The classical thermonuclear runaway model of the nova outburst predicts
that not all the material accreted on the white dwarf is ejected by
the outburst and subsequent radiation driven wind (e.g.
Starrfield 1989). A
substantial fraction of material returns to quasistatic equilibrium forming an
envelope 
around the white dwarf. This
H-rich material burns steadily until the fuel is exhausted. During this time
the nova radiates at Constant Bolometric Luminosity (hereafter CBL). This phase
lasts as long as there is H-rich material burning on top of the white
dwarf. The duration of this phase should be of the order of the nuclear time
of the envelope (hundreds of years), but up to now observations have shown
that the CBL phase is much shorter
(e.g. Orio & Ögelman 1996;
Orio et al. 1997).
Therefore there must be some
mechanism which removes efficiently material from the envelope.
A line-driven wind ensuing months after the outburst (Starrfield, private
communication), frictional drag energy during the common envelope phase
(e.g. Mc Donald et al. 1985) and the magnetic field of the WD
(Orio et al. 1992b) have
been proposed to explain the decrease of mass of the envelope.
Kato &
Hachisu (1994)
have shown that the introduction of the new OPAL
opacities in the computations of optically thick winds in novae can reduce
the predicted duration of the CBL phase to a few years or less, depending
on the mass of the WD.
In summary, theoretical models predict that post-outburst novae will radiate
at a very high bolometric luminosity (slightly below the Eddington limit)
until hydrogen depletion finally causes the shell source to shut down.
The time scale for this to occur, predicted to be in the range of 1 year
to > 10 years, is correlated with the system parameters
and, above all, in all models is inversely dependent on the white dwarf mass.
As mass is lost and the envelope shrinks, the effective temperature
increases until exceeding 105 K, and the peak emission
moves to the UV and then to the EUV and soft X-ray ranges and therefore in
these spectral regions the nova turn-off should be better observable.
Most theoretical predictions imply that during the CBL phase soft X-ray
luminosities are in the range 
, and the
expected luminosities in the IUE band are
(e.g. Prialnik 1986).
The determination of the length of the post-outburst nuclear burning phase for a significant number of objects will help to understand the factors that determine the active lifetime of classical novae.
Apart from GQ Mus and V1974 Cyg, the most recent novae are generally
not bright in supersoft X-rays at times ranging from a month
until 
200 years after the outburst
(Orio et al. 1997).
So it is not unlikely that all the hydrogen is depleted soon
after the outburst in most novae. Systematic long exposure X-ray
observations of novae with the ROSAT PSPC during the first couple of years
after the outburst have been performed for five objects: GQ Mus
(Ögelman et al. 1993;
Shanley et al. 1995), V1974 Cyg
(Krautter et al. 1996;
Balman et al. 1997); LMC 1992,
V838 Her (Lloyd et al. 1992;
Szkody & Hoard 1994) and
V351 Pup (Orio et al. 1996). Only the first three were
monitored in X-rays repeatedly allowing a measurement of the turn-off
time. GQ Mus was detected by EXOSAT in X-rays in 1983-1984, without
spectral resolution, (Ögelman et al. 1984, 1987) and later as a
super-soft X-ray source by ROSAT in 1992
(Ögelman et al. 1993).
In January 1993 it started to
decline and the X-ray flux decreased below the ROSAT sensitivity threshold
by September of that year. Shanley et al. (1995) derive a turn-off time
of 
years. Krautter et al. (1996) derive for V1974 Cyg a
turn-off time of only 18 months from the ROSAT data, in agreement with the
change in the ratio of the UV Nitrogen lines (Shore et al. 1996).
Nova LMC 1992 instead never became a supersoft X-ray source, probably
because all the hydrogen was depleted in the outburst
(Orio et al. 1997).
V838 Her and V351 Pup were detected as hard X-ray sources
in the ROSAT range, but they did not show the super-soft X-ray,
black-body like spectrum and the high luminosity expected for hydrogen
burning after the outburst.
In principle soft X-rays or EUV observations are able to directly measure
, but the interstellar extinction levels of most novae are
often too high to allow such data to be usefully obtained.
It is also possible that intrinsic absorption of the ejected shell prevents
the detection of extreme ultraviolet (EUV) or supersoft X-rays, and
it might be negligible after a year or more only for masses
of the order or below 1 
(Yungelson et al. 1996).
Because of its spectral range, the ROSAT PSPC was the best instrument to monitor the late phase of classical novae. New information in the near future might be added only by the X-ray satellite Beppo-SAX, which however did not perform an all-sky-survey and allows less serendipitous detections. The database of X-ray observations of the hydrogen burning white dwarfs therefore consists primarily of observations which have already been made. A practical approach to observational constraints on post-outburst novae, which usually are at least moderately extincted, involves a combination of X-ray and UV spectrophotometry to constrain the luminosity directly radiated from the hot stellar remnant, as well as multi-wavelength measurements of the luminosity re-radiated by the nova ejecta. For this reason it is essential to reexamine the IUE data in a systematic way in order to better understand the evolution of classical novae after the outburst. Ultraviolet data provide one relatively direct window onto novae during their late outburst phases. The IUE spectra also provide information on the UV color temperature of the continuum, which is an additional point for comparison with the models, and show the physical conditions of the shell around the nova.
This paper presents an observational study of these issues through measurements of the the decay of the ultraviolet flux of a sample of post-outburst classical novae. Even if the bolometric luminosity cannot be precisely inferred from the UV tail of the spectrum, the UV observations constrain the duration of the late constant bolometric luminosity phase, often yielding upper limits on the flux that usefully describe the evolution. We analyze the IUE observations made at late stages after the outburst, performed by us or retrieved from the archive, and correlate the post-outburst UV spectral properties with the optical characteristics of the outburst. All the IUE spectra used in this work have been processed with the new IUE processing software (NewSips, see Nichols & Linsky 1996).