Phase Curves of Hot Neptune LTT 9779b Suggest a High-metallicity Atmosphere

For the Spitzer data, we first used POET to calculate aperture photometry for both IRAC channels with a range of aperture sizes, from 2.0 to 6.0 IRAC pixels and with inner and outer sky annulus radii of 7 and 15 pixels. Interpolated partial-pixel aperture photometry (with an oversampling factor of 5) accounted for fractional pixels in the photometric calculations. To precisely track the location of the star we fit a Gaussian profile to the stellar profile in each frame, using a constant term to model each frame's background flux. The raw flux, position, and point-spread function (PSF) width are plotted in Figures 1 and 2 for the 3.6 μm and 4.5 μm data, respectively.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

POET models and removes IRAC's well-known systematic correlation between a target's measured flux and the position of the source within a pixel. Without properly accounting for this intra-pixel sensitivity effect, weak exoplanetary signals are (at best) difficult to accurately extract. POET accounts for this systematic effect using its bilinearly interpolated subpixel sensitivity (BLISS) mode (Stevenson et al. 2012a). In BLISS mapping the main model hyperparameters are the choice of grid size for the subpixel sensitivity map, the astrophysical light-curve model (e.g., phase curve, eclipses, transits), and any additional light-curve models to account for additional systematics (e.g., exponential ramp, correlation with PSF width, or other long-term trends), as well as the choice of photometric aperture size.

We followed past POET analyses of IRAC data by choosing the hyperparameters that minimize the standard deviation of the normalized residuals (SDNR)—essentially minimizing the scatter on the residuals to the full light-curve fit. However, to avoid overfitting we also followed POET's recommendation to select the set of hyperparameters that minimizes SDNR but for which a nearest-neighbor interpolation does not outperform the BLISS interpolation. After testing a range of models, aperture sizes, and grid sizes (from 0.003 to 0.03 pixels), our final analysis identified the optimal set of parameters for the 4.5 μm channel as a five-pixel photometric aperture, a 0.017-pixel grid scale, an astrophysical model including a transit, secondary eclipse, and sinusoidal phase curve, and a systematic model consisting of a quadratic correlation with PSF width. As described below, this model does a good job at accounting for the instrumental and astrophysical signals in the IRAC2 photometry; we achieve consistent final results with analyses using four- or six-pixel apertures. In Figure 3 we show the calibrated IRAC2 photometry and best-fit light-curve model. Figure 4 shows that the residuals to our fit bin down approximately as white noise. Based on the WISE infrared flux of the star (Cutri et al. 2014) and the IRAC system throughput (IRAC Instrument and Instrument Support Teams 2012), we calculate that our 4.5 μm photometry reaches 1.3× the expected photon-noise limit.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

However, for the 3.6 μm IRAC1 photometry our models that explain this channel's instrumental and astrophysical variations imply a thermal phase curve with an implausibly large amplitude. These optimal models include a 2–3 pixel aperture, a BLISS grid size of 0.005 pixels, a linear ramp at the start of the first Astrophysical Observing Request (AOR), a correlation with PSF parameters, and a transit, secondary eclipse, and sinusoidal phase curve. In Figure 5 we show the calibrated IRAC1 photometry and best-fit light-curve model. Even after calibration the light curve shows high scatter, little or no sign of a secondary eclipses at the start of the observation (Eclipse 5 in the analysis of Dragomir et al. 2020), and the best-fit phase-curve model would nominally hint at negative planetary flux between the transit and first eclipse. Figure 6 shows that even when allowing nonphysical phase-curve models, the residuals to our light-curve fits bin down substantially slower than would be expected in the presence of white noise alone, further indicating the presence of systematic correlated noise not captured by the POET analysis. We calculate that our systematics-corrected 3.6 μm photometry still reaches only 2.3× the expected photon-noise limit on short timescales, and substantially worse on longer timescales. Although the IRAC1 transit is recovered at high confidence, due to this data set's intractable problems and high noise levels we do not further consider the 3.6 μm eclipses (which is discussed in more detail by Dragomir et al. 2020) or phase curve.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Although the main goal of this analysis is LTT 9779b's phase curve (which we detect at 4.5 μm), we also observed two transits (one in each IRAC band). The eclipse analysis is described in more detail by Dragomir et al. (2020), who analyzed our eclipse observations together with four additional eclipses in each IRAC channel. Their measurement of the eclipse depth is necessarily more precise than ours would be, due to the larger number of eclipses analyzed. In our analysis we therefore hold the eclipse depth fixed at their derived value, 375 ppm (their analysis shows that the two eclipses shown in Figure 3 have a depth consistent at 1.1σ with the overall depth measured from all six eclipses). For other physical parameters (except where noted below), we use the system parameters of Jenkins et al. (2020).

Based on the stellar parameters from Jenkins et al. (2020), we used the 3.6 μm and 4.5 μm quadratic limb-darkening coefficients from Claret et al. (2013). The average limb-darkening coefficients corresponding to T_eff = 5500 K and $\mathrm{log}g=4.5$ are u₁ = 0.106 and u₂ = 0.121 (at 3.6 μm) and u₁ = −0.02 and u₂ = 0.33 (at 4.5 μm). As noted previously, our IRAC2 primary transit occurs during a break in our Spitzer observations. This break makes the transit somewhat more challenging to model, so we elect to keep u₂ fixed at the tabulated value and allow u₁ to float within the range of values calculated by Claret et al. (2013), i.e., in the range from −0.13 to +0.09.

For our phase-curve modeling, we use a simple sinusoidal model with functional form

$$\begin{eqnarray}&&f(\phi )=1-A\cos (\phi -{\rm{\Delta }}\phi ),\end{eqnarray} \tag{ 1 }$$

where ϕ is the orbital phase (zero at transit, ∼0.5 at secondary eclipse), A is the normalized (fractional) semiamplitude, and Δϕ is the phase offset (i.e., the planetary longitude corresponding to maximum flux). Since our final signal-to-noise ratio on LTT 9779b's IRAC2 phase curve is relatively low, more complicated models (e.g., including higher-order harmonics) were not justified by the data.

We also modeled the Pre-Search Data Conditioning (PDC) light curve of LTT 9779b (Smith et al. 2012; Stumpe et al. 2012, 2014) using the methodology presented by Daylan et al. (2019) to infer the planet's phase modulation components at optical wavelengths.

We performed a global fit of the full PDC light curve using allesfitter (Günther & Daylan 2019, 2020), an inference framework for joint modeling of light-curve and radial-velocity data. We modeled the light curve as the sum of the stellar baseline emission occulted by the planet during the primary transit, the planetary baseline emission occulted by the star during the secondary eclipse, the planetary modulation in the form of a cosine at the orbital period that peaks at superior conjunction, the ellipsoidal variation in the form of a cosine at half the orbital period that peaks at quadrature phases, and an additive constant offset to absorb any normalization bias. Although the planetary modulation can be subject to a phase shift, we nevertheless assumed a phase shift of zero (the value expected for reflected light and for thermal emission from such a hot planet). This is because the resulting planetary modulation amplitude was small, leaving the optical phase shift unconstrained. We also assumed a circular orbit. We modeled the data (which could consist of thermal emission, reflected starlight, or both) as a linear combination of the following components:

• 1.  
  Planetary dayside: $\tfrac{{A}_{d}}{2}(1-\cos 2\pi \phi )$ outside the secondary eclipse and 0 otherwise.
• 2.  
  Planetary nightside: $\tfrac{{A}_{n}}{2}(1+\cos 2\pi \phi )$ outside the secondary eclipse.
• 3.  
  Ellipsoidal variation, ${A}_{e}\sin 4\pi \phi$, i.e., a sinusoid at twice the orbital period that peaks at quadrature (0.25 and 0.75) phases.
• 4.  
  An additive constant offset to absorb any normalization bias.

A_d, A_n, and A_e parameterize the amplitudes of these components.

We sampled from the posterior distribution of the global model parameters using uniform priors. The posterior median phase-curve components are shown in Figure 7, where the black points denote the light-curve phase-folded at the posterior median period and binned. The blue line shows the posterior median model fitted to the data. The colored dashed lines show the individual components of the posterior median model. The stellar baseline, planetary baseline, planetary modulation, and the ellipsoidal variation are shown in Figure 1 with orange, olive, magenta, and red colors, respectively. We obtain a secondary eclipse depth of ${59}_{-21}^{+24}$ ppm (Dragomir et al. 2020), a phase-curve amplitude of ${25}_{-13}^{+14}$ ppm, and a nightside emission of ${33}_{-20}^{+24}$ ppm. The photometric precision we achieve is in line with expectations from the demonstrated performance of TESS: the star is TESS mag = 9, so the expected photometric precision is roughly 100 ppm hr^−1/2 (Ricker et al. 2014). The phase curve plotted in Figure 7 has been averaged down into 100 bins, each containing roughly one hundred fifty 2 minute cadence measurements—i.e., 5 hr. So the expected 1σ precision is $100/\sqrt{(}5)\approx 45\,\mathrm{ppm}$, consistent with our achieved precision. Figure 8 shows how the residuals to our TESS analysis bin down over time, which suggests residual correlated noise of no more than 15% on the timescale of LTT 9779b's transit duration.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Our IRAC transits are mainly useful for refining the planet's orbital ephemeris and reducing the uncertainty on its orbital period from that reported in the initial TESS analysis of Jenkins et al. (2020) and the updated analysis of Dragomir et al. (2020). Our 3.6 μm and 4.5 μm transit analyses yield mid-center times of T₀ = 2458781.13997 ± 0.00032 and 2458783.51684 ± 0.00053, respectively (BJD_TDB). The IRAC1 and IRAC2 transits occurred 539 and 542 orbits, respectively, after the discovery epoch (Jenkins et al. 2020), which reported T₀ = 2458354.21430 ± 0.00025 and P = 0.7920520 ± 0.0000093. With our new Spitzer transit times, a weighted least-squares fit gives an updated mid-transit time of 2458783.51636 ± 0.00027 (BJD_TDB) and a refined period of 0.79207022 ± 0.00000069 days, a measurement 13.6× more precise than the discovery period.

Our 3.6 μm and 4.5 μm transit depths are R_p/R_* = 0.0431 ± 0.0011 and 0.0452 ± 0.0017, respectively, consistent with the TESS transit depth (Jenkins et al. 2020). Neither the infrared nor the optical transit measurements are sufficiently precise to usefully constrain LTT 9779b's atmosphere in transmission, because the planet's expected scale height is relatively small (Jenkins et al. 2020).

Our phase-curve model (Equation (1)) can be used to determine the hemisphere-averaged flux and brightness temperature from the planet's daysides and nightsides. To effect this conversion, we use the BT-Settl library of synthetic stellar spectra (Allard 2014) to properly model the non-blackbody stellar flux. We perform this calculation in a Monte Carlo framework in order to properly account for the uncertainties in both the secondary eclipse and phase-curve amplitude. As expected, we also find that the large uncertainties on the phase-curve parameters dominate the much smaller uncertainties in the stellar parameters.

We calculate 4.5 μm dayside and nightside temperatures of T_day = 1800 ± 120 and T_night = 700 ± 430, with upper limits on the nightside brightness temperature of <1350 K and <1650 K at 95.4% and 99.73% confidence (2σ and 3σ), respectively. Note that our measurement of the dayside brightness temperature is somewhat more precise than that derived from the secondary eclipse analysis of Dragomir et al. (2020), since our analysis also includes the full phase curve.

Combining the posterior distributions of the dayside and nightside temperatures, we find a day–night temperature contrast of ΔT = 1110 ± 460 K. The distribution of ΔT is symmetric but non-Gaussian, with 95.4% and 99.73% (2σ and 3σ) confidence intervals of ±720 K and ±1030 K, respectively. We also measure the planet's phase offset (the longitude of maximum flux) to be −10° ± 21° east of the substellar point. Our full set of phase-curve measurements from our Spitzer analysis are listed in Table 2.

In our TESS light-curve analysis, we sampled from the posterior distribution of A_d, A_n, and A_e (see Section 3.3) using uniform priors. The posterior median of these components are plotted against the TESS photometry in Figure 7, along with the posterior median of the total model. As described by Dragomir et al. (2020), we find a TESS secondary eclipse depth of ${55}_{-21}^{+24}$ ppm (a roughly 3σ detection). However, the planet's phase variation is not detected at high significance: we find a phase amplitude in the TESS bandpass of just 29 ± 17 ppm. Further TESS observations of the system, anticipated in the TESS extended mission, would naturally improve the precision of these measurements.

To better interpret our observations of LTT 9779b, we calculated 3D general circulation models (GCMs) of this planet utilizing the SPARC/MITgcm (e.g., Showman et al. 2009). To reflect the range of possible observed atmospheric compositions we used two different model configurations, with atmospheric elemental abundances at the solar level and 30× the solar level. Note that our simulations do not include the effect of atmospheric condensates, which some studies suggest could be important for planets even at these high temperatures (e.g., Parmentier et al. 2016). In particular, atmospheric aerosols would likely increase the geometric albedo in the TESS bandpass and so bring our GCM predictions into better agreement with the data at these shorter wavelengths.

Figure 9 compares the 4.5 μm and TESS phase curves predicted by our GCMs with our observations, and some relevant values for comparison are also enumerated in Table 3. At optical wavelengths, our GCMs predict lower flux than we observe with TESS at all orbital phases. At 4.5 μm, when compared to the observations the 30× solar GCM predicts a hotter dayside and smaller phase offset, while the solar-abundance GCM predicts a somewhat cooler dayside, smaller day–night temperature contrast, and larger phase offset. The trend seen in Figure 9 of a phase-curve amplitude that increases with atmospheric metallicity mirrors that seen in similar simulations for the similarly sized, but much cooler, warm Neptune GJ 436b (Lewis et al. 2010). Nonetheless, neither of our GCMs provides an especially good match to our observations, which exhibit a much colder nightside and smaller (consistent with zero) phase offset. Indeed, if anything the phase curve from the solar-abundance GCM is a better match to our observations than the 30× solar model. And although our constraint on the phase offset is fairly loose, our measured dayside flux and phase-curve amplitude both disagree with those predicted by the GCMs.

Zoom In Zoom Out Reset image size

Table 3.  Phase-curve Properties

Bandpass	Source	Amplitudea	Offsetb	FP/F* (ppm) at:
		(ppm)	(deg)	ϕ = 0	ϕ = 0.5
TESS	data	29 ± 17	⋯c	${33}_{-20}^{+24}$	${59}_{-21}^{+24}$ d
	Solar abundances	1.0	77	2.5	2.6
	30× Solar	7.8	41	4.7	10.7
Spitzer 4.5 μm	data	358 ± 106	−10 ± 21	17 ± 123	375 ± 62d
	Solar abundances	113	60	241	291
	30× Solar	261	23	272	511

Notes.

^aPeak-to-valley phase-curve amplitude. ^bEastward shift of the phase-curve maximum from the substellar point. ^cZero phase offset assumed; see Section 3.3. ^dFrom Dragomir et al. (2020).

Download table as:  ASCIITypeset image

In Figure 10, we also compare our measurements of the planet's dayside and nightside emission to the phase-resolved emergent spectra predicted by the GCMs. These reinforce the impressions conveyed by the comparison in Figure 9 and Table 3: at 4.5 μm the solar-abundance GCM overpredicts the nightside flux but underpredicts the dayside flux, while the 30× solar GCM overpredicts the planet's flux at all phases; our (aerosol-free) GCMs underestimate the flux emitted in the TESS bandpass by 2.1σ and 2.4σ (for the solar and 30× solar-abundance GCMs, respectively). On balance, the 30× solar GCM matches our observations somewhat better though (unlike Figure 9) the comparison in Figure 10 includes the 3.6 μm dayside (eclipse) measurement and so more strongly favors the 30× solar GCM (Δχ² = 15), hinting at a supersolar metallicity for LTT 9779b's atmosphere. However, Figure 10 also demonstrates that this model (unlike the solar-abundance model) predicts a nearly isothermal dayside photosphere with only very weak spectral features, contrary to the strong absorption feature implied by the discrepant planetary brightness temperatures at 3.6 μm and 4.5 μm.

Zoom In Zoom Out Reset image size

The large phase-curve amplitude and small phase offset in our 4.5 μm phase curve both suggest that LTT 9779b has a high-metallicity atmosphere. Higher-metallicity atmospheres have higher opacities, and Neptune-size planets may have atmospheric metallicities of 100× solar or more (Fortney et al. 2013; Wakeford et al. 2017), much greater than expected for hot Jupiters. Therefore, the atmospheric layers probed by thermal emission measurements may tend to be at lower pressures for Neptunes than for Jovians, with consequently shorter radiative timescales. As a result, the phase-curve offsets are expected to be smaller and the phase amplitudes larger than for the lower-metallicity hot Jupiters (Parmentier et al. 2018). Our phase offset measurement is not precise enough for a useful comparison, but the enhanced phase amplitude that we see suggests that LTT 9779b's atmosphere may have an enhanced metallicity. The host star has a supersolar metallicity of [Fe/H] = +0.25 dex, but based on the GCM simulations shown in Figure 9 the planet's metallicity would be much more metal-rich, with [Fe/H] ≳ 1.5 dex.

We also used these brightness temperature with the radiative–balance model of Cowan & Agol (2011a) to estimate LTT 9779b's Bond albedo A_B and global heat recirculation efficiency , as shown in Figure 11(a). In this framework, equals zero or unity, respectively, for zero circulation or full day–night recirculation. We find  = 0.06 ± 0.18 (2σ upper limit of <0.49), indicating a very low level of heat redistribution consistent with the planet's nearzero phase offset. For LTT 9779b's Bond albedo we find a surprisingly high value of A_B = 0.72 ± 0.12 from our 4.5 μm data. Presumably this A_B value is so high because CO and/or CO₂ absorb strongly in this planet's atmosphere 4.5 μm (Dragomir et al. 2020), and so relatively less emission is seen in our IRAC2 observations. In this case, the high A_B we derive would reflect the limits of single-band photometry for detailed energy balance calculations. Cowan & Agol (2011a) found that a single broadband infrared brightness temperature measurement translates into a roughly 8% systematic uncertainty in the effective temperature, and hence into a roughly 32% systematic uncertainty in bolometric flux and the Bond albedo. As a check on A_B, we repeated the calculation using the 3.6 μm eclipse measurement and assuming all 3.6 μm flux is emitted on the dayside. In this case we calculate A_B = 0.27 ± 0.16 (or ${0.27}_{-0.26}^{+0.41}$ at 99.73% confidence), which would not be remarkable in the context of the many hot Jupiters that have been studied in this way (see Figure 12(f)).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

On the other hand, LTT 9779b has (R_P/a)² = 142 ppm, which together with the TESS secondary eclipse indicates a geometric albedo in the TESS bandpass of roughly 0.4 if all the optical light were the result of scattering. A 2100 K blackbody would contribute only roughly 10 ppm of thermal emission at these wavelengths (see Figure 9(b)), though such a planet's spectrum is unlikely to closely resemble a blackbody. The optical data may therefore hint at a moderately high geometric albedo, which would differ qualitatively from the nearzero geometric albedo measured for some hot Jupiters (e.g., WASP-12b; Bell et al. 2017). Nonetheless, better flux measurements are needed to confirm this point.

Finally, we used a similar modeling approach to evaluate LTT 9779b's 4.5 μm phase curve in terms of an alternative two-parameter model in which a planet's global temperature distribution depends on A_B and the ratio of radiative to advective timescales, τ_rad/τ_adv (Cowan & Agol 2011b; see the Appendix for some pedagogically useful analytic insights associated with this model). The posterior distribution of these parameters is shown in Figure 11(b), indicating A_B = 0.71 ± 0.04 (consistent with the measurement described immediately above) and a 2σ upper limit of τ_rad/τ_adv < 3.8 (consistent with expectations for such a hot planet).

To put LTT 9779b's phase curve and derived parameters in the context of other 4.5 μm Spitzer phase curves, we followed the same procedures as described above to calculate T_day, T_night, ΔT, A_B, and for all planets with published IRAC2/4.5 μm phase curves. These measurements are shown in Figure 12, along with the predictions for LTT 9779b from our GCMs. We restrict this comparison to the 16 planets with nearly circular orbits (Knutson et al. 2012; Maxted et al. 2013; Zellem et al. 2014; Wong et al. 2015, 2016; Stevenson et al. 2017; Demory et al. 2016; Dang et al. 2018; Kreidberg et al. 2018, 2019; Zhang et al. 2018; Bell et al. 2019; Keating et al. 2020; Mansfield et al. 2020a; May & Stevenson 2020).

Figure 12 allows us to examine the sample of 4.5 μm exoplanetary phase curves for possible trends between these planetary parameters. For example, Keating et al. (2019) and Beatty et al. (2019) claimed that the roughly uniform, ∼1100 K nightside temperatures of a sample of 12 hot Jupiters indicated the ubiquitous presence of nightside clouds. Our updated sample shows that this trend generally continues to hold, with the exceptions of hot Jupiters CoRoT-2b and KELT-9b (Dang et al. 2018; Mansfield et al. 2020a) and much smaller, presumed-rocky LHS 3844b (which has at most a tenuous atmosphere; Kreidberg et al. 2019). In a similar vein, Zhang et al. (2018) used Spitzer phase curves of 10 hot Jupiters to claim a coherent trend between T_irr and eastward phase offsets, which would indicate the increasing importance of magnetohydrodynamical effects at higher temperatures as well as high-altitude clouds. However, several recent measurements do not seem to support this trend, in particular the low phase offsets of Qatar-1b (consistent with zero shift at 2000 K; Keating et al. 2020) and CoRoT-2b (a westward shift of 23 ± 4° at 2100; K Dang et al. 2018).

With two exceptions, the ensemble of 4.5 μm phase-curve measurements depicted in Figure 12 shows that the parameters of LTT 9779b derived from its phase curve are generally similar to those of the population of hot Jupiters. The two exceptions, which stem from a single explanation, are that LTT 9779b exhibits an unusually low dayside temperature, and therefore has a high Bond albedo as derived from 4.5 μm observations (as discussed above), when compared to hot Jupiters with similar irradiation temperatures. These discrepancies are due to the 4.5 μm absorption feature inferred in LTT 9779b's dayside spectrum (Dragomir et al. 2020). In contrast, most hot Jupiters at these temperatures exhibit more nearly blackbody-like dayside emission spectra due to shallower photospheric temperature gradients.

Observations inside molecular bands (e.g., the CO/CO₂ band at 4.5 μm) probe shallower atmospheric layers. This implies that our 3.6 μm phase curve should show a larger phase offset relative to the 4.5 μm phase curve, since it probes deeper into the atmosphere where energy re-radiation is less rapid and advective heat transport is more efficient. It is tempting to interpret the large phase offset apparently seen in the 3.6 μm phase curve (Figure 5) as evidence of this phenomenon, but the nonphysical models required to explain this photometry (along with its high levels of correlated noise) caution against such an interpretation. Nevertheless, phase curves of this planet at additional wavelengths, probing a broader range of pressures, would provide a powerful diagnostic for measuring thermal structure and chemical abundances across the entire planet.

A particularly interesting comparison is between LTT 9779b and WASP-19b, an inflated hot Jupiter with multiband eclipse and phase-curve measurements (Wong et al. 2016, and references therein). The surface gravity and irradiation temperatures of these two planets are within 10% of each other, as shown in Table 4. Given two stars with such similar T_eff and log g, one would expect any spectral differences to result solely from different metallicities; the same argument should presumably apply to planets as well. Yet WASP-19b has an essentially featureless dayside spectrum (unlike LTT 9779b), and the brightness temperature differences in the two warm Spitzer bandpasses, T_day(3.6 μm) − T_day(4.5 μm), are greater for LTT 9779b than for WASP-19b at 99.1% confidence.²⁴ Table 4 also compares our hot Neptune to WASP-14b, another similarly irradiated hot Jupiter but with 10× greater surface gravity. The case here is similar, with the two-color brightness temperature difference greater for LTT 9779b at 98.2% confidence. Although both additional data and further modeling are warranted, the different Spitzer emission spectra of LTT 9779b and WASP-14b and WASP-19b may be explained by the tendency of higher-metallicity atmospheres to show more strongly decreasing thermal profiles at T_eq ≲ 2300 K (Mansfield et al. 2020b). Thus, this comparison with hot Jupiters provides another tentative line of evidence that our target may have a significantly different (presumably higher) atmospheric metallicity than do hot Jupiters.²⁵

Table 4.  Brightness Temperatures of Three Hot Exoplanets

Planet	WASP-14b	WASP-19b	LTT 9779b
Irradiation Temperature (K)	2640 ± 43	2990 ± 50	2760 ± 33
Planet Mass (MJup)	7.3 ± 0.5	1.114 ± 0.036	0.0922 ± 0.0025
Planet Radius (RJup)	1.28 ± 0.08	1.395 ± 0.023	0.421 ± 0.021
Surface Gravity (${\mathrm{log}}_{10}$ [cgs])	4.107 ± 0.043	3.152 ± 0.020	3.110 ± 0.044
Tday (3.6 μm) (K)	2341 ± 37	2361 ± 48	2305 ± 141
Tday (4.5 μm) (K)	2241 ± 45	2331 ± 67	1800 ± 120
Tnight (4.5 μm) (K)	1301 ± 69	1130 ± 320	700 ± 430
Tday(3.6 μm) − Tday(4.5 μm) (K)	100 ± 58	30 ± 82	510 ± 180
Tday − Tnight (4.5 μm) (K)	940 ± 78	1200 ± 330	1110 ± 460

Note.

^aMeasurements taken from Wong et al. (2015) (WASP-14b), Wong et al. (2016) (WASP-19b), Dragomir et al. (2020), and this work (LTT 9779b).

Download table as:  ASCIITypeset image

One of the key goals of the ongoing TESS survey is to identify the best exoplanet targets for detailed atmospheric characterization (Ricker et al. 2014). TESS is already discovering such systems, opening wider the door to atmospheric studies of smaller planets. Among the most exciting TESS planets are those for which existing facilities can already answer new questions about the atmospheres of new classes of objects or of individual planets. LTT 9779 is such a system.

We have presented infrared and optical phase-curve observations of this unusual hot Neptune, building on our analysis of its dayside emission spectrum through secondary eclipses in Dragomir et al. (2020). The planet's phase curve is clearly seen in our Spitzer 4.5 μm photometry (Figure 3); these data detect the signature of heat recirculation on this planet, with a 3σ confidence interval on the day–night temperature contrast of 1110 ± 1030 K. Our observations plug a glaring gap, since Figure 13 shows that no infrared phase-curve data exist for any similar planet (i.e., within a factor of 3 of LTT 9779b's mass and a factor of 2 of its temperature).

Zoom In Zoom Out Reset image size

Overall, via several lines of evidence our measurements hint at an atmospheric metallicity enhanced above the solar level. These arguments are mainly derived from the strong constraints placed on the planet's global circulation patterns from the 4.5 μm phase curve. First, the eastward phase offset of −10 ± 21° (to the west) is more consistent with that predicted by our general circulation models (GCMs) that are preferentially enhanced in heavy elements (see Figure 9 and Table 3). Such a small phase offset also indicates a low heat recirculation efficiency and relatively small ratio of radiative to advective timescales, consistent with the values seen for hot Jupiters at comparable temperatures. In addition, LTT 9779b's 4.5 μm phase-curve amplitude of 358 ± 106 ppm is also consistent with our enhanced-metallicity GCM predictions (see Figure 9 and Table 3), though that model's orbit-averaged infrared emission is elevated compared to our observations; future modeling at higher atmospheric metallicity and including the effects of aerosols may be needed to conclusively interpret our observations (see Kataria et al. 2015; Parmentier et al. 2016; Keating et al. 2019). Including the 3.6 μm dayside observation in the GCM comparison (Figure 10) also shows a better (though far from perfect) match between the observations and our higher-metallicity model.

Furthermore, the GCM predictions may also underpredict the planet's optical-wavelength emission seen by TESS, though this result is more tentative (TESS is scheduled to observe LTT 9779b for another month-long sector in late 2020, which should tighten the measurements at optical wavelengths). The optical observations may hint at a nonzero geometric albedo, while the two-channel Spitzer measurements do not tightly constrain the planet's Bond albedo.

Finally, further support for a high-metallicity atmosphere comes from our comparison of LTT 9779b's thermal phase-curve and dayside emission measurements with two similarly irradiated hot Jupiters, WASP-14b and WASP-19b (see Table 4). In particular, although the surface gravity and irradiation temperature of WASP-19b differ from those of LTT 9779b by <10%, the two planets have qualitatively different emission spectra at >99% confidence. The hot Jupiters WASP-14b and WASP-19b both exhibit approximately featureless, pseudo-blackbody emission spectra devoid of spectral features, in contrast to the absorption inferred in our hot Neptune's broadband emission spectrum (Dragomir et al. 2020); models also suggest that stronger absorption at this irradiation temperature is consistent with a higher-metallicity atmosphere (Mansfield et al. 2020b).

The primary, two-year TESS mission is nearly complete. When done, 85% of the sky will have been surveyed for planets transiting nearby stars. To date, LTT 9779b remains one of the best targets of its type—i.e., among highly irradiated hot Neptunes that are highly favorable for thermal emission measurements—and so it is likely to remain one of the most easily characterizable exoplanets in its class. Future phase-curve and eclipse observations of LTT 9779b and other similar planets will provide the impetus for the next generation of global circulation modeling of Neptune-size planets, supporting similar observations of many such objects with James Webb Space Telescope, ARIEL, and future observatories.