Atmospheric absorption bands limit the wavelength ranges over which
one can usefully observe in the near-infrared from ground-based telescopes. NICMOS, however, can observe at any wavelength where its detectors are sensitive, and hence its filter bandpasses do not conform to those of instruments used at ground-based observatories. Instead, its filters were designed to meet the anticipated scientific demands. Thus in practice NICMOS does not have filters matched to any of the ground-based photometric bands and calibration has to be performed independently. The photometric calibration of NICMOS data is discussed in this section. Cases of continuum sources, emission lines, and grism spectra will be presented.
Given the multitude of units and systems that have been used for
infrared photometry (magnitudes, Jy, W m-2 μ
, erg sec-1
, etc.) and given the lack of a standard for ground-based infrared filters, NICMOS has adopted the approach of calibration to physical units, i.e. Janskys (Jy), or Jy arcsec-2
for surface brightness. Details on how to transform different sets of units can be found in Appendix B of the NICMOS Instrument Handbook
or obtained using the NICMOS Unit Conversion Tool
(available on the STScI NICMOS Web
The NICMOS calibration pipeline provides two photometric parameters
for the conversion of countrates into fluxes. These parameters are found in the keywords PHOTFNU
in the header of the calibrated image. The definitions of these keywords are discussed in Section 3.4.4 of the "HST Data Handbook
". Very briefly, for NICMOS these are the bandpass-averaged flux densities (in Fλ
) for a source that would produce a count rate of 1 DN sec-1
is given in units of Jy sec DN-1
in units of ergs cm-2
. Because NICMOS calibrated data are given in countrate, i.e., DN sec-1
, the countrate to flux conversion is simply achieved by multiplying the countrate by the PHOTFNU
value, depending on which units are desired for the final calibrated image.
A fundamental challenge in calibrating NICMOS photometry is the
inherent uncertainty in absolute flux calibration in the near-infrared, whether from the ground or from space. Unlike the situation at optical wavelengths, there really are no absolutely calibrated spectrophotometric flux standards at near-infrared wavelengths. Ground-based photometric calibration observations are mostly limited to relative magnitude determinations in more or less standard bandpasses (JHK
) defined by atmospheric absorption windows. But the absolute flux calibration of primary infrared standards has generally been based on reference to some assumption that a particular type of standard star, e.g., solar-type stars, A0 stars, or white dwarfs, has a spectrum similar to something else that is believed to be well understood, such as the Sun or a well-calibrated stellar atmosphere model. As an example, see Campins, Rieke & Lebofsky (1985), who used the solar analog approach to derive absolute flux calibrations in the near-infrared. These calibrations are uncertain at a level of several percent, setting a fundamental limit on the accuracy of any absolute flux calibration in the near-infrared, including that of NICMOS.
In the case of NICMOS, the photometric calibration is based on
observations of solar analogs and hydrogen white dwarfs (WDs). The stars P330E (solar analog) and G191B2B (WD) define the primary calibration, with a few other stars observed as cross-checks. Bohlin (2006, astro-ph/0608715) recalibrated the absolute flux density distribution of the HST
NICMOS standards. The UV, optical and near-IR portions of their spectra were measured with high quality HST
STIS and NICMOS grism observations. These were then carefully matched to stellar atmosphere models to correct for the NICMOS nonlinearity and for extrapolation to near-infrared wavelengths not covered by the grism observations. The absolute calibration of these spectra are now thought to be accurate at the 2-4% level, the relative errors within the spectra are at the 1-2% level (Bohlin, 2007, ASPC, 364, 315)
The STScI NICMOS group has reanalyzed all photometric calibration
data taken in orbit with NICMOS during Cycles 7, 7N, 11 through 15. For further updates please refer to the NICMOS Web page at:
as well as interim sets of calibrations presented in earlier editions of the
handbook. Unlike earlier efforts, the new work uses a larger number of measurements, applies uniform and up-to-date NICMOS calibration software and reference files, and employs a more careful treatment of the NICMOS photometric aperture corrections1
(See “Aperture Correction”
) than was used for the original calibration efforts. The current best estimates of the PHOTFNU
values for NICMOS filters can be found on the NICMOS photometry Web page at:
These new values update previous numbers given in older versions of
the PHOTTAB calibration reference files, as well as an interim set of calibrations presented in an earlier edition of this Handbook
(version 4.0). The new calibrations are believed to be accurate to better than 5%, thus satisfying the absolute calibration requirements for NICMOS. For most filters, the relative
calibration accuracy should be better, <3%, with the possible exception of some narrow band filters where uncertainties in absorption line strengths for the solar analog and white dwarf models may result in greater calibration errors. The PHOTFNU
values listed in the photometric keyword tables are the error-weighted averages determined from the P330E and G191B2B calibration stars. The listed fractional error was determined form the individual uncertainties in the determination of both stars and their RMS difference. It does not include the uncertainty of the absolute calibration of the reference spectra, which is believed to be less than 2-4%.
Please note that the NIC1 and NIC2 polarizers have not yet been
recalibrated. The values given in the pre-NCS photometric keyword tables
are taken from the older PHOTTAB calibration reference file i7l12297n_pht.fits.
Likewise, the polarizer values in the post-NCS photometric keyword tables
were derived from the old Cycle 7 values, only correction for the change in sensitivity due to the higher operating temperature during post-NCS. Note also that the photometric calibration for the extremely broad band filters F140W, F150W and F175W should be used with caution. These bandpasses are so broad that the conversion from source flux to count rate will be very strongly dependent on the color or spectrum of the source.
|We list photometric zeropoint tables for both the pre- and post- NCS
eras. The higher operating temperature of the detectors after installation of NCS has significantly increased their sensitivity (15-70% depending on wavelength), which is reflected in the calibration zeropoints. The post-NCS calibration values were not available before the summer of 2004, and any data retrieved from the archive before this time will have incorrect calibration keywords in their headers.
In the header of calibrated NICMOS images, there are three additional
photometric parameters that characterize the filter used for the observation (PHOTPLAM
) and provide the ST magnitude zero point (PHOTZPT
gives the value of the pivot wavelength of the filter in Angstroms. This wavelength is source-independent and is the wavelength for which:
is the speed of light in vacuum. PHOTBW
gives the rms band width of the filter in Å (see the Synphot User’s Guide
for a detailed definition of both parameters).
The magnitude of an object can be determined in the ST system (e.g.,
based on a constant flux per unit wavelength) using the photometric zeropoint keyword PHOTZPT
(= -21.1) simply by:
is the count rate in units of DN sec-1
. On the other hand, the magnitude in Oke’s ABν
system (e.g., based on a constant flux per unit frequency) is obtained by applying the following expression:
As noted above, the NICMOS filter bandpasses do not match those of
ground-based filters, and therefore it is not straightforward to derive “standard” infrared magnitudes normalized, e.g., to the spectral energy distribution of Vega. However, it may sometimes be useful to approximate a Vega-based photometric system by using an estimate of the flux density of Vega through the NICMOS bandpasses as a photometric zero point. Because there are no direct spectrophotometric observations of Vega or other A0V standards through the NICMOS filter system, our best option is to use a spectrophotometric model for Vega and synthetically integrate it through the bandpasses. The photometric keyword tables
give bandpass averaged flux densities (in Jy) for the NICMOS filters using a model reference spectrum of Vega (Bohlin 2007, ASPC, 364, 315) available as alpha_lyr_003.fits in the database at STScI. An approximate Vega-normalized magnitude may be computed as
m = ZP
(Vega) - 2.5 log (PHOTFNU
(Vega) is the magnitude of Vega. Under the “CIT” infrared photometry scale, Vega is defined to have ZP(Vega) = 0.0, whereas in the “Arizona” system (e.g., Campins et al. 1985, AJ
, 90, 896), Vega has ZP
(Vega) = 0.02.
The photometric stability of NICMOS has been monitored throughout
the instrument’s lifetime using repeated observations of the primary standard stars through a limited set of filters in each camera. During Cycle 7/7N there is evidence for a small drift in the throughput with time, amounting to roughly 2% in this time period. This is to be expected, as the array quantum efficiency is temperature dependent, and the instrument temperature gradually warmed throughout Cycle 7 and 7N.
Since the installation of NCS (Cycles 11 and beyond) there is some
evidence for a gradual decline in sensitivity of a few percent over the last six years. This was more surprising, the detector temperatures had been kept stable according to the mounting cup temperature sensors used to stabilize the temperature with the cryo-cooler. However, the mounting cup sensors are not really directly connected to the detectors and by using the temperature-from-bias technique (Section 3.3.1
) it has now been shown that the detectors have been cooling down slowly since Cycle 11.
By correlating the count rates of standard stars to the
temperature-from-bias measurements we can determine the detector sensitivity dependence on temperature. Figure 5.1
shows an example of the correlation between detector temperature and sensitivity (flux / count rate) for a number of standard stars. This shows that the count rate is increasing for a give flux when the detector temperature increases.
For Cycle 7/7N there is a wavelength dependence on the slope of this
relation as is shown in Figure 5.2
. This is most clearly observed for Camera 2, but similar trends are observed for Camera 1 and 3, albeit with much more scatter, and for Cycle 7/7N the same wavelength dependence for all cameras is assumed. For post-NCS data no clear wavelength dependence is found and the same 1.4% per Kelvin sensitivity change is consistent with the data in all filters of all cameras.
reference file that is also used by calnica to populate the photfnu and photflam header keyword items (see Section 5.2.3
). When unitcorr=’perform’ in the header of an imaging observation and the tfbtemp header item is within the expected temperature range, calnica will automatically scale the output image to match the photflam and photfnu in the header. Scaling the image rather than the modifying the calibration header keywords was chosen to facilitate drizzling of multiple images without further need to scale the images. The scaling factor is stored in the header item zpscale. The scaling by calnica can be switched off by setting photcalc=’omit’.
The photometric values provided in the headers are obtained from
measurements of standard stars in the central regions of the detectors. Both high frequency (pixel-to-pixel) and low frequency (large-scale structures) sensitivity variations are corrected using on-orbit flats. On-orbit differential photometry characterization of NICMOS cameras 1 and 2 indicate that residual, large scale deviations amount to < 2% (rms). Much larger variations are seen for Camera 3, but are due to a different effect, intrapixel sensitivity variations, which we will discuss next.
As with many other infrared arrays, NICMOS detector sensitivity varies
across the area of each physical pixel. It is higher in the center, and lower near the edges. In practical terms, this means that for a source whose flux changes rapidly on a scale comparable with or smaller than that of a pixel, the measured countrate, and therefore the derived flux, will depend on where the center of the source lies with respect to the center of the pixel.
For NICMOS cameras 1 and 2, the PSF is sufficiently well sampled that
intrapixel sensitivity variations do not introduce a large (or at least, not a dominant) effect on point source photometry. Camera 3 images, however, are highly undersampled, and the exact position of a point source relative to the pixel grid can introduce large variations in the measured signal. Measurements made using calibration test data, as well as science data with large numbers of dither positions, show that the flux variation can be as large as 30%. The tighter the PSF is, the larger these variations will be. For that reason, “in focus” images taken during the two Cycle 7N refocus campaigns are more subject to the impact of intrapixel sensitivity variations, although the effect is still appreciable for non-campaign data as well. Similarly, stellar images through the shorter wavelength filters (where the telescope diffraction limit is smaller) show a greater variation due to this effect than do longer wavelength data. In addition, the intrapixel sensitivity variations can introduce complications for measuring image centroids, “pulling” the centroid away from pixel corners and edges toward the center of the pixel.
The effects of intrapixel sensitivity variations and possible approaches
to correcting NIC3 point source photometry are discussed in two useful references: NICMOS ISR-99-005
(Storrs et al.), available from the STScI web pages, and a paper by Tod Lauer (1999, PASP 107, p.1434). Storrs et al. derive a photometric correction for Camera 3 F110W and F160W stellar images based on a “sharpness ratio,” defined as the ratio of the flux in the peak pixel to that in the integrated PSF. This method is simple to apply and avoids uncertainties due to centroiding errors. Lauer derives a two dimensional pixel response map which can be used to correct photometry if the position of a point source can be measured accurately.
Intrapixel sensitivity variations should have substantially less effect on
spatially extended sources. In general, given well dithered data sampling many different sub-pixel positions, the variations due to intrapixel sensitivity should average out, but given the rather large amplitude of the effect for point sources (especially for short wavelength, in-focus NIC3 observations), many dither positions (>10) are probably needed before the error on the mean count rate is smaller than other sources of photometric uncertainty in NICMOS.
|Camera 3 intrapixel sensitivity variations can have a major impact on
photometry of point sources. For data taken with fewer than 10 dither positions, this uncertainty may dominate photometric errors unless steps are taken to correct it. We strongly recommend that Camera 3 users concerned with photometry of point sources read the references cited here.
The point spread function (PSF) of HST changes with time, and these
changes affect photometry using small apertures. The changes can be divided into short-term and long-term effects. The short-term effects are observed on an orbital time scale and are mainly due to thermal breathing of the whole telescope. In addition, the NICMOS PSF varies as a function of position on the array, and this can affect photometry made either with small apertures (by changing the effective aperture correction as a function of position), or using PSF fitting programs such as daophot
unless there are enough stars in the field for the program to derive a positionally dependent PSF. The focus field variations, however, are relatively small (see ISR 98-005
, "NICMOS Focus Field Variations and Focus Centering," Suchkov & Galas 1998; see also NICMOS ISR 98-018
(Suchkov & Krist) for a discussion of the effects of breathing on the PSF modeling). We have used synthetic PSFs from TinyTim (See “Aperture Correction”
) to determine the positional dependence of the aperture correction as a function of position on NIC2, where the focus field variations are largest. For F110W (the effect should be largest at short wavelengths), the maximum to minimum variation in the aperture correction for photometry measured within a 2 pixel radius is < 2%. The variations are expected to be smaller for the other cameras and for images at longer wavelengths.
During the pre-NCS era in Cycle 7 and 7N there were long term
variations in the NICMOS focus as the cryogen evaporated and the dewar relaxed. The solid nitrogen introduced stress on the instrument, and NICMOS detectors moved along the focus direction as this stress diminished. The change was especially rapid early in Cycle 7. Thereafter the focus remained largely stable for the reminder of the pre-NCS era. The dewar distortion had the greatest effect on the focus of Camera 3, pushing it beyond the range of the Pupil Alignment Mechanism (PAM). For this reason, two special Camera 3 refocus campaigns, each three weeks long, were undertaken, one from 12 January through 1 February 1998, and the second from 4 June through 28 June 1998. During these periods, the secondary mirror was moved in order to bring Camera 3 into the range where its correct focus could be reached by the PAM. From the post NCS era there has not been any long term effects on the focus for any of the cameras. The focus history of NICMOS can be determined from the focus page of the NICMOS Web site and for further updates please refer to the NICMOS Web page at:
It is often difficult to measure the total flux of a point source due to the
extended wings of the PSF, diffraction spikes, and scattered light. Such measurements are particularly difficult in crowded fields where the extended wings of sources can overlap with each other. An accurate method of measuring the integrated flux in these situations could consist of several steps:
|Simulate the TinyTim2
PSF for the particular camera-filter combination and position in the detector. TinyTim
allows the user to simulate PSFs with various focus settings, and users concerned with small focus variations should be careful to set these parameters to the values appropriate for their observations. (This will have the largest effect for NIC3 data taken outside the refocus campaigns.)
To obtain the total flux of the source, the countrate then only needs to be
multiplied by the PHOTFNU
value and by the inverse of the measured fraction obtained in step three above. The apertures and aperture corrections used to derive the calibration constants in Section 5.2.2
are listed on the NICMOS photometry Web
Empirical PSFs could also be used for the above mentioned method.
There have been no special calibration programs to obtain PSF measurements for all camera and filter combinations. A significant amount of PSF data does exist in the archive, however: empirical PSFs for the central regions of the detectors can be obtained from the calibrated images obtained for absolute photometry and photometric monitoring programs.
Many very red targets (e.g., protostars) have been observed with
NICMOS at short wavelengths (~1 μ
m). For these sources the flux at ~2.2–2.5 μ
m could be orders of magnitude larger than at ~1.0 μ
m and therefore exceptionally good out-of-band blocking would be required. Pre-launch tests indicated that for very red sources (temperature ~700 K and lower), some NICMOS filters might
have significant red leaks. The suspect filters were F090M, F095N, F097N, F108N, F110M, F110W, F113N, F187N and F190N. A limited set of in-flight tests (in calibration programs 7691 and 7904) were made during Cycle 7, observing very red stars to calibrate the possible effects of out-of-band leaks. These data have never been carefully analyzed, but preliminary indications are that actual red leaks are insignificant or non-existent. We nevertheless caution here that this analysis is only preliminary. Users interested in photometry of sources with extreme colors should check the STScI NICMOS Web
pages for updates, or examine the in-flight test photometry of red stars themselves.
The problem of the non-zero zeroth read for bright sources was
discussed in Section 3.3.2
(ZSIGCORR) and Section 4.4.2
. If significant signal from an object is present in the zeroth read, then this was not properly taken into account in the nonlinearity corrections made for data processed by the OPUS pipeline before 11 November 1997. It is advisable to reprocess the data with the most recent version of the calibration software which includes the ZSIGCORR
step to account for this zero read signal correction. When reprocessing such data, calnica
version 3.3 or greater will automatically apply this step to all MULTIACCUM mode observations if both ZOFFCORR
are also being performed. All data retrieved from the Archive via OTFR after 26 September 2001 are automatically processed with this step.
As was previously mentioned, NICMOS data are calibrated in units of
Jy or Jy arcsec-2
for flux densities and surface brightnesses, respectively. The filter bandpasses do not exactly match those of ground-based JHK filters, and therefore if you wish to derive standard JHK magnitudes from NICMOS data, then it will be necessary to apply a color term correction. The recommended NICMOS JHK-analog system is obtained using the F110W, F160W and F222M filters, although the F110W filter especially is quite different from the standard J bandpass (specifically, it is much broader and bluer). Synthetic color terms can be computed using synphot
, given real or model spectra for an astronomical source and the NICMOS and ground-based filter bandpasses (see, for example, Holtzman et al. 1995, PASP 107, 1065, where a similar procedure was used to derive color terms for WFPC2 photometry).
Also, as part of the absolute photometry program, we observed a few
blue stars (white dwarfs), intermediate color stars (solar analogs), and very red stars covering a large range in color (Table 5.1
). The calibrated data are available from the archive for users who wish to derive their own color terms to transform their HST
fluxes into any ground-based system. The ground-based photometry given in the table below is based on preliminary and incomplete measurements, and will be updated when the final NICMOS photometric calibrations are completed.
|The narrow band filters in NICMOS are intended primarily for
observations of emission or absorption lines in sources. Because the photometric conversion factors PHOTFNU
for all NICMOS filters are obtained from continuum observations of emission-line free standard stars, the flux in erg sec-1
of an emission line is given by the expression:
where FWHM is the full width half maximum of the equivalent gaussian
filter to the narrow-band filter used (see Appendix A of the NICMOS Instrument Handbook
, and we have assumed that the continuum has been already subtracted from the total flux in the filter and that the line is centered in the filter. If the emission line is not at the central wavelength of the filter, the line flux will need correction for the filter transmission curve. To estimate the variation in the absolute flux due to the positioning and width of the emission line in the filter bandpass, the synphot
can be used as shown below. See the Synphot User’s Guide
for additional help.
The examples in Figure 5.3
compute the countrate in the NIC3 F212N filter for a H2
(2.12 micron) emission line having a gaussian profile of 40 Angstroms and a peak flux of 1.0 x 10–13
. The integrated flux will then be 4.2 x 10–12
. In the first example the H2
emission line is at zero redshift and centered on the filter while in the second example the line is redshifted by 80 Angstroms. If the emission line is centered on the filter, the H2
flux will produce 7421.1 DN sec-1
while the countrate will be ~90% of this value (i.e., 6662.3 DN sec-1
) for the redshifted emission line. The expression for the Fluxline
above can be directly applied to the first case, while a correction factor 1.11 = (7421.1/6662.2) is needed in the second case.
The major source of uncertainty in the absolute spectrophotometry of a
given source comes from the variability and structure of the background. Every pixel on the NIC3 array will receive background radiation over the entire spectral bandpass of the particular grism, while the source spectrum will be dispersed. The accuracy of the background subtraction is limited by our knowledge of the spectral response of each pixel, which is somewhat different from pixel to pixel. Extracted spectra will have to be corrected for the spectral response of each pixel. The accuracy of this correction is again limited by our knowledge of the detector response.
Remember that the spectra are not aligned with the X axis of the
detector, and therefore move from one row to the next. This makes grism spectra especially sensitive to the intrapixel sensitivity variation. The NICMOSlook
software (see Section 5.7
) have a capability built in to correct for this variation which depends critically on the placement of the spectrum. The software handbook describes an iterative procedure to determine the best correction for this problem.
The absolute flux calibration of the spectral energy distribution of the
standard stars in the 0.8 to 2.5 μ
m wavelength range is known to 2–5%. Grism data reduction and calibration are discussed in more detail in Section 5.7