Data Reduction

The following process was used to reduce all data for this analysis:
  1. Temperature dependant darks were generated for camera 2 data taken after august 22,1997 to limit temperature dependant shading
  2. All image headers were updated with the most current reference files available
    Since this work was done new linearity files have been created for all cameras
    A program to create temperature dependant flat-fields is also in development (See Poster Session 12.08)
  3. Images were run through the CALNICA pipeline for initial processing
  4. A fit to the median of the columns, in the direction of readout ,was calculated to remove residual readout shading
  5. PEDSKY was used to get an estimate of the quadrant-dependant residual dc-bias and remove this additional signal from the image.
  6. Aperture photometry was performed using phot in the noao.digiphot.aphot package in IRAF and the respective aperture correction applied to get the final countrate for each star.

Further descriptions of NICMOS IRAF reduction routines may be found at stsdas .

Aperture Corrections

All previous photometric calibrations assumed aperature corrections of 1.15 for camera 1 and camera 2, and 1.075 for camera 3. These corrections have been greatly improved. TinyTim v5.0 was used to make filter dependent aperture corrections for each of the cameras.The latest version of this software package includes updated NICMOS aberrations and obscuration parameters, such as the camera rotations and improved Zernike coefficients. It has also incorporated each filter so that users no longer have to supply a text file of wavelengths and relative weights. Figure 1 shows a plot of aperature correction verse wavelength for all the cameras. On first observation, it can be seen that the original corrections did not accurately represent the differences between each of the filters.

Camera 1: Star aperature taken at 0.5 arcsec , Sky @ 2 arcsec
Camera 2: Star aperature taken at 0.5 arcsec , Sky @ 2 arcsec
Camera 3: Star aperature taken at 1.0 arcsec , Sky @ 4 arcsec

All cameras compared to total flux inside 10 arcsec

NICMOS Photometric Calibration

As with other HST instruments, the absolute photometric calibration of NICMOS uses on-orbit observations of standard stars. In the infrared, this effort is complicated by the fact that there are no suitable spectrophotometric measurements of stars that span the entire wavelength range accessible to NICMOS. Moreover, our understanding of NICMOS instrument performance evolved substantially with time during the instrument lifetime, leading to improved techniques for data processing, better calibration reference files (flat fields, linearity calibration, etc.), and new methods for handling instrument anomalies not treated by the standard pipeline. As a result, there is still room for significant improvement in the photometric calibration of NICMOS data. We have undertaken an effort to re-calibrate NICMOS photometry, and describe this program here. The improvements (compared to previous calibrations) come from several areas:

  • Use of all available on-orbit standard star data.
  • Improved (and uniform pedigree) calibration reference files (dark correction, etc.)
  • Careful data reduction to eliminate artifacts and anomalies such as the residual bias offset or "pedestal" effect
  • Use of new, improved ground-based photometry for the fundamental NICMOS standards

Spectrophotometric Models

The photometric calibration keywords are derived from the comparision of the measured NICMOS count-rates for the standard star observations to the flux density of the standard star averaged over the NICMOS bandpass. There are no ground-based spectrophotometric observations of standard stars with complete coverage over the NICMOS wavelength range, and therefore we must use accurately calibrated "surrogate spectra" instead. This was the motivation for using solar analog and white dwarf standards for the NICMOS photometric calibration. The absolute spectral energy distribution of the Sun is well known (see Colina, Bohlin & Castelli 1996 and references therein), and thus can be scaled reliably to represent the spectrophotometry for solar analog standards like P330E. DA white dwarfs like G191B2B have relatively simple stellar atmospheres, and considerable effort has gone into accurately modelling these and comparing them to UV-through-optical spectrophotometry (Bohlin, Colina & Finley 1995; Bohlin 1996; Bohlin 2000).

As described by Colina & Bohlin 1997, the infrared spectrum of P330E is represented by the solar spectrum from Colina, Bohlin & Castelli 1996, which is a composite of empirical measurements (Woods et al. 1996, Neckel & Labs 1984, Arvesen et al. 1969) and an ATLAS 9 stellar atmosphere model (Kurucz 1993) from 9600-25000A. The white dwarf G191B2B is represented by an LTE model calculated by D. Finley (described in Bohlin 2000). These spectrophotometric models are then normalized using ground-based photometry of the NICMOS standard stars. Persson et al. (1998) have obtained ground-based JHK photometry for a large set of faint infrared standard stars, including the HST/NICMOS solar analog standards. Persson (private communication) also observed G191B2B as part of the same program.

In order to normalize the standard star spectral models, we must convert Persson's JHK magnitudes to absolute flux density units. Campins, Rieke & Lebofsky (1985) provide an absolute infrared flux calibration scale using a solar analog method. However, the effective wavelengths and bandwidths of their JHK filters (which we will refer to as the Arizona system, where Vega is defined to have J=H=K=0.02) differ somewhat from those used by Persson et al. (calibrated to the CIT system, where Vega is defined to have J=H=K=0.0). In order to shift the Campins et al. absolute calibration to the Persson et al. bandpasses, we have used an ATLAS 9 atmosphere model for Vega. This model is not used for any absolute calibration, but simply to compute flux density ratios for Vega between the Arizona and Persson et al. bandpasses. These are then used to convert the Campins et al. Vega flux densities to the Persson et al. bandpasses, and hence to provide the absolute flux density calibration for the Persson et al. measurements. In this way, m = 0 is calibrated to be 1614, 998, and 639 Jy for the Persson et al. JHK bandpasses, respectively.

The P330E and G191B2B spectrophotometric models are then synthetically integrated through the Persson et al. JHK passbands, and the bandpass-averaged flux densities are converted to magnitudes for comparison to the ground-based photometry. This comparison (Figure 2) indicates that the Colina & Bohlin (1997) P330E model requires an average flux renormalization of +5% to match the JHK photometry at the 0.01 mag level. A similar pattern of residual offsets for the J, H and K bands is found for both standard stars. These residuals are only slightly larger than the stated uncertainties of the Campins et al. (1985) infrared absolute calibration zeropoints, and we ignore them here, but they suggest that remaining uncertainties in the absolute calibration of the ground-based JHK system set a limit to the final accuracy of the NICMOS photometry scale.

Figure 2: Comparison of the CDBS reference spectra for the NIMCOS standards to goundbased JHK photometry from Persson et al. (1998) The top panel shows the raw magnitude difference, while the bottom panel shows the residuals after applying flux offsets of 5% and 1% for P330E and G191b2b respectively. The residual offsets at H and K are only slightly larger than the uncertainties in the Campins et al. absolute flux calibration model( shown as error bars) but may represent the ultimate limitation to the absolute calibration of NIMCOS at the present time.

Photometric Keyword Comparison

the PHOTFNU zeropoint calibration keyword is the ratio of the standard star flux density averaged through each of the NICMOS bandpasses with the aperture-corrected average count rate measured from the standard star observations.Figure 3 shows the ratio of PHOTFNU valuies derived for all the NICMOS filters using the P330e and G191b2b observations. On average these calibrations agree to within 2% for all three NICMOS cameras, with a standard deviation of 1 to 2.5% among the various filters, but there are a few individual filters for which P330E and G191b2b yield calibrations that differ by 5-10%. These acases are still being investigated. Figure 4 compares the new PHOTFNU values from P330e only) to the values previously used by the NIMCOS calibration pipeline (reference file i7l12297n_pht.fits) In the mean, the differences are only a few percent, but a systematic trend with wavelength can be seen: this is due to the improved, wavelength-dependant aperture corrections that have now been adopted.

Figure 3:

Figure 4:

Room For Improvement

We are presently completing final determination of the NICMOS photometric calibration,examining the following points:

  1. Analysis of the remaining secondary NICMOS standards (P177D, GD153, and the red standards) for comparison with P330E and G191B2B
  2. Improved treatment of background removal at thermal infrared wavelengths
  3. Examining photometric trends with instrument temperature, with the possible goal of providing a temperature-dependant zeropoint determinization

New photometric keyword values and full documentation will be provided on the NICMOS WWW pages and in the NICMOS Data Handbook (Chapter 5) in the near future.

Return to previous page
Megan Sosey
Last modified: Mon Jan 29 18:09:27 EST 2001