The principal advantage of a bolometric system is the high sensitivity that is achieved, but these devices are also of interest because of their frequency range: at present they provide the best sensitivity for observing the microwave background outside the Rayleigh-Jeans part of the spectrum, and hence for separating the thermal and kinematic components of the Sunyaev-Zel'dovich effect using their different spectral shapes. Furthermore, the best systems consist of several detectors arranged in an array, and some provide simultaneous operation in several bands. A suitable choice of differencing between elements of the array reproduces many of the sky-noise subtraction properties of radiometric observing, and the multiband capability holds out the hope of rapid spectral measurements. Bolometric measurements of the Sunyaev-Zel'dovich effects are now becoming more common, as reliable technology becomes more widely available (Table 2).
et al. |
|BS+PS||90-300||5.0||5.0||. . .|
et al. |
|BS+PS||261||1.9||2.9||. . .|
|McKinnon et |
|BS+PS||90||1.2||4.0||. . .||Non-thermal|
|Wilbanks et |
|BS+DS||136||1.4||2.2, 4.3||26, 34||1 x 3 array|
|Andreani et |
|BS+PS||150||0.73||2.3||. . .|
|BS+PS||250||0.77||2.3||. . .|
|BS+DS||143||1.7||2.3, 4.6||30||2 x 3 array|
|Holzapfel et |
|BS+DS||143||1.7||2.3, 4.6||30||2 x 3 array|
|BS+DS||214||1.7||2.3, 4.6||30||2 x 3 array|
|BS+DS||273||1.7||2.3, 4.6||30||2 x 3 array|
|Silverberg et |
|Note. - The technique codes are BS for beam-switching, PS for position-switching, DS for drift- or driven-scanning. is the central frequency of observation: it is not possible to accurately describe the four bands used by Meyer et al. (1983) in this way, and only a range of frequencies is stated in this case. h is the FWHM of the telescope. b is the beam-switching angle (if beam-switching was used), and s is the scan length (for drift or driven scans).|
A bolometer such as SCUBA on the James Clerk Maxwell Telescope (JCMT) at a wavelength of 850 µm (near the peak of the thermal effect in intensity terms) has a sensitivity of 80 mJy Hz-1/2, with a 13-arcsec pixel size. The equivalent sensitivity in the Rayleigh-Jeans brightness temperature change of the thermal Sunyaev-Zel'dovich effect, TRJ, is about 60 mK in one second in each pixel, or about 13 mK in a 1 arcmin beam created by averaging over detector elements (compare the radiometric sensitivity of a typical radio telescope in Sec. 8.1). A few hours of observation should then suffice to detect the thermal Sunyaev-Zel'dovich effect at high sensitivity, and by using several bands (perhaps simultaneously), a coarse spectrum of the effect could be measured. Deviations from the spectrum of the thermal effect could then set limits to the velocities of clusters of galaxies - if the sensitivity of the bolometer is similar at frequencies near the zero of the thermal effect, then a velocity accuracy of about 6 x 104 km s-1 can be achieved in an hour of observation in any one 13-arcsec pixel. This can be reduced to 103 km s-1 or less with modern bolometers if the measurements are averaged over the entire face of a cluster (as, for example, in Holzapfel et al. 1997b). The fundamental limit of this technique for measuring cluster peculiar velocities may be set not by sensitivity, but rather by the background fluctuations in the CMBR which arise from primordial anisotropies. This depends on the angular spectrum of anisotropies (see Sec. 1.3).
Although the raw sensitivity of bolometer systems is high because of the large bandpasses and sensitive detector elements, a problem with the technique is the extremely high sky brightness against which observations must be made. Coupled with the varying opacity of the sky, this implies that telescopes on high, dry, sites are essential for efficient observing - balloon operations are possible, and the CalTech Submillimeter Observatory (CSO) on Mauna Kea has been used successfully. Antarctic operations are also an interesting future possibility, as is space operation with bolometer arrays. At present, the best results are obtained by differencing out atmospheric signals using bolometer arrays. This involves the use of small differencing angles, and introduces limitations on the selection of clusters that are similar to those that apply to radiometric work (Sec. 8.1). The small angular separations of the beams often causes the minimum redshift cutoff to be rather high, and the peak observing efficiency to be low (as in Chase et al. 1987, for which the fraction of the central decrement that was observable was only 0.38 for cluster CL 0016+16).
This technique is exemplified by the recent work of Wilbanks et al. (1994), who used the Caltech Submillimeter Observatory (CSO) on Mauna Kea with a three-element array to detect the Sunyaev-Zel'dovich effect from Abell 2163, a cluster of galaxies with an exceptionally hot atmosphere (Arnaud et al. 1992) and a bright radio halo source (Herbig & Birkinshaw 1998). The combination of drift-scanning and element-to-element differencing used by Wilbanks et al. achieved an excellent separation of the atmospheric signal from the Sunyaev-Zel'dovich effect and provided a measurement of the angular structure of the effect. At the wavelength of operation ( = 2.2 mm) radio source confusion is not a problem. This is not the case at microwave frequencies, where observations of the Sunyaev-Zel'dovich effect in Abell 2163 are severely affected by the radio environment near the cluster center, which includes a variable and inverted-spectrum radio source as well as the radio halo (Herbig & Birkinshaw 1998). Nevertheless, recent observations at 18 GHz with the OVRO 40-m telescope have succeeded in detecting the effect near the cluster center, at about the level seen by Wilbanks et al. (1994).
The most sensitive observations with bolometers (with SuZIE, the Sunyaev-Zel'dovich Infrared Experiment on the CSO) have been made using a drift-scan mode (Holzapfel et al. 1997a), as illustrated in Fig. 18, in order to reduce microphonic and sidelobe spillover effects to the minimum possible level. The SuZIE array consists of two rows of three elements, with the rows separated by 2.2 arcmin and the elements in each array separated by 2.3 arcmin. Array elements within a row are electronically differenced to produce continuous measurements of the brightness differences that they see. During a drift-scan each difference voltage is then proportional to the brightness difference on the sky between two locations which vary as the sky rotates past the detectors. The array is oriented with the long axis parallel to right ascension, so that the time series can be interpreted as a right ascension scan (as in Fig. 19). Repeated drift-scans, with the angle of the array changed from scan to scan, then allow repeated measurements of brightness differences at the same points on the sky.
|Figure 18. Sample SuZIE drift scans across Abell 2163 superimposed on an X-ray contour map of the cluster (Holzapfel et al. 1997a). The two rows of SuZIE detectors are separated by 2.2 arcmin, so that when the upper detectors pass over the X-ray center, the lower detectors pass south of the center. Two sets of scans are shown for each row of detectors, since the observations were alternately begun 12 and 18 arcmin ahead of the cluster center.|
A simple isothermal model of Abell 2163 (Holzapfel et al. 1997a) has = 0.62 ± 0.03 and c = 1.2 ± 0.1 arcmin (in equation 64). The 2.3 and 4.6-arcmin difference signals that SuZIE produces then correspond to peak observing efficiencies (fractions of the central Sunyaev-Zel'dovich effect seen by each 1.75-arcmin FWHM array element) of 0.31 and 0.51, respectively. With this type of observing, the signals returned by SuZIE are close to being measurements of the gradient of the Sunyaev-Zel'dovich effect on the sky, as can be seen in the data shown in Fig. 19.
|Figure 19. Sample 4.6-arcmin difference data from the 1994 SuZIE observations of Abell 2163 (Holzapfel et al. 1997a). The upper panel shows data taken across the cluster center together with the best-fitting (non-isothermal) model of the cluster Sunyaev-Zel'dovich effect based on the X-ray data (heavy line) and the same model with ± 1 errors on the amplitude (upper and lower light lines). The middle panel compares the predictions of the same model with the data taken 2.2 arcmin south of the cluster center, while the bottom panel shows the corresponding model fit to a region of blank sky well separated from the cluster but at a similar declination.|
Just as for radiometric work, it is important to check that the observing technique being used does not suffer from baseline effects from parasitic signals from the sky, the telescope, or the electronics. Control observations of regions of blank sky are used to provide such checks, as in the example of Fig. 19. In all plots in this figure, a best-fitting linear baseline has been removed, and then the data have been fitted using a model of the cluster Abell 2163: only small residual baseline effects remain, and the fits are of reasonable quality.
As with the radiometer data, it is important to remove from the data periods when the sky is opaque, or has rapidly-varying opacity, and the data must also be corrected for the line-of-sight opacity through the atmosphere. At the best millimetric wavelengths these corrections are small, just as they are for most cm-wave observations. Also, as radiometer data must be cleaned of radio interference, so bolometer data must be cleaned of cosmic ray hits. In both cases, this does not create additional difficulties because the effects are generally large and obvious.
A final similarity with radiometric work is the problem of calibrating the data into absolute temperature (or intensity) units. Again, the calibration is usually made by reference to the brightness of planets, and again the difficulty is that the planetary temperature scale is good to 6 per cent at best. Additional errors from the beam-pattern of the detectors, the bandpasses of the detector elements, and the opacity of the atmosphere add to this error, so that the intensity scale of any measurement is not known to better than about 8~per cent. The effect of this on the interpretation of the data will become apparent later.
No radio source or Galactic contamination signals are thought to be significant at the frequencies and angular resolutions at which bolometric data are taken on clusters (Fischer & Lange 1993), and dusty galaxies within the clusters should also be weak. Nevertheless, such signals are present (Smail et al. 1997), and may be enhanced by emission from distant (background) dusty, star-forming galaxies gravitationally lensed by clusters - especially by the massive clusters which produce the strongest Sunyaev-Zel'dovich effects (Blain 1998). If the bolometer array that is used has sufficient angular resolution, it should be possible to reduce this contamination by removing the individual pixels in the map that are affected, but at present only low-resolution bolometric observations of the Sunyaev-Zel'dovich effect exist (e.g., from SuZIE, with 1.7 arcmin resolution, Table 2). Higher-resolution observations of clusters (e.g., with SCUBA on the JCMT) are now possible and should allow checks for the presence of confusing sources, and then their subtraction from the lower-resolution data.
Since differencing in bolometric work usually involves switching over angles which are only a small multiple of the FWHM of the array elements, the observing efficiencies are low. This has led to the results from these experiments usually being quoted in terms of fitted central Sunyaev-Zel'dovich effects (or, equivalently, the y parameter) rather than the beam-averaged central Sunyaev-Zel'dovich effect that is usually quoted in radiometric measurements. Quoting the results as central y values has the virtue of encapsulating the combined statistics of the observational errors and the angular structure data (Fig. 19) into a single number, but it also has the drawback of not allowing the data to be re-interpreted later, as improved structural information becomes available. From the data in Fig. 19, Holzapfel et al. (1997a) find that Abell 2163 has a central Comptonization parameter y = (3.7 ± 0.4) x 10-4 if the cluster gas follows a simple isothermal model. The corresponding central Rayleigh-Jeans brightness temperature change is -1.6 ± 0.2 mK: a remarkably large Sunyaev-Zel'dovich effect, presumably because of the high temperature of the atmosphere in this cluster, although uncertainties in the model, which is based on X-ray data, cause additional ~ 10 per cent uncertainties in the values of y and the central Sunyaev-Zel'dovich effect that are derived.
|Figure 20. The mm to far-IR spectrum of Abell 2163 (from Lamarre et al. 1998). The solid line shows a best-fit model composed of dust emission and Sunyaev-Zel'dovich effects. The dashed line shows the dust contribution to the overall spectrum. The dash-dotted line shows the Sunyaev-Zel'dovich thermal effect. The insert shows the contribution of the kinematic Sunyaev-Zel'dovich effect in the mm to cm part of the spectrum.|
An interesting recent result on the spectrum of the Sunyaev-Zel'dovich effect from Abell 2163 is shown in Fig. 20 (Lamarre et al. 1998). Lamarre et al. combined data taken using several instruments into a single spectrum which shows the relative sizes of the Sunyaev-Zel'dovich effect and far-IR dust-like emission (which dominates from 100 - 1000 µm). This shorter-wavelength emission may arise from the lensed population of background starburst galaxies, from Galactic dust which happens to be brighter near the centre of the cluster, or from dust in Abell 2163 itself. If the spectrum in Fig. 20 is characteristic of other clusters of galaxies, then the interpretation of sub-mm data will need to take careful account of such contamination. This might particularly affect the measurement of the kinematic Sunyaev-Zel'dovich effect.