Tests to Make For Stellar Models

[rboston628]

Further code refactoring requires careful testing to make sure I don't make things worse. The following are some tests to run.

Note these tests are only tests of the stellar models, and do not test the modes.

Basic Tess

• negative test: SSRs can be bad enough to fail tests
• negative test: boundary expansions can be bad enough to fail tests
• SSRs scale like $N^{-4}$, according to RK4

Simple Models

These models mostly only exist as verification for SimpleWD models.

• isopycnic star equals known solution within machine precision
• isopycnic boundary expansions of $A^*$, $V_g$, $U$, $c_1$ equal to these functions near boundaries
• polytrope initialization suitably handles bad inputs
• polytrope $n=0$ equal to known solution
• polytrope $n=0$ boundary expansions are good approximations
• polytrope $n=1$ equal to known solution
• polytrope $n=1$ boundary expansions are good approximations
• polytrope $n=5$ equal to known solution
• polytrope $n=5$ central boundary expansions a good approximation
• polytrope exact errors scale like $N^{-4}$
• polytrope SSRs scale like $N^{-4}$
• polytrope solution $\theta$ scales like $N^{-4}$
• polytropes for a range of index can be constructed
• polytropes for a range of index have acceptably small SSR (say $N^{-3}$)
• polytropes for a range of index have good boundary expansions
• polytropes for a range of index, mass, radius can be constructed
• polytropes for a range of index, mass, radius have acceptable small SSR
• polytropes for a range of index, mass, radius equal "abstract" polytropes scaled to 1
• polytropes for a range of index, mass, radius have good boundary expansions
• Chandrasekhar WD initialization suitably handles bad inputs
• Chandrasekhar WDs match table of Chandrasekhar 1939
• Chandrasekhar WDs have acceptable SSR
• Chandrasekhar WDs central boundary expansion a good approximation
• Chandrasekhar WDs surface boundary expansion a good approximation
• Chandrasekhar WDs can be constructed with sigmoidal $\mu_e$
• Chandrasekhar WDs with sigmoidal $\mu_e$ have acceptable SSR, boundary expansions

MESA wrapper

• MESA initialization suitably handles bad inputs
• MESA models have at least one added grid point in each interval
• MESA models always divide intervals into binary fractions
• MESA model matches base model at knots
• MESA model has acceptable SSR
• MESA model central boundary expansion is a good approximation
• MESA model surface boundary expansion is a good approximation
• MESA model can be constructed from real MESA file v101 (use mesa_co_wd_cold.dat)
• MESA model constructed from mesa_co_wd_cold.dat has acceptable SSR
• MESA model constructed from mesa_co_wd_cold.dat has acceptable central boundary expansion
• MESA model constructed from mesa_co_wd_cold.dat has acceptable surface boundary expansion
• MESA model matches to a golden data set representing a previous fit

SimpleWD

• SimpleWD suitably handles invalid inputs
• SimpleWD can be constructed under a variety of inputs
• SimpleWD has acceptable SSR
• SimpleWD SSR scales appropriately with $N$
• SimpleWD has acceptable central boundary expansion
• SimpleWD has acceptable surface boundary expansion
• SimpleWD defines a half-grid-point
• A series of SimpleWD models can be used to approximate mass-radius relation
• SimpleWD prints out its Brunt-Vaisala frequency
• SimpleWD prints out its chemical composition
• SimpleWD prints out its opacity function

Artifacts

Have a run of the tests automatically generate the following useful graphs/tables for further diagnosing the code performance.

• graph of errors against exact solutions for $n=0, 1, 5$ polytropes
• graph of error scaling for $n=0, 1, 5$ polytropes
• graph of residuals for various polytropes
• graph of residual scaling for various polytropes
• graph of scaling of $\theta$ from various polytropes
• table of comparison against Chandrasekhar 1939 values
• graphs comparing SimpleWD against MESA values

[rboston628]

Some of the tests of stellar models are failing, due to not converging quickly enough, or not converging enough.

They have been disabled for the moment, to allow PR #7 to be closed out, since it's been open for almost a year now.

(Comment originally made Mar 24)