THE IFIRS FAQ 

SUMMARY

The goal of IFIRS is to obtain ultra-deep, wide field, diffraction limited, imagery from near- to mid-IR wavelengths, with flexible spectral resolution. IFIRS is a Michelson interferometer configured as an imaging Fourier transform spectrometer (IFTS). IFIRS is a high throughput, wide field, diffraction-limted imager. Since IFIRS is an FTS, it is both a camera and a multi-object or integral field spectrometer. As a camera IFIRS has unusual flexibility of spectral resolution, and a unique pan-chromatic imaging mode.

SPECIAL FEATURES

Currently three spectrometer architectures viable for NGST: tunable filter (e.g., Fabry Perot), dispersive (e.g, a MOS or IFU), and Fourier transform.  There is no single correct answer, because each concept has strengths and weaknesses, and is best in a particular domain. The correct choice depends on the type of the observations demanded by the science, detector performance, and the nature of the backgrounds.

The signal-to-noise performance of 3-d imaging spectrometers equipped with 2-d detector arrays is the same for all architectures, in the ideal case of photon shot noise limted operation. This is correct so long as the spectrometers are equipped with the same size arrays, and the same spatial and spectral degrees of freedom of the astronomical scene are observed.

Specific advantages of the IFTS are:

Deep imaging is acquired simultaneously with higher spectral resolution data over a broad wavelength range.  IFIRS has a pan-chromatic mode that is peculiar to a four port interferometer. The summed signal from the two output ports is unmodulated and corresponds to the total broad-band photon flux entering the instrument.  With a simple camera a pan-chromatic image can be formed with by summing individual filter images . The IFTS pan-chromatic image has a speed advantage factor that is equal to the number of filters used.

IFIRS also has a hybrid, or dispersed FTS mode. In a regular FTS, spectral information is encoded in the z-direction of the data cube, and there is no mixing of spectral and spatial information. The advantage is that a spectrum is recorded for every pixel on the sky. The penalty is that the photon shot-noise from all spectral channels is present at each frequency. This shot-noise can be reduced by masking the telescope focal plane around objects of interest with a programmable micro-mirror array, and inserting a prism into the collimated space.  The dispersed FTS mode is used to obtain the highest possible sensitivity at high spectral resolution (R=600-10,000).  The slit width does not determine the spectral resolution in the dispersed FTS mode, since the spectral resolution is derived from the interferograms. The dispersed FTS data-cube contains spectra which are tilted with respect to the z-axis. The tilt angle is the arc tangent of the ratio of spectral resolutions of the dispersive element and interferometer.  The dispersed FTS has better SNR performance than the pure FTS, and the source density of object slits is higher than for the pure multi-object spectrometer.

DESIGN

On the object side, a collimator illuminates the interferometer with parallel light. The interfering beams are collected by a camera, creating a one-to-one mapping between points in the object and image planes.  By placing a detector focal plane array at the focus of the camera, each pixel is matched to a single point on the sky.  At any given optical path difference (OPD) the image of the sky is modulated spatially by the interferometer fringe pattern, which encodes the spectral information.  By recording images of the sky at different OPDs, the spectrum of each pixel can be reconstructed. The OPD is scanned in discrete steps since FPAs are integrating detectors. Scanning the OPD in this fashion generates a data cube.  The set of signals from the pixels in each frame corresponding to a particular point in the sky forms an independent interferogram.  These interferograms are Fourier transformed individually yielding a spectral data cube composed of the same spatial elements as the image. The sampling theorem establishes the number and amplitude of OPD steps necessary to recover the spectrum at a given resolution.

The major components of IFIRS are the collimator, interferometer, cameras, and focal plane arrays and associated electronics.  Two important ancillary subsystems are the metrology system and the calibration unit.

(Note: For simplicity only the near-IR FPA arrays have been shown at the output ports. The final fold mirror is either a dichroic or flip-mirror which directs the mid-IR light to the mid-IR detectors).

The collimator is a three mirror anastigmat (TMA) which illuminates a four-port Michelson. There are two input and two output ports. One input port is fed with the sky signal, the other input can be illuminated by the calibration unit.  The interferometer consists of a 50:50 reflecting/transmitting beam splitter, and two cube-corner retroreflectors.  The appropriate beam splitter is selected using a filter-wheel mechanism. One cube-corner is located on a translation stage, which permits precision control of the OPD. The OPD is monitored using a laser diode interferometric metrology system. Each output port of the interferometer feeds identical TMA cameras. A final reflection (an articulated fold mirror or a static dichroic) directs short wavelength radiation to the near-IR FPA and long wavelength radiation to the mid-IR FPA.

What spectroscopic resolution(s) are available?

The spectral resolution of an IFTS is determined by the maximum optical path difference at which the interferogram is measured, and hence is continuously variable. IFIRS can obtain pan-chromatic imagery (R ~ 1), broad-band photometry (R ~ 10), low resolution spectrophotometry (R ~ 100), diagnostic spectroscopy (R ~ 1000) and kinematics (R ~ 10,000).

The optical path difference can be scanned from 0-0.5 cm, hence the maximum spectral resolution is 30 GHz (1 wavenumber). The spectral resolution is fixed in frequency, so the maximum resolution is 10,000 at 1 µm and 1000 at 10 µm.

What wavelength coverage/simultaneous wavelength coverage is available?

All the optical components, apart from the detectors and beam splitter, are based on reflective optics, and potentially support a very broad spectral range, from near UV to far-IR.

IFIRS has two detector channels: near-IR (InSb 0.6-5.6 µm) and mid-IR (HgCdTe 3-15 µm).  Operation is optimized for one channel by selection of the appropriate beam splitter. Beam splitters typically span a free spectral range of x6 in wavelength, and the entire band pass is observed simultaneously.  There are three high-efficiency beam splitters which span 0.6-3.4 µm, 0.96-5.4 µm, and 2-21 µm.

The long wavelength cut-off of the HgCdTe array is to be determined; 11 µm is currently feasible, operation as long as 15 µm should be anticipated, and17 µm may be possible.  These detectors can operate at the ISIM operating temperature of 30 K without additional cooling.  IFIRS can tolerate the relatively high dark current of HgCdTe since it integrates the entire mid-IR band on the detector and operates in a background noise limited regime.  Should space qualified cryocoolers be made available, then Si:As BIB arrays will be used instead to extend the wavelength coverage to 28 µm.

In the current design, we have optimized the operation for a single channel, and simultaneous near- and mid-IR observations are not supported. Replacement of the final fold-mirror with a dichroic would introduce some losses (~ 10%), and eliminate two mechanisms. Using the long wavelength beam splitter and both detectors would yield simultaneous 2-15 micron observations with good efficiency.

What is the instrument's spatial resolution?

The spatial resolution of IFIRS is determined by the combined performance of the telescope and instrument optics. This yields: The IFIRS wave front error budget includes: The pixel sampling is: The choice of pixel scale is not final. Since diffraction limited images can be constructed from a set of dithered images we will probably choose to be undersampled. Lauer (PASP, 111, 227) has shown that the reconstructed image is an exact representation of the data set with no loss of resolution at the Nyquist scale. Undersampling is advantageous for projects that require large instantaneous field of view, e.g., supernova cosmology.

What is the instrument's field of view?

The current optical design of IFIRS provides a corrected field of 5.'28 x 5.'28 (317" x 317"). This is designed to illuminate a 8192 x 8192 near-IR focal plane array. The mid-IR field of view is limited by the focal plane array (2048 x 2048) and is 2.'64 x 2.'64.

What is the minimum object spacing for multiobject observations?

The IFTS is a true integral field spectrometer. There are no slits. There is a spectrum for every pixel in the field of view. The near-IR focal plane is Nyquist sampled at 3 µm. With appropriate drizzling, spatial information can be recovered down to the diffraction limit.  Thus we expect minimum object spacing to be of the order of the size of the Airy disk (FWHM = 0."05) at 2 µm. 3-d PSF fitting ("3-d DAOPHOT") should be able to successfully extract spectra on these scales.

In simulated data of crowded fields we have successfully used optimal extraction using a 3-d kernel which has width equal to the PSF.

In many cases source crowding may be the limiting factor, e.g., in studies of stellar populations in nearby galaxies. With IFIRS, unambiguous spectral extraction can be performed down to the confusion limit.

The dispersed FTS mode is used to obtain the highest possible sensitivity at high spectral resolution (R=600-10,000),

Assuming that a micro-mirror array is available, a great deal of flexibility in terms of object selection may be had. The viewed portion of the image may be arranged so that the "objective prism streaks" from each target do not overlap. In the baseline concept, a prism disperses the 0.6-5 micron spectral range over approximately 500 pixels. Thus any number of objects that were separated in the field of view in a direction orthogonal to the dispersion direction, can be simultaneously observed. In the dispersion direction, a separation of 500 pixels, (19") is required for the near-IR channel to avoid overlap.

What is the number of objects that can be observed simultaneously?

In the near-IR channel the FPA is 8192 x 8192. If we conservatively assume that an 8x8 subarray is needed to record one object, then we can record spectra for a million objects. The mid-IR channel has 16 times fewer pixels, and therefore correspondingly fewer spectra. With an image mask the number of obects which can be observed simultaneously is reduced to ~ 17,000.

What is the speed of setup?

The instrument specific setup to begin an IFIRS observation consists of:
  1. Pure FTS mode
  2. Dispersed FTS mode
The major mechanisms (beam splitter & prism) take < 100 s to select. The time to acquire and stabilize a given optical path difference is < 2 s.

Precise absolute pointing is not required for the pure FTS mode --- only accurate tracking during a single exposure (t = 100-1000 s) is needed.

What is the expected instrument throughput as a function of wavelength?

The throughput of the FTS is limited by the reflectivity of metal optics, beam splitter efficiency, and detector QE. There are no slit losses or light scattered by gratings in pure FTS mode.

There are 12 gold reflections (~ 99% per surface) - input fold mirror, TMA collimator, fold mirror, three reflections at a cube corner, final fold mirror, and a TMA camera.  We have detailed designs for optimized 0.96-5.4 & 2-21 micron beam splitters.

The resultant average in-band throughputs are 86% and 87% respectively for the near- and mid-IR. The in-band response is flat.  If we include detector QE, the total system efficiency is 75% (InSb), and 63% (HgCdTe).

We have not designed a 0.6-3.4 micron beam splitter, but the average throughput will be ~ 80%. Gold is not optimal for 0.6-0.8 µm, and the efficiency in this range is reduced by 12%.  An alternative coating, e.g., protected silver, would recover the 0.6-0.8 micron performance at the expense of a serveral percent loss at longer wavelengths.

What are the calibration requirements?

A four port Michelson interferometer is intrinsically a superb instrument from a calibration point of view.  The measured interferogram results from the difference between spectra of sources at the two input ports.  Calibration sources placed at the second input port act as transfer standards for full radiometric calibrations performed on the ground prior to flight. The mid-IR channel is calibrated by varying the temperature of a cold blackbody at the second input that fills the field of view.  The near-IR channel is calibrated using a dilute, hot blackbody, that does not produce excessive heating of the instrument. In both cases, a full calibration of each FPA pixel's offset, spectral responsivity, and non-linearity can be conducted by varying integration times or source intensities and collecting data in the normal manner.  This capability is invaluable should the system response change due to exposure to the space environment.

The telescope and relay optics are not calibrated using the internal source, but inferred by observation of standard stars.  On orbit, observations of standard stars will be used to cross-calibrate the on-board sources.  Periodic observations of the same reference stars will then be used to monitor performance.

The absolute wavelength calibration is provided by the interferometer metrology system, which is reference to a diode laser.  The accuracy should be better than 7.5 GHz (0.25 wavenumbers).  This provides the wavelength calibration for both pure and dispersed FTS modes.

What telescope performance (e.g. pointing and tracking, image quality) is  required?

The telescope and instrument should contribute equally to the overall system wave front error budget. The pointing stability should therefore be approximately 0."003 pixel (3 sigma).  This stability need only be maintained no longer than a typical integration time (1000 s for the near-IR, 100 s for mid-IR).

Carefully controlled dithering of the line of sight, accurate to 0".003 with an amplitude of about 1", should be required to extract diffraction limited spatial information from undersampled images, to veto cosmic ray contaminated pixels, and to replace data from dead pixels.  Significant image motion between subsequent frames of the interferogram does not degrade spectral or spatial resolution. In particular, the FTS can obtain spectra of moving objects such as solar system targets.

What is the general tradeoff between the instrument's spectral resolution and coverage vs. spatial resolution/distribution?

An important trade is between band-pass and sensitivity. The FTS works broad-band, so the in-band photon shot-noise is associated with every spectral channel. A blocking filter reduces the spectral coverage and increases the SNR. For a flat photon spectrum, and shot-noise limited operation, reducing the number of spectral channels increases the SNR in each spectral channel by the square root of the bandwidth reduction factor. The SNR in the pan-chromatic image is reduced by the square root of this factor, since less light is being collected.  Simulations suggest that for many DRM programs the scientific value of more spectral channels outweighs the associated decrease in SNR (e.g., measuring photometric redshifts) , and that a likely workhorse operating point would be for R ~ 100.

When higher spectral resolution (R=600-10,000) observations of high latitude "blank fields" are required (e.g., for the galaxy evolution DRM), there are too few bright objects to warrant the full spatial multiplex advantage of the IFTS.  In this case the dispersed-FTS option gives the ultimate sensitivity at the expense of reduced number of targets (see 5 & 6).

What observing modes are expected to exist? What are your priorities for instrument modes?

In this mode the instrument functions as a camera with tuneable filters.  A menu of OPD scans will be provided yielding different resolutions (R = 1, 2, 4, 8, ...) or number of spectral channels, and integration time apodization (top hat, Gaussian, etc.). The combination of these options can be used to synthesize band-passes with different widths and profiles and ranging from broad-band filters (R ~ 5), e.g., RIJHKLM etc., to narrow band imaging (R ~ 300).

In this mode the major instrument configuration option is the choice of beam splitter. Observations in mid-IR will be obtained by exchanging the beam splitter, and acquiring data with the HgCdTe focal plane array.  As stated above, in certain cases simultaneous near- and mid-IR observations are possible.

This mode is a hybrid mode for higher spectral resolution (R=600-10,000) observations of sparsely distributed objects.  A mask is placed in an upstream image plane, centered around the object or objects of interest, (e.g., a programmable micro-mirror array) and a prism is inserted into the collimated space. This is an objective prism mode for resolutions from R=5-500 for densely distributed, but not space filling, objects, such as Galactic star clusters.  There is no slit in the upstream image plane, but the dispersing prism is in place in collimated space. If the dispersed FTS mode is implemented, then this mode comes for free.

All of the above modes are compatible with a coronographic "hole" in the upstream image plane.

What do you expect are the major issues on observing strategy (e.g. acquisition images, object selection)?

Because a spectrum is obtained for every pixel, it is not necessary to define the region of any given object to be fed to a spectrometer before the data is acquired, and indeed, after the fact, a spatially extended set of spectra can be averaged in a weighted sum tailored to maximize the sensitivity for detection of any particular feature of interest.  The major issue is simply the decision as to the mode to use from the three described in the answer to question (12), the desired spectral channel and bandpass, the spectral resolution, and the observing time.

How will the instrument performance depend on the on-orbit performance of the  detectors (e.g. cosmic ray susceptibility)?

By virtue of being able to acquire spectra with the full bandpass of light impinging on the detector elements, an FTS is tolerant of detector performance degradation, both in terms of dark current and read noise. Furthermore, by taking a series of readouts to effectively monitor the apparent photo-current, large cosmic ray glitches can be removed on orbit. Also, because spectral information is derived from modulated light levels, the 1/f type of focal plane array glitch noise which is apparently very significant for instruments such as those on ISO, can be very significantly reduced. Using the observed ISO cosmic ray glitches as an example, the suppression of
glitch noise by IFIRS could reasonably be expected to exceed an order of magnitude.

Are there any major technical challenges to be faced?

Back to the Berkeley NGST page.