3.2.1. Variability of the millimetre-IR emission
The first observation of a millimetre outburst in 3C 273 is described in [Robson et al. 1983] who followed the spectral energy distribution throughout the event. This was subsequently interpreted by the model based on a shock in an expanding jet mentioned above [Marscher & Gear 1985]. Although the model is probably too simplistic to be directly applicable, it remains one of the main tools to understand the radio and millimetre variability of 3C 273 and other sources.
The millimetre observations of 3C 273 in 1986 showed [Robson et al. 1986] that the sub-millimetre flux could also decrease to levels well below that normally observed. This happenened while the infrared flux remained constant at wavelengths shorter than 10µm. The radio-millimetre emission of 3C 273 is thought to be due to synchrotron emission. Energetic synchrotron emitting electrons radiating at high frequencies loose their energy faster than less energetic electrons radiating at lower frequencies. The behaviour observed in 1986 is therefore in contradiction with expectations based on synchrotron emission. This result firmly established the presence of another component in the infrared continuum of 3C 273. The small amplitude of the variations in the near infrared and the arguments described above strongly suggest that this component is due to dust close to the sublimation temperature.
The infrared emission of 3C 273 is due to 2 very different components. On one side, the dust that has already been mentioned and on the other a rapidly flaring component that is observed only during short but violent events (see Fig. 2). This activity was observed for the first time in 1988 [Courvoisier et al. 1988]. The flux variations observed then are such that, assuming isotropic emission, the luminosity changes are about 6 . 1040ergs s-2. The polarization during the flare (few percent) was much larger than during quiescent periods. Both the strong variations and the high polarization imply that this flaring component is of synchrotron origin. Using the variability timescale in the K band and assuming that the emission is due to electrons cooling through synchrotron emission, [Courvoisier et al. 1988] deduced that the magnetic field was of the order of 1 Gauss and the Lorentz factor of the electrons emitting the flare of the order of 104. It was later established that these flares may be at the origin of new components in the VLBI jet of 3C 273 (see Sect. 6.2). A similar flare probably occurred in 1990 (see Fig. 2), as can be seen in the long term light curves. It is not possible to study the duty cycle or the frequency of this activity, as the flares are very short and the flux during the flares is extremely variable. The flares are therefore easily missed in long term sets of observations which do not have a sufficiently dense sampling to systematically catch the events.
It is interesting to note that the energy radiated during the synchrotron flare of 1988 is of the order of 1051ergs (an isotropic flux of 20 mJy in a band width of 1µ in the near infrared at a distance of 1Gpc for about 1 day) and to consider whether the radiated energy could have been stored in the magnetic field. The energy available in a magnetic field of about one Gauss as deduced in [Courvoisier et al. 1988] over a volume of a few light days across is of the order of 1046ergs, insufficient to explain the flare. Another possible energy source is the kinetic energy of mass ejection. Assuming a mildly relativistic velocity of 0.1c decelerated in about one day as the flare energy source one estimates that about 6 . 1026g must be decelerated and produce synchrotron radiation with a 100% efficiency to explain the observed luminosity variations. This is the mass accreted by the central black hole every second (see Sect. 8). These estimates would need to be modified if the radiating material was moving at relativistic velocities and emitting non-isotropically.
The millimetre activity linked to the rapid flares is also very violent. The infrared variations do not formally require that the emitting electrons have a bulk relativistic motion. The millimetre emission associated with these flares cannot, however, be understood in terms of synchrotron cooling by a static source in a constant magnetic field [Robson et al. 1993]. Indeed the millimetre activity of 3C 273 is particularly complex with many flares of different characteristic times. The spectrum of individual flares and their evolution cannot, therefore, be confidently extracted.