Jason Eastman

Harvard just denied my summer student hire "due to current and potentially impending federal funding freezes."

She's an amazing scientist with a bright future that just got a little dimmer, and we're all worse for it.

Long overdue, but I've deactivated my twitter account and will be posting exclusively here going forward. I exported my old tweets and may retweet my most useful (instructional) ones here in the coming weeks.

This is huge! This should make GPs within EXOFASTv2 fairly easy to implement.

Workflow

If I broke your favorite thing, let me know.

xkcd.com/1172/

A few things I think are obsolete and were probably never used have been removed. Most significantly, I removed REFINESTAR, which was not included in my testing, difficult to maintain, and I think is better done by the user when it's occasionally useful anyway.

The inputs have been grouped and ordered more logically. Tutorials and example templates are probably still the easiest way to get started.

The documentation in $EXOFAST_PATH/exofastv2.pro was very outdated, and occasionally in conflict with $EXOFAST_PATH/mkss.pro. Overlapping documentation in mkss.pro has been moved to exofastv2.pro, which has been updated.

$EXOFAST_PATH/getdata.py (python!) is a new utility function that uses lightkurve to retrieve TESS, Kepler, or K2 data and format it for EXOFASTv2. It can optionally undeblend the TESS LCs so you can re-deblend them based on multi-star SED models.

For high SNR LCs with a known planetary eccentricity (where sigma_rho_LC ~< 5%), rstar is better constrained by the density (see Mahajan et al, 2024) rather than MANNRAD.

@amann.bsky.social 's MANNRAD and/or MANNMASS uses the Mann+ 2015 and Mann+ 2019 relations to constrain the radius and/or mass of M dwarfs, respectively, based on the apparent K-band magnitude (and derived absolute K-band magnitude).

We can also now model ramps (intended for JWST data). See /FITRAMP.

If the bandname can be interpreted as a number with "_" instead of "." (e..g, 8_0), it will be assumed to be the wavelength in microns, and will be plotted at the end. This, along with TDELTAVS, enables transmission spectroscopy modeling.

NOCLARET is now an NTRANSITS array, and the transit bandname no longer needs to be among the allowed names (if CLARET is disabled for the corresponding transit).

Note to the note: if correlations between linked parameters are not perfectly linear, extremely tight bounds can dramatically slow convergence times.

Note: only fitted parameters can be fixed and you cannot do math on them or specify prior widths or bounds based on other parameters (yet). If you want to link derived parameters, specify a Gaussian width that is small relative to the error.

You could also model EBs by linking the "planet" mass and radius to the stellar mass and radius.

Arielle Frommer has a paper coming out soon with more details about that.

This could lead to significantly better age constraints on all stars (and maybe even detailed cluster modeling). An isochrone is created for each subset of stars with the same age and initial metallicity, with that subset of stars plotted on top.

You can link parameters together in the prior file. If you're modeling a bound binary, you'll want to link their initial metalicities, distances, extinctions, and ages:

initfeh_1 initfeh_0 0
distance_1 distance_0 0
av_1 av_0 0
age_1 age_0 0

The MISTSEDFILE and SEDFILE models now plot the atmosphere and residuals (inspired by Sam Yee). Note: the plotted MISTSEDFILE atmospheres come from the NEXTGEN models, not the MIST BC's C3K models. This inconsistency is purely aesthetic.

Note: the inferior "FLUXFILE=" SED model which uses a tophat for each filter response does not support multiple SEDs or differential photometry. I expect the FLUXFILE SED model development to be frozen, as it's not much slower and far less accurate than the new one.

Thanks to Sam Yee for EXOFASTv2's first substantive pull request that backports the SEDFILE model based differential photometry to the "MISTSEDFILE=" SED model!

A separate error scaling term is fit for each spectrophotometry data set, and every fit I've seen has large errors. It's unclear to me whether this is due to errors in the model atmospheres or in the absolute flux calibration of the spectrum. Probably both.

This new model supports an arbitrary number of spectrophotometry files (SPECPHOTPATH=). MKTICSED automatically grabs Gaia spectrophotometry, but you can also supply your own (e.g., from JWST).

There is a new (third) SED model that uses NEXTGEN atmospheres and detailed filter transmission curves (which can be supplied by the user). Specify SEDFILE= (instead of MISTSEDFILE= or FLUXFILE=) to use this model. This is great for AO bands not supported by MIST BC tables.

The SED can be used to automatically deblend a transit lightcurve (see SEDDEBLEND).

By adding a 5th column in the SED photometry input file, the SED photometry can be specified as a single star (e.g., 0), a blend of any subset of stars (e.g., 0,1), differential magnitudes (0-1,2). If this column is missing, photometry is modeled as a blend of all modeled stars.

You can now model an arbitrary number of stars (NSTARS=). Each star can have its own SED and evolutionary models (updated examples and tutorials coming, but feel free to reach out)

The internal model was always self consistent, so relative TTVs, eccentricities, all computed with EXOFAST are not impacted.

This means transit times reported prior to this release are systematically early by ~a/c (semi-major axis divided by speed of light, modulo eccentricity), or ~30 seconds for a typical Hot Jupiter, and much larger for longer period planets (8 minutes for Earth).

Now we call this timestamp TJD_TDB (Target Julian Date in Barycentric Dynamical Time) and we separately compute and report a transit time that has been transformed back to BJD_TDB, the Barycentric Julian Date in Barycentric Dynamical Time.