Telluric Contamination effects challenges and solutions in the

Telluric Contamination effects, challenges, and solutions in the context of EPRV Sharon X. Wang, Jason T. Wright, Cullen Blake, Pedro Figueira, Nuno Santos, Peter Plavchan, Andreas Seifahrt, Claire Moutou, François Bouchy

summary of current status ● Telluric lines have ~m/s level impact on RV precision (either optical or NIR). ● We can model most telluric lines to ~ 2%. (radiative transfer + atmospheric model + line list) ● Telluric lines are sources of contamination instead of wavelength calibrator for EPRV.

telluric absorption in optical+NIR Smette et al. (2015) 500 nm 1μm 2. 6μm

pervasive micro lines in the optical 4430 Å 4850 Å 5040 Å 5410 Å 5750 Å line depth 0. 2 -- 2% Wavelength [Å] Cunha et al. (2014)

telluric: calibrator, contamination Griffin & Griffin (1973) suggested 10 m/s possible with O 2 (start of quest for PRVs? ) Now 5 -20 m/s in both optical and NIR using tellurics lines as calibrators (e. g. Barnes+2012, Blake+2010, Figueira+2010 a). Studies in the optical show long-term stability at 10 m/s over 10 years with HARPS (Figueira+2010 b) and that it is possible to improve upon this value using simple empirical models (e. g. Figueira+ 2012) EPRV definitely requires better calibrators and proper modeling of telluric lines (e. g. Bean+2010, upcoming instruments in NIR).

optical (Th-Ar): alias induced by micro tellurics Cunha et al. (2014)

Artigau et al. , 2014, SPIE, 9149 Assumption : telluric absorption spectrum = sum of a finite number of chemical species that vary in relative strength = linear combination of individual absorbance with weight varying with airmass and weather. Library of individual absorbance (real telluric standards (hot stars) observed at a large variety of airmass and water column or TAPAS synthetic spectra) + principal component analysis to identify independently varying absorbers. Non-zero velocity of one absorber seen as an second component equal to the spectral derivative Science spectra = least-square fit of 1) absorption-free spectra + 2) linear combination of absorbance. Requirement = large spectral range + high resolution

HARPS spectra restricted to 634 – 669 nm Without excluding telluric = 10. 3 m/s RMS Excluding telluric (18% of the domain) = 3. 6 m/s RMS PCA subtracted spectrum = 1. 6 m/s RMS Artigau et al. (2014)

optical (iodine): alias induced by micro tellurics simulation 2Å chunk telluric 0. 4% absorption clean HD 185144 Wang et al. in prep

NIR: modeling “macro” telluric lines Mauna Kea, PWV=3 mm, Airmass = 1 (from Gemini webpage) A different challenge: τ>>0. 1 at many wavelengths Several different species (H 2 O, CH 4, CO 2)

NIR: modeling “macro” telluric lines Seifahrt+2010 Bean+2010 Blake+2011

NIR: modeling “macro” telluric lines Seifahrt+2010 High optical depth - approximations that work at low optical depth not OK Detail line shape matters more, line mixing matters Water, water, everywhere! Multiple, overlapping species Blake+2011 HITRAN errors mater more Good News? CO 2, CH 4 are well behaved, only small seasonal variations in concentration!

modeling the telluric lines: approaches, codes available Physical: TAPAS (Bertaux et al. 2014, web) Tel. Fit (Gullikson et al. 2014) Molecfit (Smette et al. 2015) TERRASPEC (Bender et al. 2012, cbender@psu. edu) ATRAN (IR only, Lord 1992 + Gemini, web) Write your own using LBLTRM(Clough+2005)+HITRAN (Rothman+2012) Empirical: telluric standards (Vacca+2003) taking spectra at different airmass (Wallace+2011, solar atlas) PCA (Artigau et al. 2014)

Empirical models FTS Template - τ>0. 1, multiple species Forward modelling using FTS telluric template from Livingston & Wallace 1991 Blake et al. (2010)

challenges & pressing questions? How good are we? ● How well can we model telluric lines? o What are the current best codes? How well does each do? Do they agree? o What do we do about the challenging lines? o What’s the current limiting factor? What’s next? Empirical+physical? More physics? ● How well do we need to model them? o How well are we doing now? o What’s their impact on RV precision? o What’s the requirement for achieving 10 cm/s?

challenges & pressing questions? How good are they? ● How stable (in terms of RV) are the telluric lines? o The current knowledge? o What to be done? Different species? ● How stable do we need them to be or need to model? o What’s the RV limit for using them as calibrators? o What’s the requirement for achieving 10 cm/s? o Include telluric RV as free parameter? Species?

part 1/4 ● How well can we model telluric lines? o What are the current best codes? How well does each do? Do they agree? o What do we do about the challenging lines? o What’s the current limiting factor? What’s next? Empirical+physical? More physics?

● How well can we model telluric lines? o What are the current best codes? How well does each do? Do they agree? Cunha et al. (2014) Seifahrt et al. (2010)

Part 1/4 ● How well can we model telluric lines? o What are the current best codes? How well does each do? Do they agree? o What do we do about the challenging lines? o What’s the current limiting factor? What’s next? Empirical+physical? More physics?

Limitations of current modeling (roughly in order of impact) 1. First order problems in molecular databases (e. g. HITRAN) 1. Missing lines 2. Uncertainties/errors in line position, line strength, line shift and broadening parameters. Note: data quality in HITRAN is not homogeneous, neither for the same species nor for different species in a narrow wavelength range! 2. Limitations of current modeling codes (e. g. LBLRTM) from treatment of correct line profiles (i. e. speed dependence/line mixing effects, Dicke narrowing). Standard is Voigt profile, but would need Rautian or Galatry profile for certain species with strong line mixing. (see, e. g. , Hartman et al. 2008) 3. Second order problems in molecular databases (e. g. HITRAN) 1. Missing data or approximate treatment of line coupling and mixing. Important e. g. for O 2, CO 2, N 2, H 2 O, etc.

Limitations of current modeling (roughly in order of impact) - cont. 4. Limited knowledge of atmospheric conditions 1. Standard atmospheric profiles probably fine for well mixed species, but absolute column densities for, e. g. , CH 4, or CO 2 need to be adjusted. 2. Water vapor is problematic in standard atmospheric profiles due to high absolute variability and variable vertical distribution (different temperature, pressure broadening effects etc. ). -> Water vapor spectrum for same PWV column can look different depending on vertical distribution. 5. Wind-induced line-shifts 1. Can be substantial, especially for observatories under jet stream with light path into the jet stream (i. e. , azimut and airmass dependent). See Figuera et al. (2010)

Improvements to current modeling 1. Invest manpower and/or money into codes and (more importantly) groups providing data for HITRAN database -> unlikely to happen. 2. Build a simplified (and unphysical) astro-HITRAN with line data retrieved from fitting high-SNR, high-resolution stellar spectra from instruments with stable and/or well known LSFs (e. g. HARPS, or gas-cell aided spectrographs). Need to be done over large number of spectra taken under different conditions to disentangle stellar and telluric features. Successful example: CRIRES program of Bean et al. 2010. 3. Improve knowledge of local conditions of the atmosphere. Particularly important for H 2 O and for wind-speed issues. -> Launch weather balloons every night? -> Potential use of LIDAR techniques for remote retrieval? What about GPS or microwave radiometers?

Part 2/4 ● How well do we need to model them? o How well are we doing now? o What’s their impact on RV precision? o What’s the requirement for achieving 10 cm/s?

Part 3/4 ● How stable (in terms of RV) are the telluric lines? o The current knowledge? o What to be done? Different species?

● How stable (in terms of RV) are the telluric lines? o The current knowledge? o What to be done? Different species? Figueira et al. (2010) O 2 lines in HARPS data HARPS - Telluric O 2 line RVs relative to star: ~2 m/s with simple model over years, 5 m/s “out of the box”

● How stable (in terms of RV) are the telluric lines? o The current knowledge? o What to be done? Different species? Balthasar+1982 - Telluric O 2 lines can be modeled to 3 m/s Caccin+1985 - Pressure-induced asymmetries+winds < 5 m/s Smith 1982 - ~5 m/s on Arcturus using O 2 lines What about H 2 O lines?

Part 4/4 ● How stable do we need them to be or need to model? o What’s the RV limit for using them as calibrators? o What’s the requirement for achieving 10 cm/s? o Include telluric RV as free parameter? Species?

Bibliography (to be completed) http: //bit. ly/eprv_telluric_references Stay tuned for summary of the session!

extra slides feel free to add. be sure to give citations and add notes!

optical (iodine): micro telluric lines everywhere! using TERRASPEC

optical (iodine): alias induced by micro tellurics

Recipe: How to clean your deconvolved stellar template Ingredients: high SNR stellar template spectra (taken with no iodine), bracketing telluric standard observations through nearby B star (optional but desirable), bracketing nearby B star spectra through iodine cell (to provide spectral PSF). Prepare: telluric model generator which can generate models under different atmospheric conditions (e. g. varying water vapor), tuned to your observatory. Your favorite deconvolution algorithm. Cook: 1. Fit the B star telluric standards to get atmosphere model. 2. Fit the B star + iodine spectra with iodine atlas + telluric model to get the correct PSF. 3. Deconvolve the stellar template spectra with PSF. 4. Divide out telluric model from deconvolved stellar template (note: first solve for absolute stellar RV!).

● How well can we model telluric lines? o What are the current best codes? How well does each do? Do they agree? 3 -5% correction with Tel. Fit modeling correction with telluric standards Gullikson et al. (2014)

● How well can we model telluric lines? o What are the current best codes? How well does each do? Do they agree? 2% Smette et al. (2015)

NIR: modeling “macro” telluric lines Bean et al. (2010)

Theoretical Telluric Spectrum Terraspec: Chad Bender

Telluric Line RV Shifts Telluric Lines are Asymmetric! Optical Depth Pressure Shifts: ΔRV< 20 m/s Winds: ΔRV< 20 m/s Frequency

Altitude (km) Atmospheric Pressure ΔP/P Pressure with Altitude 1σ Variation in Pressure Δ~0. 01 For H 20, CO 2, CH 4, O 2 • 1% Pressure Change is ΔRV=3 m/s at 1 micron • Depends on Zenith Angle φ:

Atmospheric Winds Alt=20 Km Alt=5 Km Jet Stream East Average East-West North-South Vertical Average Winds January March Radial Wind Speed June September

Monte Carlo Simulations 100 night atmosphere models over 12 months Models for mountain site in New Mexico 10 random sight lines each night, AM<1. 2 • • • T=e-τ Telluric Regions Barycenter O 2 0. 685 -0. 695μm H 2 O 0. 9 -1. 0 μm CO 2 1. 59 -1. 62μm CH 4 2. 28 -2. 3μm Wavenumber • Yearly average atmosphere model subtracted: EW wind, NS wind, pressure profile • Assumption: H 2 O well-mixed

Simulation Results RV residuals after correcting for zenith angle and yearly average wind profile

Altitude of Line Formation Histograms of altitudes where lines reach optical depth τ=1