Next Contents Previous


The HBL PKS 2155-304 is the UV-brightest BL Lac and one of the X-ray-brightest blazars so it is an obvious target for multiwavelength monitoring. There have been three intensive campaigns on PKS 2155-304, in 1991 November, 1994 May, and 1996 May, covering optical through X-ray wavelengths. (A fourth, predominantly X-ray, campaign occurred in 1996 November.) Here I discuss results from the first two campaigns and their implications for blazar jets.

The first campaign consisted of daily observations with IUE and with optical and IR telescopes for the full month of 1991 November, plus a shorter period of near-continuous observations with IUE (4.6 days) and Rosat (3.5 days) in the middle of the month. Only the intensive monitoring resolved the fast variations; such rapid UV variations had not previously been seen and were quite unexpected. Flux changes by ~ 10% or more are common on time-scales of half a day.

The X-ray, UV, and optical light curves from this period, shown in Figure 4, are well-correlated, with at most a 2-3-hour lag of UV photons with respect to X-rays, and their fractional amplitudes are independent of wavelength. (Note that the optical light curve is from data taken with the IUE FES [Fine Error Sensor], a modest instrument compared to ground-based systems but in this case far superior because of its continuous temporal coverage, which is next to impossible from the ground.)

Figure 4: Multiwavelength light curves from intensive monitoring of the BL Lac object PKS 2155-304 in 1991 November (Edelson et al. 1995). X-ray data are from the Rosat PSPC; UV data are from the IUE SWP (short wavelength) and LWP (long-wavelength) spectrographs; optical data are from the FES monitor on IUE. The emission is closely correlated at all wavelengths, and the X-rays lead the UV by ~ 2-3 hours. (Copyright American Astronomical Society, reproduced with permission.)

These light curves are intriguing for several reasons. They established for the first time that X-ray and UV flux were apparently coming from the same process. This would be expected if the smooth UV-X-ray spectrum were produced by the synchrotron process but the wavelength independence of amplitude was puzzling. In the context of synchrotron radiation, higher energy (X-ray-producing) electrons lose energy faster than lower energy (UV-producing) electrons and thus cause larger amplitude variations within a fixed volume. The relative achromaticity of the 1991 data raises the possibility that another process caused the variations. Finally, the series of bumps suggest possible recurrent behavior.

For these reasons, a second campaign was carried out in 1994 May (Fig. 5), with 12 days of continuous observations of PKS 2155-304 with IUE, 9 days with EUVE (which was integrating long enough to get a good spectrum), and with ASCA which, because it was newly launched, was on target for only 2 days.

Figure 5: Normalized multiwavelength light curves from intensive monitoring of the BL Lac object PKS 2155-304 in 1994 May (Urry et al. 1997). Most of the X-ray data are from ASCA with a few early points from the Rosat HRI; EUV data are from EUVE; UV data (with SWP and LWP interspersed) are from IUE. The broad flare seen in the middle of IUE monitoring seems related to an EUV flare one day earlier and to the sharp X-ray flare two days earlier. In addition, extremely rapid (unresolved) variations at the beginning of the IUE observation have doubling time-scales of 1 hour, faster than previously observed at UV wavelengths and comparable to the fastest time-scales seen in X-rays. Compared to 1991, the flares are much larger, the lags are much longer (although the correlation is less definite because of less X-ray coverage) and the flare amplitude declines with increasing wavelength.

The ASCA data show a sharp X-ray flare with amplitude of roughly a factor of 2. The few Rosat observations preceding the ASCA observation show that similar amplitude X-ray flares must have occurred throughout this period. The broad flare in the middle of the IUE light curve can plausibly be associated with the X-ray flare 2 days earlier, particularly given the similar EUV flare which lies between UV and X-ray both temporally and in wavelength. The amplitude of this flare declines from ~ 100% in the X-ray to ~ 50% in the EUV and ~ 35% in the UV. The formal lag of UV with respect to X-ray is 1.7 days, and of UV with respect to EUV is 1.1 days. Assuming the association of these three flares is correct, we find striking differences from the 1991 results. Here the flares are much larger, the lags are much longer (although the correlation is less definite because of less X-ray coverage), and the flare amplitude depends strongly on wavelength.

In addition, extremely rapid (unresolved) variations at the beginning of the IUE observation have doubling time-scales of 1 hour, faster than any UV variations previously observed in AGN, and comparable to the fastest X-ray variations. The LWP integration times were half those of the SWP, which plausibly explains the larger amplitude in the LWP light curve. Even so, these rapid variations are badly undersampled even with the LWP.

Figure 6 shows an expanded view of the beginning of the UV and EUV light curves, with V-band data superimposed. (Unfortunately, the FES was no longer working well as a photometer for such faint objects.) The UV dip is echoed in the V-band polarization; although the latter data are more poorly sampled than the UV, they start earlier, showing that the event is indeed a series of dips beginning at the quiescent level. The sampling is too poor to say whether a similar dip occurs in total V-band flux, but in general the optical and UV fluxes track each other well, as was found in the first campaign (Urry et al. 1993). The EUVE light curve has been shifted horizontally by +1.1 days, which is the peak of the cross correlation between the UV and EUV light curves (most of the power in the cross-correlation comes from the flare several days later). It does not show as rapid or large-amplitude flaring, which could indeed have been seen (the EUV bin size is 2/3 the LWP integration time), but similar structure is apparent in the shifted curve.

Figure 6: Expanded view of the first part of the 1994 monitoring campaign. Filled points -- Normalized UV data for both SWP (circles) and LWP (squares); stars -- Normalized EUV data (in 1000 s bins, shifted vertically by 0.35 and horizontally by +1.1 days); open points -- Normalized V-band (circles) and polarized (squares) flux densities. The UV flux rises by a factor of 2 (in the LWP, for which the integration time was half that of the SWP), dips again and recovers. The polarized V-band flux, though more poorly sampled than the UV, shows the same dip and recovery. The shifted EUV light curve shows similar structure though no exact corresponding events.

The 1994 light curves are clearly different from the 1991 light curves. The central flare is roughly symmetric but broadens with wavelength, as the amplitude decreases. Either the physical state of the emitting region has changed considerably or two different mechanisms cause the variations. The latter possibility is attractive since it is hard to find a single mechanism that can cause energy-independent variations at one epoch and energy-dependent variations at another. For example, the 1991 fluctuations might be caused by microlensing in an intervening galaxy (Treves et al 1997); there is strong Lyalpha absorption midway to the BL Lac (Bruhweiler et al. 1993) although no galaxy is detected along the line of sight (van Gorkom et al. 1996).

The central event in the 1994 light curves is consistent with a synchrotron flare, in the sense that the variability is of larger amplitude and shorter time-scale at higher energies. That the plasma is homogeneous can probably be ruled out since in that case the flare should begin simultaneously at all wavelengths, although the decay would be faster at shorter wavelengths. This was not observed. The delays between wavebands are comparable to the time-scales of the flare, as expected if a disturbance (e.g., shock or compression wave) were propagating through an inhomogeneous emitting region. Such a region could arise naturally as magnetic fields and electron densities vary globally or in the steady-state situation behind a standing shock. However, the decay times appear to be slightly longer than the lag times, which is difficult to understand if the lags and/or the spatial stratification are dominated by the radiative time-scale. For further discussion, see Urry et al. (1997).

Next Contents Previous