As telescope apertures increase, fundamental limits
for doing efficient slit-based spectroscopy are being reached.
We propose that lens arrays feeding fibre bundles be used in the
focal plane of large telescopes to counteract these problems and
in the process add the capability of integral field spectroscopy.
A description of the first phase of the SPIRAL (Segmented Pupil/Image
Reformatting Array Lenses) project is presented. This system will
provide spatially mapped spectroscopy with high spectral resolution
and high throughput. An alternative mode in which the lens array
is used to segment the pupil rather than the sky is also described.
Finally, we briefly discuss our future instrument development
plans.
Keywords: spectrographs, optical fibres, single
object spectroscopy, integral field spectroscopy
The Phase A SPIRAL prototype is being built jointly
by the Institute of Astronomy in Cambridge and the Anglo-Australian
Observatory (AAO). The objectives of the phase-A study are to
develop the construction techniques for a new type of fibre-fed
spectrograph and to test it on the AAT by carrying out competitive
scientific observations. The design philosophy for SPIRAL is that
every part of the system from the focal plane of the telescope
through to the detector is optimised to give the best overall
performance - SPIRAL is not simply an add-on to an existing spectrograph.
We have also kept the phase-A programme simple in order to achieve
our goals as quickly as possible.
The SPIRAL concept offers high efficiency, high spectral
resolution 3D spectroscopy on a 4m telescope such as the AAT and
because it feeds a large 2D area of sky into the spectrograph
it offers accurate spectrophotometry. But perhaps the biggest
advantage of SPIRAL is that it can be easily transferred to 8m
class telescopes. This is in marked contrast to designing a traditional
slit-based spectrograph1 for an 8m telescope which
requires a reduction of the slit width (slit losses) or a larger
beam size (cost) or a reduction in spectral resolution (oversampling).
The option of increasing the beam size to avoid slit losses and
maintain spectral resolution requires gratings larger than the
standard commercially available ones and an exceptionally fast
camera.
The SPIRAL phase-A prototype consists of four basic
optical systems: the fore-optics, the lenslet array, the fibre
feed and the spectrograph. The first 3 of these are being made
in Cambridge while the spectrograph is being made at the AAO.
There are 3 separate options for the fore-optics module each providing
either a sky or pupil image with different image scales. The lenslet
array is coupled directly to the fibre feed. The spectrograph
is a simple Littrow arrangement offering high throughput and a
relatively high spectral resolution compared to a conventional
Cassegrain spectrograph.
In principle for integral field spectroscopy one
can place an array of microlenses directly in the focal plane
of the telescope. However in practice there are several advantages
in having some fore-optics between the focal plane of the telescope
and the lens array. These include being able to build a lens array
at a convenient scale, allowing pupil-segmentation and making
the system from the lens array onwards essentially independent
of the telescope and therefore potentially useful on many telescopes
including 8m class telescopes.
Two lenses, as shown in Figure 1, image the sky onto
the lens array. Each fibre sees a different region of an extended
object, such as a small galaxy in the case of our figure. The
image scale at the f/8 Cassegrain focus of the AAT is 6.6"
mm-1.
Trying to match the fibres to the seeing so that
each lens sees 0.5" without using fore-optics would mean
using lenses with a diameter of 75m. Although lens arrays with
such small lenses are available commercially they are difficult
to align precisely with the fibres and they introduce focal ratio
degradation (FRD) because the pupil image formed at the fibre
is not telecentric across the entire face of the fibre. Both of
these effects arise because the microlenses are only slightly
larger than the core diameters of the fibres themselves.
We have chosen to make our lens array out of hexagonal
lenslets that measure 4mm across and so the fore-optics in this
mode deliver an image scale of 0.125" mm-1 giving
a magnification of 53x and 0.5" per lenslet. The first doublet
lens of the fore-optics provides the magnification and the larger
field lens ensures that the exit pupil of the fore-optics is at
infinity. There is a small pupil image just behind the first lens
and this is conjugate with all of the fibres receiving light from
the lenslet array. There is a critical magnification scale for
which the fibre core diameter is matched to the pupil image size
and this has a corresponding size for the projection of the sky
on to the lenslet. If the pupil image is smaller than the fibre
core then no light is lost so fore-optics that provide spatial
sampling at the critical scale or finer are allowed. In practice
our fore-optics are at this critical limit with a small allowance
for misalignments and image quality.
2.2 PUPIL IMAGING (Point Source Spectroscopy)
Here, only one lens is used to project the image
of the pupil (i.e. the primary mirror) onto the lens array. (See
figure 2). Although in this case spatial mapping is lost (each
fibre looks at the same area of sky) by adjusting the power of
the lens the effective area of the sky seen by each fibre and
hence the spectrograph can be changed. The appendix at the end
of this paper explains in more detail how this mode works. For
SPIRAL phase-A we will have two fore-optics options one in which
the pupil image fills the lenslet array and the other in which
the pupil image is approximately half the size of the lens array.
In the first case each fibre sees 3.6" and in the second
case the fibres see 2.2". This means that we can match the
spectrograph to the current seeing conditions without having to
sacrifice resolution or loss of throughput.
The array is composed of 37 hexagonal lenslets which
are close packed to form a large hexagonal pattern as shown in
Figure 3. Each lenslet is made of PSK53A glass and is 4mm across
the corners. The front face of the lenslet array is AR coated.
The lenslets can be easily manufactured by traditional means and
allow the fibres to be individually aligned with very high precision
during manufacture. The lenslets are glued onto two cylindrical
pieces of glass that form an achromatic doublet with the focal
plane on the back face of the cemented glass block. By forming
a solid glass block with the lenslets attached we provide an extremely
rigid base for the fibres to be positioned and permanently fixed..
We chose the input f-ratio of the fibres, Ffibre, to
be /5 as this minimises any FRD and also leads to a simple spectrograph
design. The smaller the diameter of the fibres the higher the
spectral resolution that can be obtained as this defines the width
of the input slit of the spectrograph. By choosing a fibre diameter
of 50m we also match the fibres with the detector pixel size (Tek
10242 with 24m pixels) which enables us to use a Littrow
spectrograph.
A 2mm diameter steel ferrule holds seven smaller
steel tubes in a close packed formation (Figure 4). Three tubes
on a line in turn contain finer ferrules with a single 50m fibre
glued inside each one. The central fibre is aligned with the optical
axis of the hexagonal lenslet and the ferrules are glued in place
on the glass flat with UV setting glue. The two extra fibres act
as sky subtraction fibres in the pupil imaging mode, avoiding
the need for a sky background exposure. The lens array and fibre
ferrules are protected by a machined housing which sits at the
Cassegrain focus of the telescope behind the fore-optics.
As the fibres are over 18m long (and very fragile!)
we need to protect them from the environment - to this end they
are placed in a flexible steel-wound conduit. This has a 'strain
relief box' which contains a single loop of the optical fibres,
allowing the length of the conduit to change without stressing
the fibres themselves.
The fibres are rearranged into a vertical slit which
then feeds the Littrow spectrograph. They are held in place between
two brass blocks, one of which has grooves cut into its surface.
After the two blocks are glued together and the fibres threaded
into place the block face is then polished, along with the fibres,
and an AR coated optical flat is cemented across the fibre ends.
The central (object) fibres are arranged in the centre of the
slit and the sky fibres are above and below them. There is also
a particular relationship between the input and output end positions
for the object fibres which ensures that fibres that are adjacent
on the slit are also adjacent on the sky (see figure 5). The object
fibres are well separated on the detector for the integral field
mode whereas the sky fibres are close-packed to allow on-chip
binning.
As mentioned above we chose a fibre size of 50m so that we could use a Littrow spectrograph with fcam = fcoll and have a monochromatic slit image which is well matched to 2 detector pixels. The lenslets in the array feed the fibres at f/5 so we have designed a spectrograph which can accept light at f/4.8 to allow for some FRD (which is comparatively low at these f-ratios). The design we have adopted is a large Petzval lens with a focal length of 720mm which acts as both the collimator and the camera and provides a beam size of 150mm diameter (see figure 6). The spectrograph is very efficient because it is fully transmitting with no central obstruction and the grating is used in the Littrow configuration. The Petzval lens has 6 air-glass surfaces which are AR coated. This design was chosen because it could be built quickly and inexpensively yet it still provides excellent image quality and a large beam size so that high spectral resolution can be obtained. The high spectral resolution capability of SPIRAL is a feature we are keen to emphasise. The system does however suffer from chromatic aberration so low dispersion spectroscopy with a large wavelength range cannot be focused properly across the
CCD. High dispersion spectroscopy can be done at
any wavelength from 370-900nm by adjusting the CCD tilt and focus.
The dispersion with a 1200g/mm grating in first order is 10.17
mm-1 giving R=17,400 and a wavelength coverage of
250 at 8500. Even higher spectral resolution can be obtained at
higher spectral orders.
At the time of writing the lenslet array and the
fibre bundle have been manufactured. The next stage is to mount
the fibres on the back of the lenslet array and complete the fibre
slit. The spectrograph optics are currently being manufactured
at INAOE, Puebla, Mexico. First light is scheduled for November/December
1996 on the AAT.
We are currently planning the phase-B system which
we expect to be given approval once we have successfully demonstrated
the phase-A system. Phase-B will have an array with approximately
400 hexagonal lenslets arranged in a pattern with an aspect ratio
of 4:1. The spectrograph will use a 2048x4096 pixel thinned CCD
and not suffer from chromatic aberrations, offering a full octave
of wavelength coverage at R=2000 or reduced wavelength coverage
at spectral resolutions up to 25,000 (with an echelle grating).
The total wavelength range will be 350nm-1000nm. We are also considering
the possibility of zoom systems for the fore-optics. An important
aspect of phase-B will be the development of data inspection and
data reduction software. Phase-B will not be a fully-fledged facility
instrument in order to keep down costs and reduce time scales.
Phase-C of this project will be to upgrade the phase-B hardware
to a facility instrument.
We wish to thank Damien Jones for the design of the
spectrograph, Jim Pritchard for the engineering drawings and François
Piché for his help with laying out 120 very long optical
fibres.
1. R.G. Bingham, "Grating Spectrometers and
Spectrographs Re-examined", Q. Jl. R. astr. Soc. 30,
395-421 (1979)
The diagrams on the right illustrate the principle
of the pupil mode. By varying the power of the fore-optic lens
the size of the image of the pupil on the lens array can be adjusted.
This means that each fibre effectively sees a smaller size telescope,
Dsub
In (a), the pupil illuminates 36 lenses. As the pupil
is an image of a 3.9m primary mirror, each lenslet is now receiving
light from an area of Dsub =0.60m.Putting this value
into the equation above gives a sky area of 3.6".
In (b), there are now 18 fibres receiving light from
the pupil image and these fibres are accepting light from a segment
of the primary mirror diameter 1.0m, giving a fibre sky area of
2.2"
Finally,(c) has 7 illuminated fibres with Dsub
=2.0m. and sky =1.1".
So, by varying the power of the fore-optics the amount
of sky the fibres subtend can be matched to the current seeing
conditions, thus maximising the signal to background that the
signal fibres receive. By replacing the lens with a zoom lens
the matching could be smoothly adjusted for any seeing.
As a pupil image is cast onto the lens array, an
image of the sky is cast onto the fibres - so the two sky fibres
can now be used to provide sky subtraction without the need for
beam switching.
The advantages for single object spectroscopy are
great - no longer do you lose light on the edges of the slit jaw
and there is no longer any slit that needs to be widened to match
the seeing - hence no loss of resolution!