4.3 Factors Limiting Flux and Wavelength Accuracy

4.3.1 Flux Accuracy

The accuracy to which the absolute flux calibration of STIS spectroscopic data at slit center can be trusted is limited by several factors including:

• The accuracy of the absolute sensitivity calibration of the grating and central wavelength setting. The on-orbit absolute sensitivity calibration is determined by observing a standard star, with known absolute flux calibration, well centered in both wavelength and cross dispersion in a large slit. The STIS spectrum of this star is then extracted over a standard aperture extraction box and the sensitivity required to return the known flux from the star is determined as a function of wavelength. The standard aperture extraction box is large enough to be relatively insensitive to spacecraft jitter and breathing but small enough that the signal-to-noise of a typical stellar spectrum will not be degraded. STIS calibration accuracies are defined for the standard aperture extraction box; the standard boxes are mode dependent and are given in the XTRACTAB reference file.
• The accuracy of the calibration of the time dependence of the sensitivity for grating and wavelength region being used. This calibration is typically accurate to within 1% rms (see STIS ISR 2004-04, STIS ISR 2006-04, STIS ISR 2014-02, and STIS ISR-2017-06 for details).
• The accuracy of the calibration of the aperture throughput for the aperture being used for the science relative to the aperture that was used for the absolute sensitivity calibration.
• The accuracy to which the source is centered in the slit.
• The size of the extraction aperture used to measure flux and the accuracy to which the cross dispersion profile is known in the mode in which one is observing. Because the corrections for the aperture extraction can be large (e.g., 30% in the near-infrared and the far-UV) this effect can be important.
• Bias and background subtraction can add considerable additional uncertainty for faint sources or spectra with significant variations in flux, particularly for the echelle modes.
• Grating scatter, which can play an important role, particularly for the G230LB and G230MB gratings when used with red targets (see footnote to Table 4.1).
• For echelle modes: the accuracy of the correction of the behavior of the echelle blaze function with time and location of the spectra on the detector and the accuracy of the scattered light correction (see STIS ISR 2002-01 and 2018-01).

Additional uncertainties arise for flux measurements not at slit- and field-center. These uncertainties are relevant when, for example, one wishes to determine relative fluxes in an extended source along the long slit or when one uses POS-TARGs to move a target along the long slit. They include:

• The variation in slit throughput along the slit. The slits have 5 micron variations along their widths (corresponding to ~0.02 arcsecond), which for a 0.1 arcsecond wide slit on a diffuse source, would produce a 20% variation in flux. For a point source the variation would be more in the 5% range for that same slit. There are also dust specks with appreciably higher opacity along some places in some slits.
• The accuracy to which the vignetting along the cross-dispersion direction is known for the grating and central wavelength.
• The accuracy to which the low order flat field along the dispersion direction is known off of field center for the grating and central wavelength.

4.3.2 Wavelength and Spatial Accuracies

The accuracy with which the wavelength scale is known in the calibrated STIS spectrum will depend on several factors:

• The accuracy of the dispersion solutions, which governs the accuracy to which relative wavelengths are known in a given spectrum.
• The accuracy of the wavelength zero point, which governs the accuracy to which relative wavelengths are known across spectra.
• The accuracy to which the source is centered in the science slit (a pixel of miscentering corresponds to a pixel in absolute wavelength space).

The dispersion solutions used to calibrate STIS on-orbit data were derived on the ground during thermal vacuum testing. On-orbit tests confirm the applicability and accuracy of the ground dispersion solutions for on-orbit data, producing relative wavelength accuracies of 0.2 pixels across the spectrum for first-order gratings at the prime settings and appreciably better in some instances. For the echelle modes, at the prime settings, the accuracies are roughly 0.2 pixels for wavelengths in the same order and approximately 0.5 pixels for the entire format. The intermediate wavelength settings have roughly twice these errors. The accuracy of the dispersion solutions is well maintained across the spatial extent covered by the first-order modes. However, the illumination of the CCD detector by the line lamp used for wavecal exposures suffers somewhat from vignetting at the top and bottom of the detector. Fortunately, the effect of this at the location of the E1 pseudo apertures, has been found to be insignificant, although a few observing modes may have a slightly lower accuracy (details are presented in STIS ISR 2005-03).

Due to the lack of repeatability of the Mode Select Mechanism (STIS's grating wheel), the projection of the spectrum onto the detector in both wavelength and space will vary slightly (1 to 10 pixels) from observation to observation if the grating wheel is moved between observations. In addition, thermally induced motions can also affect centering of the spectrum. The calstis pipeline removes the zero point offsets using the contemporaneous wavecals (see Section 3.4.23). The wavelength zero point in calibrated data (the _sx2, _x2d, _x1d, and _sx1 files) is corrected for these offsets and should have a wavelength zero point accuracy of ~0.1–0.2 pixels (better when the wavecal is taken through small slits, worse for those taken through wider slits). This accuracy should be achieved, so long as contemporaneous wavecals were taken along with the science data, distributed at roughly one hour intervals among the science exposures, and assuming the target was centered in the slit to this accuracy or better.

The accuracy of the zero point pipeline calibration in the spatial direction is slightly less, roughly 0.2-0.5 pixels. This is because the finding algorithm, which must locate the edges of the aperture for short slits and the edges of the fiducial bars on the slits for the long slits, is less robust. Observers need to be aware of possible offsets between spectra in the spatial direction, particularly when deriving line ratios for long slit observations of extended targets taken with different gratings.

A source can lie off-center in an aperture because of errors in the ACQ or ACQ/PEAK procedure, because of errors in the defined displacement from the acquisition aperture to the science aperture, and because of drift of the target over time. The component of error in the AXIS1 direction produces an uncalibrated shift in wavelength. The ACQ for a point source is generally accurate to 0.01 arcsec. An ACQ/PEAK can improve the accuracy, but can also degrade it if, for example, the signal to noise is poor. See Section 5.2.6 for a guide to the interpretation of acquisition data, and Chapter 8 in the STIS Instrument Handbook, for a full discussion of acquisition procedures. The error in the defined displacement from the acquisition aperture to the science aperture has generally been insignificant, except for errors in the original definition of the E1 apertures. These long-slit pseudo-apertures, which place the target high on the detector near the readout amplifier to minimize CTE effects, were introduced on 2000-Jul-03 and revised on 2003-Aug-04. The defined AXIS1 positions of apertures 52X0.2E1, 52X0.5E1, and 52X2E1 were revised by shifting them -0.73 pixels; those of the 52X0.05E1 and 52X0.1E1 were revised by shifting them -0.55 and -0.70 pixels, respectively. Before the revision, E1 aperture exposures made after an ACQ or after an ACQ/PEAK at the center of the detector were systematically miscentered by the full amount of these shifts, causing spectral features to appear at longer wavelengths. Exposures made after an ACQ/PEAK with the same E1 aperture were free of the error, and exposures made after an ACQ/PEAK with a different E1 aperture were off-centered by the relative error, which is usually ~0. Target drift is generally insignificant over the course of an orbit when two guide stars are used, but is larger and variable when one guide star is used. See Section 5.2.6 and Section 8.1.4 in the STIS Instrument Handbook.