INFRARED RADIATION OF THE NIGHT SKY AND NITROGEN DISSOCIATION IN THE IONOSPHERE

I. A. Khvostikov

Discovery of a new radiation. — Determination of the wavelength of the new radiation. — On atomic nitrogen: the results of Dophe, Bernard, and Kaplan. — Bernard and Dophe’s discussion. — The spectrum of polar auroras at low latitudes. — The probability of forbidden-line radiation. — On the dissociation of nitrogen in the upper layers of the atmosphere. — The degree of dissociation of nitrogen molecules in the region of polar auroras. — On the infrared radiation of atomic nitrogen. — Identification of the infrared radiation of the night sky. — The (0,0) band of the first positive system of nitrogen.

1. DISCOVERY OF A NEW RADIATION

In 1944–1945, a report was published by the American astrophysicists Stebbins, Whitford, and Swings1,2 on the presence of a previously unknown infrared radiation of the night sky. This discovery is undoubtedly of primary importance. It points to the existence of new, important properties in the glow of the upper layers of the atmosphere (the luminosity of the night sky). The infrared radiation of the night sky, possessing remarkably high intensity, creates almost insurmountable difficulties for photometry of nebulae and faint stars in infrared rays. In order that the stellar world not be completely closed to astronomers in certain regions of the infrared spectrum, the development of complex special methods will be required. Finally, the discussion of the nature of the new glow is closely connected, as we shall see below, with the important question of the composition and physical state of the air in the upper layers of the Earth’s atmosphere.

In 1940, Stebbins and Whitford at the Mount Wilson Observatory began a program of spectrophotometry of stars and nebulae with the aid of a photoelectric apparatus sensitive to visible and infrared rays. The photometry was carried out in 6 spectral zones from 3530 Å to 10,300 Å, separated by light filters. The photocell was placed at the focus of a 60- and 100-inch reflector; moreover, for measuring the brightness of extended objects, the photocell was stopped down so that the field of view was 8′.6 for the 60-inch and 5′.1 for the 100-inch reflector.

At the very first observations of nebulae (this was nebula M 31) through an infrared light filter (effective wavelength of the transmitted portion of the spectrum \(10\,300\ \text{\AA}\)), the galvanometer deflections proved unexpectedly large. Control pointings of the instrument at neighboring dark regions of the sky (to determine the magnitude of the background) showed that intense infrared radiation of the night sky is present everywhere, making impossible any measurements of objects with such a small surface brightness as nebulae. Even the measurement of faint stars using the smallest diaphragms proved difficult. The measurements given in Table I may serve as an example. These data refer to a star of type dF7; stars of this type give approximately the same galvanometer deflections in all regions of the spectrum (with the given light filters and photocell). The figures in the row “sky” were obtained with a larger diaphragm than the figures in the row “star”; the diaphragms were selected so that in the visible part of the spectrum approximately equal galvanometer readings were obtained both from the star and from the neighboring dark region of the sky.

Table I

Comparison of a star and the background of the night sky

Filter	Ultraviolet \(\lambda=3530\ \text{\AA}\)	Violet \(\lambda=4220\ \text{\AA}\)	Blue \(\lambda=4880\ \text{\AA}\)	Green \(\lambda=5700\ \text{\AA}\)	Red \(\lambda=7190\ \text{\AA}\)	Infrared \(\lambda=10300\ \text{\AA}\)
Star Galvanometer deflection in mm	8	8	10	8	8	8
Sky Galvanometer deflection in mm	8	5	5	8	13	112

The intensity of the infrared radiation of the sky cannot fail to be surprising, so great is it. As is known, the spectra of the night sky, carefully studied over the course of 30 years and described in detail in their time³ in the pages of Advances in the Physical Sciences, contain two lines of especially great intensity: these are the green line \(\lambda=5577\ \text{\AA}\) and the red line \(\lambda=6300\ \text{\AA}\). Both are emitted by atomic oxygen in the terrestrial atmosphere (forbidden transitions) in layers situated, apparently, no lower than 120 km.

These lines have always attracted attention by their intensity: with a fast spectrograph aimed at any part of the sky, they can be photographed on sensitive panchromatic plates in a few minutes. The exposure must be increased by 5–10 times or more in order to obtain any other lines or bands on the spectrogram. The newly discovered infrared radiation of the night sky exceeds the green line in intensity, as Table I shows, by more than an entire order of magnitude.

How could it happen that such powerful radiation of the night sky remained unnoticed for so long? The reason is that, until very recently, light receivers with sufficiently high sensitivity in the corresponding part of the infrared spectrum (around \(1\mu\))—the sensitivity needed for measuring very weak radiations of the night sky—had not been used.

Recently, however, the French astrophysicists Dufay and others pointed to an article of theirs, published in Cahiers de Physique in 1942, in which, allegedly, the existence of the new infrared radiation of the night sky had already been reported. However, the remark made there is rather indefinite. Much earlier, L. A. Kubetsky noted the strong infrared radiation of the night sky; in 1939, in Crimea, together with the Optical Laboratory of the Geophysical Institute of the Academy of Sciences of the USSR, he carried out measurements of the sky glow by means of the “integral-balance” method he had developed, using the secondary-electron “Kubetsky tube.” Unfortunately, at that time L. A. Kubetsky confined himself to qualitative observations, since in 1939 the new method he had indicated, which opened up great possibilities in the measurement of the weakest radiations, was still at the initial stage of development.

2. DETERMINATION OF THE WAVELENGTH OF THE NEW RADIATION

Initially, Stebbins, Whitford, and Swings assumed that the infrared effect was caused by the line \(8600\ \text{\AA}\), which had been discovered by Slipher[^4]. But this idea had to be abandoned, since calculations showed that in this case, according to the transmission curves of the light filters, the galvanometer deflection with the red filter should have been greater than with the infrared one. The use of a spectrograph to determine the wavelength of this radiation is made very difficult by the small dispersion of fast spectrographs and the low sensitivity of photographic plates in the infrared part of the spectrum. By an ingenious method, Stebbins, Whitford, and Swings succeeded in determining the wavelength with sufficient accuracy, using only a photocell and light filters. The method employed by these authors, feasible only with that high photometric precision,

which is ensured in careful photoelectric measurements. For a Schott UG6 filter, 2 mm thick, the transmission is:

\[ \begin{array}{rcl@{\qquad}rcl} 8500\,\text{\AA} & \ldots & 0.01 & 11500\,\text{\AA} & \ldots & 0.79\\ 9500\,\text{\AA} & \ldots & 0.25 & 13000\,\text{\AA} & \ldots & 0.92\\ 10500\,\text{\AA} & \ldots & 0.59 \end{array} \]

If the intensity of the sky radiation is measured first through this filter, and then through 2 and 3 layers of the same filter, then from the change in intensity one can already judge approximately the wavelength of the radiation.

Indeed, if, for example, \(\lambda = 9500\,\text{\AA}\), then upon doubling the thickness of the same light filter the intensity should fall by a factor of 4, whereas for \(\lambda = 10500\,\text{\AA}\) it should fall by less than a factor of two. Preliminary measurements of this kind for the new infrared radiation of the night sky yielded a value of about \(10400\,\text{\AA}\). After this, more accurate measurements were made, for which the light filters were calibrated with the aid of three radiations of known wavelength: \(10336\,\text{\AA}\), \(10407\,\text{\AA}\), and \(10478\,\text{\AA}\). Table II gives the results of the calibration of the UG6 light filters, where 2/1, 3/1, and 3/2 denote, respectively, the ratio of the galvanometer deflections when measuring through 2 layers of the light filter to the deflection with 1 layer, through 3 layers to 1 layer, and through 3 layers to the deflection with 2 layers.

Table II

Calibration of filters

	\(10336\,\text{\AA}\)	\(10407\,\text{\AA}\)	\(10478\,\text{\AA}\)
Measured ratio \(^{2}/_{1}\)	\(0.508 \pm 0.024\)	\(0.564 \pm 0.010\)	\(0.589 \pm 0.020\)
Measured ratio \(^{3}/_{1}\)	\(0.264 \pm 0.000\)	\(0.318 \pm 0.011\)	\(0.361 \pm 0.012\)
Calculated ratio \(^{3}/_{2}\)	\(0.520 \pm 0.025\)	\(0.564 \pm 0.022\)	\(0.614 \pm 0.029\)

Measurements of the night sky through these same light filters gave:

\[ \begin{aligned} 2/1 &\ldots 0.565 \pm 0.007,\\ 3/1 &\ldots 0.348 \pm 0.005, \end{aligned} \]

and after dividing the second by the first:

\[ 3/2 \ldots 0.616 \pm 0.012. \]

From these data the authors concluded that the new infrared radiation of the night sky is an emission line (or narrow band) with wavelength:

\[ \begin{array}{lr} \text{From } 2/1 \ldots\ldots\ldots\ldots & 10\,427\ \text{\AA} \\ \text{From } 3/1 \ldots\ldots\ldots\ldots & 10\,452\ \text{\AA} \\ \hline \text{Mean value} \ldots & 10\,440 \pm 25\ \text{\AA} \end{array} \]

Concerning this final value of the wavelength, Stebbins, Whitford, and Swings write: “We have found nothing better than to take the mean of the two values, and to regard the probable error as equal to the difference between them.”

How can the new line be identified? In the question of identification, an unusual situation had arisen, both in substance and with respect to the position taken by Stebbins, Whitford, and Swings themselves. Into this question there entered another, no less important question: the existence in the upper layers of the atmosphere of dissociated (atomic) nitrogen. We must dwell on this point specifically; moreover, the discussion that has developed in recent years concerning the presence of atomic nitrogen is of independent interest and has fundamental significance for the physics of the upper atmospheric layers.

3. ON ATOMIC NITROGEN: RESULTS OF DUFAY, BERNARD, AND KAPLAN

It is known⁵ that the study of the spectra of the luminosity of the night sky and of aurorae leads to the conclusion that the upper layers of the atmosphere, like those near the ground, have the nitrogen-oxygen composition of air. But if the spectra just mentioned indicate the presence of nitrogen in a molecular state, then the oxygen radiation belongs to the atomic state. The theory of the ozone layer leads to definite ideas about the dissociation of oxygen molecules by ultraviolet rays⁵. Apparently, above 100 km oxygen is almost entirely dissociated.

Under such circumstances it would surprise no one if it turned out that nitrogen too is present in the upper layers of the atmosphere in a state of greater or lesser dissociation, and that therefore lines of atomic nitrogen \(N_1\) are found in the spectra of the night sky and aurorae.

At one time, in the thirties, indications appeared of the possibility of connecting certain lines of the spectrum of the night sky and aurorae with the radiation of atomic nitrogen⁶. But a more careful study of the data showed that each such identification either had to be rejected, or, in any case, could be made only very tentatively. Gradually it became

...the impression began to take shape that, contrary to initial expectations, atomic nitrogen does not reveal itself in the spectra of the night sky and auroras. But the idea of atomic nitrogen did not leave many researchers, although it had been established that under the action of ultraviolet radiation the molecule \(N_2\) is more easily ionized than dissociated. Of course, this is an essential question for the physics of the upper layers of the atmosphere. The search for lines of atomic nitrogen in fact continued. In 1939 the well-known investigator of the luminosity of the night sky, the French astrophysicist Dufay, published an article in which he asserted that lines of atomic nitrogen are present in the glow of the sky \(^{7}\). This article provoked a sharp discussion; opinions diverged. The discussion has not ended to this day. Dufay and his supporters insist on their results; many other specialists reject them completely.

Fig. 1. Scheme of the energy levels of the nitrogen atom \(N_1\).

The nitrogen atom has, above the ground level \({}^{4}S\), two metastable levels \({}^{2}D\) and \({}^{2}P\), whose exact position remained unknown for a long time. In 1938 a scheme of energy levels was rigorously established \(^{8}\) (Fig. 1), according to which the forbidden transitions from the \({}^{2}P\) and \({}^{2}D\) levels are accompanied by the emission of lines with wavelengths:

\[ \begin{aligned} {}^{2}P &\to {}^{4}S \qquad &&\lambda = 3466.5\,\text{\AA},\\ {}^{2}D^{5}_{1/2} &\to {}^{4}S \qquad &&\lambda = 5200.1\,\text{\AA},\\ {}^{2}D^{3}_{1/2} &\to {}^{4}S \qquad &&\lambda = 5197.8\,\text{\AA},\\ {}^{2}P &\to {}^{2}D \qquad &&\lambda = 10400\,\text{\AA}. \end{aligned} \]

Dufay \(^{7}\) initially used the old (1929) data of Compton and Boyce \(^{9}\), according to which the wavelengths of the indicated lines had somewhat different values: \(3470\,\text{\AA}\); \(5206\,\text{\AA}\) (a doublet with a separation of about \(2\,\text{\AA}\)) and \(10407\,\text{\AA}\).

As early as 1934 Dufay, having for the first time obtained the spectrum of the night sky in the ultraviolet region \(^{10,11,12}\) for \(\lambda < 3600\,\text{\AA}\), discovered an intense line at \(3472\,\text{\AA}\), which at that time he attributed to the band \((3—4)\) of the second positive system of molecular

of nitrogen, \(\lambda = 3469\ \text{Å}\). Later measurements by Gauzit \(^{13,14,15}\), who obtained many spectrograms of the night sky on two different spectrographs, gave the value \(\lambda = 3471\ \text{Å}\). In 1936 Kaplan suggested \(^{16}\) that this line belongs to atomic nitrogen and corresponds to the transition \({}^{2}P \to {}^{4}S\). This hypothesis soon seemed to receive support as a result of Kaplan’s laboratory experiments \(^{17}\), in which he observed the 3471 Å line in the afterglow spectrum of “active” nitrogen; under the same conditions the intensity of the bands of the Vegard–Kaplan system was very great. In 1938 Bernard reported that in the spectra of certain diffuse aurorae which he photographed in Tromsø, a line at 3470.3 Å appears simultaneously with the Vegard–Kaplan bands; in Bernard’s opinion, it cannot be identified with the 3469 Å band, the second positive system of molecular nitrogen \(^{18}\).

But the refinement of the energy levels of atomic nitrogen changed the wavelengths for the three indicated forbidden lines: the \({}^{2}P \to {}^{4}S\) line should have a wavelength of 3466.5 Å, not 3470 Å. Kaplan \(^{19}\), repeating his measurements of the line under laboratory conditions, obtained, using a spectrograph of greater dispersion, the more accurate value 3466.3 Å, and Bernard \(^{20}\), correcting his original measurements, found in turn \(3466.5 \pm 1\) Å. Thus the presence of the \({}^{2}P \to {}^{4}S\) line of atomic nitrogen in certain types of aurorae was confirmed. (In 1945 Bernard \(^{22}\), replying to the critical remarks of Vegard \(^{21}\), who had expressed doubt as to the possibility of obtaining an accuracy of \(\pm 1\) Å on a spectrogram with such small dispersion, repeated the measurement of these spectrograms and reported that the wavelength is \(\lambda = 3466.5 \pm 0.5\) Å.)

However, the presence of this line in the spectra of night-sky luminosity remained problematic. Bernard allowed \(^{20}\) that the discrepancy of 5 Å for the night-sky line might be the result of a large error in Duffay’s measurements of night-sky spectrograms, but Duffay rejects such an assumption, referring \(^{7}\) to the fact that comparison of his wavelength measurements for several dozen lines in the ultraviolet region with the measurements of Gauzit and Arnulf \(^{23}\) showed agreement for most of the lines within 1 Å. Duffay considers it more probable that the strong 3471 Å line of the night sky cannot be identified with the \({}^{2}P \to {}^{4}S\) line, and that the latter is either altogether absent from the night-sky spectrum or is extremely weak.

Since the \({}^{2}P \to {}^{4}S\) line of atomic nitrogen has been found in auroral spectra, it should be assumed that the \({}^{2}D \to {}^{4}S\) line may also be present. But Bernard pointed out that such a line is charac-

when it is not observed,^20 which is also confirmed by Vegard’s photographs. But as early as 1929 Slipher and Sommer reported that in the spectrum of an aurora, which they observed on June 7, 1928, at Flagstaff^24 (Arizona), a line of 5206 Å had been found, which they tentatively identified with the line \({}^{2}D \to {}^{4}S\). If, according to the new data (see above), this line should have a wavelength of 5199 Å (the mean for the doublet), then an error of 7 Å is not, in Dufay’s opinion,^7 excessive, if one takes into account the small dispersion in this part of the spectrum of the fast \((F : 1.1)\) spectrograph used by Slipher and Sommer. For a long time, however, the observation of Slipher and Sommer remained the only one.

In 1939 Babcock published^25 the results of his measurements of the night-sky spectrum, made at the Lick Observatory and at Mount Wilson Observatory, but without any identification. In these materials Dufay found data^7 that “forced him to return to the earlier question.” On Babcock’s registergram, between 5350 Å and 5150 Å, there are the lines:

\[ \begin{array}{rcl} 5317\ \text{Å}, & \text{intensity} & 1\\ 5242\ \text{Å}, & \text{”} & 1\\ 5194\ \text{Å}, & \text{”} & 2 \end{array} \]

The dispersion is \(385\ \text{Å}/\text{mm}\). “Under these conditions,” writes Dufay,^7 “a measurement error of \(0.01\ \text{mm}\) gives an error of 4 Å for the wavelength... The line 5194 Å coincides, within the limits of the measurement error, with the line \({}^{2}D \to {}^{4}S\) of atomic nitrogen... This numerical coincidence is not a sufficient basis for a confident identification. But here is an important circumstance: Babcock marked an intensity of 2 for the line 5194 Å on only one of four spectrograms, and this spectrogram was taken during an aurora (a low-latitude aurora), visible during part of the exposure period. On the other three spectrograms the intensity of the line is much less. It seems very probable that this line is usually weak in the night-sky spectrum, but is enhanced during auroras.”

In this region of the spectrum there are many bands of the first positive system of molecular nitrogen, one of which \((17-2)\) is very close in wavelength (5196 Å) to the line observed by Babcock (5194 Å). In the cited paper Dufay gives a number of considerations against the supposition that the line observed by Babcock corresponds to the indicated band. His general conclusion is as follows:^7

“Thus, twice in the spectrum of low-latitude auroras radiation has been observed, close in position to the line of atomic nitro-

of the $^2D \to {}^4S$ line. The deviation from the position of this line (approximately $+7\ \text{\AA}$ and $-5\ \text{\AA}$) does not seem incompatible with the accuracy of the measurements, and no other interpretation of this radiation has been found. The presence of the $^2D \to {}^4S$ line in the spectrum of some auroras may be regarded, at least, as quite probable, but it requires confirmation by further observations.

This line may also appear in the spectrum of the night sky, according to Babcock’s observations, but usually it is very weak.

This opinion is supported by what was said above about the $^2P \to {}^4S$ line, which has not been detected in the spectra of the night sky, but whose presence in the spectra of certain types of diffuse polar auroras was established by Bernard. The intense $3471\ \text{\AA}$ night-sky line therefore has another origin.

4. THE BERNARD—DUFOUR DISCUSSION

In 1941 Bernard published a long paper[^26], “Atomic Nitrogen in the Upper Layers of the Atmosphere,” in which serious objections were raised to many of Dufour’s arguments.

In Kaplan’s laboratory experiments mentioned above (§ 3)[^17], an afterglow of nitrogen was observed, in whose spectrum, among many bands of the Vegard–Kaplan system of molecular nitrogen, the line[^19] $\lambda = 3466.3\ \text{\AA}$ was observed, which Kaplan identified with the atomic-nitrogen line $^2P \to {}^4S$. Bernard, as we already know, found this line in the spectra of several diffuse polar auroras that he studied in Norway; according to his determination[^27] $\lambda = 3466.5 \pm 1\ \text{\AA}$. Subsequently—

Table III

Increase in the intensity of certain lines and bands in the spectrum of polar auroras in the upper part of the auroras relative to the lower part (according to Bernard)

Wavelength (in Å)	Identification	Intensity (in arbitrary units), lower part of auroras	Intensity (in arbitrary units), upper part of auroras
3914	$\mathrm{N_2^+}$ (negative system)	200	200
3785	$\mathrm{N_2}$ Vegard–Kaplan (1—11)	0	5
3603	$\mathrm{N_2}$ Vegard–Kaplan (0—10)	0	4
3503	$\mathrm{N_2}$ Vegard–Kaplan (2—11)	0.5	6
3466.5	$\mathrm{N_1}\ {}^2P \to {}^4S$	1.2	14
3428	$\mathrm{N_2}$ Vegard–Kaplan (1—10)	1	9

Bernard established that this line is not connected only with the indicated type of aurorae, but that it is present in the spectra of all polar aurorae. Comparing spectrograms corresponding to the lower and upper edges of aurorae with well-pronounced outlines (arcs and draperies), Bernard noted^28 that the Vegard–Kaplan bands and the line 3466.5 Å, being very weak below, turn out to be relatively intense in the upper part of the aurora. The results of his photometric estimates^26 are collected in Table III. These data testify to the identity and parallelism in the changes of intensity of the Vegard–Kaplan bands and of the line 3466.5 Å. On the average the intensity increases by about a factor of 10 in passing from the lower part of the aurora to the upper.

In connection with attempts to identify the line \(^{2}P \longrightarrow {}^{4}S\) of atomic nitrogen with one of the lines of the night-sky spectrum, Bernard gives a summary of all measurements in this part of the spectrum (Table IV). These are the measurements of Gozi^15, Dufay^10, Arnulf^23, Bernard, Arnulf, Cavassilas and Dejardin^29, Hofmeister^30, carried out during the period 1934–1939.

Table IV

Summary of measurements of the night-sky spectrum in the region of the atomic nitrogen line, \(\lambda = 3466.5\) Å

Gozi Old measurements	Gozi New measurements	Dufay	Arnulf	Arnulf, Bernard, Cavassilas and Dejardin	Hofmeister
3489 Å ?	3490 Å 3	3489 Å 3	3487 Å 2	3488 Å 4	3489 Å 4
3471 4	3471 3	3472 4	. . . . . . .	. . . . . . .	. . . . . . .
. . . . . . .	. . . . . . .	. . . .	. . . . . . .	3468 1	. . . . . . .
3462 1	. . . . . . .	. . . .	3460 1	3462 1	. . . . . . .
3454 1	3456 1	. . . .	. . . . . . .	3456 1	. . . . . . .
. . . . . . .	. . . . . . .	. . . .	. . . . . . .	3452 0	. . . . . . .
3446 ?	3444 1	3446 2	. . . . . . .	3446 1	3448 5

Bernard notes^26 that “only Dufay and Gozi observed an intense line near 3471 Å, whereas all the other investigators reported, on the contrary, a line \(\lambda = 3489\) Å.” Bernard further indicates that

line \(3468\ \text{\AA}\) was observed rarely; it is noted on only three spectrograms out of a total number of several dozens.

Bernard\(^{26}\) casts doubt on the existence of the line \(\lambda = 3471\ \text{\AA}\). Bernard’s general conclusion concerning the line of atomic nitrogen \({}^{2}P \to {}^{4}S\) is as follows: “If one excludes the weak line \(\lambda = 3468\ \text{\AA}\), whose reality cannot yet be guaranteed, not one of the intense lines of the night sky can be attributed to the transition \(N_1({}^{2}P \to {}^{4}S)\).”

Bernard also arrives at a negative conclusion with respect to another line of atomic nitrogen, \({}^{2}D \to {}^{4}S\). The photographic determination of the wavelength \(\lambda = 5206\ \text{\AA}\), made by Slipher and Sommer for a line which they associated with this transition\(^{24}\), was made—as may be judged from the reproduction of the spectrograms—by using, as reference wavelengths, the green line \(5577\ \text{\AA}\) and the bands \(4710—4278\ \text{\AA}\) of nitrogen, both equally overexposed and broadened. Under such conditions, the inaccuracy of the determination may be considerable and does not permit any reliable identification to be made.

Later Vegard\(^{31}\) published a spectrogram of aurorae, in all respects similar to the spectrogram of Slipher and Sommer, but having helium lines as the comparison spectrum. In this paper Vegard carefully discusses the accuracy of the measurements and assigns to the line observed by Slipher, Sommer, and himself the wavelength \(\lambda = 5238\ \text{\AA}\) with an accuracy of \(2—3\ \text{\AA}\). New measurements, made with a much greater dispersion of the spectrograph, did not noticeably change this result: \(5238.5\ \text{\AA}\) in Vegard, \(5241\ \text{\AA}\) in Vegard and Granath. Thus, between the wavelength of the auroral line and the line of atomic nitrogen \({}^{2}D \to {}^{4}S\) there is a discrepancy of \(40\ \text{\AA}\), which rules out any possibility of identification.

As we already know (see § 3), Dufay\(^{7}\) also refers to a spectrogram of the night sky obtained by Babcock\(^{25}\). On this spectrogram the line \(\lambda = 5194\ \text{\AA}\) was measured with good accuracy; Dufay is inclined to identify it with the line of atomic nitrogen \({}^{2}D \to {}^{4}S,\ \lambda = 5199\ \text{\AA}\). But, in Bernard’s opinion, this assumption cannot be accepted for two reasons:

1. The line \(5194\ \text{\AA}\) has another, better-founded identification. Let us assume, together with Dufay, that the error in determining the wavelength is \(4\ \text{\AA}\), and try to analyze the identification.

In the first column of Table V are collected the wavelengths of the auroral spectrum between 5400 and 5000 \(\text{\AA}\). In the second column are given the wavelengths of the night-sky spectrum according to Babcock. In the third column is given the most plausible identification\(^{26}\). On the basis of these data

Table V

Spectrum of auroras and the night sky in the region 5000–5400 Å

Auroras (Bernard)	Night sky (Babcock)	Interpretation
5405 Å 1	......	\( \mathrm{N}_2\) (1 pos.) (11—6) 5406 (5407; 5402; 5388)
5368 1	5374 Å Q	\( \mathrm{N}_2\) (1 pos.) (12—7) 5371 (5373; 5367; 5354)
5330 2	5317 1	\( \mathrm{N}_2\) (\(A \to X\)) (2—16) 5329; \( \mathrm{N}_2\) (1 pos.) (13—8) 5337 (5339; 5334; 5323)
5301 1	5317 1	\( \mathrm{N}_2\) (1 pos.) (14—9) 5303 (5307; 5302)
5275 2	......	\( \mathrm{N}_2\) (1 pos.) (15—10) 5270 (5275)
5236 1	5242 1	\( \mathrm{N}_2\) (\(A \to X\)) (5—18) 5232; \( \mathrm{N}_2\) (1 pos.) (16—11) 5238 (5244)
5211 0.5	......	\( \mathrm{N}_2\) (1 pos.) (17—12) 5216 (5214; 5209; 5196)
5194 2	5194 2	\( \mathrm{N}_2\) (\(A \to X\)) (1—15) 5191.1
5167	......	\( \mathrm{N}_2\) (1 pos.) (18,13) 5175 (5184; 5179; 5167)
5138 1	5142 0	\( \mathrm{N}_2\) (1 pos.) (19—14) 5145
5094 2	......	\( \mathrm{N}_2\) (\(A \to X\)) (4—17) 5094.8
5063 2	5060 1	\( \mathrm{N}_2\) (\(A \to X\)) (0—14) 5061.5

it may be stated that throughout this interval the spectrum is identified well and unambiguously as the bands of the first positive system of nitrogen \( \mathrm{N}_2\) (the sequence \(\nu' - \nu'' = 5\)) and the Vegard–Kaplan bands \( \mathrm{N}\) (\(A \to X\)), whose great significance for auroras and the luminosity of the night sky was established long ago and in an evident way. The radiation 5194 Å should logically be identified with the \(A \to X\) band, \(\lambda = 5191.1\) Å, observed, moreover, by Cabannes (\(\lambda = 5185\) Å) in the spectrum of the night sky \(^{32}\).

1. Babcock’s spectrograms do not contain the line \({}^{2}P \to {}^{4}S\), \(\lambda = 3466.5\) Å. According to theoretical data (transition probabilities, see the next paragraph), it is impossible to observe the line \({}^{2}D \to {}^{4}S\) without the line \({}^{2}P \to {}^{4}S\) being simultaneously found with much greater intensity. Babcock’s table of wavelengths contains data down to 3548 Å; however, his spectrograms show still very noticeable blackening down to 3300 Å, and their examination reveals no intense line in the entire interval from 3400 to 3500 Å. “Under such circumstances,” writes

Bernard\(^{26}\),—in any case, the line in the region of 5200 Å cannot be identified with the transition \(^{2}D \to {}^{4}S\).

Let us note that in the detailed work by Elvey, Swings, and Linke on the spectra of the night sky, published in 1941, it is stated\(^{38}\): “Recently many considerations have been published concerning the presence in the spectra of the night sky of the forbidden line of atomic nitrogen \(^{2}P \to {}^{4}S\), \(\lambda = 3466.5\) Å. But undoubtedly in our spectrograms there is no line of any noticeable intensity with such a wavelength, since our line \(\lambda = 3460 \pm 1\) Å is narrow, and an error of 6.5 Å is excluded in this part of the spectrum.”

It should also be added that in 1939 Kaplan obtained, under laboratory conditions, the line \(^{2}D \to {}^{4}S\), but at the same time he observed with relatively high intensity (“with high relative intensity”) the line \(^{2}P \to {}^{4}S\), which is theoretically more probable\(^{34}\).

Bernard’s general conclusions are as follows\(^{26}\):

a) The forbidden lines of atomic nitrogen are not observed in the spectra of the night-sky glow.

b) In aurorae only the line \(^{2}P \to {}^{4}S\), \(\lambda = 3466.5\), has been identified.

In 1943 Dufay again appeared with a major work on this question\(^{37}\). He notes that there is agreement in the views of himself and Bernard in the sense that the line \(^{4}S \to {}^{2}P\) clearly does not reveal itself in the spectra of the night sky. “From Bernard’s text,” writes Dufay, “one may conclude that I hold another opinion, but it is enough to refer to my first work to become convinced that this is not there at all. The possibility of attributing the 3471 Å line to the spectrum of the night-sky glow was proposed by Kaplan\(^{16}\) and Gauzit\(^{41}\).”

Concerning Bernard’s doubts about the presence of the 3471 Å line in the night-sky spectrum, Dufay points out that his repeated measurements, together with Dejardin\(^{42}\), of all available spectrograms in the ultraviolet region confirmed the presence of a fairly intense line (intensity 2 on a 5-point scale) with wavelength 3469 Å among a complex group of lines in the region of the strong radiation 3488 Å (intensity 4). It is not impossible, writes Dufay\(^{37}\), that this line is \(^{2}P \to {}^{4}S\), but the coincidence should be considered closer for the band \(3 \to 4\) of the second positive system of \(N_2\).

As for the line \(^{2}D \to {}^{4}S\) in the spectrum of the night sky and the possibility of attributing the line 5194 Å observed by Babcock\(^{25}\) either to this transition, as Dufay\(^{7}\) had previously proposed, or to the \(1 \to 15\) band of the Vegard–Kaplan system, as Bernard\(^{26}\) believes, Dufay writes that “Babcock’s line is situated approximately equally (equidistantly)

relative to this molecular band and the atomic line \({}^2D \longrightarrow {}^4S\), so that both interpretations are equally possible (or equally doubtful).”

In any case, Dufay concludes\(^{37}\): “the presence of the atomic nitrogen line \({}^2D \longrightarrow {}^4S\) in the spectra of the luminosity of the night sky has not been established.”

The situation is quite different, in Dufay’s opinion, with regard to spectra of aurorae. On this question, in 1941, new important data were obtained on the spectra of “aurorae at low latitudes,” which we shall now consider.

5. THE SPECTRUM OF “AURORAE AT LOW LATITUDES”

Aurorae are sometimes observed also at lower latitudes. This usually occurs during strong magnetic storms and coincides with an increase in solar activity (for example, at the epoch of the maximum of the 11-year period). The year 1938 was rich in such phenomena, and in that year many interesting observations were made\(^{37}\), initiating the systematic study of “aurorae at low latitudes,” as they came to be called. Measurements showed that their altitude is considerably greater than that of “ordinary” aurorae.

During 1941, twice (on March 1 and September 18), aurorae of rare brightness were observed in middle latitudes. In both cases their spectra were obtained: in southern France by Dufay, Gauzit, and Cheng Mao-Lin\(^{38}\), and in Arosa (Switzerland) by Götz\(^{39}\). In all cases a line near 5200 Å was found.

On March 1 Dufay, Gauzit, and Cheng obtained spectra simultaneously on two spectrographs: one spectrum on a low-dispersion spectrograph (800 Å/mm), from which \(\lambda = 5201\) Å was obtained, and another on a spectrograph with greater dispersion (450 Å/mm), which gave 5206 Å. On September 18 Dufay and Cheng, on these same spectrographs, obtained this line on seven spectrograms, on three of which the wavelength could be measured\(^{40}\). They found (the comparison spectrum consisted of mercury lines): 5199.5, 5198.8, and 5198.6 Å; the mean was 5199.0 Å, with the error in determining the mean considered to be less than 1 Å.

Dufay considers\(^{37}\) the presence of the atomic nitrogen line \({}^2D \longrightarrow {}^4S\) (5199 Å) in the spectra of aurorae at low latitudes, which arise at very great altitudes, to be proved.

Götz, for this same line, obtained on a small quartz spectrograph, first found \(\lambda = 5194\) Å and was inclined to identify it with the calcium line 5189 Å, but subsequent more accurate measurements

brought it to the value \(5198\ \text{Å}\), and he inclined toward identifying it with a line of atomic nitrogen[^39].

Considering that, on the basis of the additional data indicated, the existence of the nitrogen line \({}^{2}D \to {}^{4}S\) had been proved, Dufay, in his last work, again discusses in detail1 the spectra of aurorae obtained on June 7, 1928, by Slipher and Sommer[^24] at Flagstaff (Arizona), in which these authors, as we know, found the line \(5206\ \text{Å}\). Contrary to the opinion of Bernard[^26] and Vegard[^31], Dufay considers it proven from all the comparisons he makes (he reproduces

Fig. 2. Spectrum of an “aurora of low latitudes,” obtained by Dufay and Cheng Mao-Lin on September 18, 1941, in southern France.

in his paper 4 Slipher–Sommer spectrograms, Vegard’s spectrogram, and his own 2 spectrograms of an “aurora of low latitudes” of September 18, 1941; one of these latter is reproduced by us in Fig. 2), that the line observed by Slipher and Sommer coincides with the line \(5199\ \text{Å}\) observed by him, i.e., also belongs to the spectrum of atomic nitrogen.

As we have already indicated, “auroras of low latitudes” are characterized by a very great height, and the final point of view reached by Dufay is thus formulated by him in the résumé of his latest\(^ {37}\) work:

“In auroras the line \({}^{2}P \longrightarrow {}^{4}S\) (3466.5 Å) of atomic nitrogen was identified by Bernard in Norway, and the line \({}^{2}D \longrightarrow {}^{4}S\) (5199 Å)—by the author and Cheng Mao-Lin in France. This latter line has been observed many times in the spectra of ‘auroras of low latitudes.’ Its presence appears to be characteristic of certain types of auroras distinguished by a very great height; in the spectra of these auroras the line \({}^{2}P \longrightarrow {}^{4}S\) is either very weak or altogether absent.”

Such is the factual side of the matter in the important question of the presence of atomic-nitrogen lines in the spectra of auroras and of the night sky. These facts touch upon an important problem in the physics of the upper atmospheric layers, connected with the possible dissociation of nitrogen. In the following paragraphs some questions relating to this will be considered.

6. PROBABILITY OF EMISSION OF FORBIDDEN LINES

Both of the nitrogen lines under discussion, \({}^{2}P \longrightarrow {}^{4}S\) (3466.5 Å) and \({}^{2}D \longrightarrow {}^{4}S\) (5199 Å), correspond to forbidden transitions of the nitrogen atom. This circumstance in itself cannot be regarded as excluding the possibility of emission of these lines by the atmosphere with appreciable intensity, since the brightest lines in the spectrum of the night-sky glow are precisely forbidden lines of oxygen atoms (the green line \({}^{1}S_{0} \longrightarrow {}^{1}D_{2}\), \(\lambda = 5577.3\) Å, and the red triplet \({}^{1}D_{2} \longrightarrow {}^{3}P_{2,1,0}\), \(\lambda = 6300.2—6363.9—6391.7\) Å). However, it does make desirable a more careful theoretical discussion of the question, all the more so because the very presence of atomic-nitrogen lines, unlike the oxygen lines, is, as we have seen, not clear in all respects.

As is known, “forbidden transitions” are not such in the absolute sense, but their probability is much smaller than the probability of ordinary, non-forbidden transitions. The selection rules that determine forbidden transitions are valid only for radiation associated with the dipole moment (dipole radiation). However, along with the dipole moment the atom possesses a quadrupole moment and moments of higher orders, which also give rise to some radiation (quadrupole radiation), though considerably weaker. The probability of transitions corresponding to dipole and quadrupole radiation can be calculated.

The calculation of the probability of emission of the forbidden lines of atomic nitrogen was carried out in 1939 almost simultaneously and independently by Nicol\(^ {43}\) and Pasternack\(^ {44}\). The numerical results of their calculations pri-

are given in Table VI, where \(A\) denotes the transition probability (the number of transitions per 1 sec.); \(1/A\) is, evidently, the lifetime of the given level. On the basis of these data it is easy to calculate the ratio of the intensities of the lines \({}^{2}P \to {}^{4}S\) and \({}^{2}D \to {}^{4}S\). If by \(N_a\) we denote the number of nitrogen atoms in the metastable state \({}^{2}P\) and by \(N_b\) the number of atoms in the metastable state \({}^{2}D\), then the ratio of the intensities

Table VI

Transition probabilities for forbidden lines of atomic nitrogen according to Nicol and Pasternak

Transition	Nicol \(A\)	Nicol \(1/A\)	Pasternak \(A\)	Pasternak \(1/A\)
\({}^{2}P \to {}^{2}D,\ \lambda=10400\ \text{Å}\)	—	—	\(0.507\ \mathrm{sec}^{-1}\)	\(2\ \mathrm{sec.}\)
\({}^{2}D \to {}^{4}S,\ \lambda=5200\ \text{Å}\)	\(3.1\cdot10^{-5}\ \mathrm{sec}^{-1}\)	\(9\ \mathrm{hours}\)	\(5.3\cdot10^{-5}\ \mathrm{sec}^{-1}\)	\(5.3\ \mathrm{hours}\)
\({}^{2}P \to {}^{4}S,\ \lambda=3466\ \text{Å}\)	\(10^{-2}\ \mathrm{sec}^{-1}\)	\(100\ \mathrm{sec.}\)	\(1.25\cdot10^{-1}\ \mathrm{sec}^{-1}\)	\(80\ \mathrm{sec.}\)

of the lines \(\lambda_a=3466\ \text{Å}\) and \(\lambda_b=5200\ \text{Å}\) is determined by the expression:

\[ \frac{I_{3466}}{I_{5200}} = \frac{N_a}{N_b}\cdot \frac{A_{3466}}{A_{5200}}\cdot \frac{\lambda_b}{\lambda_a}. \]

According to Nicol’s data this is:

\[ \frac{I_{3466}}{I_{5200}} = 490\cdot\frac{N_a}{N_b}, \]

and according to Pasternak’s data:

\[ \frac{I_{3466}}{I_{5200}} = 360\cdot\frac{N_a}{N_b}. \]

In 1940 Pasternak published a more complete calculation of the transition probabilities of forbidden lines[^45]. He calculated transition probabilities for the metastable states \({}^{2}P\), \({}^{2}P\), \({}^{4}P\), \({}^{2}D\), and \({}^{3}D\) in connection with the emission of forbidden lines in the spectra of nebulae, in which they play an important role. These data of Pasternak are at present the most accurate; in Table VII we give his data for atomic nitrogen, and also for atomic oxygen, since we shall have to touch upon this latter in connection with the problem under discussion.

According to these data, for the nitrogen lines we have:

\[ \frac{I_{3466}}{I_{5200}} = 269\cdot\frac{N_a}{N_b}. \]

If the number of atoms \(N_a\) in the excited state \({}^{2}P\) were equal to the number of atoms \(N_b\) in the excited state \({}^{2}D\), then the ratio of the inten-

sities of these two lines would be equal to 269. Since in the spectra of aurorae the line 3466 Å has a small intensity, the line 5200 Å, being 269 times weaker, can hardly be detected in the spectrum. But what in reality is the ratio \(\frac{N_a}{N_b}\)? In order to try to form at least some judgment about this, it is necessary to turn to the question of the possible mechanism of formation of nitrogen atoms in the upper layers of the atmosphere, i.e., the mechanism of dissociation of nitrogen molecules.

Table VII

Transition probabilities for forbidden lines of nitrogen and oxygen atoms
(according to Pasternak’s refined data)

Atom	Transition	Wavelength in Å	\(A\) (sec\(^{-1}\))
N\(_1\)	\({}^2P_{3/2}\to{}^4S_{3/2}\)	3466,4	0,0067
N\(_1\)	\({}^2P_{1/2}\to{}^4S_{3/2}\)	3466,4	0,0027
N\(_1\)	\({}^2D_{5/2}\to{}^4S_{3/2}\)	5200,7	0,000021
N\(_1\)	\({}^2D_{3/2}\to{}^4S_{3/2}\)	5198,5	0,000014
O\(_1\)	\({}^1S_0\to{}^1D_2\)	5577,3	2,2
O\(_1\)	\({}^1S_0\to{}^3P_1\)	2972,3	0,090
O\(_1\)	\({}^1S_0\to{}^3P_2\)	2958,3	0,00080
O\(_1\)	\({}^1D_2\to{}^3P_0\)	6391,7	0,0000022
O\(_1\)	\({}^1D_2\to{}^3P_1\)	6363,9	0,0026
O\(_1\)	\({}^1D_2\to{}^3P_2\)	6300,2	0,0078

7. ON THE DISSOCIATION OF NITROGEN IN THE UPPER LAYERS OF THE ATMOSPHERE

Turning to the scheme of the energy levels of the nitrogen molecule (Fig. 3), we see that the dissociation of an unexcited nitrogen molecule, i.e., one in the state \(X\,{}^1\Sigma\), requires an energy of 7.38 electron-volts, which considerably exceeds the dissociation energy of the oxygen molecule (5.09 eV). If the photodissociation of oxygen molecules can be produced by radiation \(\lambda < 2500\) Å, then for the dissociation of nitrogen molecules much shorter-wave radiation is required, \(\lambda < 1680\) Å. Owing to the rapid decrease of the energy of solar radiation toward short wavelengths, dissociation of nitrogen molecules by solar rays must be regarded as much more difficult in comparison with the dissociation of oxygen molecules; the latter, according to Chapman’s well-known photochemical theory of atmospheric ozone and night-sky luminosity, plays the principal role in phenomena in the high layers of the atmosphere\(^{5}\).

Dissociation of excited nitrogen molecules in the state \(A^3\Sigma\) can proceed much more easily: it requires no more than 3.56 electron-volts. Since the state \(A^3\Sigma\) is metastable, the probability of two-step excitation—first with a transition from the normal state \(X\) to the excited state \(A\), and then dissociation—cannot be considered negligible. It should be noted that the Vegard–Kaplan bands \(A^3\Sigma \to X^1\Sigma\) are the most intense in the spectra of the night-sky glow, which directly indicates the presence of a large number of nitrogen molecules in the state \(A\).

Fig. 3. Scheme of the energy levels of the nitrogen molecule \(N_2\).

We arrive at an analogous conclusion as a result of analyzing the form of the potential curves of the nitrogen molecule. It turns out (see, for example, the paper by Gozzi \(^{46}\), published in 1944) that their form makes absorption from the ground level, accompanied by dissociation of the molecule, very improbable (the probability of ionization of the nitrogen molecule considerably exceeds the probability of its dissociation). But the probability of dissociation as a result of absorption from an excited level is considerably greater. The following scheme of the process is conceivable:

\[ \tag{1} N_2(A) + 3.56\ \text{eV} \to N({}^4S) + N({}^2D). \]

The source of the dissociation energy may be, for example, excited oxygen atoms \(O({}^1S_0)\), having an excess energy of 4.18 eV (collisions of the second kind).

However, the reaction

\[ \tag{2} O({}^1S_0) + N_2(A^3\Sigma)_{v=x} \to N({}^4S) + N({}^2D) + O({}^3P) + 0.62\ \text{eV} \]

must be considered improbable for two reasons: a) the short lifetime of the atoms \(O({}^1S_0)\), namely \(1/2\) sec. (see Table VII); b) the liberation of a considerable amount of energy in kinetic form in the amount \(E_k = 0.62\) eV plus \(E_x\) (the vibrational energy of the molecules \(A^3\Sigma\)); it is known that the probability of collisions of the second kind decreases with increas-

...as regards the magnitude of the released energy. From this point of view, the following reaction involving collisions of the second kind with O \(({}^{1}D_{2})\) atoms, which have a much longer lifetime (107 sec.), appears more probable\({}^{26}\):

\[ \tag{3} \mathrm{O}({}^{1}D_{2})+\mathrm{N}_{2}(A^{3}\Sigma)_{v=x}\to \mathrm{N}({}^{4}S)+\mathrm{N}({}^{2}P)+\mathrm{O}({}^{3}P)-1.60\ \text{eV}. \]

However, in order for this reaction to be possible, an additional participant is needed to make up the energy deficit of 1.60 eV. In general, this may be a nitrogen molecule \(A^{3}\Sigma\) lying on the vibrational level \(v=10\), although the study of the Vegard–Kaplan bands in the spectra of the night sky shows that such states are not represented to any significant extent\({}^{26}\). Preliminary additional excitation of the molecule (prior to dissociation) by collision with an oxygen atom \({}^{1}D_{2}\) is possible:

\[ \mathrm{O}({}^{1}D_{2})+\mathrm{N}_{2}(A^{3}\Sigma)_{v=x}\to \mathrm{N}_{2}(B^{3}\pi)_{v=y}+\mathrm{O}({}^{3}P). \]

This reaction transfers the nitrogen molecule from the states \(A^{3}\Sigma(0<v<10)\) into the states \(B^{3}\pi(3<v<12)\), the presence of which is confirmed by the structure of the 1st positive system of nitrogen.

In judging which excited states of nitrogen atoms may be more frequent, Nicole\({}^{43}\) and Bernard\({}^{26}\) consider various variants of collisions of the second kind between nitrogen atoms and other particles. According to all available data, the principal role in the high layers of the atmosphere belongs to oxygen atoms and nitrogen molecules, and collisions with them must be taken into account first of all.

A. Collisions with oxygen atoms:

\[ \tag{4} \mathrm{O}({}^{1}S)+\mathrm{N}({}^{4}S)\to \mathrm{O}({}^{3}P)+\mathrm{N}({}^{2}P)+0.61\ \text{eV}. \]

\[ \tag{5} \mathrm{O}({}^{1}S)+\mathrm{N}({}^{2}D)\to \mathrm{O}({}^{1}D)+\mathrm{N}({}^{2}P)+1\ \text{eV}. \]

\[ \tag{6} \mathrm{O}({}^{1}D)+\mathrm{N}({}^{2}D)\to \mathrm{O}({}^{3}P)+\mathrm{N}({}^{2}P)+0.77\ \text{eV}. \]

\[ \tag{7} \mathrm{O}({}^{1}D)+\mathrm{N}({}^{2}D)\to \mathrm{O}({}^{1}S)+\mathrm{N}({}^{4}S)+0.16\ \text{eV}. \]

\[ \tag{8} \mathrm{O}({}^{3}P)+\mathrm{N}({}^{2}D)\to \mathrm{O}({}^{1}D)+\mathrm{N}({}^{4}S)+0.41\ \text{eV}. \]

Reactions (4) and (5) are unlikely because of the short duration of the \({}^{1}S_{0}\) state and the large yield of kinetic energy. Reactions (6) and (7), especially (7), are more probable, since very little kinetic energy is released in it (0.16 eV). Finally, reaction (8) must play a noticeable role, despite the release of a considerable amount of kinetic energy (0.41 eV), since the main part of the oxygen atoms is in the ground state \({}^{3}P\).

B. Collisions with nitrogen molecules, for which two states must be taken into consideration: the ground state \(X^{1}\Sigma\) and the metastable state \(A^{3}\Sigma\).

\[ \tag{9} \mathrm{N}_{2}(X^{1}\Sigma)_{v=x}+\mathrm{N}({}^{4}S)\to \mathrm{N}_{2}(X^{1}\Sigma)_{v=x'}+\mathrm{N}({}^{2}D). \]

I. A. KHVOSTIKOV

The magnitude of the excitation energy of the atoms (2.38 electron-volts) makes this reaction possible only for \(x > 8\).

\[ \tag{10} \mathrm{N_2}(X^1\Sigma)_{v=x}+\mathrm{N}(^4S)\to \mathrm{N_2}(X^1\Sigma)_{v=x'}+\mathrm{N}(^2P). \]

Excitation of atoms \(^{2}P\) requires an energy of 3.56 electron-volts, so that the reaction is possible for \(x > 12\). In the spectra of aurorae there are bands of the Vegard–Kaplan system, confirming the presence of \(X^1\Sigma\) states \((7<v<18)\). In view of the long lifetime of such molecules, reactions (9) and (10) are sufficiently probable\(^{26}\).

\[ \tag{11} \mathrm{N_2}(X^1\Sigma)_{v=x}+\mathrm{N}(^2D)\to \mathrm{N_2}(X^1\Sigma)_{v=x'}+\mathrm{N}(^2P). \]

Here the additional excitation energy for transferring atoms \(^{2}D\) to the \(^{2}P\) state is equal to 1.18 electron-volts, so that \(x > 4\) is sufficient. This reaction is very probable.

\[ \tag{12} \mathrm{N_2}(X^1\Sigma)_{v=x}+\mathrm{N}(^2D)\to \mathrm{N_2}(X^1\Sigma)_{v=x'}+\mathrm{N}(^4S). \]

The return of \(^{2}D\) atoms to the ground level releases an energy of 2.38 electron-volts, which goes into exciting vibrational levels. If before the collision \(x=0\), then \(x'=8\). This reaction is also very probable.

\[ \tag{13} \mathrm{N_2}(A^3\Sigma)_{v=y}+\mathrm{N}(^4S)\to \mathrm{N_2}(A^3\Sigma)_{v=y'}+\mathrm{N}(^2D). \]

The reaction is possible for \(y > 16\), but, as the spectra show, such states are extremely rare.

\[ \tag{14} \mathrm{N_2}(A^3\Sigma)_{v=y}+\mathrm{N}(^4S)\to \mathrm{N_2}(A^3\Sigma)_{v=y'}+\mathrm{N}(^2P). \]

This reaction is impossible, since by means of vibrations the molecule cannot provide the energy of 3.56 electron-volts necessary for transferring \(^{4}S\) atoms to the \(^{2}P\) level.

\[ \tag{15} \mathrm{N_2}(A^3\Sigma)_{v=y}+\mathrm{N}(^2D)\to \mathrm{N_2}(A^3\Sigma)_{v=y'}+\mathrm{N}(^2P). \]

The reaction is very probable, since it can occur at \(y=7\).

\[ \tag{16} \mathrm{N_2}(A^3\Sigma)_{v=y}+\mathrm{N}(^2D)\to \mathrm{N_2}(A^3\Sigma)_{v=y'}+\mathrm{N}(^4S) \]

or

\[ \tag{17} \mathrm{N_2}(A^3\Sigma)_{v=y}+\mathrm{N}(^2D)\to \mathrm{N_2}(B^3\pi)_{v=z}+\mathrm{N}(^4S). \]

Both of these reactions are equally significant, since the energy of 2.38 electron-volts released when \(^{2}D\) atoms return to the \(^{4}S\) level is simply absorbed by the nitrogen molecule in the form of vibrational energy (16), or partly in the form of vibrational energy and partly in the form of a transition to a higher electronic level (17). In the first case the final state of the molecule is \(A^3\Sigma_{v=y'}\) with \(y'>16\); in the second it is the state \(B^3\pi_{v=z}\) with \(z>5\).

What conclusions, then, can be drawn from the above brief analysis of the principal reactions? Of 14 reactions of second-kind collisions between nitrogen atoms, nitrogen molecules, and oxygen atoms, only one

reaction (9) produces atoms \(N({}^{2}D)\), three reactions (10), (11), and (15) produce atoms \(N({}^{2}P)\), and seven reactions (7), (8), (11), (12), (15), (16), (17) destroy the excited states \(N({}^{2}D)\).

“Under such conditions,” writes Bernard\(^{26}\), “the number of atoms \({}^{2}D\) present in the upper layers of the atmosphere is much smaller than the number of atoms in the state \({}^{2}P\), and, consequently, the ratio of the intensities of the lines \({}^{2}P \to {}^{4}S\) and \({}^{2}D \to {}^{4}S\):

\[ \frac{I_{3466}}{I_{5200}} = 269\,\frac{N_a}{N_b} \]

for \(N_a > N_b\) is much greater than 269.”

Nicole\(^{43}\) also arrived at analogous conclusions, though without such detailed justification.

8. DEGREE OF DISSOCIATION OF NITROGEN MOLECULES IN THE REGION OF THE AURORAE

Measuring the brightness of individual spectral lines at different altitudes of the aurorae allowed Bernard\(^{26}\), and subsequently Dufay, to estimate the degree of dissociation of nitrogen. Their method is based on comparing the intensity of nitrogen lines with oxygen lines and on the assumption that the atmosphere is “mixed,” i.e., that the ratio of the number of molecules

\[ \frac{O_2}{N_2} = \frac{1}{4} \]

at all altitudes, and that above 100 km oxygen is completely dissociated.

Bernard found\(^{26}\) from his measurements that the line \(N({}^{2}P \to {}^{4}S)\) has intensity 1.2 in the lower part of the aurorae and 14 in the upper. On the same intensity scale the intensity of the green oxygen line \(O({}^{1}S \to {}^{1}D)\) is approximately 400. Assuming that the intensity \(I = NAh\nu\) (\(N\) is the number of atoms, \(A\) is the transition probability, \(h\nu\) is the energy of the emitted quantum), for the ratio of intensities one may write:

\[ \frac{I_a}{I_b} = \frac{N_a}{N_b}\cdot \frac{A_a}{A_b}\cdot \frac{\lambda_b}{\lambda_a}. \]

For the line \(\lambda_a = 5577\ \mathring{\mathrm A}\), \(A_a = 2\ \text{sec}^{-1}\), and for \(\lambda_b = 3466\ \mathring{\mathrm A}\), \(A_b = 1.25 \cdot 10^{-2}\ \text{sec}^{-1}\), therefore

\[ \frac{I_a}{I_b} = 100\,\frac{N_a}{N_b}, \]

which gives, for the lower edge of the aurorae, \(\frac{I_a}{I_b} = 400\) and \(N_a = 4N_b\), and for the upper edge \(\frac{I_a}{I_b} = \frac{400}{40}\) and \(N_b = 3N_a\). If, in addition to the assumptions indicated above, one further assumes that the number of excited atoms \({}^{1}S\) or \({}^{2}P\) remains proportional to the total number of atoms, then it follows that in the lower region of the aurorae the degree of dissociation of nitrogen is \(1/16\) of the number of \(N_2\) molecules, and in the upper part is \(3/4\).

Dufay1, from his measurements, found complete dissociation of nitrogen molecules in the upper part of the aurorae.

Thus, it may be considered established that atomic nitrogen is present in the region of aurorae, and in the very upper layers of the atmosphere the degree of dissociation of nitrogen is apparently very great.

9. ON THE INFRARED RADIATION OF ATOMIC NITROGEN

Is the presence of atomic nitrogen characteristic only of the period of the aurora, or does it occur constantly?

In this question the decisive word must belong to the spectra of the luminosity of the night sky. If atomic nitrogen is constantly present in the upper layers of the atmosphere, and moreover not only in the region of aurorae but at all latitudes, then it is natural to expect the presence of nitrogen lines in the spectra of the luminosity of the night sky. However, we have seen (§§ 3 and 4) that the assumptions concerning the presence in these spectra of the nitrogen lines 3466 and 5200 Å have not yet received indisputable proof. But if we look again at the table of transition probabilities for the lines of atomic nitrogen (see Table VI), one cannot fail to note that the transition \(^{2}P \to {}^{2}D\), with the emission of the infrared line \(\lambda = 10\,400\) Å, stands out with the greatest probability, considerably exceeding the probability of the other transitions. The presence of atomic nitrogen should be revealed first of all by the radiation of this line. Unfortunately, this region of the spectrum is very difficult to investigate.

Thus the following interesting situation arises: if it were possible to carry out proper investigations in the region of 10 400 Å, then the problem of atomic nitrogen could at once be advanced. Such a situation developed by 1942–1944. And it so happened that precisely by this time Stebbins, Whitford, and Swings had managed to penetrate into the indicated region of the spectrum of the radiation of the night sky by means of the photoelectric installation on Mount Wilson. What did they find?

We already know (see §§ 1 and 2) that radiation was discovered with a wavelength very close to 10 400 Å, namely, \(10\,440 \pm 25\) Å. Moreover, according to the first determinations of Stebbins, Whitford, and Swings, precisely the wavelength value that was “needed” was obtained at once: \(\lambda = 10\,400\) Å.

Of course, under such circumstances there arose a great temptation immediately to publish a preliminary communication on the discovery of the infrared radiation of atomic nitrogen. But Stebbins, Whitford, and Swings took another course. They delayed publication until the completion of more accurate wavelength measurements with the aid of three auxiliary lines 10 336 Å, 10 407 Å, and 10 478 Å, which made it possible to “bracket” the line under investigation, \(\lambda = 10\,400\) Å, and consequently,

determine its wavelength more reliably. Of course, such caution and care are commendable. As a result of more careful measurements of the wavelength, Stebbins, Whitford, and Swings came to the conclusion that this line is not connected with the spectrum of atomic nitrogen and proposed an entirely different identification.

Anticipating somewhat, it should be said that the identification proposed by Stebbins, Whitford, and Swings is very interesting and was developed by them seriously. But one must first note that the skepticism of these authors with regard to the hypothesis of atomic nitrogen has so far been substantiated by them, in our opinion, to a lesser degree than is suggested by their formulations in the paper cited. Now, when the question of final identification cannot yet, as we shall try to show below, be considered settled, and the problem of atomic nitrogen in the upper layers of the atmosphere will undoubtedly still be the subject of many works, discoveries, and broad discussion for a number of years, it is appropriate to note that the above-mentioned skepticism of Stebbins, Whitford, and Swings has, in our impression, a certain underlying motive.

In the discussion on atomic nitrogen that has developed since 1939, the American authors did not take a direct part, although one of the important questions in the polemic was precisely the analysis of the spectrograms published by Babcock (see §§ 3–4). But in essence the attitude of American astrophysicists toward the proposals concerning atomic nitrogen was undoubtedly negative. For example, in the work of Elvey, Swings, and Linke on the spectra of the night sky (1941) it is said*): “A careful study of our material has convinced us that many of the announced lines are in reality not present on our spectrograms... Recently an extensive discussion has been published on the presence in the spectra of the night sky of the forbidden line of atomic nitrogen 3466.5 Å. But lines of such a wavelength and of any noticeable intensity are certainly absent...” One gets the impression that the American astrophysicists, having discovered the line 10 400 Å, which might have proved to be the principal proof of the hypothesis of atomic nitrogen that they had previously rejected, found themselves in some embarrassment, but then mobilized all resources in order nevertheless to show the untenability of this hypothesis.

However, let us turn to the factual material.

10. IDENTIFICATION OF THE INFRARED RADIATION OF THE NIGHT SKY

Stebbins, Whitford, and Swings indicate two possible methods of identifying the line \(\lambda = 10\,440 \pm 25\) Å. The first is the familiar, long-predicted radiation of atomic nitrogen; the forbidden mul-

*) Elvey, Swings and Linke, Astrophys. Journ. 93, No. 2 (1941), pp. 337–338.

undergoes \(^{2}P \to {}^{2}D\) (see Fig. 1). The second variant is connected with the band spectrum of molecular nitrogen: the first positive system of nitrogen, \(B^{3}\Pi \to A^{3}\Sigma\). The first variant is discussed briefly in the paper, and this variant is quickly rejected. The second variant is analyzed in detail, and, it must be admitted, new and very important considerations are adduced. It is another matter how convincing their final conclusion is.

Stebbins, Whitford, and Swings give a table of atomic nitrogen lines (Table VIII), in which they also indicate the values of the excitation potential in electron-volts and the probabilities of the corresponding transitions.

Table VIII

Lines of atomic nitrogen

Transition	Wavelength in Å	Excitation potential	Transition probability
\(^{2}D \to {}^{4}S\)	5197.8	2.4	0.000021
\(^{2}D \to {}^{4}S\)	5200.1	2.4	0.000014
\(^{2}P \to {}^{4}S\)	3466.5	3.6	0.0094
\(^{2}P \to {}^{2}D\)	10407.3	3.6	0.21
\(^{2}P \to {}^{2}D\)	10397.8	3.6	0.25

Although the most intense line should be the infrared doublet \(10397.8\)—\(10407.3\) Å with a mean wavelength of \(10402.7\) Å, the same upper level is the initial one for the emission of the line \(3466.5\), which should have an intensity only \((0.21 + 0.25) : 0.0094\) times, i.e. 49 times, smaller; meanwhile such a line in the spectrum of the night sky either is altogether absent or, in any case, can only barely be noticed on spectrograms. A line 50 times more intense than \(3466\) Å would be weak in comparison with the green or red oxygen line \(O_{1}\), but the green line exerted no influence on the photoelectric measurements described, whereas infrared radiation has a very large effect on the photocell. It should be supposed that it is not the radiation \(N_{1}\) that produces the powerful effect in the region \(10440\) Å. “Moreover,” write Stebbins, Whitford, and Swings\(^{2}\), “the measured wavelength of the intensity maximum of the night-sky radiation is a value definitely greater than that of the atomic nitrogen line, and the difference of wavelengths appears greater than the error of measurement.”

Thus, the argumentation of Stebbins, Whitford, and Swings is based, on the one hand, on photometric data and, on the other, on wavelength measurements. But it seems to us that in both cases the authors overstate the grounds from which their conclusions proceed. Indeed, as regards the photometric measurements, it would be more correct to say that in the work of Stebbins, Whitford, and Swings there are no measurements of the kind from which conclusions of a photometric character could be drawn. What quantities, what quantitative data do they have for their conclusions? These data are as follows: 1) the figure “49”—the line 3466 Å must be “only” 49 times less intense than the line 10 400 Å. 2) The green line 5577 Å had no effect on the photoelectric measurements, whereas the effect of the infrared line was very large (see Table I). But in order to draw quantitative conclusions it is necessary to take into account the difference in the spectral sensitivity of the photocell for \(\lambda = 5577\) Å and \(\lambda = 10400\) Å and the difference in the transmission of the light filters. Neither the one nor the other is taken into account in the work; meanwhile it is known that an oxygen-cesium photocell has its maximum sensitivity in the infrared region of the spectrum and that light filters isolating a narrow band in the green part of the spectrum inevitably have a small transmission (15–30%) at the maximum, whereas infrared filters are distinguished by high transparency.

It seems to us that to reject the atomic-nitrogen variant on the basis of photometric estimates of the kind indicated is premature, owing to the inadequacy of the latter. As for the wavelength measurement, the value obtained by Stebbins, Whitford, and Swings, \(10440 \pm 25\) Å, differs by 33 Å from the theoretical value \(10407\) Å of one of the components of the infrared doublet and by 38 Å from the mean value of the wavelength of the doublet, \(10402\) Å. This is greater than the stated measurement error, \(\pm 25\) Å. But how was the magnitude of this error determined? The authors themselves write²: “We found no better way than to take the mean of the two values obtained, 10427 and 10452, and to estimate the probable error as a quantity of the same order as the difference of the two determinations.” Undoubtedly, this question needs further study. Perhaps it will prove possible to obtain a spectrogram in this region.

Let us proceed to the second variant of the identification of the infrared radiation of the night sky.

11. THE (0,0) BAND OF THE FIRST POSITIVE SYSTEM OF NITROGEN

This band was studied under laboratory conditions by Pottker⁴⁷, and also by Frost and Oldenberg⁴⁸. Measurements with a thermometer on a prism spectrograph gave a maximum at 10 420 Å, while with the aid of a dif-

with a diffraction grating, 4 peaks were found between \(10350\ \text{Å}\) and \(10460\ \text{Å}\), of which the 2 most intense are at \(10410\ \text{Å}\) and \(10430\ \text{Å}\). In photographic measurements the individual lines were distinguishable, but it was not possible to measure them accurately.

Frost and Oldenberg give a part of the spectrum of the \((0,0)\) band, taken with a concave diffraction grating with a focus of 21 feet; in the photograph many individual resolved lines are visible, but the authors cited do not give an analysis of the band.

One may try to estimate the intensity of the maximum of the \((0,0)\) band on the basis of the well-measured \(N_2\) bands in the visible part of the spectrum \(^{49}\), but, as Stebbins, Whitford, and Swings \(^{2}\) rightly point out, such a method is difficult and unreliable, especially because of the specific influence of the temperature of those high layers of the atmosphere where the infrared radiation of the night sky arises. Such an estimate gives the wavelength of the maximum \(\lambda = 10450\ \text{Å}\).

Comparing all the data indicated, one may conclude that there is satisfactory agreement between the wavelength of the night-sky radiation and the \((0,0)\) band of molecular nitrogen \(^{2}\).

In intensity the \((0,0)\) band should be one of the strongest bands of the entire first positive group, since it corresponds to transitions with \(\Delta v = 0\). The absence of the \(N_2\) band in the absorption spectrum of solar rays is due to the high energy (6.1 eV) of the lower level \(A^3\Sigma\).

However, if the intense radiation of the night sky is attributed to the \((0,0)\) band, then in the spectrum of the night sky other bands of the first positive group must also be sufficiently intense, at any rate if one judges on the basis of data on the glow of nitrogen in a gas discharge in laboratory investigations. Some components of the night-sky spectrum in the region from \(5000\ \text{Å}\) to \(6000\ \text{Å}\) are tentatively attributed to bands of the first positive group, but this radiation of the night sky is extremely weak, and much in the identification is still very doubtful. In addition, Stebbins, Whitford, and Swings note that they were unable to detect on their photoelectric apparatus any radiation of noticeable intensity at \(\lambda = 8910\ \text{Å}\), where the maximum of the \((1,0)\) band is located. In their opinion this means that only a small number of \(N_2\) molecules are in the excited state \(v' = 1\) of the \(B^3\Pi\) level. Consequently, the identification of the infrared radiation of the night sky with the band of molecular nitrogen compels one to suppose a special excitation mechanism, specifically enhancing the \((0,0)\) band relative to other vibrational transitions. Stebbins, Whitford, and Swings put forward certain considerations in this direction.

The heat of dissociation of molecular nitrogen has usually been taken \(^{50}\) to be \(D(N_2)=7.383\) eV (from data on predissociation phenomena),

Recently, however, indications have been obtained\(^{51}\) of a higher value, \(D(\mathrm{N}_2)=9.764\) eV. Stebbins, Whitford, and Swings use in their calculations \(D(\mathrm{N}_2)=7.383\) eV. It should be noted that the mechanism they propose for the excitation of the \((0,0)\) band depends entirely on the correctness of the adopted value of \(D(\mathrm{N}_2)\). In Fig. 3 a scheme of the energy levels of the \(\mathrm{N}_2\) molecule was given. The value \(7.38\) eV coincides almost exactly with the energy of the excited state \(v'=0\) of the electronic level \(B^3\pi\), which is the upper level of the first positive emission system of \(\mathrm{N}_2: B^3\pi \to A^3\Sigma\). The heat of dissociation of the \(\mathrm{N}_2\) molecule is only slightly greater than the excitation energy of the level \(B^3\pi\), \(v'=0\), but less than the excitation energy of the level \(B^3\pi\), \(v'=1\).

Let us now suppose that during the day \(\mathrm{N}_2\) molecules, or other molecules containing N atoms, are dissociated into N atoms owing to absorption of far ultraviolet radiation. These N atoms recombine during the night. But the recombination reaction

\[ \mathrm{N}+\mathrm{N}+\mathrm{M}\to \mathrm{N}_2+\mathrm{M}^{\mathrm{exc}} \]

requires a third participant, M, in order that the laws of conservation of energy and momentum be satisfied. Since \(\mathrm{N}_2\) molecules are (according to the “mixed-atmosphere” hypothesis) the principal constituent of the atmosphere at all heights, it is they that can often be the third partner M in the reaction. If the dissociation energy, 7.38 eV, released in the recombination \(\mathrm{N}+\mathrm{N}\), is transferred to the third particle \((\mathrm{N}_2)\), then by this means the level \(B^3\pi\), \(v'=0\), can be excited. The state \(v'=1\), or an even higher vibrational state of the electronic level \(B^3\pi\), can be excited by such processes only provided additional kinetic energy is used. But the large velocities of N atoms or \(\mathrm{N}_2\) molecules required for this must be exceedingly rare even in the relatively warm upper layers of the atmosphere. Therefore, if the mechanism described is correct, then we should expect that only the band arising in the transition from the level \(v'=0\) will be intense. The second in intensity should be the band at \(\lambda=12253\) Å, corresponding to the \((0,1)\) transition, and the third—the \((0,2)\) band at \(\lambda=14790\) Å. Both these bands could not have been detected by the photoelectric measurements of Stebbins, Whitford, and Swings, since the photocell is not sensitive in this far region of the spectrum.

The hypothesis set forth by Stebbins, Whitford, and Swings deserves serious attention. Until more definite data are obtained, it must certainly be considered on a par with the hypothesis of atomic nitrogen. But in our opinion it cannot yet be regarded as better founded than the hypothesis of atomic nitrogen, for the following reasons.

a) The hypothesis of the origin of the new infrared line due to molecular nitrogen depends entirely on the adopted value of the heat of

dissociation \(D(\mathrm{N}_2)=7.38\) eV. Increasing this figure even by several tenths of an eV would overturn the hypothesis, since in that case the absence in the night-sky glow of another intense infrared line, \(\lambda=8910\) Å, would be incomprehensible. Meanwhile, more recent determinations by Gaydon\(^{51}\) give an entirely different value for the heat of dissociation, \(D(\mathrm{N}_2)=9.76\) eV.

b) The available data on the wavelength of the new infrared line, in terms of their accuracy, do not make it possible to distinguish the atomic nitrogen line \(10397—10407\) Å from a molecular band, for which laboratory measurements give, in one case, a maximum at \(10420\) Å, and in another case 4 peaks, of which 2 are more intense, at \(10410\) and \(10430\) Å.

Finally, one cannot fail to note an internal contradiction in the reasoning of Stebbins, Whitford, and Swings. They deny the existence of atomic nitrogen in the upper layers of the atmosphere, but the hypothesis they put forward would lose its meaning if there were no atomic nitrogen in the atmosphere. Indeed, in that case the reaction

\[ \mathrm{N}+\mathrm{N}+\mathrm{M}\longrightarrow \mathrm{N}_2+\mathrm{M}^{\mathrm{exc}} \]

would be impossible.

Further substantial progress in the question under discussion may be achieved if it proves possible to obtain good spectrograms in the region \(10300—10500\) Å. It is not excluded that in this case both emissions of nitrogen in the region of \(10400\) Å would be detected: both atomic and molecular.

The new radiation of the night sky in the infrared region of the spectrum is a very interesting and important object for further investigations, the significance of which is increased by the fact that this radiation is apparently connected with the fundamental problem of atomic nitrogen in the high layers of the atmosphere.

CITED LITERATURE

1. J. Stebbins, A. E. Whitford and P. Swings, A strong infrared radiation from molecular nitrogen in the night sky, Phys. Rev. 66, No. 7/8, 225—226 (1944).

2. J. Stebbins, A. E. Whitford and P. Swings, A strong infrared radiation from molecular nitrogen in the night sky, Astrophys. Journ. 101, 39 (1945).

3. I. A. Khvostikov, The glow of the night sky, UFN, 17, issue 2, 121—174 (1937).

4. V. M. Slipher, Monthly Not. 93, 666 (1939).

5. I. A. Khvostikov, Essays on atmospheric physics, UFN, 19, issue 1, 49—83 and issue 2, 145—194.

6. Georges Dejardin, Light of the night sky, Rev. of Modern Physics, 8, No. 1, 1—21 (1936).

1. Jean Dufay, Sur la présence de l’azote atomique dans la haute atmosphère, Annales d’Astrophysique, 2, No. 2 (1939).

2. M. Nicolet, Naturwiss. 51, 839 (1938).

3. K. T. Compton and J. C. Boyce, Phys. Rev. 33, 145 (1929).

4. J. Dufay, Comptes Rendus, 198, 107 (1934).

5. J. Dufay, Observation du spectre d’émission du ciel nocturne dans l’ultra-violet. Journ. d. Phys. et le Rad., 5, 523—526 (1934).

6. J. Dufay, Public. Observatoire Lyon, 1, No. 9 (1934).

7. J. Gauzit, L’extrémité ultraviolette du spectre du ciel nocturne, Comptes Rendus, 199, 29 (1934).

8. J. Gauzit, L’extrémité ultraviolette du spectre du ciel nocturne, Journ. de Phys. et le Rad. 5, 527—532 (1934).

9. J. Gauzit, Annales d’Astrophysique, 1, 334 (1938); Public. Observatoire Lyon, 2, 253 (1938).

10. L. Kaplan, 4-ème Rapport de la Commission pour l’étude des relations entre les phénomènes solaires et les phénomènes terrestres, 140 (1936).

11. J. Kaplan, Nature, 141, 645, 1139 (1938).

12. R. Bernard, Nature, 141, 1140 (1938).

13. J. Kaplan, Phys. Rev. 54, 541 (1938).

14. R. Bernard, Bull. Soc. Française de Physique, No. 426, 157 (1938).

15. L. Vegard, Geofysiske Publikasjoner, XIII, 5, 11 (1941).

16. R. Bernard, À propos de la longueur d’onde de la raie de l’azote atomique dans l’aurore boréale, Journ. Phys. et le Radium (VIII), 6, No. 12, 29 (1945).

17. A. Arnulf, Comptes Rendus, 202, 1412 (1936).

18. V. M. Slipher und Z. A. Sommer, Naturwiss. 17, 862 (1929).

19. H. W. Babcock, Public. Astron. Soc. Pacific, 51, No. 299, 47 (1939).

20. René Bernard, L’azote atomique dans l’atmosphère supérieure, Annales d’Astrophysique, 4, No. 1 (1941).

21. R. Bernard, Journ. de Phys., 8, 157 (1939); Phys. Rev. 55, 511 (1939).

22. R. Bernard, Comptes Rendus, 208, 1831 (1939).

23. A. Arnulf, R. Bernard, D. Cavassilas et Déjardin, Comptes Rendus, 208, 1331 (1939).

24. C. Hoffmeister, Das Spektrum des Nachthimmels und des Sodiakallichts nach Beobachtungen auf der Astronomischen Station bei Windhuk, Südwestafrika, Zeits. für Astrophysik, 19, No. 2, 116—131 (1939).

25. L. Vegard, Naturwiss. 17, 980 (1929).

26. J. Cabannes, Sur l’origine des radiations émises par le ciel nocturne, Journ. de Phys. et Rad., 5, 607—613 (1934).

27. C. T. Elvey, P. Swings and Walter Linke, The spectrum of the night sky, Astrophys. Journ. 93, No. 2, 337—348 (1941).

28. J. Kaplan, Bull. Amer. Phys. Soc., 14, No. 4, 16 (1939).

29. S. Störmer, Photographische Höhenmessungen und Spektre des grossen Nordlichtes vom 25. bis 26. Januar 1938, Naturwiss. 26, 633 (1938).

30. Z. Harang, Das Polarlicht und die Probleme der höchsten Atmosphärenschichten. Probleme der Kosmischen Physik, Bd. XX, Leipzig (1940).

31. Jean Dufay, Les raies interdites de l’atome d’azote dans les spectres des aurores et du ciel nocturne, Annales d’Astrophysique, 6, No. 3—4, 81—89 (1943).

32. J. Dufay, J. Gauzit et Tcheng Mao-Lin, Cahier de Physique, 1, 61 (1941); Public. Observatoire, Lyon, 3, No. 3.

33. P. Götz, Naturwissenschaften, 29, 690, 1941; Zeitschrift f. Angewandte Meteorologie, 58, 337 (1941).

34. J. Dufay et Tcheng Mao-Lin, Bull Soc. Astron. Populaire Toulouse, 32, 252 (1941); Comptes Rendus, 213, 692 (1941); Cahier de Physique, 8, 51 (1942); Public. Observatoire Lyon, 3, No. 8.

35. J. Gauzit, Nature, 142, 572 (1938).

1. J. Dufay et G. Déjardin, Cahier de Physique, 12, 36 (1942); Public. Observatoire Lyon, 3, No. 11.
2. M. Nicolet, Le problème d’azote atomique dans l’atmosphère supérieure, Ciel et Terre, 55, 257 (1939); Institut Royal Météor. de Belgique, Mémoire, T. IX.
3. S. Pasternack, Publ. Astr. Soc. Pac., 51, No. 301, 160 (1939).
4. S. Pasternack, Transition probabilitie of forbidden lines, Astrophys. Journ., 92, No. 2, 129—155 (1940).
5. J. Gauzit, The composition of the upper atmosphere according to the dissociation of oxygen and nitrogen molecules, Bulletin American Meteorol. Society, 25, 245—250 (1944).
6. A. H. Poetker, Phys. Rev., 30, 812 (1928).
7. A. A. Frost and O. Oldenberg, Phys. Rev., 48, 66 (1935).
8. Pearse and Gaydon, The identification of molecular spectra, London (1941).
9. G. Herzberg und H. Sponer, Zeits. f. Phys. Cheng., 26, 1 (1934).
10. A. G. Gaydon, Nature, 153, 407 (1944).

1. Dufay. ↩↩