BT-Settl: Low-Temperature Atmosphere Models

• BT-Settl is a grid of atmosphere models for cool stars, brown dwarfs, and planetary-mass objects that combines updated molecular opacities, revised solar abundances, and a self-consistent cloud model.
• It employs radiation-hydrodynamic simulations to capture cloud formation, dust settling, and vertical mixing, significantly improving spectral fits in optical and near-IR bands.
• Its applications in stellar isochrones and spectral analyses reveal both advanced predictive capabilities and systematic challenges in modeling molecular opacities and cloud treatments.

Searching arXiv for recent and foundational BT-Settl papers to ground the article in the literature. BT-Settl is a grid of model atmospheres and synthetic spectra for very low mass stars, brown dwarfs, and extending into planetary-mass regimes. It is computed with the PHOENIX atmosphere code and was developed as an update to the earlier NextGen and AMES-Dusty/AMES-Cond families by combining updated molecular opacities, revised solar abundances, and a physically motivated treatment of cloud formation, sedimentation, and mixing derived from radiation-hydrodynamic simulations (Allard et al., 2010). In stellar-population work it also functions as a “modern cool-star atmosphere” alternative to ATLAS9 when stellar interior models are transformed into observable magnitudes and colors, because it “carefully treat[s] molecular absorption lines for cool stars” and is especially relevant for low-temperature stars (Thompson et al., 2014).

1. Developmental context and model scope

BT-Settl was introduced to address the atmospheric regime in which older grids cease to provide a physically continuous description. In the framing of the model paper, NextGen reproduced warm M-dwarf atmospheres reasonably well, but increasingly overestimated $T_\mathrm{eff}$ along the lower main sequence and did not handle dust/cloud formation in the later M, L, and T regimes. AMES-Dusty and AMES-Cond were useful limiting cases—Dusty assumed dust remains suspended and affects the spectrum strongly, while Cond assumed dust forms but rapidly settles out of the observable atmosphere—but they were not a full physical solution. BT-Settl was developed to bridge this gap by modeling the formation, growth, settling, and mixing of dust clouds self-consistently, so that the sequence from very low mass stars through L and T dwarfs could be explained without ad hoc tuning of cloud parameters (Allard et al., 2010).

The model grid is described in the abstract as spanning

$$100{,}000\,\mathrm{K} > T_\mathrm{eff} > 400\,\mathrm{K},\quad 5.5 > \log g > -0.5,\quad [\mathrm{M/H}] = +0.5 \text{ to } -1.5,$$

with the reference solar abundances of Asplund et al. (2009). The summary section broadens the metallicity coverage further and includes various $\alpha$-element enhancements. The paper’s central claim is that the new solar abundances allow an improved reproduction of the photometric and spectroscopic properties of very low mass stars and, for the first time, a smooth transition between stellar and substellar regimes, together with an improved explanation of the M–L–T spectral transition (Allard et al., 2010).

2. Physical ingredients and numerical construction

Three ingredients define the BT-Settl formulation in the foundational presentation. The first is the BT2 water vapor opacity line list, introduced to improve the treatment of the strong H$_2$O absorption that shapes the near-IR spectra of cool dwarfs. The second is the revised solar abundance scale of Asplund et al. (2009), with particular emphasis on the lower oxygen abundance, which reduces excess molecular opacity and improves agreement with observed very low mass star spectra and temperatures. The third is a cloud model with supersaturation and radiation-hydrodynamic-based mixing, so that dust formation is treated with condensation and sedimentation physics plus mixing inferred from dynamical simulations rather than by arbitrary parameters (Allard et al., 2010).

A central advance is the use of radiation hydrodynamic 2D simulations of cool atmospheres, employing PHOENIX opacities in multi-group binning and forsterite dust cross sections. These simulations track advection, condensation, and sedimentation, and they reveal gravity waves as a key driver of cloud formation and atmospheric mixing in cool brown dwarf atmospheres. Around and below $T_\mathrm{eff}\sim 2200$ K, cloud layers can become optically thick enough to trigger cloud convection, while overshoot contributes to the redistribution of the largest grains. The simulations were used to derive a rule for the velocity field as a function of atmospheric depth and effective temperature, and this rule is described as relatively insensitive to gravity; it is then used in BT-Settl to determine the microturbulence velocity, the diffusion coefficient, and the advective mixing of molecules as functions of depth (Allard et al., 2010).

The dust model follows the expected condensation sequence as the atmosphere cools: refractory oxides and ceramics, silicates—especially forsterite $(\mathrm{Mg_2SiO_4})$—salts, and at still lower temperatures, ices. The cloud prescription is based on condensation and sedimentation timescales in the Rossow sense, but with the supersaturation pressure computed from pre-tabulated equilibrium chemistry rather than from a fixed approximate value. Later BT-Settl descriptions for young M and L dwarfs and for X-Shooter brown-dwarf analyses specify that 180 condensate species are included in the chemistry and 55 grain species contribute directly to radiative transfer, with mixing-length convection, overshoot and gravity-wave mixing, and non-equilibrium chemistry for CO, CH$_4$, CO$_2$, N$_2$, and NH$_3$ [(Manjavacas et al., 2014); (Manjavacas et al., 2015)].

3. BT-Settl in stellar colors, isochrones, and M-dwarf atmospheres

In stellar-evolution applications, BT-Settl is used at the atmosphere boundary and bolometric-correction stage rather than in the interior calculation itself. In the M35 isochrone comparison, the interior evolution calculations were not changed; instead, the atmosphere boundary/bolometric-correction step was standardized across different isochrone families so that the effects of stellar structure physics could be separated from those of atmosphere physics. BT-Settl was paired with Dartmouth, $$100{,}000\,\mathrm{K} > T_\mathrm{eff} > 400\,\mathrm{K},\quad 5.5 > \log g > -0.5,\quad [\mathrm{M/H}] = +0.5 \text{ to } -1.5,$$0, Padova, and PARSEC isochrones and compared with deep $$100{,}000\,\mathrm{K} > T_\mathrm{eff} > 400\,\mathrm{K},\quad 5.5 > \log g > -0.5,\quad [\mathrm{M/H}] = +0.5 \text{ to } -1.5,$$1 photometry for M35, using WIYN/MOSAIC $$100{,}000\,\mathrm{K} > T_\mathrm{eff} > 400\,\mathrm{K},\quad 5.5 > \log g > -0.5,\quad [\mathrm{M/H}] = +0.5 \text{ to } -1.5,$$2 optical data and NEWFIRM plus 2MASS in the near-IR. The cluster was placed at a distance of 870 pc with $$100{,}000\,\mathrm{K} > T_\mathrm{eff} > 400\,\mathrm{K},\quad 5.5 > \log g > -0.5,\quad [\mathrm{M/H}] = +0.5 \text{ to } -1.5,$$3 and $$100{,}000\,\mathrm{K} > T_\mathrm{eff} > 400\,\mathrm{K},\quad 5.5 > \log g > -0.5,\quad [\mathrm{M/H}] = +0.5 \text{ to } -1.5,$$4, with age choices of 250 Myr for Dartmouth, 200 Myr for $$100{,}000\,\mathrm{K} > T_\mathrm{eff} > 400\,\mathrm{K},\quad 5.5 > \log g > -0.5,\quad [\mathrm{M/H}] = +0.5 \text{ to } -1.5,$$5, and 178 Myr for Padova and PARSEC (Thompson et al., 2014).

The observational result is sharply mass dependent. For any isochrone set and atmosphere model, the observed data are accurately reproduced for all stars more massive then $$100{,}000\,\mathrm{K} > T_\mathrm{eff} > 400\,\mathrm{K},\quad 5.5 > \log g > -0.5,\quad [\mathrm{M/H}] = +0.5 \text{ to } -1.5,$$6 M$$100{,}000\,\mathrm{K} > T_\mathrm{eff} > 400\,\mathrm{K},\quad 5.5 > \log g > -0.5,\quad [\mathrm{M/H}] = +0.5 \text{ to } -1.5,$$7. Below $$100{,}000\,\mathrm{K} > T_\mathrm{eff} > 400\,\mathrm{K},\quad 5.5 > \log g > -0.5,\quad [\mathrm{M/H}] = +0.5 \text{ to } -1.5,$$8 M$$100{,}000\,\mathrm{K} > T_\mathrm{eff} > 400\,\mathrm{K},\quad 5.5 > \log g > -0.5,\quad [\mathrm{M/H}] = +0.5 \text{ to } -1.5,$$9, Padova and PARSEC consistently produce higher temperatures than observed. Dartmouth and $\alpha$0 isochrones with BT-Settl atmospheres reproduce optical data accurately, and the offsets seen with ATLAS9 shrink; the authors interpret this as evidence that the more careful inclusion of molecular features in BT-Settl gives a better color–temperature relation for cool dwarfs in the optical. The near-IR result is different: all isochrone/atmosphere combinations, including BT-Settl-based Dartmouth and $\alpha$1, produce $\alpha$2 colors that are too blue and $\alpha$3 magnitudes that are too faint relative to the observed M35 main sequence for stars below $\alpha$4. The paper speculates that molecular contributions to stellar spectra in the near-infrared may not be fully explored (Thompson et al., 2014).

A complementary validation on field M dwarfs used 152 optical spectra across the M-dwarf sequence. BT-Settl synthetic spectra reproduce the slope of the optical spectral energy distribution and most of its features, and the fitting procedure yields effective temperatures accurate to about $\alpha$5 K. The study identifies remaining deficiencies: the CaOH band at 5570 Å and AlH and NaH hydrides in the blue part of the spectra are still missing, with additional residual issues in MgH and in TiO and VO opacities around 8200 Å. The resulting $\alpha$6-spectral type relation is consistent between the NTT and SSO samples, and BT-Settl colors and isochrones are in good agreement with observed $\alpha$7, $\alpha$8, and $\alpha$9, although disagreement increases in $_2$0 and $_2$1 at the latest types (Rajpurohit et al., 2013).

4. Brown dwarfs, condensate clouds, and non-equilibrium chemistry

Ross 458C provides a benchmark test of BT-Settl in the late-T regime. Using near-infrared photometry and spectroscopy together with warm-Spitzer IRAC photometry, the object was typed as T8.5p, and an age prior of $_2$2 Gyr—more specifically roughly $_2$3–1 Gyr—was adopted from the primary star Ross 458A. From the luminosity and evolutionary interpretation the paper derives $_2$4 K, $_2$5, and a mass of approximately $_2$6. BT-Settl and Saumon & Marley both provide fits consistent with these empirical constraints, but BT-Settl gives the marginally better fit overall (Burningham et al., 2011).

The main distinction in that comparison appears in the 3.5–5 $_2$7m region, especially the IRAC [4.5] band. BT-Settl reproduces both [3.6] and [4.5] fluxes well, whereas Saumon & Marley underpredicts [3.6] and overpredicts [4.5]. The explanation given is BT-Settl’s treatment of non-equilibrium CO/CO$_2$8 chemistry. The paper explores the CE, R0, R1, and R2 chemistry cases and concludes that the stronger 4.3 $_2$9m CO$T_\mathrm{eff}\sim 2200$0 absorption in BT-Settl arises from its non-equilibrium CO$T_\mathrm{eff}\sim 2200$1 chemistry rather than simply from different opacities. The same study also finds that BT-Settl models incorporating a condensate cloud model better match the near-infrared spectrum than the Saumon & Marley cloudy model with $T_\mathrm{eff}\sim 2200$2, although the K-band flux remains difficult and BT-Settl tends to underestimate it somewhat (Burningham et al., 2011).

In medium-resolution X-Shooter spectra of late-L to T dwarfs, BT-Settl 2014 reproduces the majority of the observed spectral energy distributions and the evolution of the equivalent widths of Rb I 794.8 nm and Cs I 852.0 nm with spectral type. It also captures the weakening of K I 1253 nm for the young low-gravity object 2M0355. However, it usually fails to reproduce the shape of the H band for L and T dwarfs, does not reproduce the evolution of the Na I 818.3/819.5 nm and K I 1253 nm equivalent widths for field objects, and is limited by incomplete CH$T_\mathrm{eff}\sim 2200$3 and FeH opacities (Manjavacas et al., 2015).

Young late-M and L dwarfs expose related cloud limitations. For seven young M9.5–L3 objects observed with VLT/ISAAC, BT-Settl 2010 and 2013 fit the spectra and the 1–5 $T_\mathrm{eff}\sim 2200$4m SED of the L0–L3 dwarfs for temperatures between 1600–2000 K, and the L0–L3$T_\mathrm{eff}\sim 2200$5 dwarfs cluster near $T_\mathrm{eff}\sim 2200$6 K. Yet the models fail to reproduce the triangular H-band profile and the near-IR slope of some targets. The preferred super-solar metallicity solutions in several cases are interpreted not as abundance evidence but as indicative of a lack of dust, in particular at high altitude, in the cloud models; the modeling of the vertical mixing and of the grain growth was therefore identified as a target for revision in the next version of BT-Settl (Manjavacas et al., 2014).

5. Large-sample ultracool-dwarf analyses and inferred systematics

A major large-sample assessment used optical-to-mid-IR spectral energy distributions for 1054 ultracool dwarfs and planetary-mass objects. In that work BT-Settl was not the primary source of the bolometric luminosity; instead, it was used to estimate the unobserved optical and mid-IR flux contributions when the observed SED did not fully cover 0.1–2000 $T_\mathrm{eff}\sim 2200$7m. The fitted grid was solar metallicity, with $T_\mathrm{eff}\sim 2200$8–3500 K in steps of 100 K and $T_\mathrm{eff}\sim 2200$9–5.5 dex in steps of 0.5 dex. The study found that the goodness of fit is best for T dwarfs, then L dwarfs, then M dwarfs, and that field objects fit slightly better than young objects; 171 objects were flagged with $(\mathrm{Mg_2SiO_4})$0, comprising 80 M dwarfs, 83 L dwarfs, and 8 T dwarfs (Sanghi et al., 2023).

The strongest atmospheric-model systematics in that sample occur at the M/L transition boundary. For the most directly comparable self-consistent pair, BT-Settl versus BHAC15, BT-Settl generally underestimates $(\mathrm{Mg_2SiO_4})$1, with discrepancies up to $(\mathrm{Mg_2SiO_4})$2 K, and overestimates radius by up to $(\mathrm{Mg_2SiO_4})$3 at the M/L transition. Surface gravity differences are also large: bluer M/L transition dwarfs show BT-Settl underestimates of $(\mathrm{Mg_2SiO_4})$4 by up to 2 dex, whereas redder M/L transition dwarfs show overestimates of about 0.5 dex. Averaged over spectral type, atmospheric-model-derived and evolutionary-model-derived parameters can differ by as much as 400 K in $(\mathrm{Mg_2SiO_4})$5, 0.4 $(\mathrm{Mg_2SiO_4})$6 in radius, and 1.2 dex in $(\mathrm{Mg_2SiO_4})$7. The paper explicitly suggests that BT-Settl lacks sufficient dust opacity in the $(\mathrm{Mg_2SiO_4})$8–2200 K regime (Sanghi et al., 2023).

A second large benchmark analysis used low-resolution SpeX prism near-IR spectra for 90 late-M and L dwarfs in nearby young moving groups, the Pleiades, and the Hyades, fit within a Bayesian nested-sampling framework. Here BT-Settl was scaled by $(\mathrm{Mg_2SiO_4})$9, permitting direct inference of radius from the fitted solid angle and known distance. The authors note that a more sophisticated emulator framework could not be used because BT-Settl is not monotonic at low temperatures ($_4$0 K), which breaks the smooth PCA/GP interpolation assumption. Relative to evolutionary-model benchmarks, the BT-Settl likelihood surface develops two systematic attractors near the M/L boundary: one at $_4$1 K and $_4$2 dex, implying masses of roughly 150–1400 $_4$3, and another at $_4$4 K and $_4$5 dex, implying masses of about 0.02–3 $_4$6 and implausibly young ages (Hurt et al., 2023).

The spectral residuals in that analysis are equally diagnostic. BT-Settl tends to overpredict the peak $_4$7- and $_4$8-band flux for objects near the M–L boundary, with residuals at the 5–15% level or worse, while the $_4$9 band is more mixed. The authors interpret the $_2$0 overprediction as evidence that the dust content included in the model atmospheres is insufficient to match the observations. When an interstellar medium-like reddening law is added as a phenomenological proxy for atmospheric dust extinction, the fits improve substantially and the largest reddening coefficients occur at the M–L transition, but the inferred physical parameters remain systematically biased (Hurt et al., 2023).

6. Scientific significance, recurrent limitations, and directions for revision

Across the literature summarized here, BT-Settl’s primary scientific significance is that it makes atmosphere physics an explicit, testable component of low-temperature stellar and substellar modeling. In open-cluster isochrone work, switching from ATLAS9 to BT-Settl visibly improves optical color–magnitude agreement for cool stars while leaving the interior evolution unchanged, demonstrating that atmosphere models matter independently of the stellar interior model (Thompson et al., 2014). In the foundational atmosphere-grid paper, the model is presented as achieving improved photometric and spectroscopic fits to very low mass stars, a smooth transition between stellar and substellar regimes, and a better explanation of the M–L–T spectral sequence (Allard et al., 2010).

The recurrent limitations are equally clear and are concentrated in opacity completeness and cloud structure. In M-dwarf optical spectra, the missing CaOH band at 5570 Å and missing AlH and NaH hydrides in the blue remain explicit deficiencies, with additional issues in MgH, TiO, and VO (Rajpurohit et al., 2013). In young late-M and L dwarfs and in X-Shooter L/T spectra, the H-band profile and some near-IR slopes are not reproduced, with missing FeH opacity and incomplete CH$_2$1 opacity identified as contributors [(Manjavacas et al., 2014); (Manjavacas et al., 2015)]. In cluster and ultracool-dwarf population studies, BT-Settl-based colors can be too blue in the near-IR, and atmospheric parameters become systematically biased near the M/L transition, where multiple studies converge on the conclusion that the cloud/dust treatment lacks sufficient dust opacity or insufficient high-altitude dust [(Thompson et al., 2014); (Sanghi et al., 2023); (Hurt et al., 2023)].

This pattern suggests that BT-Settl is most successful when the dominant atmospheric processes are already represented by its updated opacities, revised abundances, and RHD-calibrated mixing, and least successful in transition regimes where cloud thickness, vertical mixing, grain growth, and molecular opacity incompleteness jointly control the emergent spectrum. A plausible implication is that the model’s enduring value lies both in the fits it achieves and in the systematics it exposes. The literature repeatedly points to future work on low-mass stellar atmospheres, especially molecular opacity in the near-infrared, vertical mixing, grain growth, and cloud prescriptions tailored to ultracool atmospheric dust, as the path toward reconciling BT-Settl-based models with the full optical-to-mid-IR behavior of cool stars, brown dwarfs, and planetary-mass objects [(Thompson et al., 2014); (Manjavacas et al., 2014); (Hurt et al., 2023)].