :: s
   integer, intent(out) :: ierr
   ierr = 0
   call star_ptr(id,s,ierr)
   if(ierr/=0) return

   ! here is an example for adding an extra history header item
   ! also set how_many_extra_history_header_items
   ! names(1) = 'mixing_length_alpha'
   ! vals(1) = s% mixing_length_alpha

end subroutine data_for_extra_history_header_items

integer function how_many_extra_profile_header_items(id)
   integer, intent(in) :: id
   integer :: ierr
   type (star_info), pointer :: s
   ierr = 0
   call star_ptr(id, s, ierr)
   if (ierr /= 0) return
   how_many_extra_profile_header_items = 0
end function how_many_extra_profile_header_items

subroutine data_for_extra_profile_header_items(id, n, names, vals, ierr)
   integer, intent(in) :: id, n
   character (len=maxlen_profile_column_name) :: names(n)
   real(dp) :: vals(n)
   type(star_info), pointer :: s
   integer, intent(out) :: ierr
   ierr = 0
   call star_ptr(id,s,ierr)
   if(ierr/=0) return

   ! here is an example for adding an extra profile header item
   ! also set how_many_extra_profile_header_items
   ! names(1) = 'mixing_length_alpha'
   ! vals(1) = s% mixing_length_alpha

end subroutine data_for_extra_profile_header_items

Removal of inlist_massive_defaults

The file inlist_massive_defaults has been removed from star. Copies of the inlist can now be found in the following test cases:

  • 25M_pre_ms_to_core_collapse
  • 25M_z2m2_high_rotation
  • adjust_net
  • black_hole
  • envelope_inflation
  • example_ccsn_IIp
  • example_make_pre_ccsn
  • magnetic_braking
  • split_burn_20M_si_burn_qp
  • split_burn_big_net_30M
  • split_burn_big_net_30M_logT_9.8

Other changes

  • The routines {alloc,move,store,unpack}_extra_info were removed from (These routines were used to store/retrieve information from photos.) If you have existing run_star_extras code that includes these routines, it will continue to function. However, in new run_star_extras code, the recommended way to store/retrieve data is using the other_photo_read and other_photo_write hooks. Examples can be found in the conductive_flame and brown_dwarf test suite cases.
  • The controls xtra_coef_os_* and xtra_dist_os_* which could be used to modify mesh_delta_coeff in overshooting regions have been removed. The same functionality is available using the other_mesh_delta_coeff_factor and an example implementation is given in the agb test suite case.
  • The output-related control alpha_bdy_core_overshooting and related history options core_overshoot_{Hp,f,f0,hstep,r0} and {mass,radius}_bdy_core_overshooting have been removed.
  • The star_data module was split out of the star module. The source file describing the contents of the star_info data structure is now located at star_data/public/
  • If not provided with an argument, the ./re script will now restart from the most recent photo (determined by filesystem modification time).
  • Added star_control pre_ms_relax_to_start_radiative_core to existing star_control pre_ms_relax_num_steps to provide option for creating a pre-main sequence model just after the end of the fully convective period. The relaxation steps from raw pre-ms model to end of fully convective are done using simple control setting selected for robustness. After the relaxation is complete, the actual inlist parameter settings are used.
  • Added a new hook other_accreting_state to allow the user to specify the specific total energy, pressure, and density of the accreting material. These properties are used by eps_mdot to compute the contribution of accretion to the energy equation. By default (when this hook is not used), these properties are all taken from the surface cell.

Changes in r12115

This section describes changes that occurred since r11554. The changes were originally described by this post to the MESA Users’ mailing list.

Backwards incompatible changes

Changes to atmospheres

There has been a major overhaul of the atmosphere controls and related code. This improves consistency between the atmosphere and interior calculations and offers more flexibility to users. To learn more, please consult the user guide available at

Changes to s% xtra variables

The MESA star pointer provides a set of extra variables that can be used in run_star_extras.f and are automatically saved and restored during retries and backups. The old variables were

  • s% xtra1, s% xtra2, …, etc. for floats,
  • s% ixtra1, s% ixtra2, …, etc. for integers, and
  • s% lxtra1, s% lxtra2, …, etc. for logicals (booleans).

These have now been collapsed into arrays (e.g., s% xtra(:)). If you use these variables in your run_star_extras.f, you will need to enclose the variable number in brackets. E.g., s% xtra1 becomes s% xtra(1), s% ixtra17 becomes s% ixtra(17), etc.

The new scheme allows you to define integers with meaningful names that can make it more obvious how an xtra variable is used. For example, if you end up storing some integrated quantity in s% xtra(11), you could define i_my_integral = 11 and then refer to the value as s% xtra(i_my_integral).

The ppisn test suite case provides an example of this usage.

Other changes

Changes to WD atm tables

There are now 2 options for white dwarf atmosphere tables:

  • WD_tau_25: the original WD atmosphere table option for DA (H atmosphere)

WDs; also found and fixed a bug in the header of this file that was causing it to use only a small portion of the actual table

  • DB_WD_tau_25: new table for DB (He atmosphere) WDs

Changes to header format

The header format is now taken from the star_history_*_format and profile_*_format variables defined in controls.defaults. This addresses the bug caused by the compiler version string exceeding the allowed length of a header column found by some users with the MESA SDK and running on macOS. The default is now 40 characters but this can be set to a larger (or smaller) value in &controls.

In analogy to the routines in run_star_extras.f, run_binary_extras.f now has the routines


that allow the user to add custom header items to the binary history output.

New overshooting controls

We have introduced new, easier to use controls for overshooting, based on convection-zone matching criteria.

Use overshoot_new = .true. to use the new controls.

Note that in a future release, these new controls will become the default. Therefore, when you start new inlists, we recommend that you use these new controls.

In the new set of controls, for each convective boundary it is possible to define an overshoot_zone_type, overshoot_zone_loc and an overshoot_bdy_loc as well as values for the overshooting parameters.

The permitted values are the following:

  • overshoot_scheme: 'exponential', 'double_exponential' or 'step'
  • overshoot_zone_type: 'burn_H', 'burn_He', 'burn_Z', 'nonburn' or 'any'
  • overshoot_zone_loc: 'core', 'shell' or 'any'
  • overshoot_bdy_loc: 'bottom', 'top' or 'any'

The following controls assign values for the diffusive or step overshooting parameters:

  • overshoot_f
  • overshoot_f0
  • overshoot_f2

The following example will apply exponential overshoot, with f = 0.128 and f0 = 0.100, at the bottom of non-burning convective shells; and exponential overshoot, with f = 0.014 and f0 = 0.004, at all other convective boundaries.

overshoot_scheme(1) = 'exponential'
overshoot_zone_type(1) = 'nonburn'
overshoot_zone_loc(1) = 'shell'
overshoot_bdy_loc(1) = 'bottom'
overshoot_f(1) = 0.128
overshoot_f0(1) = 0.100

overshoot_scheme(2) = 'exponential'
overshoot_zone_type(2) = 'any'
overshoot_zone_loc(2) = 'any'
overshoot_bdy_loc(2) = 'any'
overshoot_f(2) = 0.014
overshoot_f0(2) = 0.004

Other examples are illustrated in the gyre_in_mesa_rsg and high_mass test_suite cases.

Changes in r11554

This section describes changes that occurred since r11532. The changes were originally described by this post to the MESA Users’ mailing list.

The release was principally made to quickly fix some memory leaks in r11532. Several users saw long-running jobs killed due to exhaustion of system memory. Thanks to Avishai Gilkis for the report.

This release also sets the star_job control num_steps_for_garbage_collection = 1000. Periodically MESA will free some memory from data structures that are no longer needed but have not been deallocated yet. At present, this only targets the EOS tables. (Implemented by Rob Farmer)

The header of MESA history/profile files now includes information about the compiler used and start date of the MESA run. (Implemented by Aaron Dotter)

             1           2        3                        4           5
version_number    compiler    build         MESA_SDK_version        date
         11554  "gfortran"  "8.3.0"  "x86_64-linux-20190313"  "20190314"

Changes in r11532

This section describes changes that occurred since r10398. The changes were originally described by this post to the MESA Users’ mailing list.

RSP is a new functionality in MESAstar that models the non-linear radial stellar pulsations that characterize RR Lyrae, Cepheids, and other classes of variable stars. See the rsp_* examples in the test suite.

We significantly enhance numerical energy conservation capabilities, including during mass changes. For example, this enables calculations through the He flash that conserve energy to better than 0.001%. Most test cases now have this enabled, for instance 1.3M_ms_high_Z, 25M_pre_ms_to_core_collapse, and wd as examples.

To improve the modeling of rotating stars in MESA, we introduce a new approach to modifying the pressure and temperature equations of stellar structure, and a formulation of the projection effects of gravity darkening. The latter are controlled by the grav_dark options in history_columns.list; see high_rot_darkening for an example of its use.

A new scheme for tracking convective boundaries, called Convective Pre-Mixing (CPM), yields reliable values of the convective-core mass, and allows the natural emergence of adiabatic semiconvection regions during both core hydrogen- and helium-burning phases. Examples for this can be found in the inlists provided with the mesa 5 paper.

We have updated the equation of state and nuclear reaction physics modules.

There are an increased number of warnings for when MESA goes beyond the validity of the input physics (for instance the nuclear reactions rates from REACLIB are ill-defined when logT>10.0). These warnings are controlled by the warn_* options.

The definition of eps_nuc has slightly changed (see MESA V, Section 3.2) in order to be suitable for use with the new energy equation. If you are running models using the dLdm form that includes eps_grav, you should consult the controls option include_composition_in_eps_grav and its associated documentation.

A new set of tests (gyre_in_mesa_*) demonstrate how to call GYRE on the fly during a MESA run.

The astero module now allows users to define model parameters (my_param[123]) that will be optimised in a similar way to the standard options (mass, Y, FeH, alpha, f_ov). These are defined in the subroutine set_my_params in run_star_extras.f in a similar way to how users can define their own observables (my_var[123]).

The astero module now has controls normalize_chi2_* that allow the user to decide whether or not to normalize each component of \(\chi^2\) by the number of terms that contributed to that component.

The format of the OP_MONO opacity table cache has changed. If you have used these files in a previous version of MESA then you should do:


before installing MESA. If you use multiple MESA versions, this means that you cannot share the cache file between old and new versions. Therefore, you should make sure to use a different cache file in each case. This may be more easily accomplished using the controls option op_mono_data_cache_filename rather than the environment variable.

The version of GYRE bundled with MESA has been updated to version 5.2.

Binaries can now model “twins”, where we can skip the calculation of the companion as its assumed to be identical to the primary. This is controlled by the binary_job parameter *_model_twins_flag.

There is a new way to treat convection in a model, via the convective_velocity_flag. This adds an equation to solve the velocity of convective motion, instead of using the value derived from MLT. This is useful for models evolving on fast timescales and is a replacement for min_T_for_acceleration_limited_conv_velocity.

Two new test cases (hydro_Ttau_solar and hydro_Ttau_evolve) demonstrate the use of mixing length parameters and T(τ) relations calibrated to 3D radiation-coupled hydrodynamics (RHD) simulations computed by Trampedach et al. (2014). More details are provided in Mosumgaard et al. (2018). MESA also includes low-temperature opacity tables that match those used in the 3D RHD simulations, which can be used by setting kappa_lowT_prefix = 'lowT_rt14_ag89'.

There have been many bug fixes and performance enhancements to MESA. Reports of bugs or suggested improvements are welcome on the mesa-users mailing list.

A reminder to please share your inlists and run_star_extras on upon publication of your science papers!

Changes in r10398

This section describes changes that occurred since r11554. The changes were originally described by this post to the MESA Users’ mailing list.

Equation of State: PTEH, DT2, and ELM (Bill)

Several new options for the mesa/eos have been added, all aiming for more accurate partials for the Newton solver. All of these new eos options use bicubic spline interpolation in tables of lnPgas, lnS, and lnE as a way to get numerically accurate 1st and 2nd partial derivatives with respect to lnRho and lnT. The partials are directly calculated from the interpolating bicubic polynomials to give numerical accuracy, but this comes at a cost in thermodynamic consistency since the actual thermodynamic relations can only be approximated by bicubic splines.

The new eos options are called “PTEH”, “DT2”, and “ELM”. The PTEH tables are created using the approach of Pols, Tout, Eggleton, and Han (1995) as implemented by Paxton (2004) in a program derived from Eggleton’s stellar evolution code (1973). PTEH extends the mesa/eos coverage to lower densities than allowed by OPAL (down to 10^-18 g cm^-3) and higher metallicity than covered (OPAL stops at Z = 0.04 while PTEH covers all Z). When PTEH is enabled, it is used for low densities and for high Z in cases that for lower Z would be handled using data from OPAL/SCVH tables. In the old MESA EOS we fell back to HELM to provide approximate results for the cases now covered by PTEH.

The mesa/star default controls enable PTEH for both low densities and high Z.

use_eosPTEH_for_low_density = .true.
use_eosPTEH_for_high_Z = .true.
Z_for_all_PTEH = 0.040d0
Z_for_any_PTEH = 0.039d0

The two remaining new eos options, DT2 and ELM, provide high resolution tables in logRho and logT for values from mesa/eos for OPAL/SCVH values and for HELM respectively. These cover a subset of the standard eos domain with standard eos results for logPgas, logS, and logE in a form suitable for bicubic spline interpolation in order to give 1st and 2nd partials with high numerical accuracy. However, since use of DT2 and ELM will give decreased thermodynamic consistency that might not be compensated for by better residuals, these are both disabled by default in mesa/star.

use_eosDT2 = .false.
use_eosELM = .false.

Opacities (Josiah, Aaron)

The opacity module (kap) underwent some internal restructuring. The kap module now exposes only a single kap_get interface instead of separate kap_get_Type1 and kap_get_Type2 subroutines. This has two user-visible consequences.

  • The control kappa_type2_logT_lower_bdy was removed. That control was no longer needed, as the existing control kappa_blend_logT_lower_bdy now also applies to Type2 opacities. All other related opacity controls (e.g., use_Type2_opacities) remain unchanged in name and behavior.
  • Previously, there were separate “other” hooks for Type1 and Type2 opacities. Now, there is only one hook, other_kap_get. It has the call signature of the previous Type2 hook, which is a super-set of the arguments to the Type1 hook (see star/other/other_kap.f90).

In previous versions opacities where clipped to the edge values of the tables when logR=logRho-3logT+18<-8. This has been replaced for a blend to Compton opacities between logR=-7.5 and logR=-8.

Element Diffusion (Evan)

Fixed a bug in the ionization treatment for diffusion in the pressure ionization routine. This was due to a typo in the original paper that presented the ionization scheme. Restored the missing factor of rho^1/3 thanks to a later presentation of this same scheme (Dupuis et al. 1992) and a note here.

Added a user control (D_mix_ignore_diffusion) for when to ignore element diffusion in surface or core mixing regions. Previously, diffusion would be turned off for surface mixing regions of ANY strength, even very weak mixing where diffusion might still be relevant. Now this control is set to a D_mix of 10⁵ (cm²/s), so that mixing that will obviously overwhelm diffusion (like convection) will turn it off, but weaker mixing won’t.

Gravity Darkening (Aaron)

Added options to include gravity darkening, in the form of projected (surface-averaged) luminosities and effective temperatures of the star viewed along the equator and pole, to the history file. Assumes the star is an oblate spheroid; see here for more info.

grav_dark_L_polar !Lsun
grav_dark_Teff_polar !K
grav_dark_L_equatorial !Lsun
grav_dark_Teff_equatorial !K

Isomers (Frank, Josiah, Bill)

The isomers of ²⁶Al can now be added to a reaction network. To use them, include the isomers in your network specification file. Two examples include

    h  1  1     ! hydrogen
   he  4  4    ! helium
   mg 25 25 ! magnesium
   al26-1      ! ground state
   al26-2      ! meta-stable excited state


include ''

One may use either al26 or al26-1 and al26-2. Reaction rates for the ²⁶Al isomers with other isotopes are picked up from the JINA reaclib file. Reaction rates for al26-1 <-> al26-2 are from Gupta & Meyer (2001) and located in data/rates_data/rate_tables along with the new default rate_list.txt file.

User-Beware: if you want a local rate_tables directory, <>, and you want the ²⁶Al isomers, then the two al26-1 <-> al26-2 rate files must be copied from their default location to your local rate_tables directory and your local rate_list.txt modified to include these two rates.

Installation Debugging (Rob, Josiah)

There is a new command, $MESA_DIR/help which outputs system information we need when debugging installation issues and/or MESA crashes. ./install will now also log its output to a file $MESA_DIR/build.log, if you have an installation issue please include this file when reporting an issue to mesa-users.

Miscellaneous improvements (Rob, Josiah)

You can now use the MESA_INLIST environment variable to set the name of the main inlist file when using MESA binary.

The output cadence of MESA binary has been tweaked to that its behavior is the same as MESA star. (If you use the same options, you should get output at the same steps.)

There is now a flag b% need_to_update_binary_history_now, which if set forces binary history output to occur at the current step.

Run_star_extras (Aaron)

Put calls to extra_header_items back into standard_run_star_extras and provided working examples of how to call all of them. These are useful for adding extra information to the history and profile headers beyond what is provided by default, such as including mixing_length_alpha in the history file header.

Building with Other Compilers

MESA currently does not compile with ifort. Other non-SDK compilers that are known to work (at the bit-for-bit level): Gfortran 7.3.1 (fedora 27)