Near Infrared Integral Field Spectroscopy at UKIRT

Roger Haynes¹, Jeremy Allington-Smith¹, Robert Content¹, David Lee^1,2

1 : Durham University Astronomical Instrumentation Group, University of Durham, South Road, Durham DH1 3LE, UK

2 : Now at the Anglo-Australian Observatory, PO Box 296, Epping, NSW 2121, Australia

The IFU reformats a field of 6" x 4" into a pseudo-slit which is reimaged onto the cold slit inside CGS4. CGS4 then disperses the light as normal to produce a longslit spectrum. The field is divided into 72 hexagonal elements measuring 0.6 arcsec between opposite faces (Fig. 1). There are no gaps between the elements so the filling factor is essentially 100%.
 

Figure 1 : Input field format showing how the elements are reformatted at the slit.

Because the system is warm (it cannot be put inside the CGS4 cryostat) it is designed for use in the J and H bands, although some useful performance is expected in the K-band.

The IFU was built by the Astronomical Instrumentation Group (AIG) of Durham University, making use of parts of the SMIRFS multiobject system (Haynes et al.1995).  Although this is a fully-functional scientific instrument, it was intended to serve as a prototype for the more ambitious Thousand-Element Integral Field Unit (TEIFU) to work with the ELECTRA and NAOMI Adaptive Optics system on the Herschel telescope and for the GEMINI Multiobject Spectrographs (GMOS).

What is integral field spectroscopy?

How does it work?

The input lenslets perform two functions. Firstly they maximise the filling factor by funnelling all the light within the field of each element onto the fibre core, thereby avoiding the deadspace between cores.  Secondly, they convert the telescope f/36 beam to a faster focal ratio suitable for efficient transmission though the fibres.  The output lenslets convert the beam back to f/36 as required by CGS4.

The requirement to work with an existing beam-fed spectrograph without having the option to modify it in any way, imposed constraints on the design of the IFU. In practice this limits the theoretical throughput to ~ 50%.  Another restriction is that, with the short camera, the long slit projects onto only 72 pixels. This meant that we had to limit
the number of elements to 72 with a consequent compromise between field of view and sampling.

Proper sampling by the IFU is achieved when the image produced by the telescope is critically sampled by the IFU at the input, i.e. when the image FWHM is 1.2 arcsec or more.  To first order, the sampling by the detector of the output of each element is irrelevant. Furthermore, any overlap between the light emerging from adjacent IFU elements at the output has negligible effect on spatial resolution provided that elements which are adjacent at the slit originate from adjacent regions of sky. This is indeed the case with the SMIRFS-IFU (Fig. 1).  Despite this, the combination of overlaps and undersampling does cause some problems. This is because flexure between CGS4 and the IFU makes it desirable to obtain separate flatfields for the fibres and the detector and this is difficult to obtain if the IFU output is undersampled.  However, this is only the case when using the short camera: the IFU will be used exclusively with the long camera from now on.

Commissioning results

• The transmission of the IFU alone is 47\% in the J-band when averaged over all elements. This implies that the better elements achieve the theoretical expectation of 53%. In the H-band the value was somewhat lower (33%), perhaps due to a problem in the alignment of the IFU pseudo-slit with the CGS4 slit. It is also possible that cloud affected the IFU exposures, particularly the single exposure taken in the H-band. For this reason, since the CGS4-only observation were taken in good conditions, we should regard the transmission figures as lower limits.

• When uniformly illuminated, the flatfield shows variations with an RMS of 23%. However, much of this is due to four damaged fibres which have anomalously low response. Excluding these reduces the RMS to 14%. With the coarse sampling of the short camera, we cannot uniquely determine if the remaining variation is due to a real variation in element transmission or (in part) a periodic variation in the element spacing at the output.  This issue should be resolved with the better sampling provided by the long camera.  Whatever the source of these variations, they can be calibrated out to high precision (limited by photon noise) using dome or sky flats.

• The point-spread function defined by the output from each element is consistent with the theoretical expectation of 1 pixel FWHM, although the measured PSF is dominated by CGS4. The effective slit width is 1.1 pixels. With the long camera, these widths will be approximately doubled.

• With beam-switching (via a telescope nod), good sky subtraction can be achieved, as shown in Fig. 2.

Figure 2 : Raw data for NGC7469. [FeII] and Pa-beta emission is visible.

Figure 3 : Left - Reconstructed J-band image of the Seyfert-2 galaxy NGC7469. The nucleus is at the right edge (Colour coding: blue= low intensity, yellow=high intensity). Right - Spectra from selected elements showing the [FeII] and Pa-beta lines.

Figure 4 : J-band intensity (left) and [FeII]/Pa-beta flux ratio (right) in NGC7469. Dark colours indicate high intensity and high line ratio. The nucleus is at the right-centre of the field. The drop in relative [FeII] flux coincides with the starburst ring. The data are masked to only show regions of high signal/noise.

Figure 5 : Location of the IFU field compared with an HST WFPC2/PC image of NGC7469 (M. Malkan, UCLA). The image is on a logarithmic scale.

Conclusion

Please contact Jeremy Allington-Smith or Roger Haynes at the Durham AIG (j.r.allington-smith, roger.haynes @durham.ac.uk) if you wish to use the IFU or for more detailed information from the commissioning run
(see also star-www.dur.ac.uk/~jra/ukirt\_ifu.html).

Acknowledgements

In Durham, we thank George Dodsworth, John Webster, Ray Sharples and David Robertson for engineering and programmatic assistance. We also thank Chris Done and Reynier Peletier in Durham for help with data reduction and analysis.

References

Haynes, R. Sharples, R.,& Ennico, 1995, SPECTRUM, 7, 4

Genzel et al 1995, ApJ, 444, 129.
 
 

Back to the Contents