Limitations on Detecting Faint Broad Lines: Low-level Instrumental Ripples

Since the CO molecule is a good tracer of star formation in galaxies, studying this molecule at cosmological distances is important for understanding the evolution of young galaxies. In this note we will briefly report on our recent experience with the JCMT attempting to observe CO in distant radio galaxies. We will show that, at least in some modes of operations, faint baseline ripples are the limiting factor in these observations and can behave in a way indistinguishable from a real astronomical signal even when various LO settings are observed.

Scientific Rationale

To date, the two objects at high redshift that have a confirmed detection of a CO line are the ultraluminous IRAS galaxy F10214+4724 at z = 2.3 (Solomon et al. 1992) and the cloverleaf quasar (Barvainis et al. 1994). The good news is that these two detections demonstrate convincingly that studying CO in the early universe is possible. The bad news is that the IRAS galaxy F10214+4724 is the only IRAS galaxy known at these distances. Studying a sample of such objects is therefore not (yet) possible. Since the cloverleaf quasar is a gravitationally lensed quasar for which the amplification factor is not well known, the estimate for the intrinsic CO luminosity of this object is uncertain.

The significant samples of quasars and radio galaxies that exist are a good starting point for a further search for CO in the early universe. The expected width of the CO emission lines is a few hundred km/s (FWHM) and the bandwidth of the receivers and backend combination as used in major millimetre observatories is limited to about 1000 km/s. It is therefore important that objects to be studied have well determined redshifts, preferably to better than a few hundred km/s. Since the FWHM of the emission lines of radio galaxies is of order 1000 km/s, a factor of 5 - 10 smaller than in quasars, there is some advantage in using radio galaxies rather than quasars for a survey of CO in the distant universe.

By focusing on the steepest radio continuum spectra we have found more than 30 of the 60 known radio galaxies at z > 2 (e.g. Miley et al. 1992, Röttgering et al. 1994). We have started a programme to detect CO emission from distant radio galaxies from this sample using the James Clerk Maxwell Telescope. The results will be reported in the PhD thesis of Rob van Ojik.

Figure 1: The average of all the CO(7-6) spectrum centred at -450 km/s from the optical redshift of 0211-122, binned at 50 channels.

One of the most intriguing objects from this sample is the radio galaxy 0211-122 at z = 2.34. It is an ultra-steep spectrum radio source and identified with an object with an R-magnitude of 22.7. For a distant radio galaxy its optical spectrum is highly anomalous; it has very faint Lya emission and strong NV lambda 1240 emission. The spectrum resembles that of the IRAS galaxy F10214+4724. We have suggested that 0211-122 is undergoing a vigorous starburst possibly induced by the passage of the radio jet. Such a vigorous starburst would produce enough dust to attenuate the Lya emission (Van Ojik et al. 1994).

Observations

Last December we had an extensive run at the JCMT to detect CO in distant radio galaxies. Since there was good evidence that 0211-122 contained enormous amount of cold gas, this object was our prime target to detect CO(7-6) emission from the early universe.

The set-up of the whole system was standard. We used the receiver A2 and the DAS digital backend with 2 overlapping bands of 500 MHz each. These two bands consist of 4 overlapping subbands. The total band width is then 750 MHz, although the noise increases somewhat towards the edges of the bands. We investigated different merging parameters, but the standard DAS-merge seemed to give better results. About 5% of the data had bad baselines and were discarded in the subsequent reduction. Only a dc offset was removed before averaging, i.e. no linear or higher order baselines were removed.

We employed standard beam-switching with a throw of 60 arcseconds. The secondary was chopped at a frequency of 2 Hz and the telescope was moved every 30 seconds in such a way that the off-source position was on alternate sides of the source. The theoretical noise (in Ta*) averaged over 64 channels in an 8 hour integration is 0.6 mK. The expected level of the CO emission in distant radio galaxies is not really known. However, in the case where CO at the JCMT was detected in high redshift galaxies it was at the 3- 5 mK level, regardless of the transition. This would indicate we would have a fair chance of detecting CO in these distant radio galaxies.

The main limitation in this kind of observations is that the expected width of the lines (at least a few hundred km/s) is comparable to the width of broad, low level ripples in the baseline of the spectra. It might not be too surprising that these ripples are present, since with typical system temperatures of 400 K the expected signal is at a level of about at least 5 orders of magnitude fainter.

The technique that is being used to check the reality of possible detections of faint lines is to switch the LO setting every time after the required noise levels have been obtained. If the detection is real, then each individual spectrum should show the presence of a faint line, at the same Doppler-shift/velocity.

During the first shift we observed CO(7-6) using the redshift as determined from the optical emission lines. A peak was detected at approximately -150 km/s from this redshift. The next night we centred the frequency setting at this value and found an additional and even stronger peak at -400 km/s. We then spent a total of one shift at -450 km/s. The resulting spectrum from this shift seemed to show a clear 2.5 mK CO detection with a width of 150 km/s (FWHM) (see Figure 1).

However, we were worried that this possible detection and the detection at -150 km/s might be due to a baseline ripple. We therefore decided to continue to observe at different velocity settings to investigate this. We spent a total of 6 shifts on 0211-122 during good weather conditions (CSO tau between 0.03 and 0.06). The total integration time was approximately 40,000 seconds, divided over 5 different LO settings. In Figure 2 we show 7 different spectra smoothed with a gaussian of 200 km/s (FWHM). The baseline ripple shows a sinusoidal behaviour with a period of 300 km/s and a peak to peak amplitude of 2 mK. The main point of this note is to emphasise that the baseline appears to be stationary; shifting the LO frequency does not shift the ripple. This makes it very difficult to distinguish between these ripples and a true signal at the level of a few mK. At the moment it is not clear what is causing the ripples. The frequency spacing of the ripple corresponds to a standing wave over a path of about half a metre. This led Richard Hills to suggest that they might be due to a coupling between the LO plate and the receiver, since their distance is of this order. Further investigations are necessary to determine its nature.

Figure 2: 7 different spectra centred at 5 different Vlsr from the CO(7-6) transition of 0211-122, smoothed with a gaussian of 200 km/s.

Conclusions

By showing these data we hope to accomplish two things. First, we quantify an important limitation of the present instrumentation on the JCMT for detecting weak broad spectral lines. Relevant programmes include observations of clusters of galaxies, absorption systems, quasars and radio galaxies. The expected spectral features from these objects will also be faint and broad and therefore the cautionary note given here will certainly apply. Only because of the large number of shifts at our disposal were we able to explore a relatively large range in frequency, thereby showing the ripple. It is clearly essential to do this in programmes of this type.

Secondly, we hope that these data, together with the astrophysical importance of such work, will provide additional drivers for developing wide-band spectrometers and specific observing techniques for carrying out programmes of this type.

Acknowledgements.

We would like to emphasise that these observations could not have been done without the very good support of the staff at the James Clerk Maxwell Telescope. Special thanks to Fred Baas, Chris Purton and Remo Tilanus. Furthermore, we would like to thank the Netherlands TAG who had the courage and the enthusiasm to allocate a very significant amount of time for such a difficult project. We further acknowledge support from an EEC twinning project.

References:

Barvainis R., Tacconi L., Antonucci T., Alloin D., Coleman P., 1994, Nature, 371, 586.

Miley G., Röttgering H., Chambers K., Hunstead R., Macchetto F., Roland J., Schillizi R., van Ojik R., 1992, ESO Messenger, 68, 12.

Röttgering H. J. A., Lacy M., Miley G., Chambers K., Saunders R., 1994, A &A, 108, 79.

Solomon P. M., Downes D., Radford S. J.E., 1992, ApJ, 398, L29.

van Ojik R., Röttgering H., Miley G., Bremer M., Macchetto F., Chambers K., 1994, A &A, 289, 54.