Citing tengri

If tengri appears in a publication, cite the methods paper (in preparation) and the upstream codes providing physics, grids, and samplers. tengri.print_components_bibtex(result) prints BibTeX for every model, SSP, and inference backend that ran in your fit, keeping the acknowledgement in sync with the components you used. (tengri.cite_all() with no argument returns every citation registered in tengri, regardless of your fit.)

The tengri methods papers

@ARTICLE{Cooray_2026,
  author  = {{Cooray}, Suchetha},
  title   = {{tengri}: Differentiable SED fitting with Information-Field-Theory
             star formation history priors. I. Framework and mock recovery},
  journal = {in preparation},
  year    = {2026}
}

Three papers are planned, and the citation keys are stable now even though the metadata is not:

  • Paper I: Cooray (2026), framework and mock recovery. Registry key tengri, BibTeX key Cooray_2026.

  • Paper II: Cooray (2026), stochastic SFHs with IFT correlated-field priors and hierarchical population inference through geoVI. Registry key tengri_paper2, BibTeX key Cooray_2026a.

  • Paper III: Cooray (2026), application to deep JWST surveys. Registry key tengri_paper3, BibTeX key Cooray_2026b.

DOIs, volumes, and ADS bibcodes are backfilled on acceptance. See CITATION.bib for the entries as shipped, or call tengri.paper_citation().

Acknowledgements

Tengri is built on a long stack of open-source astronomy and Bayesian inference projects. We are grateful to the authors and maintainers of each one: their work shapes what tengri can do, and tengri brings it together in one place.

Everything below is registered in src/tengri/citations/references.bib, the single source of truth. tengri.cite("dsps") returns the same record this page quotes, and a CI guard fails the build when an entry in that file has no acknowledgement here. Papers are grouped by the role they play in the forward model, and listed oldest first within each group.

Framework and inference

  • JAX: autodiff, JIT, and XLA compilation. Bradbury et al. (2018).

  • optax: gradient-based optimizers behind the MAP routes.

  • NIFTy: original NIFTy signal-inference library, foundation for the correlated-field SFH prior. Selig et al. (2013), arXiv:1301.4499.

  • Information Field Theory: the theoretical foundation for tengri’s correlated-field SFH priors. Enßlin (2019), arXiv:1804.03350.

  • Bayesian Model Averaging. Hoeting et al. (1999), doi:10.1214/ss/1009212519.

  • Elliptical Slice Sampling: MCMC for Gaussian-prior models. Murray, Adams & MacKay (2010), arXiv:1001.0175.

  • Importance-sampling evidence estimation. Perrakis, Ntzoufras & Tsamardinos (2014), arXiv:1106.5578.

  • Generalized Pareto fit by profile likelihood, used for the Pareto-smoothed importance-sampling tail. Zhang and Stephens (2009), doi:10.1198/tech.2009.08017.

  • Saddle-free Newton steps, which move past saddle points using the Hessian’s negative curvature. Dauphin et al. (2014), arXiv:1406.2572.

  • Pareto-smoothed importance sampling: stabilizes the importance weights and checks their reliability with the Pareto k diagnostic. Vehtari et al. (2024), arXiv:1507.02646.

  • Predictive model comparison by WAIC and Pareto-smoothed leave-one-out cross-validation, with the paired standard error of an elpd difference. Vehtari et al. (2017), doi:10.1007/s11222-016-9696-4.

  • The Barker proposal: gradient-based MCMC robust to a step size that is wrong for one direction’s scale. Livingstone & Zanella (2022), arXiv:1908.11812.

  • Optimal scaling of Langevin diffusions: the 0.574 acceptance rate the Barker and MALA backends tune to. Roberts & Rosenthal (1998), doi:10.1111/1467-9868.00123.

  • Student-t noise as a Gaussian scale mixture, whose reweighted least squares finds the mode of the heavy-tailed marginal over nuisance coefficients. Lange et al. (1989), doi:10.1080/01621459.1989.10478852.

  • Pathfinder: fast approximate-posterior sampler. Zhang et al. (2022), Journal of Machine Learning Research 23(306), 1–49.

  • NIFTy.re: information field theory, geoVI and MGVI. Edenhofer et al. (2024), arXiv:2402.16683.

  • BlackJAX: NUTS, HMC, MCLMC and friends. Cabezas et al. (2024), arXiv:2402.10797.

  • Ray Tracing Sampler: noise-robust MCMC for high-dimensional fits. Behroozi (2025), arXiv:2510.25824.

  • Nested Slice Sampling: vectorized nested sampling for Bayesian evidence. Yallup, Kroupa & Handley (2026), arXiv:2601.23252.

  • Nested sampling: evidence and its uncertainty sqrt(H / n_eff), with n_eff = n_live for single-point deletion. Skilling (2006), doi:10.1214/06-BA127.

Stellar populations, isochrones, and spectral libraries

  • Starburst99: ionizing-spectrum synthesis. Leitherer et al. (1999), arXiv:astro-ph/9902334.

  • BaSeL: theoretical stellar spectral library. Westera et al. (2002), arXiv:astro-ph/0111175.

  • BC03: stellar population synthesis, used as a grid generator. Bruzual & Charlot (2003), arXiv:astro-ph/0309134.

  • STELIB: empirical stellar spectral library. Le Borgne et al. (2003), arXiv:astro-ph/0302334.

  • BaSTI: stellar evolution isochrones. Pietrinferni et al. (2004), arXiv:astro-ph/0405193.

  • Padova: stellar evolution isochrones. Marigo et al. (2008), arXiv:0711.4922.

  • FSPS: flexible stellar population synthesis, used as a grid generator. Conroy, Gunn & White (2009), arXiv:0809.4261; and the reference implementation, Conroy & Gunn (2010), arXiv:0911.3151. Its stellar remnants follow Renzini & Ciotti (1993), doi:10.1086/187068.

  • Carbon-star spectral library extending the FSPS TP-AGB stars redward of K. Aringer et al. (2009), arXiv:0905.4415.

  • MILES: empirical stellar spectral library. Falcón-Barroso et al. (2011), arXiv:1107.2303.

  • PARSEC: stellar evolution isochrones. Bressan et al. (2012), arXiv:1208.4498.

  • python-fsps: the Python interface used to generate the SSP grids. Foreman-Mackey et al. (2014), doi:10.5281/zenodo.12157.

  • Circumstellar AGB dust-shell SEDs in FSPS (add_agb_dust_model). Villaume et al. (2015), arXiv:1504.00900.

  • PARSEC massive-star tracks (the metallicity limit of the PARSEC SSP grids). Chen et al. (2015), arXiv:1506.01681.

  • MIST: stellar isochrones. Choi et al. (2016), arXiv:1604.08592; isochrone construction method, Dotter (2016), arXiv:1601.05144.

  • BPASS: binary stellar population and spectral synthesis. Eldridge, Stanway et al. (2017), arXiv:1710.02154.

  • DSPS: differentiable stellar population synthesis, the engine tengri evaluates SSPs with. Hearin et al. (2023), arXiv:2112.06830.

  • ProGeny: stellar population spectra generator. Robotham & Bellstedt (2025), arXiv:2410.17697.

  • APOGEE Milky Way disk metallicity distribution functions: the measured MDF widths sourcing the met_logzsol_scatter prior interval. Hayden et al. (2015), arXiv:1503.02110.

Cosmology

  • Massive-neutrino energy-density fitting formula (eq. 26), used in CosmoParams’ radiation-inclusive E(z) so PLANCK18 matches astropy.cosmology.Planck18’s luminosity distance and age(z). Komatsu et al. (2011), arXiv:1001.4538.

Initial mass functions

Star formation histories

  • Dense Basis non-parametric SFH reconstruction. Iyer & Gawiser (2017), arXiv:1702.04371.

  • Prospector model description; source of the continuity SFH prior. Leja et al. (2017), doi:10.3847/1538-4357/aa5ffe.

  • Non-parametric SFH priors: Dirichlet, continuity, and bursty continuity. Leja et al. (2019), arXiv:1811.03637.

  • Stochastic SFH power-spectral-density model for main-sequence scatter. Caplar & Tacchella (2019), arXiv:1901.07556.

  • Forensic SED reconstruction of cosmic star formation and metallicity history, by galaxy type. Bellstedt et al. (2020), doi:10.1093/mnras/staa2620.

  • PSD-governed SFH variability in cosmological simulations. Iyer et al. (2020), arXiv:2007.07916.

  • Current-episode star formation (constant rate over the last Delta t, spliced over a smooth history; recent, recent_ssfr). Chevallard & Charlot (2016), doi:10.1093/mnras/stw1756.

Nebular emission

  • MAPPINGS III: radiative shock and nebular models. Allen et al. (2008), arXiv:0805.0204.

  • CLOUDY: the photoionization code behind the nebular grids. Ferland et al. (2017), arXiv:1705.10877.

  • FSPS nebular emission grids baked into the with-nebular SSP files. Byler et al. (2017), arXiv:1611.08305.

  • Density-bounded H II regions and Lyman-continuum escape in the emission-line spectra of metal-poor galaxies, the context of the Gutkin+2016 backend. Plat et al. (2019), arXiv:1909.07386.

  • Cue: neural emulator for nebular emission lines. Li et al. (2025), arXiv:2405.04598.

  • Galactic H-alpha background implies pervasive diffuse ionized gas (DIG mixing motivation). Reynolds (1984), doi:10.1086/162190.

  • Synthesizer: synthetic observables package, source of the Cloudy AGN NLR/BLR grids and much nebular and SSP machinery. Lovell et al. (2025), doi:10.33232/001c.145766; and the software paper, Roper et al. (2026), doi:10.21105/joss.09436.

Dust attenuation

  • Milky-Way extinction curve (CCM). Cardelli et al. (1989), doi:10.1086/167900.

  • Starburst attenuation law. Calzetti et al. (2000), arXiv:astro-ph/9911459.

  • Two-component birth-cloud plus ISM attenuation. Charlot & Fall (2000), arXiv:astro-ph/0003128.

  • Optical-depth-dependent diffuse-ISM attenuation curve, the chevallard13 law. Chevallard et al. (2013), “Insights into the content and spatial distribution of dust from the integrated spectral properties of galaxies”, MNRAS 432, 2061, doi:10.1093/mnras/stt523.

  • Clumpy-medium radiative transfer attenuation. Witt & Gordon (2000), arXiv:astro-ph/9907342.

  • FM90 ultraviolet extinction parameterization (c1, c2, c3, c4, x0, gamma) used by gordon03_smcbar. Fitzpatrick & Massa (1990), doi:10.1086/191413.

  • Milky Way foreground extinction law (spline anchors and R_V dependence). Fitzpatrick (1999), doi:10.1086/316293.

  • SMC, LMC, and MW extinction curves. Gordon et al. (2003), arXiv:astro-ph/0305257.

  • Four-coefficient analytical extinction curve, tengri’s li08 law. Li et al. (2008), arXiv:0808.4115.

  • Modified Calzetti with variable UV slope and bump, the CIGALE form. Noll et al. (2009), arXiv:0909.5439.

  • Variable UV-slope and UV-bump attenuation law. Kriek & Conroy (2013), arXiv:1308.1099.

  • High-redshift attenuation curve from MOSDEF. Reddy et al. (2015), arXiv:1504.02782.

  • Modified-Calzetti attenuation with variable slope, the DSPS default. Salim et al. (2018), arXiv:1804.05850.

  • Redshift-dependent median attenuation curves from MUFASA radiative transfer, tengri’s narayanan_z law. Narayanan et al. (2018), arXiv:1805.06905.

Dust emission

  • Optically-thin modified-blackbody dust emission. Hildebrand (1983).

  • Luminosity-dependent IR SED templates. Chary & Elbaz (2001), doi:10.1086/321609.

  • IR SED template family indexed by IR luminosity. Dale & Helou (2002), doi:10.1086/341632.

  • Silicate-graphite-PAH dust emission. Draine & Li (2007), arXiv:astro-ph/0608003.

  • PAH emission feature spectra as Drude profiles. Smith et al. (2007), arXiv:astro-ph/0610913.

  • Thermal continuum emission from dust grains. Draine (2011), Physics of the Interstellar and Intergalactic Medium, Ch. 22.

  • Modified-blackbody plus power-law FIR SED model. Casey (2012), arXiv:1206.1595.

  • CMB-heating and CMB-contrast corrections for high-redshift dust. da Cunha et al. (2013), arXiv:1302.0844.

  • THEMIS: amorphous-hydrocarbon dust foundation. Jones et al. (2013), arXiv:1411.6293; and the global dust modeling framework, Jones et al. (2017), arXiv:1703.00775.

  • Two-parameter IR, submillimeter, and radio SED templates. Dale et al. (2014), arXiv:1402.1495.

  • Updated Draine & Li templates calibrated on Andromeda. Draine et al. (2014), arXiv:1306.2304.

  • Tabulated dust-continuum and PAH template library (T_dust, PAH mass fraction). Schreiber et al. (2018), arXiv:1710.10276, doi:10.1051/0004-6361/201731506.

  • BOSA dust SED templates parameterized by total IR luminosity and specific SFR. Boquien & Salim (2021), doi:10.1051/0004-6361/202140810.

  • Warm plus cold two-temperature energy-balance dust split. Kokorev et al. (2021).

  • PAHspec emission grid over grain size distribution, ionization, and starlight intensity. Draine et al. (2021), doi:10.3847/1538-4357/abff51.

  • Astrodust plus PAH unified grain model. Hensley & Draine (2023), arXiv:2208.12365.

AGN

Radio and X-ray

IGM and the Galactic foreground

  • IGM attenuation, retained as the legacy law. Madau (1995), doi:10.1086/175332.

  • IGM transmission as implemented by CIGALE. Meiksin (2006), arXiv:astro-ph/0512435.

  • Voigt-Hjerting approximation for damped Lyman-alpha profiles. Tepper-García (2006), arXiv:astro-ph/0602124.

  • IGM Lyman-series attenuation. Inoue et al. (2014), arXiv:1402.0677.

  • CGM damping-wing transmission, which removes a photometric-redshift bias at z > 7. Asada et al. (2025), arXiv:2410.21543.

  • 3D Galactic dust map used for Milky-Way extinction preprocessing. Edenhofer et al. (2024), arXiv:2308.01295.

Photometry and spectral indices

  • AB magnitude system and absolute spectrophotometric zero point. Oke & Gunn (1983), doi:10.1086/160817.

  • Photon-counting AB magnitude definition, as used by the FSPS getmags routine. Fukugita et al. (1996), doi:10.1086/117915.

  • Photon-counting bandpass AB flux and the K-correction formalism. Hogg et al. (2002), arXiv:astro-ph/0210394.

  • Photonic passbands and pivot wavelength. Bessell & Murphy (2012), arXiv:1112.2698.

  • Filter preintegration for differentiable photometry on GPUs. Zacharegkas et al. (2025), arXiv:2506.19919.

  • D4000 and Lick indices, and the BPT star-forming/AGN selection. Kauffmann et al. (2003), arXiv:astro-ph/0205070.

Comparison and reproduction

Tengri implements a superset of the union of the physics these codes cover, so the reproduction studies configure the public dict API to mimic each of them to the extent the models overlap. They are not ports, and residual differences are expected.

Inheriting credit automatically

import tengri
fitter = ...
result = fitter.run("mcmc_nuts")
tengri.print_components_bibtex(result)   # BibTeX for every component that ran

print_components_bibtex walks the model and inference call graph and emits BibTeX for everything that contributed (including the SSP grid), the easiest way to keep a paper’s acknowledgements in sync with the fit. (For the table form, use tengri.cite_components(result).)