Spectrographs: Integral Field Units
E
N C Y C L O P E D I A O F
A
S T R O N O M Y A N D
A
S T R O P H Y S I C S
spatial dimension and are generally used for high spectral
resolution spectroscopy in the infrared. An exception is
the imaging FTS called Bear operated at CFHT (Maillard
1995).
Basic principle of an integral field unit
In contrast to scanning spectrophotometers, IFUs manage
to cram the full three-dimensional data in a single exposure
on the detector. An IFU is made of two successive stages:
the spatial stage whose function is to reformat the field of
view and the spectral stage whose function is to disperse
and focus the light on the detector. The latter is nothing
else than a classical spectrograph that deserves no special
comment (but see
SPECTROSCOPE/SPECTROGRAPH
). The spatial
stage is the most critical part. There are currently three
types of IFU (figure 2): lenslet unit, fibers unit and slicer
unit. They differ in the geometrical arrangement of the
spatial elements.
An important parameter of an IFU is the packing
efficiency:
this is the number of spatial elements
multiplied by the length of a spectrum in pixels divided
by the total number of pixels of the detector. An ideal
IFU will have a packing efficiency of 100%. For example
with a detector of 2048
2
pixels, one is limited to 45
× 45
spatial elements if each spectrum covers a full column or
line of the detector. This explains why IFUs usually have
a relatively small spatial field of view, even with the large
detector format available nowadays. In the real world,
however, packing efficiency is limited by the geometrical
packing and the optical quality of the spatial and spectral
stages.
Lenslet IFU
Lenslet IFUs use an array of lenses to sample the field
of view. The light intercepted by each lenslet is focused
in a spot called the micropupil which is an image of
the telescope pupil. Micropupils are then dispersed by
the spectrograph in a conventional manner. The ratio of
the lenslet diameter to the micropupil diameter must be
large (typically 50). It is this demagnification that saves
space on the detector to store the spectral dimension.
A slight rotation between the dispersion direction and
the microlens array orientation avoids spectral overlap
in one dimension. In the other dimension, a wide-band
interference filter limits the spectral range to a finite length
to prevent overlap. The lenslet size is typically of the order
of mm which does not generally match the sampling scale
at the telescope focal plane (for example, the lenslet size
should be as small as 14
µm to sample 0.1 arcsec at the focal
plane of a 3.6 m
f/8 telescope). An enlarger, preceding
the lens array, is thus added to adapt the spatial sampling
to the expected spatial resolution. Throughput of this
type of IFU is generally good: square or hexagonal lenslet
shapes provide a 100% covering efficiency and the lenslet
can be made in optical glass and coated. A drawback
is that a significant fraction of spectra are truncated if
spectra are too long: in practice, the maximum spectral
length should be less than 25% of the detector format in
the dispersion direction to minimize this effect. The total
packing efficiency on the detector is limited by the need to
separate each spectrum from its neighbors as neighboring
pixels do not share the same wavelength. When using a
low spectral resolution grating at the order
n, the n + 1 and
n - 1 orders could cause some overlap with other spectra.
In that case using a prism should be considered, especially
at blue wavelengths where this effect is more pronounced.
The first lenslet IFU was called TIGER (Bacon et
al 1995) and has been operated at the CanadaFrance
Hawaii 3.6 m telescope (CFHT) between 1987 and 1996.
Starting in 1997 it has been replaced by OASIS (figure 3):
a lenslet IFU with 1600 hexagonal lenses dedicated to the
CFHT adaptive optics bonnette. Since TIGER, a few other
instruments have been realized with the same design: one
for the SAO 6 m telescope (Afanasiev and Sil'chenko 1991),
another for the Kyoto three-dimensional spectrograph for
the 1.88 m Okayama telescope (Ohtani 1995) and recently
the SAURON spectrograph for the 4.2 m William Herschel
Telescope.
Fiber IFU
Fiber IFUs use optical fibers arranged in a close-packed
bundle at the telescope focal plane and then reformatted
into a pseudo-slit, which is then fed into the spectrograph.
This presents an advantage compared with lenslet IFUs in
the sense that the spectra can be as long as the detector
format allows. On the other hand, fiber IFUs suffer from a
lower efficiency owing to limited packing efficiency at the
entrance (generally less than 75% owing to geometrical
loss and cladding) and
FOCAL RATIO
degradation.
The
latter is due to diffusion by imperfections within the fiber
and diameter variation along it. This effect introduces
some light loss and becomes more important with large
f numbers. This strongly constrains the designs of the
spectrograph and the spatial stage. Fiber IFUs need to
be calibrated carefully to control the fiber-to-fiber point
spread function and transmission variations.
The first fiber IFU that has been extensively used at
a telescope was SILFID at CFHT (1986) (Vanderriest and
Lemonnier 1988), later replaced by MOS-ARGUS (1993).
Other development took place at the Kitt Peak 3.5 m
telescope with DensePak (1988) (Barden and Wade 1988)
and at the William Herschel 4.2 m telescope with Hexaflex
(1989) (Arribas et al 1991), 2D-FIS (1994) (Garcia et al
1994) and INTEGRAL (Arribas et al 1997). All of these
instruments worked in the visible.
Adding lenses in front of the fibers helps to increase
the packing efficiency and to solve the problem of focal
ratio degradation. Each lenslet forms an image of the
telescope pupil at the entrance of each fiber in the bundle.
At the output, a linear array of lenslets forms a pseudo-
slit fed to the spectrograph. Working examples are the
SMIRFS at the 4 m UK infrared telescope (Haynes et al
1998) and SPIRAL at the 4 m Anglo-Australian telescope
(Kenworthy et al 1998). There are a number of fiber IFUs in
development for the 810 m class telescope (e.g. VIRMOS
IFU for VLT, GMOS IFU for Gemini); most of them include
lenslets.
Copyright © Nature Publishing Group 2001
Brunel Road, Houndmills, Basingstoke, Hampshire, RG21 6XS, UK Registered No. 785998
and Institute of Physics Publishing 2001
Dirac House, Temple Back, Bristol, BS1 6BE, UK
2