The near-infrared spectrum is strongly affected by terrestrial atmospheric absorption. This is corrected for, and the extracted object spectra placed on an absolute flux scale, by dividing by the observed spectrum of a flux calibrator star and multiplying by the absolute flux distribution of the calibrator. Ideally, each object measurement would be accompanied by a similar measurement of a nearby featureless flux calibrator taken as close as possible in time to the object observation and with the same spectral resolution. In practice, the flux calibrator will be at some distance from the object on the sky and will contain intrinsic absorption features. Our approach is to flux calibrate with the best available flux calibrator and then to correct for any shortcomings in the flux calibrator by dividing by a `flat spectrum' star. It is necessary to record spectra of the flux calibrator with the same slit width as the object in order to accurately cancel terrestrial atmospheric absorption, and it is advisable to also record spectra of the flux calibrator with a very wide slit if absolute flux calibration is required. The narrow slit measurement is used for the initial flux calibration and cancellation of terrestrial atmospheric absorption features, and the wide slit spectrum is used later to derive a correction for slit losses.
The choice of suitable flux calibrators is often difficult due to the
range of intrinsic absorption features present in stellar spectra.
Theoretical spectra for the Kurucz 
model atmospheres
are plotted in Figures 34 to 37 of
Appendix H. While these may not accurately reproduce
molecular features, they are certainly a good guide to the types and
strengths of absorption features present in the near-infrared spectra
of main-sequence stars. Early-type stars have the smoothest continua
and should be used when the features of interest cover a broad
wavelength range. However, early-type stars have hydrogen absorption
lines which must be accurately measured, especially if these are
features of interest in the object spectra. F and early G dwarfs have
relatively weak hydrogen lines and are sufficiently common to allow
examples to be found near most objects. However, their spectra in the
J band are contaminated by weak absorption lines which may be
problematical. Dwarf stars later than mid G have CO first-overtone
absorption beyond 2.3 
m and weak absorption throughout their
near-infrared spectra. These should be avoided for the purpose of
flux calibration. However, late K and early M dwarfs lack significant
hydrogen absorption so they can be usefully employed in measuring the
strength of hydrogen absorption in early-type flux calibrators.
The above considerations were used in forming the list of
spectroscopic flux calibrators in Table 24 of
Appendix G. Most stars in this list have
accurately determined near-infrared photometry on a well-defined
photometric system, and as such are useful flux calibrators. Ideally,
the flux distributions of these stars would be known. However, this
is not the case. Instead, we are forced to model the shape of the
flux distributions for these stars. The redgspec task uses two
model types; a blackbody distribution parameterised by a color
temperature, and approximations to the Kurucz flux
distributions parameterised by the model effective temperature. An
interstellar extinction can also be applied to the model distribution.
Blackbody models can be used for early-type stars, but Kurucz models
should be preferred for mid- and late-type stars where the continuum
distributions deviate significantly from blackbodies. The model
distributions are normalised at K using a magnitude calculated from
the model flux distribution, the CASPIR K filter profile, the
predicted transmission function of the CASPIR anti-reflection
coatings, and the theoretical atmospheric transmission function shown
in Appendix M.
Model flux distributions are calculated using the cspflux task. This task has the parameters listed below. The fctype parameter defines whether blackbody or kurucz models are used. The kmag parameter defines the K magnitude normalization to be used, and should be set to the K magnitude of the flux calibrator star. The temp parameter specifies the blackbody color temperature, or the Kurucz model effective temperature to be used. The av parameter defines the visual extinction in magnitudes to be applied to the model flux distribution. The cspflux task outputs the J, H, and K magnitudes, J-K and H-K colors, and K band normalisation constant for the specified model. Different model parameters should be tried until a satisfactory fit to the J, H, and K magnitudes of the flux calibrator star has been found. The model J-K and H-K colors listed in Table 1 help in converging on this solution.
I R A F
Image Reduction and Analysis Facility
PACKAGE = caspir
TASK = cspflux
fctype = k Type of flux calibration to use
kmag = 4.7 K magnitude of flux calibrator
temp = 5500 Adopted stellar temperature
(av = 0.) Visual extinction.
(verbose= yes) Verbose output?
(mode = ql)
Table 1: Model Near-Infrared Colors
The redgspec task searches the file
caspirdir$fluxstds.dat to locate model parameters for the flux
calibrator star using the object name header entry as the search
parameter. Entries in this file have the format shown below where the
columns are the object name in uppercase characters, the model type
(b for blackbody, k for Kurucz), the K magnitude, the
temperature, the 
value, and the K band normalisation constant.
Entries for new flux calibration stars should be added to this file
when suitable model parameters have been found using the cspflux
task. Model parameters are prompted for if they are not found in the
caspirdir$fluxstds.dat file.
HD216009 b 7.947 10000 0. 4.4555796155286E-12 Y5117 k 3.078 4000 0. 1.7494077560988E-12 Y5584 k 3.382 4000 0. 1.3221814130991E-12 BS8477 k 4.700 5500 0. 4.5171744088401E-13
The model flux densities are in units of 
in
erg/s/cm 
/Å for consistency with other IRAF spectral reduction
packages. The absolute flux calibration of the magnitude system is
based on the normalization derived by Bersanelli, Bouchet, & Falomo
(1991, A&A, 252, 854) and effective wavelengths for the CASPIR
broadband filters calculated for a 10000 K blackbody spectrum. These
effective wavelengths and the zero magnitude flux densities are listed
in Table 2.
Table 2: Adopted Zero Magnitude Flux Densities
Flux calibration is achieved by dividing the object spectrum by the observed spectrum of the flux calibrator and multiplying by the absolute flux spectrum modelled for the flux calibrator. This is done in the redgspec task by setting the fluxcal flag and specifying the base name of the flux calibrator file in the fluxspec parameter. No correction is applied for airmass differences between the object and flux calibrator; it is assumed that the observations were made at similar airmasses in order to optimise terrestrial atmospheric absorption correction. The flux calibrated object spectrum is stored in a file named by appending .fos to the base spectrum name and the flux calibrated sky spectrum is stored in a file named by appending .fss to the base spectrum name. Comparison spectra are not flux calibrated.