The accuracy of Giraffe measurements of radial velocity

The accuracy of Giraffe measurements of radial velocity in young clusters Richard Jackson – Keele University in collaboration with Rob Jeffries and Amy Dobson - Keele Jim Lewis and Sergey Koposov - Casu

The accuracy of Giraffe measurements of radial velocity in young clusters Gamma Velorum cluster To study cluster kinematic substructure We need to know Lithium rich (1) Typical uncertainty in RV (using MAD) Jeffries, Jackson, Cottaar et al. 2013 Frequency One distribution or two? ? RV (km/s) (2) Tail of the uncertainty distribution in RV (3) Any bias in RV with SNR and/or Teff

Empirical “Poisson” uncertainty in RV - for short term repeats Spectra vs SNR and vsini Measured uncertainty of repeat observations Derive a normalised uncertainty -independent of SNR and vsini

Empirical “Poisson” uncertainty in RV - for short term repeats Spectra vs SNR and vsini Measured uncertainty of repeat observations Normalised uncertainty Reduce to normalised uncertainty independent of SNR and vsini

Fix C = 26. 5 in 8 clusters Gama 2 Vel Cha_I rho_oph RV/ 2 x SNR 8463 repeats NGC 2264 RV/ 2 x SNR/(1+vsini 2/C 2) Empirical “Poisson” uncertainty in RV - for short term repeats Fix Bav=5. 7 NGC 2547 NGC 6633 IC 4665 + field stars (Corot sample) Uncertainty normalised to where B varies with Teff and B/C 2 constant RV/ 2 x SNR/(1+vsini 2/C 2) NGC 2516 Find B(Teff) log. Teff

Uncertainty in wavelength calibration - for long term repeats Change day 1 to 2 plate 1 Change day 1 to 2 plate 2 Simcal 0. 22 & 0. 23 Uncertainty 0. 08 & 0. 09 Simcal -0. 10 & -0. 17 Uncertainty 0. 15 & 0. 08 Change day 2 to 3 plate 1 Change day 2 to 3 plate 2

Uncertainty in wavelength calibration - for long term repeats Change day 1 to 2 plate 1 Change day 1 to 2 plate 2 Simcal 0. 22 & 0. 23 Uncertainty 0. 08 & 0. 09 Simcal -0. 10 & -0. 17 Uncertainty 0. 15 & 0. 08 Change day 2 to 3 plate 1 Change day 2 to 3 plate 2 SIMCAL offset in wavelength scale varies through night In similar way for all filters. Appears to be a “mechanical offset” - independent of SNR & vsini

Total uncertainty in RV - for long term repeats between OBs Normalised uncertainty Cumulative probability Poisson uncertainty (2 spectra per OB) Total uncertainty between OBs (2047 repeats) A = 0. 28 km/s B = 5. 70 km/s C = 26. 5 km/s Normalised uncertainty Poisson term Wavelength term

Total uncertainty in RV - for long term repeats between OBs Normalised uncertainty Total uncertainty between OBs 2047 repeats A = 0. 28 km/s Poisson term Wavelength term B = 5. 70 km/s C = 26. 5 km/s Normalised uncertainty Fraction of population Cumulative probability Poisson uncertainty (2 spectra per OB) Tail of empirical uncertainty between OBs Tail of Poisson uncertainty RV / (normalisation factor)

Total uncertainty in RV - for long term repeats between OBs Normalised uncertainty Cumulative probability Poisson uncertainty (2 spectra per OB) Total uncertainty between OBs 2047 repeats A = 0. 28 km/s Poisson term Wavelength term B = 5. 70 km/s C = 26. 5 km/s Normalised uncertainty Comparison empirical and Velclass uncertainties -Velclass uncertainties are ~ 2 higher.

Absolute accuracy of RVs observed in HR 10, HR 15 N and HR 21 23 standards Standard stars RV of standards from Soubiran et al. 2013 rms uncertainty 0. 04 km/s

Absolute accuracy of RVs observed in HR 10, HR 15 N and HR 21 23 standards Standard stars Field stars RV = 0. 32 km/s RV = 0. 09 km/s RV of standards from Soubiran et al. 2013 rms uncertainty 0. 04 km/s Data set GE_SD_CR ~1500 Corot targets

Variation in absolute RV between HR 10 and HR 21 filters Apparent bias in Velclass RVs (for HR 21) a function of SNR and/or Teff?

Possible causes of difference in RV measured by Velclass and CCF method Target HIP 066032 (Spt K 2) Velclass method Measured difference in RV RV template hi-res synthetic 50 -4. 0 -0. 0 (COELHO 2005) CCF method Differences in templates used

The accuracy of Giraffe measurements of radial velocity in young clusters 1. Comparison of repeat measurements of RV in young clusters shows the measurement uncertainty can be normalised to a simple function SNR and vsini with a weaker dependence stellar properties. that of 2. The measurement uncertainty at higher SNR is dominated by a fixed uncertainty of ~0. 28 km/s due to changes in wavelength calibration between setups. 3. Analysis of RVs of standard stars shows a bias of ~0. 4 km/s for measurements made using HR 21 (but no significant bias for HR 10 and HR 15 N). The most likely cause is a mismatch between target spectra and synthetic RV templates. 4. Where possible stellar properties used to select RV templates should be fixed for each target - based on the best available estimates of Teff, log. G and Fe/H. 5. Following the planned change to GAIA/Phoenix synthetic spectra revised RVs should be re-analysed for evidence of bias with Teff and/or SNR in all filters.