1. From the NGSL to Absolute Flux Sara Heap, NASA/Goddard Space Flight Center Don Lindler, Sigma Space Corporation Phase 1: NGSL observations + in situ calibration Phase 2: NGSL observations + calstar flux calibration Phase 3: Future NGSL-like observations

2. Phase 1: NGSL Observations & In Situ Calibration

3. Next Generation Spectral Library Teff log g log Z 6341 4.19 -1.12 (this study) 6114 4.07 -1.79 (Clem et al. 2004) FEATURESBENEFITS Wide coverage: 0.2-1.0  Define new spectral indices; get wide-baseline SED, colors Absolute flux calibrationDo wide-band spectrophotometry; measure spectral breaks Moderate spectral resolutionMeasure spectral indices; Check model spectra Large number of program stars: 378 304 are good observations 248 have good fits to their spectra Trace systematic trends of spectral indices HD 16031 HD 2665 (G5 IIIw) T eff =5004 log g=2.27 log Z= -1.96 ELODIE 4

4. The NGSL snapshot survey involves STIS spectra at three different grating settings

5. Reduction of the NGSL spectra Improvements over the Standard STScI Pipeline Calibration made by Don Lindler New spectral trace files were produced using the average spectral y-position versus x-position for the 52X0.2E1 aperture for all of the NGSL stars. CALSTIS adjust the trace by a constant offset for each individual observation. Improved background subtraction. ST ScI pipeline uses only a lower background taken 300 pixels below the spectrum. We use both an upper and lower background taken much closer to the spectrum (30 pixels). Larger extraction slit height (11 instead of 7) to improve overall photometric at the cost of some loss in S/N. Custom wavelength dispersion coefficients created for the E1 aperture position. Custom sensitivity curves were created for the 52X0.2E1 aperture using observations of BD+75D325 centered in the aperture. Correction for mis-centering of the target within the slit using the G750L fringe flat to determine the target’s position. Zero-point wavelength correction using stellar lines. (no Wavecals were taken with the observations)

6. Correction for Fringing

7. G230LB in-order grating scatter

8. Problem #1: Charge Transfer Inefficiency

9. NGSL Observing Strategy A new location on the aperture wheel, called E1, was defined. The E1 pseudo- aperture locations allows spectra to be taken easily near row 900. By moving a spectrum closer to the readout, the number of parallel charge transfers is reduced by about a factor of four, with a comparable reduction expected in the losses due to charge transfer inefficiency during the readout. “Use of the E1 aperture positions is strongly recommended for all but the very brightest (> 50 e-/pixel/s source count rate) or most extended spectroscopic point sources (> 5" diameter)”. Readout registers

10. Problem #2: Offset of Star in Slit

11. Wide-baseline SED’s must be corrected if the star is off-center in the 0.2” slit When the target star is close to the edge of the slit, a greater fraction of red light is blocked than blue light. At long wavelengths (~8000 Å), the PSF is asymmetrical so that the spectrum of off- center targets has a characteristic “bowing”. bowing Observations 0.2” slit 4 pixels

12. One Effect of Offset from Slit Center: Vignetting

13. Wide-baseline SED’s must be corrected if the star is off-center in the 0.2” slit When the target star is close to the edge of the slit, a greater fraction of red light is blocked than blue light. At long wavelengths (~8000 Å), the PSF is asymmetrical so that the spectrum of off- center targets has a characteristic “bowing”. bowing Observations 0.2” slit 4 pixels

14. Another Effect of Offset: “Bowing” of Spectra

15. Initial Solution: -dependent flux correction based on “measured” offset in slit Offset of star in slit (pixels) Observed Flux / Model Flux

16. Result of Flux Correction

17. Phase 2: Throughput calibration of the 52x2” slit using calstar, BD+75 o 325 * * * * * * Observing Procedure: Step the star across the the 52x2 slit, and at each step, get spectra (G750L, G430L, G230LB) of the star at the E1 position.

18. Results [1]

19. Results [2]

20. Results [3] Offset Uncertainty Pixels 1.0 0.0 1.0

21. The stellar parameters (L bol, T eff, ilog g, logZ, , E(B-V) ) are derived from NGSL spectra and distances via Castelli’s (2004)/Marcs (2008) model spectra and the Basti evolutionary isochrones. Make  2 fit to spectrum (  =0.20 -1.00  ) for T eff, logZ, and E(B-V) Determine L v range corresponding to V, , e  Calculate L bol from BC(T eff, log Z, E(B-V)) Determine range in distance (using new reduction of Hipparcos data, 2007, and a two sigma error for the parallax. Derive allowed range in ilogg via comparison with Basti evolutionary models Make  2 fit to spectrum for best ilog g Allowed LogL Range Allowed LogG values

22. HB stars Blue straggler? 6 HR Diagram of NGSL Stars

23. HB stars Blue straggler? 6 Calstars ?

24. T eff log g log Z Weak-line F-type stars are good calibration stars

25. Weak-line cooler stars work well too

26. Comparisons

29. Phase 3: The Future Measure Absolute Fluxes Method 1: The standard method used by Bohlin in his calstar program Calibration star at the center of the detector Method 2: The NGSL method Calibration star at the E1 position (like NGSL) Flat field taken at the E1 position through the 0.3x0.09 slit

30. E1 Pseudo Apertures: Sensitivity and Throughputs STIS FLATFIELD Calibration program: 9616 Date: 10 October 2002 Dataset: O8J5010S0 Exposure time: 300 s Aperture: 0.3x0.09 Grating: G750L Proposal AbstractThe E1 pseudo-aperture locations were implemented to allow spectra to be taken easily near row 900. By moving a spectrum closer to the readout, the number of parallel charge transfers is reduced by about a factor of four, with a comparable reduction expected in the losses due to charge transfer inefficiency during the readout. There is, however, some evidence that the detector sensitivity and focus near row 900 differ by a few percent from that near the middle of the detector. This results in errors in the extracted fluxes that may differ from aperture-to-aperture. While some limited data is already available, in order to improve the absolute accuracy of the calibration, additional observations of one of the hot white dwarf standards would be extremely useful, both to verify the other results, and to check the stability of the PSF and aperture throughputs near row 900. We will therefore observe BD+75D325 with the G230LB, G430L, and G750L through a variety of supported slits. We will also explore a variety of strategies for improving fringe flat subtraction near the E1 positions. This will include using the slit-step optional parameter to move a small slit to row 900, and adding postargs to large aperture observations to better align the source with the 52X0.1 slit used for the fringe flat. Relative aperture locations will also be checked by a series of G430L wavecals taken through a variety of apertures.; E1 Alien aperture

31. Observing Procedure [1] Move the 0.3x0.09 aperture to the E1 position Not to Scale!

32. Observing Procedure [2] Obtain a G750L spectrum of the tungsten lamp in the 0.3x0.09 aperture. The spectral flat at the bottom comes from the aperture next to the 0.3x0.09 aperture on the aperture wheel. Not to Scale!

33. Observing Procedure [3] While keeping the 0.3x0.09 aperture at the E1 position, slew the star and peak up in the aperture Not to Scale!

34. Observing Procedure [4] While holding the star fixed, move the wider 52x2 aperture into view. The 52x2 aperture is 40 pixels wide. All the flux from the star should be transmitted through the aperture. Not to Scale!

35. Observing Procedure [5] Obtain the spectrum of the star. Not to Scale!

36. Summary The NGSL is a package consisting of customized 1.Observing procedure 2.Reduction procedure 3.Analysis and modeling It is possible to obtain HST spectra with the required <1 % accuracy

37. The stellar parameters (L bol, T eff, ilog g, logZ, , E(B-V) ) are derived from NGSL spectra and distances via Castelli’s (2004) model spectra and the V-R evolutionary isochrones. The new reduction of Hipparcos data (2007) makes a big difference to the ilog g of some stars. Make  2 fit to spectrum (  =0.20 -1.00  ) for T eff, logZ, and E(B-V) Determine L v range corresponding to V, , e  Calculate L bol from BC(T eff, log Z, E(B-V)) Derive allowed range in ilogg via comparison with evolutionary models Make  2 fit to spectrum for best ilog g Allowed  logg  logg from distance

38. NGSL Data Require Custom Data Processing Fringing in the red spectral segment (G750L). We separated the flats into a fringe-flat and a broad-band flat. We then shifted the fringe-flat to match the observed spectrum via cross- correlation of the fringes and then recombined it with the broad-band flat. Aperture effects. Errors in centering the star in the 0.2” slit (4 pixels wide) lead to changes in the apparent flux and relative flux distribution since the PSF=f( ). We estimated the offset of the star from the center of the slit from the fringing data and corrected the flux for this offset. Wavelength calibration. We assigned wavelengths based on a new wavelength calibration (quadratic dispersion), and measurements of stellar features. Background estimation. The spectra were positioned near the edge of the CCD frame in order to minimize CTI effects. We therefore extracted the background flux close to the star. Grating scatter in the UV. Grating scatter produces a spurious upturn in the flux of very red stars. We have worked out a correction scheme based on scattering ~ -3 Sensitivity calibration. Based on observations of BD+75 o 325, we generated a new sensitivity curve applicable to the target aperture and background/spectral extraction parameters. [POSSIBLE ONLY AFTER MOST RECENT HST SERVICING MISSION]