Gemini Near-Infrared Integral-Field Spectrograph (NIFS)
|
|
|
Performance Model
NIFS is an unconventional and unfamiliar instrument incorporating high-order AO correction on an 8 m telescope with a near diffraction-limited integral-field unit and a moderate resolution near-infrared spectrograph. Before discussing details of the science that NIFS will address, it is necessary to inform that discussion with realistic predictions of the instrument performance. This has been done by developing a NIFS performance model which estimates the detected currents from various noise sources in the instrument and compares these to the expected signal photo-current. This is implemented as a WWW-based tool for point sources and as a detailed data simulation program, NIFSSIM, for more complex situations. The performance data presented here refer to the NIFS optical design defined in the ZEMAX file nifs_ful_40.zmx. This uses 31 slitlets each 0.1² wide with 0.05² pixels in the spatial direction to accept a 3.1²´3.3² field-of-view. The current baseline design for NIFS uses 29 slitlets each 0.1² wide with 0.04² pixels in the spatial direction to accept a 3.0²´3.0² field-of-view. The two designs differ only slightly in other respects; both designs use a Bouwers spectrograph collimator and the throughput calculation assumes refractive designs for the spectrograph cameras. The input parameters used in the performance model are summarized in Table 1.
Table 1: Model Input Parameters
|
Parameter |
Value |
Unit |
Description |
|
|
|
|
|
|
RN |
5 |
e |
Detector read noise |
|
D |
50000 |
e |
Detector well depth |
|
Tmax |
3600 |
s |
Maximum integration time |
|
Tcry |
70 |
K |
Cryostat temperature |
|
lmax |
2.60 |
mm |
Detector sensitivity cutoff |
|
Q |
0.60 |
|
Detector quantum efficiency |
|
Twin |
275 |
K |
Window temperature |
|
ewin |
0.01 |
|
Window emissivity |
|
Taos |
275 |
K |
AOS temperature |
|
eaos |
0.12 |
|
AOS emissivity |
|
taos |
0.88 |
|
AOS transmission |
|
Ttel |
275 |
K |
Telescope temperature |
|
etel |
0.044 |
|
Telescope emissivity |
|
ttel |
0.956 |
|
Telescope transmission |
|
Tsky |
260 |
K |
Sky temperature |
|
Wh |
0.4 |
arcsec |
Seeing FWHM |
|
S |
0.2,0.4,0.6 |
|
Strehl ratio at J, H, K |
2.1 Noise Sources
2.1.1 Detector Read Noise
The NIFS detector will be a Rockwell 2048´2048 HAWAII-2 array with 18 mm pixels. The read noise, RN, is expected to be ~ 9 e for a single non-destructive read (NDR). However, it should be possible to reduce this to ~ 5 e using ~ 16 NDRs (i.e., either Fowler sampling or linear fitting). We adopt the lower value. The detector well depth is assumed to be D = 50000 e. This is appropriate for the low detector reverse bias voltage that will be needed to minimize dark current. The maximum integration time will be set by sky variations and cosmic ray events. We adopt tmax ~ 3600 s initially, but this may be extended using software rejection of OH airglow emission and cosmic ray events. A cosmic ray rate of ~ 1% per hour is expected (Hall, priv. comm.).
2.1.2 Detector Dark Current
The dark current, Idc, for a PACE technology HAWAII-2 array will be similar to that of a HAWAII-1 array. However, the dark current obtained with a typical HAWAII-1 array is poorly documented. At least one device has a mean dark current of ~ 0.0175 e s‑1 pix‑1 and a standard deviation of ~ 0.0288 e s‑1 pix-1 (Rockwell Science Center WWW pages; Figure 1), although various users report higher values. HAWAII-2 arrays based on CdZnTe substrates and MBE technology are expected to reliably deliver mean dark currents this low (i.e., mean of 0.01 e s‑1 pix‑1). It is not yet clear whether NIFS will use a MBE technology array. We adopt the dark current distribution of Figure 1 and approximate the width of the distribution using random Gaussian deviates with a mean of 0.01 e s‑1 pix‑1 and a standard deviation of 0.0065 e s‑1 pix‑1, and resample Gaussian deviates falling below zero. A simulated 3600 s dark exposure is shown in Figure 2. Dark current will be a significant noise source in NIFS in the J and H bands, and the true dark currents may be significantly higher than those modeled if a MBE device is not used in NIFS (§8.1.2).
Figure 1: Dark current distribution for a HAWAII-1 array (Rockwell WWW pages).
Figure 2: Simulated 3600 s dark exposure dominated by the adopted fixed dark current pattern.
The width of the adopted dark current distribution (Figure 1) exceeds that expected from measurement errors alone. We interpret this width as due to real pixel-to-pixel dark current variations, possibly due to pixel-to-pixel capacitance variations. This systematic dark current pattern (Figure 2) will dominate dark current shot noise for typical integration times. This makes it necessary to characterize and remove an average dark current pattern in order to achieve optimal dark current-limited performance. We envisage characterizing the dark current pattern by median combining 5 to 10 dark exposures of 1 hr duration recorded during the day prior to observing. The degree to which this approach will be successful will depend on the stability of the dark current pattern over timescales of ~24 hr. We are moderately optimistic that the required stability is achievable based on discussions with Don Hall, but no quantitative data are available.
2.1.3 Systematic Detector Noise Sources
Anecdotal accounts of illumination dependent bias shifts in HAWAII-1 arrays associated with a systematic noise component have been reported for ISAAC on the VLT (Lidman, priv. comm.). If present in HAWAII 2 arrays, such noise sources will affect the performance of NIFS. No quantitative information is available, so these noise sources cannot be included in the present simulations.
2.1.4 Cryostat Thermal Emission
Each pixel views 2p steradians of the cryostat interior which is effectively radiating with unit emissivity as a blackbody at the cryostat temperature, Tcry. This emission occurs after the order blocking filter and grating so the detector sees broadband emission extending to the detector cutoff wavelength, lmax = 2.60 mm for a HAWAII-2 array. The detector quantum efficiency, Q, is taken to be that of the HAWAII-1 array published on the WWW pages of the Rockwell Science Center and shown in Figure 3. The detector quantum efficiency is ~ 60% over most of the 1-2.5 mm wavelength range. Rockwell CdZnTe/MBE technology devices have higher quantum efficiencies due to the better lattice match with HgCdTe compared to PACE devices. Quantum efficiencies of > 85% may be achieved over the 1.0-2.5 mm wavelength range with MBE devices. This gain has not been included in the calculations performed.
Figure 3: Quantum efficiency for a HAWAII-1 array (Rockwell WWW pages).
Cryostat thermal emission produces a photo-current of
e s-1 pix-1
where Apix is the detector pixel size in cm2. This photo-current for a detector cutoff wavelength of 2.6 mm is shown in Figure 4 as a function of Tcry. The cryostat thermal emission photo-current should be insignificant compared to the detector dark current so it must be below ~ 0.001 e s-1 pix-1. This requires Tcry < 135 K. The duplicate NIRI cryostat is expected to reach a temperature below 70 K.
Figure 4: Cryostat thermal photo-current versus cryostat temperature for a 2.6 mm detector cutoff wavelength (left) and 5.5 mm detector cutoff wavelength (right).
Use of a 2048´2048 HgCdTe/CdZnTe MBE technology detector with sensitivity to lmax = 5.5 mm will dramatically increase the requirement on controlling cryostat thermal emission, principally by requiring that the cryostat temperature be lower than ~ 65 K. The cryostat thermal emission photo-current as a function of Tcry for a 5.5 mm detector cutoff wavelength is also shown in Figure 4.
2.1.5 Window Thermal Emission
Dust on the cryostat window is a source of thermal emission. We assume that the window has an emissivity of ewin = 0.01 and a temperature of Twin = 275 K. Thermal emission from the cryostat window generates a photo-current of
e s-1 pix-1
where Atel is the telescope collecting area in cm2, Wpix is the solid angle on the sky subtended by one pixel along the slit, Dl is the spectral resolution per pixel, and tspe is the transmission of the NIFS spectrograph. The reformatted NIFS spectrograph slit has a width of 0.1² and each detector pixel is assumed to project to 0.05² on the sky. Spectral resolutions for the NIFS gratings are discussed in §4.6. The NIFS spectrograph transmission is described in §4.9. The assumed transmission as a function of wavelength is shown in Figure 5. The grating efficiency is the largest throughput uncertainty. All simulations assume a wavelength-independent grating efficiency of 0.65. NIFS will use moderate resolving power gratings with R ~ 5300 and may also use lower resolving power gratings with R ~ 2800 or higher resolving power gratings with R ~ 7000. The moderate resolving power and higher resolving power gratings are expected to have efficiencies closer to 80%. The efficiencies of the lower resolving power gratings may be as low as 50%.
Figure 5: NIFS spectrograph transmission excluding telescope, AO system, and detector with (solid line) and without (dotted line) anti-reflection coatings. The wavelength ranges of the moderate resolving power gratings are indicated.
2.1.6 ALTAIR Thermal Emission
The ALTAIR science path emissivity budget has been presented in the ALTAIR Preliminary Design Review documentation. We adopt an emissivity of eaos = 0.12, a transmission of taos = 0.88 (ALTAIR PDR Documentation p. 34 minus telescope contribution), and an operating temperature of Taos = 275 K. The emissivity is higher when the ALTAIR atmospheric dispersion corrector is used. This will only be necessary in the J band (§10.2) where thermal emission from the ALTAIR optics is negligible, so the emissivity of the atmospheric dispersion corrector has been ignored. Thermal emission from the ALTAIR science path generates a photo-current of
e s-1 pix-1.
2.1.7 Telescope Thermal Emission
The telescope M1, M2, and M3 mirrors are assumed to use protected silver coatings with a single surface reflectivity of 98.5% as discussed in §4.9. This gives a combined emissivity of etel = 0.044, a combined transmission of ttel = 0.956, and we adopt a typical night-time operating temperature of Ttel = 275 K. Thermal emission from the M1, M2, and M3 telescope mirrors generates a photo-current of
e s-1 pix-1.
2.1.8 Airglow Line Emission
Sky emission is significant at near-infrared wavelengths. Airglow emission, mainly due to OH molecules, is dominant shortward of ~ 2.27 mm and thermal emission from the sky contributes at longer wavelengths. Airglow line emission data are based on a tabulation of typical line strengths provided by François Piche which is based on the data of Maihara et al. (1993) and Oliva & Origlia (1992). These data have been checked for order of magnitude consistency with the atmospheric emission spectrum provided on the Gemini WWW pages, and have been extended into the K band with additional line strength and wavelength data chosen to mimic this Mauna Kea emission spectrum. The airglow emission spectrum is calculated using an emission-line profile derived from the NIFS diffraction analysis (§4.4). This ignores near-angle scatter at the grating which may prove to be significant. Near-angle scatter at the grating will be dominated by satellite lines produced by grating ruling defects.
2.1.9 Airglow Continuum Emission
A knowledge of the atmospheric continuum emission between strong airglow emission-lines is crucial for accurately predicting NIFS performance in the J and H bands. This is poorly known and difficult to distinguish empirically from instrumental scattered light. Sobolev (1978) reported an airglow continuum emission of 300-600 photon s-1 m-2 arcsec-2 mm-1 in the J band. McCaughrean (1988) estimated the airglow continuum to be 280 photon s‑1 m-2 arcsec-2 mm-1 in the H band. Maihara et al. (1993) measure a continuum of 580 photon/s/m2/arcsec2/mm at 1.665 mm but found higher values on moon lit nights. We adopt the latter value at all wavelengths and note that the precise value is not important if other continuum sources dominate. Nevertheless, dark/gray time may be required for sensitive NIFS observations in the J and H bands.
2.1.10 Scattered Airglow Emission
A significant concern is that scattered light within the cryostat will exceed the low natural continuum emission level between strong airglow emission-lines. Scatter or diffraction of line emission into the neighboring continuum will occur at some level. To model this, we adopt the (hopefully) pessimistic assumption that 10% of the total airglow line emission within the spectral band of interest is redistributed uniformly across the detector. This is sufficient to make scattered airglow emission rival dark current as a noise source.
2.1.11 Sky Thermal Emission
Thermal emission from the sky will also produce continuum emission with increasing importance towards longer wavelengths. We assume a typical sky temperature of Tsky = 260 K and base the adopted sky emissivity, esky, on Mauna Kea atmospheric transmission data, tsky, available from the Gemini WWW pages. Thermal emission from the sky generates a photo-current of
e s-1 pix-1.
In fact, this sky thermal emission may be the source of the “airglow” continuum (§2.1.9) in the J and H bands. Nevertheless, we prefer to be conservative by including both continuum sources in the NIFS performance model; this may over-estimate the sky continuum in the J and H bands by a factor of ~2.
2.2 Signal Currents
2.2.1 Point Spread Function
The signal-to-noise ratio achieved by NIFS on point sources depends on the point spread function (PSF), P(q); poorer images spread light over more detector pixels. ALTAIR will deliver a complex image to NIFS (Figure 6). We approximate P(q) by a diffraction-limited core, Pc(q), and a seeing-limited halo, Ph(q). The diffraction-limited core is modeled with an Airy function modified to allow for the telescope central obstruction (Schroeder 1987):
where J1 is the Bessel function of order one, e is the ratio of the central obstruction radius to the primary mirror radius, and
where RP is the primary mirror radius, q is the angle on the sky, and l is the wavelength. The seeing-halo is modeled by a Moffat function with index of 11/6 as suggested by Racine et al. (1999) based on results from the AO Bonnette on CFHT:
where Wh is the seeing FWHM. The contribution of each profile is set by the Strehl ratio, S, such that
and
where Sc and Sh are the total counts in the core and halo templates, respectively. We define baseline observing conditions to correspond to Strehl ratios of 0.2, 0.4, and 0.6 in the J, H, and K bands, respectively, and a seeing FWHM of 0.4².
Figure 6: Natural (left) and AO-corrected (right) images of a star obtained with the University of Hawaii AO system at K showing the diffraction-limited AO-corrected core and speckle structure in the seeing halo.
The quoted Strehl ratios are those derived from the Gemini top-down image quality error budget in median seeing conditions (Oschmann 1997, Gemini System Error Budget Plan, SPE-S-G0041). The Strehl ratio degradation is due to a combination of uncorrected seeing and optical aberrations in the telescope, ALTAIR, and the science instrument. Clearly, only the uncorrected seeing component is likely to be distributed in a manner similar to the seeing PSF. However, in the absence of more detailed information, we assume that all of the halo light is distributed in this way. This is likely to over-estimate the extent of the true halo and under-estimate speckle structure in the halo distribution. In reality, the image halo is likely to be dominated by a number of time-variable discrete peaks superposed on a lower surface brightness continuum.
2.2.2 Point Sources
The source signal current, Isig(q), is derived from the source brightness specified in magnitudes, m, and converted to flux density using the absolute flux calibration of Bersanelli et al. (1991) which is based on a Vega effective temperature of 11200 K:
W cm-2 mm-1,
e s-1 pix-1.
The stellar spectral distribution in the K band is interpolated from the FTS spectra of Kleinmann & Hall (1986). These spectra are tabulated at approximately twice the highest resolution of NIFS data. We generally use their spectrum for the M1 III star 75 Cyg.
2.3 Performance Tools
2.3.1 Performance Web Tool
A simplistic web-based tool is available at http://www.mso.anu.edu.au/nifs/performance.html for making approximate signal-to-noise ratio calculations for point sources only (see Figure 7). Default parameters should be used in most cases. The source brightness in magnitudes and the grating selection must be specified. The user clicks on the Calculate button to display contributing photocurrents and the approximate signal-to-noise ratio achieved in a 0.1²´0.1² aperture with a Strehl ratio of 0.2 in 0.4² FWHM seeing. The signal photo-current is quoted as 17% of the total stellar signal in order to approximate aperture effects on the PSF (see §2.8.2). The calculation is performed at the central wavelength of the selected grating. OH line emission is not included in detail. However, a numerical line flux can be specified. The calculated signal-to-noise ratio includes a factor of Ö2 degradation for sky subtraction during data reduction. No averaging is performed in the spectral direction. No attempt is made to treat wavelength dependent parameters in detail. In short, this tool has limited application.
Figure 7: NIFS web-based performance calculator.
2.3.2 Data Simulation Program: NIFSSIM
Detailed assessment of NIFS performance requires full simulation of NIFS data frames including all spatial and spectral dependencies and data reduction steps. The program, NIFSSIM, has been written to perform full 3D simulations of NIFS data sets. The program simulates all the NIFS noise sources discussed in §2.1 and can include signals from single stars, binary stars, galactic nuclei, disk galaxies, and QSOs. The total detected signal, Stot, is given by
e pix-1
where T is the integration time. To this is added statistical noise which is derived from random Gaussian deviates with standard deviation, N, given by
for each pixel.
NIFSSIM produces bias frames, dark frames, arc frames, sky frames, and object frames which can then be processed to simulate a complete data analysis. In its simplest form, NIFSSIM can be run in “Observe” mode to generate typical observational data using default parameters. Default parameters can be changed by editing the NIFSSIM.DAT file. NIFSSIM produces a 2048´2048 pixel image in FITS format containing spectra of 31 NIFS slitlets which each contain 66 pixels in the spatial direction and 2048 pixels in the spectral direction. This geometry differs slightly from the current baseline design for NIFS. Figure 8 shows a simulated 1800 s frame of a K = 15 mag star in 0.4² FWHM seeing with a Strehl ratio of 0.2. Spectral dispersion is in the horizontal direction. Several spectra of the star are obtained through adjacent NIFS slitlets due to the nature of the PSF. Airglow emission-lines are seen as vertical bright lines; the “staircase” slit of real data is not modeled. The brightening to the red (i.e., to the right) is due to increasing thermal background contributions.
Figure 8: Simulated 1800 s integration with the K grating on a K = 15.0 mag star in 0.4² seeing with a Strehl of 0.2.
A NIFS package of IRAF tasks is used to process these images. Data processing consists of first subtracting a sky frame with IMARITH, then reformatting the 2D image into a data cube with the NIFS package task ENCUBE. A sky-subtracted version of the Figure 8 image is shown in Figure 9. ENCUBE forms a cube with 2048 pixels in the spectral direction and 62´66 pixels in the spatial directions so that images of each cube plane have the correct aspect ratio. Images of subsections of the data cube can be made by averaging in the spectral direction using the NIFS package task IEXTRACT (Figure 10). Similarly, spectra of subsections of the image can be extracted with the NIFS package task SEXTRACT (Figure 11). These tasks have been used to derived the NIFS performance characteristics described in §2.5 below.
Figure 9: Sky subtracted version of the frame in Figure 8.
Figure 10: Spatial image of the K = 15 mag star shown in Figure 8. Note the diffraction-limited core and the 0.1²´0.05² pixels produced by the NIFS IFU. The field-of-view is 3.1²´3.3². The Strehl ratio is a modest 0.2.
Figure 11: K band spectrum of an M1 III star extracted from the central 0.1²´0.1² of the data cube shown in Figure 8. Atmospheric CO2 absorption is seen at left. Stellar CO Dv = 2 absorption bands are seen at right.
2.4 Dominant Noise Sources
The various noise sources in NIFS are wavelength dependent and so are not parameterized simply (hence the need for NIFSSIM). The web-based performance tool provides a means of identifying the dominant noise sources at the central wavelengths of each grating. These are listed in Table 2 which demonstrates that observations with NIFS will be limited by dark current and read noise in the J and H bands, and limited primarily by thermal emission from ALTAIR in the K band. K band observations that do not use ALTAIR will be limited in approximately equal proportions by read noise, dark current noise, and thermal emission from the telescope mirrors, and perhaps the uncertain sky continuum.
Table 2: Dominant Noise Sources in a 3600 s NIFS Exposure
|
Grating |
RN2 |
Sky Thermal |
OH Cont. |
Tel. Thermal |
ALTAIR Thermal |
Cryostat Thermal |
Dark Current |
|
|
(e) |
(e/pix) |
(e/pix) |
(e/pix) |
(e/pix) |
(e/pix) |
(e/pix) |
|
|
|
|
|
|
|
|
|
|
J1 |
25 |
0 |
8 |
0 |
0 |
0 |
36 |
|
J2 |
25 |
0 |
9 |
0 |
0 |
0 |
36 |
|
H |
25 |
0 |
14 |
0 |
0 |
0 |
36 |
|
K |
25 |
4 |
17 |
17 |
116 |
0 |
36 |
|
|
|
|
|
|
|
|
|
|
J |
25 |
0 |
18 |
0 |
0 |
0 |
36 |
|
HK |
25 |
9 |
38 |
37 |
254 |
0 |
36 |
2.5 Optimal Wavelength Range
NIFS will record spectra in the full wavelength range over which the transmission of the Earth’s atmosphere exceeds ~ 50%. NIFSSIM atmospheric transmission spectra for Mauna Kea in the J, H, and K bands are shown in Figure 12, Figure 13, and Figure 14. The wavelength ranges corresponding to > 50% transmission are l < 1.35 mm, 1.44 < l < 1.81 mm, and 1.95 < l < 2.50 mm for the J, H, and K bands, respectively. The background emission spectra corresponding to these wavelength ranges are shown in Figure 15, Figure 16, and Figure 17.
Figure 12: J band atmospheric transmission with the 50% transmission range shaded.
Figure 13: H band atmospheric transmission with the 50% transmission range shaded.
Figure 14: K band atmospheric transmission with the 50% transmission range shaded.
Figure 15: J band background emission spectrum with the 50% transmission range shaded.
Figure 16: H band background emission spectrum with the 50% transmission range shaded.
Figure 17: K band background emission spectrum with the 50% transmission range shaded.
2.6 Optimal Spectral Resolving Power
NIFS will record data at moderate spectral resolution and use software suppression to reject OH airglow emission during data reduction. The choice of optimal spectral resolving power depends on the required level of OH rejection and on the science requirements of the instrument. NIFSSIM has been used to calculate the fraction of each of the J, H, and K photometric bands that is blocked by OH airglow emission in spectra recorded at different resolving powers (Figure 18, Figure 19, and Figure 20); as the spectral resolving power increases the fraction of wavelength space occupied by OH emission decreases. The K band OH blocking fraction was calculated over only the short wavelength part of the band containing OH emission. What constitutes an acceptable level of OH blocking depends on the science goals. Measurements of specific spectral features at low redshift may require a low blocking fraction, whereas high redshift galaxies can often be selected in particular redshift ranges and so higher blocking fractions can be tolerated. We subjectively adopt a maximum allowable blocking fraction of 20%. This requires spectral resolving powers greater than 3200, 3800, and 2750 in the J, H, and K bands, respectively.
The spectral resolving power should be chosen to maximize the signal-to-noise ratio achieved, consistent with the required OH blocking fraction and the resolution requirements dictated by the science goals. Figure 18, Figure 19, and Figure 20 also show NIFSSIM predictions for the average signal-to-noise ratios achieved in 1800 s in a 0.1²´0.1² aperture on stars with J = 20 mag, H = 20 mag, and K = 19 mag, respectively, for observations recorded at different spectral resolving powers and then smoothed to a common resolving power of R = 1000. The signal-to-noise ratios were calculated from the mean and standard deviation of data values in spectra of a featureless input star, after correction for atmospheric absorption. In each figure, the solid line shows the signal-to-noise ratios achieved when regions containing OH emission are excluded from the calculation, while the dot-dash lines show the signal-to-noise ratios achieved with no OH rejection. Higher signal-to-noise ratios are achieved in OH-free regions of the J and H bands in spectra recorded at progressively lower resolving powers. This is due to the lower relative contributions of dark current and read noise in low resolution spectra. However, only small fractions of the bands are OH-free at these low resolving powers. The dot-dash lines are applicable in these case where OH rejection is not possible. It is apparent from Figure 18 and Figure 19 that the optimal spectral resolving powers in both the J and H bands are the minimum values that are consistent with the required OH blocking fraction and the science requirements. Observations in the K band are background-limited. Figure 20 demonstrates that the signal-to-noise ratio achieved at R = 1000 is largely independent of the resolving power of the observation (the increasing signal-to-noise ratios at higher resolving powers occur because the high background, long-wavelength end of the K band is progressively excluded from the high resolution simulations). The spectral resolving power that should be used in the K band is again defined by the required OH blocking fraction and science requirements, but the actual value has little impact on the signal-to-noise ratio achieved at R = 1000.
Figure 18: J band fractional blocking due to OH airglow emission versus spectral resolving power (left) and signal-to-noise ratio in 1800 s on a J = 20 mag star versus spectral resolving power (right) with OH lines masked (solid line) and with no OH line masking (dot-dash line).
Figure 19: H band fractional blocking due to OH airglow emission versus spectral resolving power (left) and signal-to-noise ratio in 1800 s on a H = 20 mag star versus spectral resolving power (right) with OH lines masked (solid line) and with no OH line masking (dot-dash line).
Figure 20: K band fractional blocking due to OH airglow emission versus spectral resolving power (left) and signal-to-noise ratio in 1800 s on a K = 19 mag star versus spectral resolving power (right) with OH lines masked (solid line) and with no OH line masking (dot-dash line).
2.7 Optimal Spatial Scale
NIFS will sample the spatial image with two spectral pixels per slit
width. The spatial resolution along the slit is then defined by the spatial
pixel size, while the spatial resolution in the perpendicular direction is
defined by the ~ 2´ larger slit width. The optimal
spatial pixel scale for NIFS is a compromise between adequately sampling the
diffraction-limited image core, providing sufficient field-of-view to satisfy
the science requirements, and minimizing the relative contributions of dark
current and read noise at short wavelengths. The FWHM of the
diffraction-limited image core will be ~ 0.06² in the K band and less at shorter wavelengths. Nyquist sampling this image
in the spatial direction requires a pixel size of ~ 0.03². The field-of-view using 2048
spatial pixels would then be only 1.0²´1.0². The science requirements (§3) dictate a field-of-view of at
least 3.0²´3.0² which is satisfied with a spatial
pixel size of ~ 0.05². The contribution of dark current
and read noise is significant at short wavelengths even with this pixel size.
We give highest priority to the field-of-view requirement in adopting a pixel
size of ~ 0.05² and a slit width of 0.10² as the baseline geometry. The
sensitivity limitation imposed by dark current and read noise remains a
concern.
2.8 Performance Results
2.8.1 Image Quality Effects
We consider the effect of image quality as parameterized by the Strehl ratio on the fraction of light contained within the central 0.1²´0.1² of a stellar profile. This is shown in Figure 21 for our adopted PSF and a seeing FWHM of 0.4² typical of Mauna Kea. In fact, the Strehl ratio does not uniquely define the AO-corrected PSF, especially when the effects of natural guide stars and laser guide stars are compared (Tyler & Ellerbroek 2000), but we ignore this complication. Figure 21 helps to demonstrate several effects of significance to the scientific performance of NIFS: 1) Only ~ 17% of the total stellar flux is contained within the central diffraction core for a Strehl ratio of 0.2. This is the case modeled by the web–based performance calculator and is appropriate to J band observations with ALTAIR in median seeing conditions. 2) Higher Strehl ratios lead to significantly higher sensitivities to point sources. The highest Strehl ratios will be obtained in the K band where Strehl ratios of ~ 0.6 are expected in median seeing conditions. Even then, only ~ 0.5 of the light from a point source will be contained in the central 0.1²´0.1² of the PSF. 3) High Strehl ratios will also be required to separate the spectra of spatially distinct sources. This is particularly true of K band stellar velocity dispersion measurements of galactic nuclei. Approximately 50% of the high velocity dispersion nuclear light will contaminate spectra at larger radii. The effect of this light will need to be modeled based on a knowledge of the actual PSF. It will be necessary to determine this by frequent measurements of a nearby star, by deriving it from the OIWFS output, or by modeling based on a knowledge of the ALTAIR control loop output (Véran et al. 1997). Correcting for contamination due to the residual seeing halo in partially AO-corrected NIFS data will be an important part of NIFS data reduction.
Figure 21: Fraction of PSF contained within a 0.1²´0.1² square aperture for 0.4² FWHM seeing.
2.8.2 Aperture Effects
We now consider the signal-to-noise ratios that will be obtained with NIFS on point sources using different extraction aperture sizes. This is shown in Figure 22 for the K grating and in Figure 23 for the H grating with 1800 s integration times, 0.4² FWHM seeing, and a Strehl ratio of 0.2. The aperture radii are quoted in units of 0.05² pixels. These plots show that the best signal-to-noise ratio is achieved, in general, with an extraction aperture matched to the 0.1²´0.1² diffraction core. Only for extremely bright point sources is the seeing halo sufficiently bright to warrant the use of a larger extraction aperture. The halo will be even less important in observations with higher Strehl ratios. Unless otherwise stated, we always adopt a 0.1²´0.1² extraction aperture in what follows.
Figure 22: Signal-to-noise ratio in K band integrations of 1800 s for different extraction aperture sizes. The seeing FWHM is 0.4² and the Strehl ratio is 0.2.
Figure 23: Signal-to-noise ratio in H band integrations of 1800 s for different extraction aperture sizes. The seeing FWHM is 0.4² and the Strehl ratio is 0.2.
2.8.3 Point Source Performance
We can now predict the point source performance of NIFS. We do this for a 1800 s integration time in 0.4² seeing for a range of source magnitudes and Strehl ratios. Signal-to-noise ratios per spectral pixel for the moderate resolving power J1, J2, H, and K gratings are shown in Figure 24 to Figure 27, and for the lower resolving power J and HK gratings are shown in Figure 28 and Figure 29. These data are tabulated in an Appendix (§11). Signal-to-noise ratios have been calculated both with and without rejection of OH emission-lines for the J1, J2, and H gratings. These are the gratings for which the gains in rejecting OH emission are the largest (greater than a factor of 2 in signal-to-noise ratio on faint objects with the H grating). The plots were obtained by extracting a simulated spectrum from the central 0.1²´0.1² stellar core, dividing by a normalized extracted spectrum of an identical zero magnitude star, and determining the mean and standard deviation of the pixel values in the resulting featureless spectrum.
Figure 24: SNR per spectral pixel obtained with the J1 grating in 1800 s and 0.4² seeing on point sources with different Strehl ratios. Dashed lines show the result of OH rejection.
Figure 25: SNR per spectral pixel obtained with the J2 grating in 1800 s and 0.4² seeing on point sources with different Strehl ratios. Dashed lines show the result of OH rejection.
Figure 26: SNR per spectral pixel obtained with the H grating in 1800 s and 0.4² seeing on point sources with different Strehl ratios. Dashed lines show the result of OH rejection.
Figure 27: SNR per spectral pixel obtained with the K grating in 1800 s and 0.4² seeing on point sources with different Strehl ratios.
Figure 28: SNR per spectral pixel obtained with the J grating in 1800 s and 0.4² seeing on point sources with different Strehl ratios.
Figure 29: SNR per spectral pixel obtained with the HK grating in 1800 s and 0.4² seeing on point sources with different Strehl ratios.
Based on these predictions, NIFS should achieve signal-to-noise ratios of ~ 10 per spectral pixel in a 0.1²´0.1² aperture with median seeing and the expected Strehl ratios of 0.2 at J, 0.4 at H, and 0.6 at K in a single 1800 s exposure on point sources with J = 18.8, J = 18.4, H = 18.8, and K = 17.8 mag using the J1, J2, H, and K gratings, respectively, and J = 18.8 and K = 18.3 mag using the J and HK gratings, respectively.
2.8.4 Extended Continuum Sources
The detected signal from a uniform extended continuum source is not reduced by incomplete AO-correction as it is for a point source. Rather, the detected signal is contaminated by light from adjacent regions. The signal-to-noise ratio achieved in this situation can be estimated from the results shown in Figure 24 to Figure 29 for a Strehl ratio of 1.0. The signal-to-noise ratios in these figures were measured with a 0.1²´0.1² square aperture. Therefore the stellar magnitudes can be converted to equivalent surface brightnesses in mag arcsec-2 by subtracting 5 mag from the magnitudes in the figures (in fact, only ~ 0.8 of the light from a point source is included in a 0.1²´0.1² aperture for a Strehl ratio of 1.0 [Figure 21]; we ignore this loss).
Based on these surface brightness predictions, NIFS should achieve signal-to-noise ratios of ~ 10 per spectral pixel in a 0.1²´0.1² aperture for single 1800 s exposures on uniform, extended, continuum sources with surface brightnesses of J = 15.4, J = 15.0, H = 14.8, and K = 13.5 mag arcsec-2 using the J1, J2, H, and K gratings, respectively, and 15.3 and 13.7 mag arcsec-2 using the J and HK gratings, respectively.
2.8.5 Extended Emission-Line Sources
The sensitivity to extended line emission depends on both the line width and the background continuum surface brightness. The signal-to-noise ratio in a 0.1²´0.1² aperture achieved on extended line emission can also be estimated from the results shown in Figure 24 to Figure 29 for a Strehl ratio of 1.0 by comparing the emission-line flux to the noise in the background continuum. The emission-line surface brightnesses required to make a 10s per spectral pixel measurement of an extended emission-line with FWHM=100 km s-1 in a 0.1²´0.1² aperture and 1800 s integration time are listed in Table 3. Such a spectrum would be suitable for measuring the profile of a 100 km s-1 wide emission-line at the full velocity resolution available with each grating.
Table 3: 10s per spectral pixel emission-line sensitivities in 1800 s integrations for 100 km s-1 line width for different background continuum surface brightnesses.
|
J1 grating |
J2 grating |
H grating |
|||
|
mJ |
R = 5090 |
mJ |
R = 6100 |
mH |
R = 5340 |
|
(mag arcsec-2) |
(W cm‑2 arcsec‑2) |
(mag arcsec-2) |
(W cm‑2 arcsec‑2) |
(mag arcsec-2) |
(W cm‑2 arcsec‑2) |
|
9.0 |
1.6´10-21 |
9.0 |
1.3´10-21 |
9.0 |
8.3´10-22 |
|
10.0 |
1.1´10-21 |
10.0 |
8.4´10-22 |
10.0 |
5.1´10-22 |
|
11.0 |
7.0´10-22 |
11.0 |
5.5´10-22 |
11.0 |
3.2´10-22 |
|
12.0 |
4.5´10-22 |
12.0 |
3.5´10-22 |
12.0 |
2.1´10-22 |
|
13.0 |
2.9´10-22 |
13.0 |
2.3´10-22 |
13.0 |
1.3´10-22 |
|
14.0 |
2.1´10-22 |
14.0 |
1.6´10-22 |
14.0 |
9.5´10-23 |
|
15.0 |
1.4´10-22 |
15.0 |
1.3´10-22 |
15.0 |
7.4´10-23 |
|
16.0 |
1.2´10-22 |
16.0 |
1.0´10-22 |
16.0 |
6.8´10-23 |
|
|
|
|
|||
|
K grating |
J grating |
HK grating |
|||
|
mK |
R = 5340 |
mJ |
R = 2790 |
mK |
R = 2530 |
|
(mag arcsec-2) |
(W cm‑2 arcsec‑2) |
(mag arcsec-2) |
(W cm‑2 arcsec‑2) |
(mag arcsec-2) |
(W cm‑2 arcsec‑2) |
|
9.0 |
4.9´10-22 |
9.0 |
1.1´10-21 |
9.0 |
3.4´10-22 |
|
10.0 |
3.3´10-22 |
10.0 |
7.0´10-22 |
10.0 |
2.3´10-22 |
|
11.0 |
2.2´10-22 |
11.0 |
4.5´10-22 |
11.0 |
1.6´10-22 |
|
12.0 |
1.8´10-22 |
12.0 |
3.0´10-22 |
12.0 |
1.2´10-22 |
|
13.0 |
1.4´10-22 |
13.0 |
2.0´10-22 |
13.0 |
1.1´10-22 |
|
14.0 |
1.3´10-22 |
14.0 |
1.5´10-22 |
14.0 |
1.0´10-22 |
|
15.0 |
1.2´10-22 |
15.0 |
1.2´10-22 |
15.0 |
9.3´10-23 |
2.9 Guide Star Availability
NIFS will be operated in three modes; with the ALTAIR natural guide star system, with the ALTAIR laser guide star upgrade, and on its own without ALTAIR. Observations that achieve the best image quality using NIFS with the ALTAIR natural guide star system will require an optical AOWFS guide star brighter than R ~ 15 mag as well as an OIWFS guide star. The AOWFS guide star must be located within ~ 20² of the science object. The limiting magnitude for OIWFS guide stars is not known precisely. It is expected to be significantly fainter than the AOWFS limit because the OIWFS needs only to track slow flexure changes when used with the ALTAIR natural guide star system. The OIWFS guide star must be located within the 120² diameter ALTAIR field-of-view. NIFS observations with the ALTAIR laser guide star upgrade will require a brighter OIWFS guide star from which fast tip-tilt and focus corrections can be derived. NIFS observations that do not use ALTAIR at all can access OIWFS guide stars over the full 180² diameter field accepted by NIFS. The OIWFS senses only slow flexure changes in this mode so fainter OIWFS guide stars will suffice.
Sky coverage issues for AO on Gemini have been discussed by Ellerbroek & Tyler (1998). They plot guide star density functions based on the Bahcall & Soneira (1980) model of the Galaxy (Figure 30). The AO corrected field of ~ 20² diameter corresponds to ~ 2.4´10-5 deg2, so crudely we consider guide star densities of ~ 4´104 deg‑2 to be required for complete sky coverage. Actual guide star densities to R ~ 15 mag are closer to 102 deg-2 (Figure 30), so an AO natural guide star sky coverage of ~ 0.3% is expected. Sky coverage fractions for different Strehl ratios in the J, H, and K bands have been calculated in detail by Ellerbroek & Tyler (1998) and give a similar result for optimal performance (Figure 31). The 120² diameter ALTAIR field over which OIWFS guide stars are accessible corresponds to 8.7´10-4 deg2, so guide star densities of ~ 103 deg-2 are required for complete sky coverage. Full sky coverage at 30° Galactic latitude would require an OIWFS guide star limit of R ~ 18 mag. The OIWFS field for NIFS observations without ALTAIR corresponding to 3.5´10-3 deg2. Guide star densities of ~300 deg-2 are required for complete sky coverage, so an OIWFS flux limit of R ~ 18 mag would provide nearly complete sky coverage to 90° Galactic latitude.
Figure 30: Guide star density functions for 30° Galactic latitude and 90° Galactic latitude (Ellerbroek & Tyler 1998).
Figure 31: Sky coverage probabilities at 90° Galactic latitude for natural guide stars (lower curves) and laser guide stars (upper curves) in median seeing, 0° zenith distance, and typical windshake (Ellerbroek & Tyler 1998).
These sky coverage estimates are pessimistic in that some science objects can also be used as their own AO guide object. In order to more directly assess the availability of guide stars for the types of object NIFS will observe, the Gemini Observing Tool has been used to identify AO guide stars and OIWFS guide stars within the ALTAIR field for a range of likely NIFS targets (Table 4). The result suggests that ~ 50% of nine Galactic objects can definitely be measured with ALTAIR and NIFS, with eight of the nine being measurable if the OIWFS is not used (i.e., if flexure between ALTAIR and NIFS is not monitored). The ninth object could be observed with a laser guide star. Only ~ 35% of the extragalactic NIFS targets will definitely be observable using ALTAIR. This may rise to ~ 60% if the OIWFS is not needed to monitor flexure, and to ~ 78% when the laser guide star facility is also available. This is more encouraging than the random star density calculations suggest (but some objects are being measured with no OIWFS star). Nevertheless, NIFS with ALTAIR will be better suited to observing programs for which objects are chosen for their proximity to guide stars rather than for their unique astrophysical characteristics. Peripheral wavefront sensor (PWFS) guide stars are available for all targets in Table 4.
Table 4: Guide Star Availability for Specific NIFS Targets
|
Target |
Galactic Latitude |
NGS AO Guide Star |
ALTAIR OIWFS Star |
Direct OIWFS Star |
NGS AO Suitability |
LGS AO Suitability |
|
Galactic |
||||||
|
Sgr A* |
0.0 |
Yes |
Yes |
Yes |
4 |
4 |
|
T Tau |
-20.9 |
Yes |
No |
Faint |
? |
? |
|
HL Tau |
-20.0 |
Yes |
Yes |
Yes |
4 |
4 |
|
R Mon |
1.3 |
Yes |
Faint |
Yes |
? |
? |
|
Z Cma |
-2.6 |
Yes |
Yes |
Yes |
4 |
4 |
|
Lk Ha 101 |
-9.0 |
Yes |
Yes |
Yes |
4 |
4 |
|
MWC 349 |
0.5 |
Yes |
No |
Yes |
? |
? |
|
S 269 |
-1.7 |
Yes |
Yes |
Yes |
4 |
4 |
|
S 106 |
-0.6 |
No |
Yes |
Yes |
6 |
4 |
|
Nearby Galaxies |
||||||
|
M31 nucleus |
-21.6 |
Yes |
Yes |
Yes |
4 |
4 |
|
M32 nucleus |
-22.0 |
Yes |
No |
Yes |
? |
? |
|
M33 nucleus |
-31.3 |
? |
Yes |
Yes |
? |
4 |
|
IC 342 nucleus |
10.6 |
Yes |
Yes |
Yes |
4 |
4 |
|
Starburst Galaxies |
||||||
|
M82 |
40.6 |
No |
No |
Yes |
6 |
6 |
|
M83 |
32.0 |
No |
No |
Faint |
6 |
6 |
|
NGC 253 |
-88.0 |
No |
No |
No |
6 |
6 |
|
NGC 1808 |
-35.9 |
No |
No |
Yes |
6 |
6 |
|
NGC 7252 |
-56.1 |
Yes |
Faint |
Faint |
4 |
? |
|
NGC 7552 |
-65.2 |
No |
Yes |
Yes |
6 |
4 |
|
I Zw 18 |
44.8 |
No |
Yes |
Yes |
6 |
4 |
|
Seyfert Galaxies |
||||||
|
NGC 1068 |
-51.9 |
Yes |
No |
No |
? |
? |
|
NGC 1097 |
-64.7 |
No |
? |
Yes |
6 |
? |
|
NGC 1275 |
-13.3 |
No |
Yes |
Yes |
6 |
4 |
|
NGC 4151 |
75.1 |
Yes |
No |
Yes |
? |
? |
|
NGC 6764 |
18.2 |
Yes |
Yes |
Yes |
4 |
4 |
|
NGC 7469 |
-45.5 |
Yes |
Yes |
Yes |
4 |
4 |
|
Cen A |
19.4 |
No |
No |
Faint |
6 |
6 |
|
Ultra-Luminous Infrared Galaxies |
||||||
|
Arp 220 |
53.03 |
Yes |
No |
No |
? |
? |
|
NGC 6240 |
27.3 |
Yes |
Yes |
Yes |
4 |
4 |
|
NGC 3690 |
55.0 |
Yes |
Yes |
Yes |
4 |
4 |
|
NGC 1614 |
-34.4 |
Yes |
No |
No |
? |
? |
|
Radio Galaxies |
||||||
|
Cygnus A |
5.8 |
Yes |
Yes |
Yes |
4 |
4 |