K. Leighly et al., EVIDENCE FOR X-RAY FLUX AND SPECTRAL MODULATION BY ABSORPTION IN NGC-6814 .1. THE NATURE OF THE MOST RAPID VARIABILITY, The Astrophysical journal, 421(1), 1994, pp. 69-86
The Seyfert 1 galaxy NGC 6814 was observed using the Japanese X-ray as
tronomy satellite Ginga in 1990 April and October. The rapid variabili
ty characteristically associated with this source was reconfirmed. Spe
cifically, three dips were found in the April observation in which the
flux dropped to nearly zero in similar to 300 s. The doubling timesca
le was similar to 50 s. A similar but separated drop and rise in flux
was observed in the October data, different in that the flux did not d
ecrease completely to zero. A detailed analysis of the data around the
structures of most rapid variability found spectral variability and l
ags in flux between different energy bands. Lags were on the order of
a few to tens of seconds for the April data, and on the order of tens
to a couple of hundred seconds for the October data. The sense of the
lags was such that during flux decreases the hard flux lagged, while d
uring flux increases the soft flux lagged. Associated significant appa
rent hardening of the spectrum at low flux was observed in the April d
ata. Apparent hardening of the spectrum also occurred in the October f
lux decrease, to a photon index of Gamma = 0.85; however, the spectrum
softened at lowest flux to the index of the predecrease level, Gamma
= 1.54. In the April dips, the line flux was found to decrease signifi
cantly. A marginal decrease in line flux was also observed in the Octo
ber data. The variability of the line flux reconfirmed the result of K
unieda et al. (1990), who found that the line producton region must be
within similar to 300 light-seconds from the source. To explain the o
bservational results, a variable-absorption model was proposed, in whi
ch the column density was assumed to vary as a function of time. The t
ime dependence of the column was determined in two ways. First, the fu
nctional dependence was assumed to be exponential, and model parameter
s were derived by fitting explicitly to the lag data obtained from the
October observation. Second, a partial-covering spectral model was as
sumed, and the column density implied by the change in flux was found
for both observations. The exponential folding time of the column dens
ity implied during the October flux decrease could be reliably determi
ned to be about 100 s. The spectral variability could be qualitatively
explained by the variable-absorption model, since superposition of co
ntinuously variable columns can result in an apparent hardening of the
spectrum when measured by moderate-resolution instruments. The differ
ences between the April and October observations could be explained if
the two model parameters, the fraction of unabsorbed flux and the gra
dient with respect to time of the column density, were adjusted. Other
physical processes, including intrinsic spectral changes and warm abs
orber models, which could account for the fastest variability, could b
e ruled out by the results of the data analysis. A primary constraint
found for geometrical models is that the material which is doing the a
bsorption must have relatively low ionization (xi < 100). This result
leads to severe constraints on a general orbiting cloud model, requiri
ng high densities (n similar to 10(16) cm(-3)) and sheetlike geometry.