344 
and lowest at Capetown. Visual measurements of the 
sky near Polaris, reported by Fessenkoff [14], indicated 
that the brightness increased by a factor of 4 from 
latitudes +44° to +80°, but the result may have been 
influenced by polar aurorae; it is not in accord with the 
observatiens of Fig. 3. 
Variation with Zenith Angle. The variation of the 
visual brightness of the night sky with zenith angle is 
also shown by the data of Fig. 3, which as has been 
pointed out refer mainly to the oxygen green line 5577 A. 
It is seen that the variation of the brightness with 
zenith angle Z was about the same for latitudes from 
—17° to +63°, and that the brightness increased from 
about 130 muL at the zenith to a maximum of about 
230 muL at about 15° above the horizon for a clear 
atmosphere; with slight haze the maximum near the 
horizon became less pronounced, and with more haze it 
disappeared [17]. The variation was observed at the 
Pic du Midi Observatory, France, to be somewhat dif- 
ferent for red and green wave lengths [1], the sky inten- 
sity observed through a red filter being about 2.2 times 
as bright at Z = 80° as at the zenith, and through a 
ereen filter being about 1.7 times as bright at Z = 80° 
as at the zenith. The difference was probably due to the 
greater scattering of the lower atmosphere for green 
wave lengths than for red wave lengths. 
Altitude 
The altitudes of the regions in which the high atmos- 
pheric emissions originate are not known with certainty. 
The method which has been used has been well de- 
scribed [17] and need not be detailed here. It is based on 
the observed variation of the night-sky intensity with 
zenith angle, and when applied to the night-sky ob- 
servations led to erratic and discrepant values from 
fifty to several hundred kilometers. The uncertainties 
were due to nonuniformity of the emissions and to in- 
completely worked-out corrections for atmospheric at- 
tenuations. But even with better correction formulas no 
improved or more trustworthy values of the altitudes 
were determined [17]; and in no case was the atmos- 
pheric attenuation measured at the same time that the 
sky intensity was observed. Swings [26] stated the 
situation thus: 
There are great irregularities in the distribution of the emis- 
sion over the sky. These irregularities appear consistently 
when simultaneous exposures in different azimuths at the 
same zenith distance are compared. No layer of uniform 
brightness exists, but rather a set of bright clouds, which move 
around and change brightness. The only remaining hope is 
that, by taking a sufficiently large number of observations, the 
erratic fluctuations will average out. 
From 1941 to 1944 Dufay and Tcheng Mao-Lin [11] 
in France made several series of measurements with 
their spectrograph of the ratios of the intensities of 
various wave lengths at the zenith to the intensities near 
the horizon. From averages of the observations with 
approximate corrections for estimated atmospheric at- 
tenuation, they obtained an altitude of 103 km for 
5577 A, 80 km for 5892 A, and 181 km for 63800 A. In 
THE UPPER ATMOSPHERE 
1948 Roach and Barbier [22] carried out surveys of the 
night-sky intensity in California with recording photo- 
cell equipment and interference filters for isolating 
narrow regions of the spectrum. By using average 
values, with correction for scattered starlight and esti- 
mation of atmospheric attenuation, they obtained an 
altitude of 100 km for 5577 A and about 300 km for 
5892 A. Plans are under way to carry out direct photo- 
cell measurements from rockets to determme whether 
the nocturnal radiations arise in regions accessible to 
the rockets. 
The altitude of the yellow sodium line 5892 A in 
twilight was determined in another way based on the 
fact that these radiations decreased suddenly in in- 
tensity at twilight when the region in which they 
originated was shielded by atmospheric ozone from the 
direct ultraviolet rays of the sun. Calculations made by 
Penndorf [20] from the observations of Vegard and 
Tensberg [27] in February and March 1939, at Troms6 
and Oslo, indicated definitely that the altitude of the 
sodium emission at twilight in those places was in the 
region from 80 to 116 km. 
At present one should keep in mind the probability 
that the various night-sky radiations may arise at differ- 
ent levels, and that the levels may change in the course 
of the night and may vary with the latitude and season. 
Further, it seems reasonable to think that the night-sky 
emissions originate mainly in the atmospheric levels of 
the aurorae and the ionosphere, that is, from about 80 
to 300 km, for it is in this region that strong photo- 
chemical effects occur. 
Theory 
The spectral identifications show that most of the 
night-sky emissions come from the two most abundant 
gases of the atmosphere, oxygen and nitrogen; a small 
portion comes from sodium. No one has questioned the 
general notion that the nocturnal emissions derive their 
energy from the sun, but no processes or quantitative 
details have been agreed upon because information is 
lacking about radiation transitions and atmospheric 
composition. In fact, without exception, all theoretical 
processes, qualitative or quantitative, thus far proposed, 
have been weighted down with a wealth of criticism [6]. 
The excitation energy of the line systems 5577 O 1, 
6300 O 1, 5893 Na 1 are 4.2, 2.0, and 2.1 ev, respectively; 
and of the band systems Vegard-Kaplan N2, Herzberg 
O2, Schumann-Runge O2, and Lyman N» are 7.0, 4.7, 
6.2, and 8.7 ev, respectively. Three theoretical sugges- 
tions for the source of the energy in the high atmosphere 
are that it arises from (1) the association energies of 
atomic oxygen and nitrogen which are 5.09 and 7.38 
ev, respectively; (2) the energies of ionization, which are 
about 13.5 and 14.5 ev for the first ionization potentials 
of atomic oxygen and nitrogen, respectively; (3) parti- 
cles of solar origin or particles swept up from the dust of 
interplanetary space. It appeared that (1) was sufficient 
for the line systems but for none of the band systems 
except that of Herzberg (perhaps the Shumann-Runge 
and the Lyman band systems need no longer be con- 
sidered because, as has been said, they are not listed in 
