A feature-rich transmission spectrum for WASP-127b

TL;DR

This study presents the first transmission spectrum of WASP-127b, revealing potential atmospheric components like TiO, VO, and Na, and suggests a cloud-free atmosphere suitable for future detailed characterization.

Contribution

First low-resolution transmission spectrum of WASP-127b, indicating possible atmospheric constituents and a cloud-free atmosphere, aiding future atmospheric studies.

Findings

Detection of a strong Rayleigh slope at blue wavelengths

Hint of Na absorption in the spectrum

Possible presence of TiO and VO absorption features

Abstract

WASP-127b is one of the lowest density planets discovered to date. With a sub-Saturn mass ($M_{\rm p}=0.18 \pm 0.02 M_J$) and super-Jupiter radius ($R_{\rm p}= 1.37 \pm 0.04 R_J$), it orbits a bright G5 star, which is about to leave the main-sequence. We aim to explore WASP-127b's atmosphere in order to retrieve its main atmospheric components, and to find hints for its intriguing inflation and evolutionary history. We used the ALFOSC spectrograph at the NOT telescope to observe a low resolution ($R\sim330$, seeing limited) long-slit spectroscopic time series during a planetary transit, and present here the first transmission spectrum for WASP-127b. We find the presence of a strong Rayleigh slope at blue wavelengths and a hint of Na absorption, although the quality of the data does not allow us to claim a detection. At redder wavelengths the absorption features of TiO and VO are the best explanation to fit the data. Although higher signal-to-noise ratio observations are needed to conclusively confirm the absorption features, WASP-127b seems to posses a cloud-free atmosphere and is one of the best targets to perform further characterization studies in the near future.

Figures

Tables

Table 1. Table 1: Derived system parameters for white light curve analysis

Parameter	Value
$P$ [days]	4.178062 (fixed)
$e$	0 (fixed)
$T_{\mathrm{mid}}$ [${\mathrm{BJD}}_{\mathrm{TDB}}$]	2457808.60283 $\pm$ 0.00031
$i$ [∘]	88.2 ${}_{-0.9}{}^{}{}_{}{}^{+1.1}$
$a/R_{\star }$	7.95 ${}_{-0.27}{}^{}{}_{}{}^{+0.19}$
$R_p/R_{\star }$	0.1004 $\pm$ 0.0014
$u_1$	0.365 $\pm$ 0.057
$u_2$	0.258 ${}_{-0.85}{}^{}{}_{}{}^{+0.88}$

Table 2. Table 2: Goodness of fit for different atmospheric models

Model	415–885 nm		415–655 nm
	${\chi }^2$	$P\left({\chi }^2\right)$	${\chi }^2$	$P\left({\chi }^2\right)$
Pure Rayleigh Scattering (RS)	22.11	0.453	4.89	0.936
Flat	18.83	0.656	8.43	0.645
1$\times$solar, Na/K, clear	40.41	0.010	26.59	0.005
0.1$\times$solar, Na/K, clear	30.64	0.104	14.71	0.196
1$\times$solar, Na/K/TiO/VO, 5$\times$RS	24.97	0.299	10.61	0.477
0.1$\times$solar, Na/K/TiO/VO, clear	23.95	0.350	13.40	0.268
0.1$\times$solar, Na/K, 5$\times$RS	23.29	0.386	5.79	0.887
0.1$\times$solar, Na/K/TiO/VO, 5$\times$RS	13.60	0.915	5.51	0.904
0.1$\times$solar, TiO/VO, 5$\times$RS	11.73	0.963	4.45	0.955

Table 3. Table 3: Transmission spectrum values obtained with NOT/ALFOSC

#	Wavelength (nm)		$R_p/R_{\star }$	$R_p/R_{\star }$
	Center	Width	(With 2nd order)	(Corrected)
1	425	20	0.0977 $\pm$ 0.0045	0.0977 $\pm$ 0.0045
2	445	20	0.0984 $\pm$ 0.0029	0.0984 $\pm$ 0.0029
3	465	20	0.1022 $\pm$ 0.0025	0.1022 $\pm$ 0.0025
4	485	20	0.1011 $\pm$ 0.0018	0.1011 $\pm$ 0.0018
5	505	20	0.0986 $\pm$ 0.0024	0.0986 $\pm$ 0.0024
6	525	20	0.0996 $\pm$ 0.0019	0.0996 $\pm$ 0.0019
7	545	20	0.0976 $\pm$ 0.0022	0.0976 $\pm$ 0.0022
8	565	20	0.0950 $\pm$ 0.0029	0.0950 $\pm$ 0.0029
9	585	20	0.0990 $\pm$ 0.0017	0.0990 $\pm$ 0.0017
10	605	20	0.0977 $\pm$ 0.0019	0.0971 $\pm$ 0.0019
11	625	20	0.0959 $\pm$ 0.0025	0.0955 $\pm$ 0.0027
12	645	20	0.0975 $\pm$ 0.0026	0.0970 $\pm$ 0.0028
13	665	20	0.0994 $\pm$ 0.0019	0.0991 $\pm$ 0.0020
14	685	20	0.1009 $\pm$ 0.0025	0.1006 $\pm$ 0.0026
15	705	20	0.1009 $\pm$ 0.0024	0.1002 $\pm$ 0.0026
16	725	20	0.1033 $\pm$ 0.0023	0.1018 $\pm$ 0.0023
17	745	20	0.1019 $\pm$ 0.0029	0.1006 $\pm$ 0.0030
18	775	20	0.0988 $\pm$ 0.0037	0.0973 $\pm$ 0.0043
19	795	20	0.1019 $\pm$ 0.0020	0.1000 $\pm$ 0.0020
20	815	20	0.0962 $\pm$ 0.0029	0.0940 $\pm$ 0.0029
21	835	20	0.0965 $\pm$ 0.0025	0.0946 $\pm$ 0.0027
22	855	20	0.0966 $\pm$ 0.0026	0.0948 $\pm$ 0.0030
23	875	20	0.0981 $\pm$ 0.0040	0.0971 $\pm$ 0.0043

Equations

$${\mathcal{B}}_{\mathrm{w}}=$$

$${\mathcal{B}}_{\mathrm{spec}}(\lambda )=$$

Full text

*Aims. *We aim to explore WASP-127b’s atmosphere in order to retrieve its main atmospheric components, and to find hints for its intriguing inflation and evolutionary history.

*Methods. *We used the ALFOSC spectrograph at the NOT telescope to observe a low resolution ($R\sim 330$, seeing limited) long-slit spectroscopic time series during a planetary transit, and present here the first transmission spectrum for WASP-127b.

*Results. *We find the presence of a strong Rayleigh slope at blue wavelengths and a hint of Na absorption, although the quality of the data does not allow us to claim a detection. At redder wavelengths the absorption features of TiO and VO are the best explanation to fit the data.

*Conclusions. *Although higher signal-to-noise ratio observations are needed to conclusively confirm the absorption features, WASP-127b seems to posses a cloud-free atmosphere and is one of the best targets to perform further characterization studies in the near future.

1 Introduction

The atmospheres of exoplanets are a unique window to investigate the planetary chemistry, which can help improve our understanding of planetary interior properties and provide links to planet formation and migration histories (e.g., Guillot, 2005; Fortney et al., 2007; Fortney & Nettelmann, 2010; Öberg et al., 2011; Mousis et al., 2012; Madhusudhan et al., 2014, 2016). Transmission spectroscopy retrieves the absorption and scattering signatures from the atmosphere at the planetary day-night terminator region. These signatures are only imprinted on the stellar light when it is transmitted through the planetary atmosphere during a transit, and they can be extracted through the differential method when compared to out-of-transit measurements. Such studies have been carried out by many ground-based large telescope and space telescope, in a wide range of spectral resolutions (e.g., Charbonneau et al., 2002; Snellen et al., 2010; Bean et al., 2010; Sing et al., 2016), resulting in robust detections of Na, K, H2O, CO, and scattering hazes (see the inventory listed in Bailey 2014 and Sing et al. 2016). A recent HST+Spitzer survey led by Sing et al. (2016) performed a comparative study on ten hot Jupiters covering $0.3-5\mu m$. This diverse hot Jupiter sample reveals a continuum from clear to cloudy atmospheres, and suggests clouds/hazes as the cause of weakened spectral features.

As the investigated sample increases, it is fundamental to construct a spectral sequence for exoplanets, for a global picture of population characteristics and formation/evolution scenarios, as we have achieved for stars and brown dwarfs. In the near-future the JWST will provide spectral resolutions at high SNR with a large wavelength coverage $0.6-28\mu m$ that can distinguish among different atmospheric compositions. However, ground-based observations can also complement JWST by extending the wavelength range to $\lambda<600nm$, which is critical to examine spectral signatures arising from Rayleigh scattering, Na, or TiO/VO (Murgas et al., 2014; Nortmann et al., 2016; Chen et al., 2017a, b). The ideal starting point are low density planets, which are more likely to host extended atmospheric envelopes that can produce stronger transmission signals if cloud-free.

WASP-127b (Lam et al., 2017), with a mass of $0.18\pm 0.02M_{J}$ and a radius of $1.37\pm 0.04R_{J}$, is the puffiest, lowest density, planet discovered to date. It has an orbital period of 4.18 days, and orbits a bright parent star (V = 10.2), which makes it a very interesting object for atmospheric follow-up studies.

WASP-127b’s host star is a G5 star which is at the end of the main-sequence phase and moving to the sub-giant branch (Lam et al., 2017). Moreover, the unusually large radius (compared to its sub-Saturn mass) cannot be explained by the standard coreless model (e.g., Fortney et al., 2007), and places it into the short-period Neptune desert, a region between Jovian and super-Earth planets with a lack of detected planets (Howard et al., 2012; Mazeh et al., 2016). Several inflation mechanisms have been proposed to explain this inflation, including tidal heating, enhanced atmospheric opacity, Ohmic heating, and/or re-inflation by host star when moving towards the RGB phase (Leconte et al., 2010; Batygin & Stevenson, 2010; Batygin et al., 2011; Lammer et al., 2013; Rauscher & Menou, 2013; Spiegel & Burrows, 2013; Wu & Lithwick, 2013; Lopez & Fortney, 2016), although no concluding observations have yet been established to favour one or the other. Therefore, the formation and evolution mechanisms of WASP-127b are very intriguing, given its transition size between these two classes of planets.

2 Observations and data reduction

We observed one transit of WASP-127b on the night of February 23rd, 2017, using the Andalucia Faint Object Spectrograph and Camera (ALFOSC) mounted at the 2.5 m Nordic Optical Telescope (NOT) at ORM observatory. ALFOSC has a field of view of 6’.4x6’.4 and a 2048x2048 E2V detector with a pixel size of 0”.2. The observation was carried out in the long-slit mode using a 40” wide slit to avoid flux losses, and placing both WASP-127 and a reference star simultaneously aligned into the slit. The reference star TYC 4916-897-1 is located 40”.5 away from WASP-127 and it is about one magnitude fainter (V = 11.2) over the observed spectral range. Grism #4 was used covering simultaneously the spectral range from 320-960 $nm$. Observations started at 23:45 UT and ended at 05:36 UT, resulting in a time series of 746 spectra. Exposure times were set to 20s. The transit of WASP-127b ($T_{14}$) started at 00:19 UT and ended at 04:38 UT, resulting in 554 spectra taken within transit. The night was clear, with a relatively stable seeing of around 0”.5 during the full observation. The airmass changed from 1.35 to 1.19, then to 2.45.

Data reduction was carried out using the approach outlined in Chen et al. (2016, 2017a) for similar OSIRIS long-slit data taken with the GTC. The one-dimensional spectra (see Figure 1) were extracted using the optimal extraction algorithm (Horne, 1986) with an aperture diameter of 13 pixels, which minimized the scatter for the white-color light curves created from various trial aperture sizes. The time stamp was centered on mid-exposure and converted into the Barycentric Dynamical Time standard (BJDTDB; Eastman et al., 2010). Misalignment between the target and reference stars in the wavelength solutions and any spectral drifts were corrected. Then the requested wavelength range of a given pass-band was converted to a pixel range, and the flux was summed to generate the time series.

A broad-band (white-color) light curve was integrated from 395 nm to 945 nm, excluding the range of 755–765 nm to eliminate the noise introduced by the oxygen-A band (Parviainen et al., 2016), and used to derive the transit parameters in Figure 2. Moreover several narrow band light curves were constructed to study the wavelength-dependence of the transit depth and derive the transmission spectrum (see Figure 1 for the band ranges).

3 Light-curve analysis

The light-curve data were modeled in the approach detailed in Chen et al. (2016, 2017a). In brief, the light-curve model contains two multiplicative components. One component describes the astrophysical signal, which adopts the analytic transit model $\mathcal{T}(p)$ proposed by Mandel & Agol (2002). The other component describes the systematics of telluric or instrumental origins in a fully parametric form or in a semi-empirical form, which is designated as the baseline model $\mathcal{B}(c_{i})$.

The transit model $\mathcal{T}(p)$ was parameterized as orbital period $P$, inclination $i$, scaled semi-major axis $a/R_{\star}$, planet-to-star radius ratio $R_{\rm p}/R_{\star}$, mid-transit time $T_{\rm mid}$, and limb-darkening coefficients $u_{i}$, where a circular orbit was assumed. The orbital period $P$ was fixed to 4.178062 days as reported by Lam et al. (2017). A quadratic limb-darkening law was adopted and conservatively constrained by Gaussian priors of width $\sigma=0.1$, whose central values were calculated from the ATLAS atmosphere models following Espinoza & Jordán (2015) with stellar parameters $T_{\rm eff}=5750$ K, $\log g=3.9$, and $[\rm{Fe/H}]=-0.18$.

The baseline model $\mathcal{B}(c_{i})$ consisted of a selected combination of auxiliary state vectors, including spectral and spatial position drifts ($x$, $y$), spectra’s full width at half maximum (FWHM) in the spatial direction ($s_{y}$), airmass ($z$), and time sequence ($t$). The Bayesian information criterion (BIC; Schwarz, 1978) was used to find the baseline model that can best remove the systematics. For the white-color light-curve, the model

[TABLE]

gave the lowest BIC value. The second best model yields a value of $\Delta\mathrm{BIC}=3.3$ higher. For the spectroscopic light-curves, the model was chosen in a semi-empirical form:

[TABLE]

which inherited a common-mode component $\mathcal{S_{\mathrm{w}}}$ determined from the white-color light-curve. The common-mode systematics $\mathcal{S_{\mathrm{w}}}$ were derived after dividing the white-color light-curve by the best-fitting transit model $\mathcal{T}(p)$.

The Transit Analysis Package (TAP; Gazak et al., 2012), customized for our purposes, was employed to perform the Markov chain Monte Carlo analysis. The correlated noise was taken into account by the wavelet-based likelihood function proposed by Carter & Winn (2009). The overall transit parameters were determined from the white-color light-curve, whose best-fitting values and associated uncertainties were calculated as the median and 1$\sigma$ percentiles of the posterior probability distributions and listed in Table 1. For the spectroscopic light-curves, only the planet-to-star radius ratio $R_{\rm p}/R_{\star}$, the limb-darkening coefficients $u_{i}$, and the baseline coefficients $c_{i}$ were fit, while the other transit parameters were fixed to the ones determined from the white-color light-curve. The wavelength-dependent radius ratios are presented in Table 3. The white-color and spectroscopic light-curves are shown in Fig. 2 and 4, respectively.

4 Results and Discussion

4.1 Second order contamination

When using the grism #4 with ALFOSC, second order contamination can be present due to the overlap in the detector of different diffraction orders (Stanishev, 2007). To check this issue, on March 21st 2017, we performed consecutive observations of WASP-127 with the grism #4, with and without the second order blocking filters #101 (GG475) and #102 (OG515). We find that for WASP-127, second order contamination of the stellar flux appears at 1% level at 655 $nm$ and rises nearly monotonically reaching 10% at 900 $nm$. Following the approach of Stanishev (2007), we were able to directly remove the second order component of the blue light from the first order stellar spectra, and then to derive a new transmission spectrum. As shown in Fig. 3, this correction makes the transit depths slightly smaller at red wavelengths ($\lambda\gtrsim 600$ nm), which agree with the original ones well within the error bars and still show the same relative spectral shape.

4.2 Transmission spectrum

To interpret the transmission spectrum of WASP-127b, a series of atmospheric models with an isothermal temperature structure were generated using the Exo-Transmit code (Kempton et al., 2016). Various metallicities, chemical compositions (with or without the presence of Na, K, TiO, VO), and weather conditions (clear, hazy, or cloudy) were considered. We also analytically calculated a pure Rayleigh scattering model following the approach of Lecavelier Des Etangs et al. (2008), and used a simple flat straight line to represent the gray absorbing clouds.

It is clear from Fig. 3 that the transmission spectrum is not flat, but on the contrary it has strong spectral features. It is not surprising that we can detect spectral features even using a relatively small aperture telescope, given that one atmospheric scale height, $H$, of WASP-127b corresponds to approximately 2500 $km$ (equivalent to a signal of $510\leavevmode\nobreak\ ppm$) assuming an H-He atmosphere, and that the amplitude of a given spectral signature can typically achieve about $\leavevmode\nobreak\ 5H$ (e.g., Seager, 2010).

At the bluer wavelengths, the spectrum shows a decreasing slope with $\lambda$, which seems to indicate the presence of Rayleigh scattering. A hint of Na absorption is seen (although statistically insignificant), with the band centered on the Na doublet presenting a larger $R_{\rm p}/R_{\star}$ value than the surrounding bands. Unfortunately, the analysis of narrower pass bands around Na did not provide more information but increasing noise (not shown). No K absorption is seen. Toward the red, strong absorptions from TiO and VO molecules seem to dominate the spectral shape.

Fitting the different models to the whole spectral range (415-885 $nm$) or the blue spectral range free of second order (415–655 $nm$), the one with the minimum $\chi^{2}$ is always precisely the model including only TiO/VO, and with an enhanced Rayleigh slope indicative of some haze in the atmosphere (see Figure 3, and Table 2 for $\chi^{2}$ fitting results). We find that the best fitting models are metal poor, which is interesting because the host star is also metal poor ([Fe/H]= $-0.18\pm 0.06$).

Given the relatively cool equilibrium temperature of WASP-127b ($T_{\rm eq}=1400$ K; Lam et al., 2017), the tentative inference of the TiO/VO molecules is somewhat unexpected and intriguing. For planets with equilibrium temperatures lower than $\sim$1900 K, TiO could be cold trapped in the deep atmospheric layers when the temperature-pressure profile crosses the condensation curve (Showman et al., 2009; Parmentier et al., 2013, 2016). Several other possibilities could also account for TiO/VO’s absence in the upper atmosphere (Spiegel et al., 2009). Until now only two very hot Jupiters, that is WASP-121b ($T_{\rm eq}=2400$ K; Evans et al., 2016) and WASP-48b ($T_{\rm eq}=1956$ K; Murgas et al., 2017), have shown evidence of TiO/VO in the transmission spectrum. If the presence of TiO/VO were true in WASP-127b’s relatively “cool” atmosphere, one possible scenario to avoid the cold trap could be that the stellar irradiation is directly deposited into WASP-127b’s deep interior which thereby changes the deep temperature profile (e.g., Perna et al., 2010; Batygin & Stevenson, 2010; Batygin et al., 2011; Huang & Cumming, 2012; Rauscher & Menou, 2013; Spiegel & Burrows, 2013; Lopez & Fortney, 2016).

5 Conclusions

We have observed one transit of WASP-127b, an inflated, sub-Neptune mass planet. Because of its low density, the observed atmospheric scale height signals are large, and even with the NOT telescope we could retrieve its transmission spectrum. After considering the possible effects of second order contamination in the spectra, the spectrum shows the presence of a strong Rayleigh-like slope at blue wavelengths and a hint of Na absorption, although the quality of the data does not allow us to claim a detection. At redder wavelengths the absorption features of TiO and VO are the best explanation to fit the observed data. While the SNR is small, these findings are enough to conclude that the atmosphere of WASP-127b is either completely or partially cloud-free.

The brightness of its host star, a close-by comparison star, its extraordinary inflation, and its short period, all contribute to make WASP-127b a prime target for further followup with ground- and space-based facilities, including the JWST, which will be able to confirm our findings and extend them into the infrared regime. Finding the physical mechanism(s) responsible for this inflation will help us understand how this type of planets evolve and how their fate is tied to that of their host star.

Acknowledgements.

This article is based on observations made in the Observatorios de Canarias del IAC with the NOT telescope operated on the island of La Palma by the NOTSA in the Observatorio del Roque de los Muchachos (ORM). This work is partly financed by the Spanish MINECO through grants ESP2013-48391-C4-2-R, and ESP2014-57495-C2-1-R. G.C. acknowledges the support by the National NSF of China (Grant No. 11503088) and the Nat. Sci. Found. of Jiangsu Province (Grant No. BK20151051). DLP is supported by the UK’s STFC and a Royal Society Wolfson Merit award.

Appendix A Appendix A: Spectro-photometric data

Observed color light curves are shown in Figure 4 and the derived transit depths at each spectral pass band are given here in Table 3.