214 PUBLICATIONS OF THE

ABSORPTION OF LIGHT IN THE GALACTIC SYSTEM

By Robert J. Trumpler

For more than a century astronomers have interested them-

selves in the question: Is interstellar space perfectly trans-
parent, or does light suffer an appreciable modification or loss
of intensity when passing through the enormous spaces which
separate us from the more remote celestial objects? Any effect
of this kind is generally referred to as "absorption of light in
space," whatever the peculiar physical process assumed for its
cause. Various hypotheses have been proposed for the latter.
The older views attributed such absorbing properties to the
hypothetical ether itself; but at present we think rather of a
much rarefied invisible material medium and admit that the
latter is not necessarily of uniform distribution throughout all
space. According to prevailing physical theories, light passing
through such a material medium will be affected in various
ways: Aside from possible refraction and dispersion effects,
light may be absorbed by free atoms or molecules ; it may be
scattered by free electrons, atoms, or molecules, or by solid
particles of extremely small size; and finally light may be ob-
structed by larger bodies, such as meteorites. The space ab-
sorption of light is thus intimately related to the question of the
presence, distribution, and constitution of dark matter in the
universe.
Let us briefly review the observable phenomena which may
give information on this question :
1. General Absorption.—By this term we designate the loss
of starlight on its passage from the star to the observer. If such
loss exists, the apparent brightness of a star will not decrease
inversely proportional to the square of its distance, but more
rapidly. This will make itself felt in the statistical determina-
tion of the space distribution of stars from star counts of suc-
cessive magnitude intervals. It is further to be noted that a
general absorption will affect all photometric distance determi-
nations which are based on a comparison of absolute and ap-
parent magnitudes. Distances derived by such methods (spec-

© Astronomical Society of the Pacific · Provided by the NASA Astrophysics Data System
 ASTRONOMICAL SOCIETY OF THE PACIFIC 215

troscopic parallaxes, variable star parallaxes, etc.) should then

differ systematically from the results of other methods not af-
fected by absorption (statistical distances from proper motions,
apparent diameters of star clusters or nebulae, etc.).
2. Selective Absorption.—If the loss of light is not the same
for all colors, but varies with the wave-length, we speak of a
selective absorption. Its consequence is that the apparent color
of a star changes with its distance from the observer.
3. Monochromatic Absorption, or the observation of inter-
stellar absorption lines in stellar spectra.—Evidence that a cer-
tain spectral line is not produced in the atmosphere of the star
but by atoms contained in the space between star and observer
may be gained in two ways :
a) There should be an increase with distance in the inten-
sity of the line for stars of the same spectral type and lumi-
nosity.
b) The Doppler shift of such line will generally differ from
that of the stellar absorption lines, and it should appear sta-
tionary in the case of spectroscopic binaries.
According to the investigations of O. Struve, J. S. Plaskett,
Eddington, and others, we have good reason to conclude that the
Κ line of calcium in stars of types OS to B3 is of interstellar ori-
gin and that ionized calcium atoms are scattered through space
within our galactic system, taking part in its rotational motion.
4. Obscuration Effects.—Among these, we have in the first
place to mention the so-called dark nebulae. They are noticed
either as well-defined nearly starless patches in the middle of
rich Milky Way star fields, or as dark passages apparently pro-
jected on bright diffuse nebulae. The view that these forma-
tions are caused by local obscuration or absorption of light is
rather generally accepted, and some astronomers are even in-
clined to consider the dark division of the Milky Way between
Scorpio and Cygnus as of a similar origin.
In the second place there is the well-known fact that practi-
cally no globular clüsters or spiral nebulae are visible near the
galactic equator. This suggests that some of these distant ob-
jects are obscured by an absorbing medium in our Milky Way
system which is strongly concentrated to the galactic plane.

© Astronomical Society of the Pacific · Provided by the NASA Astrophysics Data System
 216 PUBLICATIONS OF THE

5. Dispersion of Light, which means that in a material me-

dium the velocity of light varies with the wave-length. If an
effect of this kind is present, the light variation of a distant
eclipsing binary should not be observed simultaneously in all
colors ; there should be a phase difference between the light
variations of the various colors. This is known as the Nord-
mann-Tikhoff phenomenon. Kienle,1 discussing the observa-
tional data relating to it, comes to the conclusion that there is
no reliable evidence of a measurable dispersion of light in inter-
stellar space and that we should theoretically not expect it un-
less accompanied with a large absorption effect.
The study of open star clusters2 has brought to light new
results indicating the existence of a general and selective ab-
sorption, and the present paper is mainly concerned with a dis-
cussion of these two phenomena.
The attention of astronomers was first directed to the pos-
sible existence of a general absorption of light in space by sta-
tistical investigations on the space distribution of the stars. If
we make the assumptions that the stars are uniformly dis-
tributed through unlimited space and that light is not subject
to a general space absorption, calculation would lead us to the
conclusion that the stars counted in every magnitude interval
should be 3.98 times as numerous as in the preceding interval,
and that the whole night sky should be bright with starlight.
Since the latter is not the case and since the observed star num-
bers increase at a smaller ratio, we are forced to reject either
one of the two assumptions or both. The star counts alone can-
not tell us which interpretation is correct, but for various other
reasons most astronomers now favor the view that the stars are
thinning out at greater distances from the Sun. Nevertheless,
it is of interest to inquire what amount of general absorption
would be necessary to reconcile the observed star counts with a
uniform space distribution of the stars. Halm,3 in his second
paper treating the star counts of the whole sky, finds a value of
2ψ1 per 1000 parsecs; Schalén,4 from a study of faint A- and
1
Jahrbuch d. Radioaktivität und Elektronik, 20, 14, 1923.
2 s
Lick Obs. Bull., 14, 154, 1930. M.N.R.AS., 80, 162, 1919.
*A.N., 236, 249, 1929.

© Astronomical Society of the Pacific · Provided by the NASA Astrophysics Data System
 ASTRONOMICAL SOCIETY OF THE PACIFIC 217

B-type stars in the Milky Way, derives a value of 0ψ5 per 1000
parsecs. Halm's figure should rather represent an upper limit,
as a general increase in star density at greater distances from
the Sun is very unlikely. A local increase in star density in
the region of the Milky Way star clouds, however, seems pos-
sible, and Schalén's figure may therefore be too small.
While the statistical treatment of star counts can only fur-
nish an upper limit or an estimate for a general absorption, the
apparent diameters of open star clusters offer an exceptionally
favorable opportunity to test its actual existence and determine
its numerical amount.
In Lick Observatory Bulletin, 14, 154, 1930, the writer de-
termined the distances of 100 open clusters by means of the
magnitudes and spectral types of their members. These "photo-
metric distances" are based on the inverse square law, and are
found in Tablé 3, column 8, of the paper referred to. In view
of the great diversity in cluster formations it seemed a priori
likely that the real space dimensions are not the same for all
of them but that they depend on the constitution of each cluster.
The open clusters were therefore classified according to the
degree of star concentration toward the center and according to
the number of stars contained in them. As expected, the linear
diameter in parsecs, computed from the photometric distance
and the apparent angular diameter, was found to be correlated
with both of these characteristics. Making the assumption that
clusters of similar constitution have on an average everywhere
the same dimensions, it is possible to determine the distance of
a cluster also by comparing its angular diameter with the mean
linear diameter of the subclass to which it belongs. The units
in which these diameter distances are expressed are, of course,
arbitrary; they depend on the values adopted for the mean
linear diameters of the subclasses.
A comparison between the diameter distances so obtained
and the photometric distances is illustrated in Figure 1 (p. 218).
Each cluster is plotted according to its photometric distance as
abscissa and its diameter distance as ordinate. The units of the
latter are here so chosen that for the mean of all clusters they
correspond to the par sec units of the photometric distance scale.

© Astronomical Society of the Pacific · Provided by the NASA Astrophysics Data System
 218 PUBLICATIONS OF THE

The smaller dots represent clusters for which the photometric

distances are less certain and received half weight. The asterisks
mark the group means (geometric) formed when the clusters

Jhamter
Jiistafut
i\
HOOO ~

. ·. ^
Jooo ,'λ ··
v2;V . .

í > Thfitom-etrxe.
iooo tooo JOOO tooo » SOOOp+necf DistnTiee
Fig. 1.—Comparison of the distances of 100 open star clusters deter-
mined from apparent magnitudes and spectral types (abscissae) with
those determined from angular diameters (ordinates). The large dots
refer to clusters with well-determined photometric distances, the small
dots to clusters with less certain data (half weight). The asterisks and
crosses represent group means. If no general space absorption were pres-
ent, the clusters should fall along the dotted straight line; the dotted
curve gives the relation between the two distance measures for a gen-
eral absorption of O1?? per 1000 parsecs.

are arranged according to their photometric distances; the

crosses mark similar group means obtained by ordering the
clusters according to their diameter distances.
If no absorption of light were present, the two distance
scales should be in agreement, and the clusters should be scat-
tered along the straight line which divides the angle of the two
co-ordinate axes. It is, however, evident that the clusters de-
viate systematically from this straight line. For the nearer
clusters the diameter distances are larger, for the distant clusters
they are smaller, than the photometric distances. This discrep-

es) Astronomical Society of the Pacific · Provided by the NASA Astrophysics Data System
 ASTRONOMICAL SOCIETY OF THE PACIFIC 219

ancy is not due to a systematic error in the estimates of angular

diameters; it persists even if we select only clusters which are
nearly of the same angular diameter but at different distances.
No error of observation, nor any effect of selection, was found
to account for it. Unless we are willing to admit that the di-
mensions of open clusters depend on their distance from the
Sun, we are led to the conclusion that the inverse square law on
which the photometric distances are based does not hold and
that a general absorption is taking place within our stellar sys-
tem. Assuming that this absorption is approximately uniform
within the region occupied by these clusters, we find its numeri-
cal value to be 0^7 per 1000 parsecs. The relation between the
diameter distances and the photometric distances which we
should expect for an absorption of this amount is shown in Fig-
ure 1 by the dotted curve. It represents the observations quite
satisfactorily, much better than the straight line.
A close investigation of the distances and diameters of open
clusters thus gives good evidence of a general absorption affect-
ing the photographic magnitudes of stars by about 0ψ7 per 1000
parsecs. This result is quite compatible with the conclusions
drawn from star counts. When we compare our figure with the
limiting value derived by Halm (p. 216), we see easily that our
general absorption will not change the fact that the space den-
sity of stars decreases in general for greater distances from the
Sun, although the decrease will be somewhat slower. A com-
parison with Schalén's value, on the other hand, would indicate
a slight local increase in star density in the region of the Milky
Way star clouds. Unfortunately, neither trigonometric or dy-
namical parallaxes nor statistical parallaxes derived from proper
motions are at present sufficiently accurate to detect so small an
absorption by comparison with spectroscopic parallaxes.
Studies of the diameters of globular clusters and spiral
nebulae, and of the surface brightness of the latter, have been
made by Shapley,5 van Rhijn,6 and Lundmark. Notwithstand-
ing the great distance of these objects, no appreciable space
absorption was found. This suggests that the general absorp-

s 6
Han\ Bull., 864, 1929. 4, 123.

© Astronomical Society of the Pacific · Provided by the NASA Astrophysics Data System
 220 PUBLICATIONS OF THE

tion which affects the open clusters is not operating throughout

the Universe, but appears to be confined to the Milky Way sys-
tem, while extragalactic space is of much greater transparency.
We have now to take up the question whether this absorp-
tion is selective, that is, whether it varies with the wave-length.
If this is the case, the color of a star should change with its
distance. The fact that the fainter stars on an average are more
reddish and have larger color indices was at first taken for an
indication of selective absorption. Since spectral types of fainter
stars became available, however, it was recognized that this is
due to an increasing percentage of low-temperature dwarf stars.
Since the color of a star depends mainly on its surface tem-
perature, a change of color by selective absorption can only be
detected if the surface temperature can be determined inde-
pendently of the color, e.g., from the spectral types. As far
as the latter are based on the relative intensities of spectral lines
they are not much affected by a space absorption and measure
the temperature by means of the ionization and excitation in the
stellar atmosphere. A selective absorption would then be noticed
by an increase with distance of the mean color index corre-
sponding to a given spectral type. Investigations of this prob-
lem were made in 1909 by Kapteyn7 for the stars of the Har-
vard Revised Photometry and later by van Rhijn8 and Jones9
for the stars of the Y erkes Aktinometry. Taking into account
also the change of color with absolute magnitude and the exist-
ence of possible systematic errors in the visual and photographic
magnitude scales, the results were the following :

Increase in Color Index

per 1000 Parsecs
Kapteyn . +0^31 ± 0^09
Jones ... +0.47 ±0.05
van Rhijn +0.15 + 0.05
Mean +0^3

Although the stars investigated were relatively near and al-

7
Contr. Mt. W. Ño. 42; Ap. 7., 30, 284 and 398, 1909.
8
Dissertation, Groningen, 1915.
9
M.N.R.A.S., 75, 4, 1914. /-

© Astronomical Society of the Pacific · Provided by the NASA Astrophysics Data System
 ASTRONOMICAL SOCIETY OF THE PACIFIC 221

though the separation of the unknowns was uncertain, these

results suggested a selective absorption of 0^3 per 1000 parsecs.
A few years later Shapley began his brilliant researches on
the globular clusters. The fact that he found many small color
indices in globular clusters more than 10,000 parsecs distant
seemed to disprove the existence of any appreciable selective
absorption and discouraged further work on the subject.
When observing the spectral types of faint stars in open
star clusters, the writer was struck by the consistent discrep-
ancy between his spectral-type estimates and the color indices
determined by other observers. Since the average relationship
between color index and spectral type for the nearer stars is
now well known, we can predict for any giant or dwarf star
of known spectral type its "normal color index." The differ-
ence between this normal color index and the one actually ob-
served is called the color excess. Table I gives for eight of

TABLE I

Color Excess in Open Star Clusters

Cluster Distance in Mean Color No. of Observer of
N.G.C. Messier Parsecs Excess Stars Color—I
1647 .. 610 +0^17 33 Hertzsprungio
.19 6 Seares11
2682 67 740 .27 81 Shapleyis
2099 37 820 .05 25 Von Zeipel and
Lindgren13
2168 35 840 .14 98 Wallenquist^
1960 36 980 . 05 40 Wallenquist15
6705 11 1340 .65 46 Shapley^
7654 52 1360 .49 43 Wallenquist17
663 .. 2170 +0.71 41 Wallenquist17

the more distant clusters the mean color excess (col. 4) derived
by combining the writer's spectral types with the color indices
10
Contr. Mt. W. No. 100; Ap. J., 42, 92, 1915.
Contr. Mt. W. No. 102; Ap. J., 42, 120, 1915.
12
Contr. Mt. W. No. 117.
13
Kgl. Svenska Vet. Akad. Handlingar, 61, No. 15, 1921.
^ Ann. BosscharSterrenwacht, Lembang (Java), 3, No. 2, 1929.
15
Meddel. Upsala, No. 32, 1927.
16
Contr. Mt. W. No. 126; Ap. J., 45, 164, 1917.
17
Meddel. Upsala, No. 42, 1929.

© Astronomical Society of the Pacific · Provided by the NASA Astrophysics Data System
 222 PUBLICATIONS OF THE

measured by the observer listed in the sixth column. The clus-

ters are arranged according to their distance from the Sun
(corrected for general absorption), and the table clearly brings
out the fact that the color excess is always positive and that it
increases with the distance. For the three most distant clusters
the excess is so large that B- and A-type stars were found to
have color indices normally associated with F- or G-type stars.
As all the color index determinations, with the exception of the
first, are based on comparisons with the North Polar Sequence,
it is out of the question to trace all these figures to observational
errors. There seems to be no alternative but to interpret the
observed color excess as being due to a selective absorption of
light in space ; this will explain not only why the color excess
is always positive but also why we find its largest values in the
most distant clusters. If we assume the absorption to be uni-
form, the color excess γ should be proportional to the distance,
and we obtain by least squares solution for its value:

γ = +0^31 ± 0^03 per 1000 par sees

This result is in good agreement with the mean of the earlier

determinations of Kapteyn, Jones, and van Rhijn; it is further
confirmed by two recent unpublished investigations which show
that it applies not only to clusters but to the fainter Milky Way
stars in general.
Miss L. Slocum, comparing the spectral types observed by
A. Maxwell18 for stars of magnitudes 11-14 in five selected
areas near the galactic circle with her color index determina-
tions of the same stars, found:

γ = -|-0ψ34 per 1000 parsecs.

Dr. P. van de Kamp, investigating the color indices of B-

and A-type stars of about ΙΟ111, obtained the following values for
the selective absorption (communicated by letter) :

γ = +0^38 db 0ψ03 Galactic lat. l0-30

+0.23 ±0.04 40-60
+0.22 ±0.05 70-120
18 Lick Obs. Bull, 13, 68, 1927.

© Astronomical Society of the Pacific · Provided by the NASA Astrophysics Data System
 ASTRONOMICAL SOCIETY OF THE PACIFIC 223

Although much further research on the change of color with

distance is desirable, we have in the various results quoted quite
definite evidence of a selective absorption of about 0ψ32 per
1000 par sees in low galactic latitudes.
Only a brief reference can here be made to some other
observational facts which may also find their explanation as
effects of selective absorption. Concerning the colors of O-type
stars the textbook of Russell-Dugan-Stewart19 makes the fol-
lowing statement: "The stars of class O (which, from the
character of the lines that appear in their spectra, must be the
hottest of all) are slightly yellower than those of class B." In
consequence of their high luminosity (absolute magnitude -4),
even the brighter stars (4-6111) are at distances of 400-800
parsecs, and, since they lie close to the galactic plane, their
color index will be increased by 0^15 to 0^25 through selective
absorption in space. B-type stars of the same apparent magni-
tude are nearer and therefore less affected by such absorption.
It is, however, possible that the small change in color between
types Β5 and B0 is also due to the same effect.
Another case is that of B- and A-type stars which are ab-
normally yellowish for their spectral class. Two lists compiled
by Hertzsprung20 contain 41 such objects and show that they
are nearly all situated in low galactic latitudes, while their small
proper motions indicáte great distance. This is exactly what
we should expect if their abnormal color is due to selective
absorption. The most bluish stars listed by Hertzsprung, on
the other hand, are mostly found in higher galactic latitudes.
The existence of a selective absorption in low galactic lati-
tudes should make itself felt in the statistical study of the mag-
nitudes and color indices of faint stars in Milky Way regions.
With decreasing apparent brightness the lower and upper limits
of observed color indices should gradually shift toward the red.
In other words, among the fainter stars we should find none of
blue-white color, but on the other hand a number of unusually
large color indices (>2ψ0). A phenomenon of this kind is

10
P. 735.
1. 204 and 217, 1923.

© Astronomical Society of the Pacific · Provided by the NASA Astrophysics Data System
 224 PUBLICATIONS OF THE

indeed noticeable in the majority of observations so far made,21

but this subject will be taken up more fully in a later paper.
There remains the task of reconciling our results on selec-
tive absorption with Shapley's observations of small color in-
dices in distant globular clusters and with similar observations
of Hubble in spiral nebulae. To avoid any contradiction we must
again draw the conclusion that the medium responsible for the
observed general and selective absorption is limited to our
stellar system and does not extend through extragalactic space.
Moreover, it must be highly concentrated to the galactic plane,
as most of the observed absorption effects are found in low
galactic latitudes, and van de Kamp's figures in fact indicate a
rapid decrease in selective absorption for higher galactic lati-
tudes. When we consider that two-thirds of all open star clus-
ters lie within 100 parsecs of their plane of concentration, it
does not seem unreasonable to assume that the absorbing me-
dium has a similar distribution, forming a relatively thin disk
or sheet extending along the galactic plane but thinning out
very rapidly at greater distances from this plane. Since the
globular clusters and spiral nebulae observed for color are situ-
ated in higher galactic latitudes, their light in reaching us passes
only the short distance of a few hundred parsecs through the
absorbing medium and will not be much affected. Whether this
medium should be thought of as a contihuum or whether it is
composed of local aggregations of which we observe the sta-
tistical effect, cannot be decided at present.
By combining the numerical data of general and selective
absorption, we derive the absorption coefficients for the photo-
graphic and visual region of the spectrum:

Absorption Coefficients per 1000 Parsecs

Rayleigh Non-selective
Observed Scattering Residual
Photographic (λ = 4300) κ2, = 0Ι?7 0^51 0^19
Visual (λ = 5500) κν = 0.38 0.19 0.19
Selective absorption = 0.32 0.32

21
See, e.g., Lick Obs. Bull., 14, 122, 1929. Dr. Krieger's color indices, however,
need a zero correction of about +0^2 because they were calibrated by means of
spectral types of faint stars before any selective absorption effect was suspected.

© Astronomical Society of the Pacific · Provided by the NASA Astrophysics Data System
 ASTRONOMICAL SOCIETY. OF THE PACIFIC 223

The space absorption decreases rapidly with increasing wave-

length, similar to the extinction in the Earth's atmosphere. This
suggests its interpretation as Rayleigh scattering, i.e., scattering
of light by particles which are small compared with the wave-
length of light. According to Rayleigh such scattering is in-
versely proportional to the fourth power of the wave-length.
This law, however, is not exactly fulfilled by our figures. If we
determine its numerical coefficient so as to fit the selective ab-
sorption, which is the more accurate of the two, we have a
small non-selective residual effect left which, if real, must be
explained by other causes (electron scattering, obstruction by
larger meteoric particles). It will be of great interest to test
by spectrophotometric measures whether the selective absorp-
tion actually follows the inverse fourth-power law. If we admit
it tentatively, we can use Rayleigh's formula to draw some con-
clusions on the constitution of our scattering medium. We shall
first calculate its space density, making an assumption about the
mass of the individual particles. It is natural that we should try
free atoms with an average atomic weight of 40 (like calcium
atoms). In this case about 1800 atoms per cubic centimeter are
required to produce the observed effect; this corresponds to a
space density of 1700 times the Sun's mass per cubic parsec.
Such a figure is quite inadmissible ; it would lead to an enormous
mass for our stellar system, and this in turn would by its gravi-
tational action produce much greater stellar velocities . than
those actually observed. In fact, we can say that the latter do
not admit a greater space density than about half a Sun's mass
per cubic parsec. If we adopt this limiting value for the density,
we can proceed in the reverse order and find that the average
mass of the particles would have to be of the order of 2 X 10-19
grams (3400 calcium atoms). Such extremely fine solid dust
particles with a diameter of about one-hundredth of the wave-
length of visual light would still be small enough to produce
Rayleigh scattering. We see, thus, that our numerical results
for the selective absorption cannot be traced to Rayleigh scat-
tering by free atoms in interstellar space ; they admit, however,
interpretation as scattering by fine cosmic dust.
It seems very probable that there is some relation between

© Astronomical Society of the Pacific · Provided by the NASA Astrophysics Data System
 226 PUBLICATIONS OF THE

the medium causing our general and selective absorption on the

one hand, and interstellar calcium and obscuration effects on the
other hand. Although we have so far made only a beginning in
the study of space absorption and interstellar matter, it may be
helpful for future research to formulate a tentative working
hypothesis which attempts to incorporate and co-ordinate the
various observational data available, and utilizes the interesting
theoretical studies by Eddington.22
Our Milky Way system seems to contain a considerable
amount of finely divided matter, noticeable by its absorption of
light. This matter appears to be made up mainly of :
1. Free atoms {Ca, Ν a, and probably others) causing inter-
stellar (stationary) absorption lines observable in the spectra of
distant stars. Eddington estimates their space density of the
order of 10~ 24
grams per cubic centimeter (one Η atom per
cubic centimeter) and shows that this is not sufficient to origi-
nate an observable amount of Rayleigh scattering.
2. Free electrons are likely to be included, since the observed
interstellar calcium atoms are ionized.
3. Fine cosmic dust particles of various sizes (average mass
of particle 10-19 grams or lárger, space density of the order of
10"23 grams per cubic centimeter) maintained in space by light
pressure of the stars and prodijpng the observed selective ab-
sorption by Rayleigh scattering.
4. Perhaps we should add also larger meteoric bodies, ob-
structing light of all wave-lengths equally, which may be re-
sponsible for a small part of the general absorption (residual
effect).
This absorbing medium is limited to our galactic system,
forming an essential feature of it; it is much concentrated to
the galactic plane extending along the latter like a thin disk
probably not more than a few hundred parsecs thick. While
its distribution follows the Milky Way in general, it is not
necessarily uniform. The observed obscuration of globular
clusters and spiral nebulae near the galactic circle then follows
as a natural consequence of the great depth of the medium in

22
Bakerian Lecture, Proc. Roy. Soc., Ill, 1926.

© Astronomical Society of the Pacific · Provided by the NASA Astrophysics Data System
 ASTRONOMICAL SOCIETY OF THE PACIFIC 227

this direction. The so-called dark nebulae or obscuring clouds

seem to be of incomparably greater opacity, and it is as yet
uncertain whether their absorption is selective or not. As they
are also most prominent in the Milky Way, they may represent
strong local condensations of the general absorbing medium or
of some of its above-mentioned constituents.

Mount Hamilton
July 19, 1930

© Astronomical Society of the Pacific · Provided by the NASA Astrophysics Data System