Full text of DOI:10.1016/S1384-1076(02)00104-5

New Astronomy 7 (2002) 185–189
www.elsevier.com/locate/newast
The incompatibilities between the standard theory of interstellar
extinction and observations
´ ´
Frederic Zagury
´
02210 Saint Remy Blanzy, France
Received 5 February 2002; received in revised form 6 March 2002; accepted 6 March 2002
Communicated by F. Melchiorri
Abstract
In this paper I review a series of observations which do not agree with the standard interpretation of the extinction curve.
The consequence is that light we receive from a reddened star must be contaminated by starlight scattered at very small
angular distances from the star. The true extinction curve is a straight line from the near infrared to the far-UV. If so, all
interstellar grains models must be questioned. Another conclusion concerns the average properties of interstellar grains
which seem much more uniform than previously thought.
2002 Elsevier Science B.V. All rights reserved.
PACS: 98.38.j; 98.38.Cp; 98.58.Ca
Keywords: Dust, extinction
1. Introduction
extinction curve to fully understand the interstellar
medium.
The standard interpretation of the extinction curve
separates the visible and UV light spectrum into
three parts: the visible, the 2000 A bump, and the
Although interstellar dust represents a very small
part of interstellar matter, it is responsible for the
extinction of starlight, it accounts for a large part of
the scattered starlight we receive from nebulae, and
it is also responsible for the thermal emission of
interstellar clouds in the infrared. Thus, interstellar
dust plays a most important role as a tracer of the
distribution and of the structure of interstellar matter.
It is also involved in the chemistry of the interstellar
medium.
The study of interstellar dust has led to several
dust models that are heavily constrained by the
average extinction curve obtained from observations
of stars through interstellar clouds. It is of primary
importance to have a correct interpretation of the
˚
far-UV region. Separate types of particles (grains or
molecules) with specific size distributions and ex-
tinction properties are assumed to be responsible for
the extinction of light in each of these wavelength
domains. The variation of the extinction curve with
the line of sight is interpreted as the effect of
different proportions of each type of particles from
cloud to cloud. All current dust models are based on
this paradigm.
In a preceding series of papers (Zagury, 2000a,b,
2001a,b, 2002) I have detailed observations which
contradict these models, question the standard inter-
pretation of the extinction curve and call for another
explanation. There are important consequences in-
E-mail address: fzagury@wanadoo.fr (F. Zagury).
1384-1076/02/$ – see front matter
PII: S1384-1076(02)00104-5
2002 Elsevier Science B.V. All rights reserved.

186
F. Zagury / New Astronomy 7 (2002) 185–189
volved, as we need to reconsider the nature of light
received from a reddened star, which, in turn, will
deeply affect the analysis and the interpretation of
the observations on interstellar matter. The applica-
tions are numerous ranging from practical aspects
(correcting the reddening for stellar distance estima-
tions), to more theoretical problems on the nature of
interstellar dust, its properties depending on environ-
ment, etc . . .
In this paper I first reviewed the principles of the
standard interpretation of the extinction curve (Sec-
tion 2). The observations which run contrary to this
interpretation are summarised in Section 3. The
implications and ways to reconcile theory and ob-
servation are discussed in Section 4.
to the combined effect of the extinction of starlight
by three types of particles, which are present, but in
variable proportions, in the interstellar clouds. The
linear in 1/l visible extinction is due to a dis-
tribution of large grains, which have a flat scattering
cross section in the UV (Fig. 1 in this paper or Fig. 2
in Greenberg (2000)). The bump region is attributed
to very small grains (VSG). Last, the far-UV rise of
the extinction curve should be due to molecules,
thought to be poly-aromatic (the PAH). ‘Large’ or
‘small’ particles refer to the wavelength domain
which is considered, since the ratio of the size of the
particle to the wavelength is the fundamental param-
eter in scattering theory.
By allowing the proportion of each type of particle
to vary from cloud to cloud, the standard theory
acquires three degrees of freedom which permits to
fit most extinction curves. But, there does not seem
to be any logic behind the grain type repartition with
environment (density, exposure to UV radiation . . . )
(Jenniskens and Greenberg, 1993).
2. The extinction curve and its standard
interpretation
The light we receive from a reddened star is
extinguished by a factor e2t , where tl is the
l
This freedom is paid with an important compensa-
tion. The particular extinction curve each type of
grains must have, the necessity to respect cosmic
abundances, tightly constrain the nature of the grains
and led to several models of grains. To date, none of
the models of interstellar dust in use (the first PAH
extinction optical depth, at wavelength l, of the
interstellar matter between the star and us. If Fl and
F0l are the flux we receive from the star and the one
we would receive if the star was not reddened:
Fl 5 F0l e2t
(1)
l
´
model of Desert et al. (1990), the unified model of Li
and Greenberg (1997), the model of Mathis (1996))
are truly satisfying.
In magnitudes:
ml 5 m0l 1 1.07tl 5 m0l 1 Al
(2)
Since F0l or m0l are not known, they are replaced by
the values observed for a non-reddened star of the
same spectral type. The extinction curve Al is then
obtained to within an additive constant generally
determined from the V-magnitudes of the stars. The
extinction curve can be normalized by E(B 2V) 5
AB 2 AV, proportional to the slope of the extinction
curve in the visible.
3. The standard theory against observations
The standard theory can be tested in several ways.
I have proposed four tests, three of which are
detailed in the following sub-sections.
3.1. The UV spectrum of nebulae
Seaton (1979) gave the average normalized ex-
tinction curve for the stars of the solar neighborhood.
The standard theory separates this curve into three
The UV spectrum of a nebula illuminated by a
nearby star is well reproduced by the product of the
spectrum of the source star and of a linear function
of 1/l (Zagury, 2000a).
Neither the large grains supposed to be responsible
for visible extinction—of which the far UV extinc-
tion cross section is nearly independent of wave-
length—, nor the small particles supposed to be
˚
parts: the visible, the 2200 A bump region, and the
far-UV. The normalized extinction curve in the
direction of a reddened star follows Seaton’s curve in
the visible, but large variations are observed in the
UV, especially in the far-UV (Bless and Savage,
1972). The standard theory attributes these variations

F. Zagury / New Astronomy 7 (2002) 185–189
187
Fig. 1. Seaton’s extinction curve (plain line) and the separate extinction curves of the large grains, the very small grains (VSG), and the
´
poly-aromatic molecules (PAH). From Desert et al. (1990). I have added (dashes) the linear extinction curve.
responsible for the UV extinction—which should
scatter starlight as 1/l4—can explain this relation
between the nebula and the star spectra.
The linearity in 1/l of the nebula to the star
spectrum ratio suggests that the scattering law valid
in the visible extends to the UV.
exponential of 1/l in the visible and in the UV
(Zagury, 2001b). Thus, the extinction law in very
low column density media is the same in the visible
and in the UV, which confirms what is already
suggested by the UV observation of the nebulae.
We also do not observe any excess of scattering in
the bump region (Zagury, 2000a), which means that
if there is a specific extinction at 2200 A, it is
3.3. The extinction curve in directions of low
reddening
˚
absorption only. But, some nebulae, associated to
low-reddened stars, do not show a bump. Therefore,
either the small grains which, according to the
standard theory, are responsible for the bump are not
present (or only in very small quantities) in low
column density clouds, or the bump is not a common
absorption process.
Increasing the reddening of the stars, the extinc-
tion law deviates from the 1/l linear extinction in
the far-UV first, whereas the linear visible extinction
law still extends to the bump region (Zagury,
2000b). The reduced spectrum of the stars is an
exponential in the near-UV. This exponential
prolongs in the UV the visible extinction of the stars’
light. The deviation from the visible extinction of the
far-UV reduced spectrum clearly appears as an
additional component superimposed on the tail of the
exponential (Fig. 2).
3.2. The extinction curve in directions of very low
reddening
The spectra of stars of same spectral type and very
low reddening, not reddened enough to have a
2200 A bump, differ one from the other by the same
Here again, the standard theory cannot explain in a
natural way the extension of the visible extinction
law to the near-UV.
˚

188
F. Zagury / New Astronomy 7 (2002) 185–189
Fig. 2. The spectrum of HD62542 divided by the spectrum of the non-reddened reference star HD32630. The visible extinction corresponds
to the exponential and extends in the near-UV and down to the bump region. The far-UV difference between the two curves clearly appears
as an excess of light superimposed on the tail of the exponential. This excess is attributed to an additional component of scattered light.
4. Discussion
If the column density is increased, scattered light
first appears in the far-UV, because extinction, hence
the number of photons available for scattering,
increases towards the shortest wavelengths. Still
increasing the reddening, the scattered starlight will
merge in the visible, provoking the departure of
Seaton’s curve from the linear extinction between
1/l | 2.5 mm21 and 1/l | 4 mm21 (Fig. 1).
There are many consequences. Firstly, the extinc-
tion curve is a straight line from the near infrared to
the far-UV. Secondly, there is a priori non reason to
suppose changes of the average properties of inter-
stellar dust from one interstellar cloud to another: the
law of starlight scattering found for the nebulae for
instance is the same in the different nebulae. If, in
the near future, spectral observations of nebulae
confirm that this law extends to the visible, we will
also be able to conclude that the albedo and phase
function of interstellar grains are wavelength-inde-
pendent from the near infrared to the far-UV. Varia-
tions of the RV 5 AV /E(B 2V) parameter are ob-
served for some stars and used as a proof of
variations of the properties of interstellar dust. These
variations will as well be explained by the slight
modification, due to the scattered starlight, of the
The observations mentioned in Section 3 con-
tradict the standard theory of interstellar extinction.
The only alternative to the standard theory was
first mentioned by Savage (1975). It implies that
when a reddened star is observed in the UV a non
negligible proportion of scattered light is re-intro-
duced into the beam of the observation.
Snow and York (1975) have rejected this hypoth-
esis from UV observations of s-Sco with two
different apertures of 89 3 38 and 0.30 3 390. There is
no difference between the UV spectra of the star
observed with one aperture or the other. However,
these observations only prove that if scattered star-
light contaminates the spectrum of reddened stars, it
must be within an angle of 0.30 from the star.
The observations of Section 3 are explained if a
non-negligible contribution of scattered light is
added to the observed direct starlight. When there is
little interstellar matter between the star and the
observer the contribution of scattered light is negli-
gible: we observe the direct starlight only, slightly
extinguished, and the observed extinction reflects the
exact extinction law of starlight by interstellar dust.

F. Zagury / New Astronomy 7 (2002) 185–189
189
slope, 2E(B 2V), of the visible extinction curve of
References
the stars (Zagury, 2001a).
It also follows that current models of interstellar
dust, formed on an overly direct interpretation of the
peculiarities of the extinction curve, are questionable.
There are two practical aspects of this research
which merit mentioning. The papers I have published
in the two preceding years show the need of acquir-
ing stellar spectra on an as large as possible wave-
length range. The simultaneous study of visible and
UV extinctions becomes necessary to understand and
to separate the different contributions of direct and
scattered starlights. This necessity of separating
direct and scattered starlight implies that the current
method of using the magnitudes of the spectra rather
than the spectra themselves is not recommended. The
traditional way would be best if there was direct
starlight only, but the separation of two additive
components is much easier from the raw spectra than
from their logarithm.
Bless, R.C., Savage, B.D., 1972. ApJ 171, 293.
Desert, F.X., Boulanger, F., Puget, J.L., 1990. A&A 237, 215.
Greenberg, J.M., 2000. The Secrets of Stardust. Sci. Am. 283 (6),
46.
Jenniskens, P., Greenberg, J.M., 1993. A&A 274, 439.
Li, A., Greenberg, J.M., 1997. A&A 323, 566.
Mathis, J.M., 1996. ApJ 472, 643.
Savage, B.D., 1975. ApJ 199, 92.
Seaton, M.J., 1979. MNRAS 187, 73.
Snow, T.P., York, D.G., 1975. Ap&SS 34, 19.
Zagury, F., 2000a. NewA 4, 211.
Zagury, F., 2000b. NewA 5, 285.
Zagury, F., 2001a. NewA 6, 403.
Zagury, F., 2001b. NewA 6, 471.
Zagury, F., 2002. NewA 7, 117.