The narrow emission-line template spectrum of NGC 5548 has
for the
first time enabled us to isolate the narrow and broad UV
emission-lines and thereby determine firm estimates for the
intensities of the broad emission-lines for both the low-state
spectrum, and the average HST/FOS archival spectrum obtained from the
intensive monitoring campaign undertaken by "AGN Watch" on April
19th 1993, when the continuum was a factor of ~4.8 brighter. A
comparison of the low-state and high-state spectra show that whilst
the narrow UV emission-lines are non-variable over timescales of
~10 months, the broad UV emission-lines exhibit large variations
in both their strength and shape. By combining a photoionization code
with a robust global optimization routine, we have determined global
best-fit parameters for the "average" physical conditions within the
broad emission-line region gas for both the low and high-state
spectra.

By using a "best-guess" estimate for the shape of the ionizing
continuum of this source, we find that a single zone photoionization
model of the BLR cannot simultaneously fit both the emission-line
ratios of the strongest UV lines and their variability
timescales. However, the line ratios and variability timescales can be
reproduced if we assume a stratified BLR, i.e.,
a BLR which has strong
gradients in density
(N_H  1/r²,
10^11.3-10^10.0). These models also suggest that the BLR gas
is in a moderately high state of ionization with log₁₀U ~-0.6 in the low-state spectrum, rising to ~0.0 in the
high-state.

We find that the observed differences in the broad emission-line
fluxes and their ratios, between the low and high-state spectra, are
not solely a consequence of changes in the ionizing continuum source
luminosity. Rather they imply in addition, a change in the spectral
energy distribution of the ionizing continuum, although changes in
either the covering fraction, or composition of the broad
emission-line region gas cannot necessarily be ruled out.

By constructing a simple two-zone model for the high-ionization lines,
we find that in order to reproduce the observed line ratios and line
equivalent widths, the gas covering fraction for this source must
necessarily be high, ~38% at a radial distance of 2 light-days,
and decreasing outward to ~32% at 10 light-days. This is
considerably larger than the typical value of ~10% quoted for
AGN, derived from the incidence of Lyman edges in high redshift
quasars. Although the statistics for the incidence of Lyman edges in
Seyfert 1 galaxies is poorly determined, our derived covering fraction
is broadly consistent with the 25% covering fraction estimate
obtained from observations with the Hopkins Ultraviolet Telescope
(HUT).


END:: epreps.stsci//prep1201