Discussion on the Ionosphere

On June 22, at the Royal Society of London, a discussion on the ionosphere took place with the participation of E. Appleton, Chapman, and others. Below we give a complete translation of the report on the discussion, published in Proc. Roy. Soc. A 141, 1933. The translation was made under the editorship of D. I. Blokhintsev.

E. V. Appleton. In opening the meeting, it is interesting to recall that seven years ago a similar meeting, devoted to a discussion of the same question, was opened by Rutherford. For me it was extremely instructive to reread the report of that meeting. I was struck, for example, by the fact that the progress made during the last seven years has substantially clarified the existing state of affairs and enables us to answer more accurately and with greater confidence some of the questions raised by the discussion. It may also be noted with great satisfaction that many of the suggestions made at that time proved to be extraordinarily fruitful and especially useful for experimental investigation in this field. First of all, it should be noted that our subject now has its own name: the previous discussion was devoted to “Electrical phenomena in the upper layers of the atmosphere,” whereas the object of the present discussion is the “ionosphere.” This term was first proposed by Watt (W. Watt). In opening the discussion, I set myself the main task of outlining the advances achieved in the various directions sketched by Rutherford. I shall touch upon the question as a whole only in general outline, leaving the following speakers to fill in the gaps. I also propose to give exclusive attention to information obtained by means of the method that has especially interested me, namely radio investigations.

In the absence of data obtained from direct in situ measurements, possible for the lower layers of the atmosphere, information concerning the nature of the ionosphere is obtained from observations: 1) of the Earth’s magnetism, 2) of luminous phenomena, such as, for example, electric discharges, auroras, meteorites, etc., and 3) from studies of the propagation of radio waves; although the first indication of a sharply expressed electrification of the upper layers of the atmosphere came from magnetic observations, the use of radio waves has on the whole proved the most fruitful. Radio methods have the great advantage that the investigation can be carried out at any time, and there is no need to await particular natural phenomena.

Radio investigation consists in sending radio signals (usually vertically upward) and observing the characteristics of the waves reflected by the ionosphere. The following quantities are measured: a) the total time of travel of a group of waves upward and downward, b) the polarization, and c) the intensity of the reflected waves. Multiplying the group time by the speed of light, we obtain the equivalent path \(P'\), greater than the actual path traversed by the waves. On the other hand, the optical path \(P\) is less than the actual path, so that where an estimate of \(P\) can be made (this is possible in some cases with the aid of extrapolation), upper and lower limits of the actual height of reflection can be given. In measurements the transmitted wave must be given a distinctive characteristic, such as, for example—

for example, a change of frequency or a change of amplitude. Of the two indicated methods, amplitude modulations give more readily interpretable data. The intervals of time subject to measurement are of the order of milliseconds and therefore do not present serious experimental difficulties. Measurements of polarization and intensity are, for ordinary wavelengths (from 20–2000 m), measurements of the variable electric energy at a distance negligible in comparison with the wavelength.

Measurements a), b), and c) have yielded information on the nature of the ionosphere which may be summarized as follows:

A. Structure. The most direct indication of ionospheric conditions is obtained by measuring the equivalent path \(P'\) for a number of frequencies \(f\). For the known values of \(f\) found, the first break in \(P'\) is of greatest interest; it is interpreted as an indication of the existence of a maximum of ionization. There is usually one break in the curve \((P', f)\), indicating penetration into the lower (region \(E\)) of

Fig. 1.

the two main regions into which the ionosphere is divided. Fig. 1 shows an example of such a curve for a quiet day, where the discontinuity in the usual curve at a frequency of 3.1 megacycles per second is clearly visible. At this frequency no reflection from the lower region \(E\) is observed, and reflection from region \(F\) begins. Complete ionization in the upper region \(F\) is usually \(3\frac{1}{2}\)—4 times greater than the ionization in region \(E\), the mean ionization values being respectively \(6.1 \cdot 10^5\) and \(1.8 \cdot 10^5\) electrons per cubic centimeter*. Variations (diurnal and seasonal) in region \(F\) are not as noticeable as in region \(E\); this difference must probably be attributed to the different pressures in the two layers. There can be no doubt that the maximum of ionization of region \(E\) is attained at an altitude of approximately 100 km above the Earth’s surface. The height of the corresponding region \(F\) (maximum) is much more difficult to estimate, but a value of about 180 km will, in all likelihood, be quite accurate.

Experiments carried out during the last twenty months at Slough in England have revealed further details concerning what may be called the fine structure of the ionosphere. The intermediate

* These values were calculated from the application of Lorentz–Hartree scattering theory with inclusion of the “polarization” term. If this term is neglected, since its inclusion is still open to discussion, the given values should be multiplied by \(^{2}/_{3}\).

the region between the \(E\) and \(F\) regions proved not to be a region of excess ionization, and even in relatively rare cases the existence in this region of still another maximum of ionization has been demonstrated. But ordinarily for England this intermediate region at midday should not be as strongly ionized as the \(E\) region.

It has also been proved that there exists a convexity or protrusion on the lower side of the \(F\) region, where the maximum of ionization increases with altitude. It should not, however, be asserted that any region situated above the \(F\) region, but having a lower ionization, could be detected by ordinary radio methods.

B. Diurnal changes, seasonal and other regular ones. All regions of the ionosphere show a sharp diurnal variation, which proves that they are normally under the influence of direct solar radiation; ionization increases during the day at a rate depending on the altitude of the Sun, while during the night it gradually decreases owing to the process mentioned below in the \(F\) region. The maximum of ionization in the \(E\) region is found at local noon, whereas the maximum in the \(F\) region occurs probably only an hour or two later (although the latter has not been sufficiently verified). The simple diurnal curves of variation on quiet days have a great similarity to the theoretical results obtained by Chapman for the ionizing effect of monochromatic radiation on a rotating atmosphere. The experimentally found ratio of the ionization at summer noon to the ionization at winter noon is approximately \(2.2\) for the \(E\) region and (with less reliability) from \(1.5\) to \(1.8\) for the \(F\) region.

Weekly measurements of the maximum noon ionization in the \(E\) region, made in 1931 and later together with other observations, less accurate but extending over a longer period of time, make it possible to suppose a direct relation between the undisturbed values of noon ionization and the cycle of sunspots. The ionization at the maximum of sunspots is apparently 50–60% greater than at the minimum of sunspots. It still remains to be decided whether this variation can be ascribed to the eleven-year cycle in the intensity of the Sun’s ultraviolet radiation.

C. Periodic tendencies and irregularities. Measurements of ionization density have convincingly shown that the ionization of both layers \(E\) and \(F\) may from time to time increase during the night, when direct solar radiation cannot exert an effect. Anomalous values for the \(E\) region have also been observed in the daytime, especially in summer, greatly exceeding the values expected from seasonal changes. The curve \(P^{1}, f\) for such an abnormal summer day is shown in Fig. 2, which should be compared with Fig. 1.

Fig. 2 shows that the ionization in the \(E\) region is so intense that even waves of so high a frequency as \(7.0\) megacycles per second do not penetrate through it. Moreover, very small changes of \(P^{1}\) with \(f\) indicate an extremely sharp gradient of ionization with height; these conditions are not local in character, as is indicated by the fact that uniform data were obtained on the same day at Slough and in the laboratory of Gallé-Stuart in Kenley. Possible causes of the abnormally high ionization of the \(E\) region are thought to include both thunderstorms and charged solar corpuscles of high velocity. It is possible that thunderstorms, as C. T. R. Wilson pointed out some ten years ago, influence the ionization of the upper layers of the atmosphere either directly by means of free electrons, or indirectly through ionization by collisions occurring at high levels owing to the increased intensity of electric fields.

Known types of anomalous nocturnal increases in ionization become all the more noticeable the farther one moves from the equator in latitude, which points to the influence of the Earth’s magnetic fields on the paths of charged corpuscles. In addition, the correlation of radio data with sunspots and magnetic disturbances, and the presence of a tendency toward periodicity, further support the firmly established view of a sharply expressed solar influence. Much speaks in favor of the usual explanation that charged corpuscles originate from the Sun, but further, more detailed investigations are needed into the possible influence of thunderstorms in geophysical phenomena. It is not entirely certain that “world disturbances” do not act to a certain degree as an intermediate mechanism, preparing charged corpuscles which are the cause of phenomena showing a tendency toward 27-day periodicity; it would be of great interest to examine thunderstorm data for this tendency, and also for the

[Graph: vertical axis—“Equivalent height \(E'\), in km”; horizontal axis—“Frequency \(f\), in months”; top label—“\(12^{00}\) mean time.”]

Fig. 2.

11-year period. The first hint in this respect is found in Naismith’s measurements and my own, concerning annual variations of ionization in region \(E\) at noon. With the assistance of Lutkin (M. Lutkin) it was found that these data, which indicate the ratio between anomalously high ionization and thunderstorm activity, likewise lead to the assumption of a 26–27-day period. In the further development of the question I am greatly indebted to Brooks (C. E. P. Brooks) of the Meteorological Office, who kindly gave me valuable references concerning the literature on the subject. The closest connection between the frequency of thunderstorms and solar activity was established by Seltzer, who examined data obtained at 229 Siberian stations for the period 1888–1924. Three maxima and four minima of sunspots are exactly reproduced on the curve of the spot-forming activity of the Sun, with the exception that the double maximum of spots in 1905–1907 is represented by a single thunderstorm maximum in 1906. Brooks found that the correlation coefficient for these data is 0.88 and established the equation:

\[ \text{Number of thunderstorms} = 10.4 + 0.11 \times \text{number of spots}. \]

Concerning the 27-day period, Seltzer writes: “Periods of 25–27 days have already been discovered by D. O. Svyatsky, as well as a period of 25.8 by Bezold (V. Bezold) and 27.5 by Ridder, very closely approaching the 27-day synodic rotation of the Sun; it may be asserted that the maxima of thunderstorms on the terrestrial globe, occurring with a 27-day periodicity, are caused by certain centers of activity lying

DISCUSSION OF THE IONOSPHERE

within the Sun. Many attempts have been made to compare the number of sunspots with the amount of precipitation; most of these comparisons were unconvincing, but Clayton summarizes them in the sense that they prove the greatest quantity of precipitation during or just after the maximum of sunspots at most tropical stations in the North Atlantic and North Pacific Oceans, southern Chile, the southern coast of Africa and Australia, and at continental stations where summer rains predominate.

D. The Nature of Ionization. In interpreting data obtained by means of radio, one should remember that the effect of ionization, producing reflection, refraction, and absorption of waves, is measured not by the number of ions \(N\), but by the ratio \(N/m\), where \(m\) is the mass of one of the charged corpuscles. On this basis we may estimate ionization by the magnitude

\[ \frac{N_e}{m_e}, \quad \frac{N_i}{m_i}, \]

where the index \(e\) refers to electrons, and the index \(i\) refers to ions. Measurements of the polarization of reflected waves prove that \(N_e/m_e\) is greater than \(N_i/m_i\). The influence of the Earth’s magnetic field makes the ionosphere an allotropic medium, and, owing to the difference in group velocity of the two components, a single radio pulse can be transformed into a double one. A similar effect occurring in the \(E\) and \(F\) regions may be expected when the prevailing process is caused chiefly by electrons. The relation between the observed polarization and the direction of propagation with respect to the Earth’s magnetic field shows that the effective electric charges have a negative sign.

E. Ionizing Agents. Experiments made in connection with the prolonged eclipse at Kanade showed that the normal cause of ionization of the ionosphere is ultraviolet radiation from the Sun. We picture the structure of the ionosphere in such a way that there exist two principal regions, each of which probably is formed from two elements in the course of a day. Probably four components are associated with ionization potentials of various atmospheric constituents, atomic and molecular.

In C possible anomalous agents were indicated. What are the charged corpuscles and their fall, according to the measurements of Naismith and mine, corresponding to magnetic disturbances and thunderstorms? They are usually the cause of ionization in the normal region \(E\) or immediately below it, but give an ionization gradient steeper than the gradient obtained at the same level from ultraviolet light.

F. Processes of Electron Capture. During the night there is observed a steady decrease in the ionization of the ionosphere, which, in the absence of anomalous influences, can be used for studying the process of electron capture. It is easier to make measurements for the upper region than for the lower one, but even for the upper region the data are contradictory. Eckersley, on the basis of a study of wireless transmission of Marconi images, concludes that the ordinary law of recombination for electrons and positive ions describes the process; however, he finds that this law is not suitable for some of my measurements and that the law of attachment of electrons to uncharged atoms agrees better with the experiments.

Further work is necessary for a final conclusion.

Chapman. I should like to express my admiration for an excellent exposition of the question made by Appleton. I regard the ozone and the ionization of the upper atmosphere as factors representing in a certain sense an absorption spectrum of solar radiation. Solar radiation evidently has various constituent parts, absorbed at various levels and forming a certain kind of spectrum, for which the absorbing atmosphere itself is the spectroscope.

Our task is to determine what the radiation is from which the spectrum is formed, and what the nature is of the agent that gives it to us.

For the purpose at hand, the absorbing medium—air—may be regarded as consisting only of oxygen and nitrogen; but the various states of these two substances, which are important for absorption, may in certain respects be numerous. Both gases undoubtedly occur in the atomic as well as in the molecular state, and some of these four forms (atomic and molecular oxygen, and atomic and molecular nitrogen) probably have other variations with different and significant absorption. This is due to the fact that atoms and molecules may be in states of excitation, some of these states being metastable; solar radiation must continuously produce excitation; although the duration of an excitation state for each individual corpuscle may be short, nevertheless for any given time a certain number of corpuscles will be in one or another state of excitation. As for ionization, each of these variations is an independent component of the atmosphere with its own absorption coefficient and its own ionization potential, smaller than that of a normal corpuscle of the same kind.

Since solar radiation with a frequency sufficient to ionize a normal atom or molecule is probably absorbed high in the atmosphere, there is a band of solar radiation of greater wavelength which cannot excite normal atoms and molecules or ionize excited atoms and molecules. This second band has, in general, much greater energy than the first, although the individual photons have less energy. The lower part of the ionized region may be ionized by a double process; however, the absorption coefficients that determine the level of each kind of absorption have not yet been established either theoretically or experimentally.

Let us note an interesting point, namely that the ion content produced by a single ionization process, i.e. at very short wavelengths, must be proportional to the intensity of the ionizing radiation $I_1$, whereas, if there is a double ionization process, its intensity must be proportional to the product of the two ionizing radiations $I_1$ and $I_2$. Therefore, if solar radiation changes during the sunspot cycle, as is quite clearly revealed by magnetic data and to the same extent by radio measurements (the probability of the latter was indicated by Appleton), then all these calls to exclude ion content variations by arguing that solar radiation is constant—until we know whether ionization in a given layer is a single or double ionization process—are unfounded. Among particles that can be more readily ionized than atoms and molecules in normal states, I may also point to negative oxygen ions formed by the attachment of electrons to oxygen particles. The ionization potential of these negative ions is probably 5–6 V, and therefore they may be ionized rather easily and form an additional source of free electrons. Negative oxygen ionization may be caused either by solar radiation, or by a direct influx of energy from the same source by means of collisions with excited particles. The color of the night sky indicates that excited particles exist throughout the night and are formed, in all probability, continuously; it is possible that this circumstance slows the decrease of electron content during the night.

The nature of the decrease in electron content represents ...

would be of great interest. Do electrons recombine with positive ions, or do they attach themselves to neutral particles? Closely connected with this question is the ratio of the number of available electrons to the number of ions; if electron attachment is considerable, then there must be more ions than electrons. Our awareness of these questions should, it seems to me, rest partly on the curves of the diurnal changes in ion content at different levels, and partly on seasonal changes. I am engaged in a detailed theoretical study of these questions and believe that the situation with respect to diurnal changes is rather complex; but changes of ions as a function of the time of year give clearer indications. If recombination predominates, then the electron content in summer \((n_s)\), as compared with the content in winter \((n_w)\), is given by the formula

\[ \frac{n_s}{n_w} = \sqrt{\frac{J_s}{J_w}} = \sqrt{\frac{T_w \sin(\theta+\delta)}{T_s \sin(\theta-\delta)}} , \]

where \(\theta\) denotes latitude, \(\delta\) the maximum solar declination \((23^\circ)\), and \(T_s\) and \(T_w\) the summer and winter values of the absolute temperature in the ionized layer. For our latitude we obtain:

\[ \frac{n_s}{n_w} = 1.84\sqrt{\frac{T_w}{T_s}} . \]

The corresponding proportion, when attachments predominate, will be:

\[ \frac{n_s}{n_w} = \frac{T_w \sin(\theta+\delta)}{T_s \sin(\theta-\delta)}, \]

or \(3.4\left(\frac{T_w}{T_s}\right)\) for our latitude*.

Appleton pointed out this difference, and the values he gave in the preface—about 2.2 for the \(E\) layer and 1.5–1.8 (with less reliability) for the \(F\) layer—in both cases indicate, as was shown, that recombination is more significant than attachment. Let us add also that if attachment is of the same magnitude as recombination in the \(F\) layer, then in the \(E\) layer it may predominate owing to the greater density of the oxygen particles to which electrons can attach. Therefore it is entirely possible that, when diurnal observations over ionization are continued and when a sufficient number of curves are taken, it will become clear that this law is also the law of recombination in the \(F\) layer.

In conclusion I shall mention Henderson’s valuable observations\(^2\) during the eclipse in Canada on August 31, 1932. They showed that \(n\) fell by about 60% during the full phase, thereby confirming Appleton’s opinion that this layer is ionized chiefly by ultraviolet light, and not, as I suggested, by corpuscles.

The possibility of a corpuscular eclipse in some other layer is still, as it seems to me, subject to further investigation. But, with Miller’s assistance, I have determined what the expected change in \(n\) during the Canadian eclipse should have been, assuming ultraviolet light to be the sole ionizing agent and assuming a degree of recombination compatible with the curves

\[ \text{* The value } \frac{T_w}{T_s} \text{ is unknown, but apparently it is close to unity.} \]

of the diurnal variation for the \(E\) layer. The result is in complete agreement with Henderson’s diagram for the eclipse and shows that there is no need to attribute any part of the remaining 40% of the electron content during the total phase to another ionizing source. Similar calculations for the \(F\) layer indicate that here** a much smaller effect of the eclipse should be expected; this was not studied during the present eclipse and should be elucidated by observations during subsequent eclipses.

Eckersley. Appleton has clearly formulated three ways in which measurements of waves reflected by the ionosphere may be made. I propose to discuss the last two methods, i.e. measurements of polarization characteristics and the intensity of the reflected waves.

In our measurements the signals were received by two loop antennas placed at right angles to one another, and were fed through a rotating coil to the receiver, and thence to the plates of a cathode-ray oscillograph.

A circularly polarized incident ray will induce equal EMF in each of the two loops, and these EMFs will be displaced in phase by \(90^\circ\).

If both antennas are correctly tuned to the incident wave, then a rotating field is formed in the ionometer. If, however, the antennas are tuned so that one current leads the EMF by \(45^\circ\), while the other lags by \(45^\circ\), and if the EMFs in the antennas are shifted by \(90^\circ\) in phase, then the currents in both antennas will coincide in phase.

It follows from this that, for a circularly polarized wave, the searching coil of the ionometer can be brought into such a position that no EMF will be induced. The zero position may thus be used for polarization to the right minus \(-45^\circ\), and for polarization to the left plus \(+45^\circ\). A linearly polarized signal has uniform strength when the angle of incidence of the ray bisects the angle between the planes of the two antennas.

With the aid of this method we obtain a means of determining the polarization of the arriving waves. An auxiliary source of high-frequency oscillations is used in conjunction with the receiver to facilitate the correct tuning of both loops and for the further measurement of the intensity of the reflected pulses.

Purely visual observations were made at wavelengths between 45 and 90 m and only by day; a 200-watt transmitter was installed at Chelmsford, and the receiver at a distance of 1.5 km at Brookfield.

One of the most characteristic features of the results obtained is the unusual variability of the so-called ionospheric weather. Each day brings a new series of conditions; a large number of systematic measurements is necessary in order to clarify the principal ones.

In many cases splitting was observed. This agrees with the magneto-ionic theory, which predicts the splitting of a plane-polarized wave into two component parts with opposite circular polarizations, since most of the electrical carriers—namely electrons—are normally polarized in opposite ways. The ordinary right-hand component was smaller, while the left-hand component was larger, although in some cases the opposite phenomenon was observed. In rare cases very definite tri-

* The parameter \(\varepsilon_0\), on which the diurnal variation depends (Proc. Phys. Soc. 43, 36, 1930), was taken equal to \(1^{1/2}\).

** Taking \(\varepsilon_0\) equal to 1 or \(1^{1/2}\).

*** Slides were shown to illustrate the effects in the \(E\) and \(F\) layers. The diagrams will be included in a subsequent paper on this question.

triplets with one right-hand and two following left-hand polarizations, so close to one another that they may be considered as belonging to one system.

Triplets were observed in five cases during the last five months, at wavelengths between 55 and 60 m. It should be noted that sometimes \(E\)- and \(F\)-reflections were received simultaneously. This definitely indicates partial penetration and partial reflection; since in some cases both right- and left-hand \(E\)-reflections are observed simultaneously with \(F\)-reflections, this fact also definitely contradicts the ray theory, which requires that the ray either penetrate completely or be completely reflected.

The next observed feature is the occasional splitting of the echo, usually just before the moment when the given signal disappears, penetrating through the layer.

This multiple splitting, observed in many cases, corresponds to only one polarization; usually it is not split, and is then split again alternately into the right-hand and the left-hand component.

This propagation effect, it seems to me, is caused mainly by dispersion (as Plendl (Plendl³) assumes) and occurs when there is a rapid change of group and phase velocities with frequency.

The rapid change is especially great for frequencies close to the limiting frequencies. This circumstance testifies that splitting is usually observed immediately before the given ray penetrates into the ionosphere.

The polarization characteristics fluctuate considerably from day to day. The results of a large number of observations indicate, however, a certain regularity in their changes.

Thus, at 60 m the left-hand and right-hand polarizations are observed equally often, as a series of observations during December, January, and part of February of this year shows. At shorter wavelengths the right-hand polarization predominates; at longer wavelengths the left-hand predominates.

RELATIVE ATTENUATION OF LEFT- AND RIGHT-HAND POLARIZED WAVES

This is represented by attenuation curves, which show that:

1. At 60 m during the daytime the right-hand and left-hand rays are attenuated approximately equally.
2. A progressive weakening of the attenuation is observed as sunset is approached, where the reflection coefficient approaches 0.5.
3. An almost stationary value of the attenuation coefficient of the right-hand component for frequencies exceeding 4–6 megacycles.
4. Absence of left-hand components at frequencies above 5 megacycles because of a lack of electrons or excessive attenuation of the right-hand ray.

On the basis of these measurements we can obtain a measurement of the coefficient of daytime reflection of these waves. Thus, for known characteristics of the transmitting antenna and transmitting current, the transmitted wave, if it propagates uniformly and is reflected at a height of 250 km without loss, should give a field intensity of about 350 \(\mu\)V/m. It should be noted that the measured fields give approximately half of this value after sunset, when the reflection coefficient should be approximately \(1/2\). For midday winter reception at 60 m a reflection coefficient of \(1/60\) is obtained. The loss of energy upon reflection, if not wholly, then for the most part is determined by collisions of electrons or positive ions, set in motion by the wave, with neutral particles in the atmosphere (not counting partial reflection and penetration, which occur in very ...

narrow frequency band). From this it follows that, for this reason, there is a fairly significant loss of energy in the \(F\) layer and that this loss is greater for the left-hand component.

The line of reasoning is as follows:

Both left- and right-hand rays pass through the \(E\)-layer again to the Earth. Both rays must pass through the entire \(E\) layer twice. For each element of the path in the \(E\) layer the specific attenuation of the right-hand ray (assuming a predominance of electrons) is approximately three times greater (for \(\lambda 60\) m) than the attenuation of the left-hand ray. Since both rays follow the same path, the total attenuation of the rays in the \(E\) layer must occur in the ratio 3 to 1 in favor of the left-hand component. The fact that, nevertheless, the attenuation of the left-hand component is equal to the attenuation of the right-hand one indicates that the reflection of the left-hand component in the \(F\) layer is much weaker than the reflection of the right-hand component.

Such a decrease in reflection for the left-hand component may be attributed to a deficiency of electrons, but partial penetration connected with a deficiency of electrons occurs in a very narrow frequency band, whereas this phenomenon of weak reflection of the left-hand component extends over a considerable frequency band.

Under these conditions the coefficient of total attenuation of the left-hand ray in the \(F\) layer is at least \(2/3\) of the total measured attenuation, i.e. approximately 36 or 40 decibels at noon. Consequently, the attenuation of the left-hand ray by the \(F\) layer is at least 24–27 decibels, i.e. the coefficient of reflection of the left-hand ray in the \(F\) layer is from \(1/10\) to \(1/22\) (for vertical transmission) at noon.

Moreover, if the loss of reflection in the \(F\) layer were due to partial penetration, we could expect an increase of attenuation in the \(F\)-layer at sunset, when the density is lower and there is a greater possibility of frequency penetration for the left-hand ray, but in fact the opposite is observed.

One can avoid the conclusion that the left-hand ray is strongly attenuated in the layer by assuming, for example, that the electric carriers act in such a way that they change the refractive index of the lower \(E\) layer; these may be positive ions, singly ionized oxygen molecules.

In that case we would have an inestimable differential absorption between the left- and right-hand components in the \(E\)-layer (since the critical frequency of rotation is extremely small), and all the attenuation over 60 m, where the right- and left-hand components are almost equal, may be attributed to the absorbing \(E\) layer. If this were so, the right- and left-hand attenuations should be equal at all wavelengths from 45 to 90 m, whereas observation shows quite the opposite: the left-hand component predominates at wavelengths longer than 60 m (frequencies of 5 megacycles), and the right-hand component predominates at shorter wavelengths. Thus, again we should have to ascribe the difference in the refraction of the right- and left-hand components to differential absorption in the \(F\)-layer.

The fact that the left-hand component can be absorbed in the \(F\) layer needs explanation. Although the specific attenuation of the right-hand ray is greater than the attenuation of the left-hand ray, a deeper penetration and a longer path of travel can easily compensate for this effect. Usually the attenuation, although not entirely proportional to the group transit time of the pulse, varies to a certain degree proportionally, and the undoubted fact of the existence of a longer group time for the left-hand ray quite definitely confirms the assumption of the possibility of greater absorption of this ray. These attenuation results are difficult to explain without some preconceived idea about the density of the air above 100 km.

If it is assumed that the dissociation process for \(O_2\) is almost complete at an altitude of 120 km, so that the oxygen atom predominates at this altitude, and that the height of the homogeneous atmosphere for this case is 23 km, then it is possible to calculate the collision frequency above this point. Such a distribution would then agree with Chapman’s view of the density at an altitude of 200 km\(^4\).

With any acceptable distribution of the ionic density, the damping in this region would in practice be reduced to zero. Bearing in mind the damping, in this region a collision frequency 50 and 100 times greater than that which would occur if dissociation at an altitude of 100 km were complete and \(H\) above this region were 20 would be required. The collision frequency would vary from \(3.5 \cdot 10^5\) at an altitude of 100 km to approximately \(2 \cdot 10^3\) at 200 km. Much weightier evidence for this type of damping is furnished by observations at large distances, confirming the existence of appreciable damping in the F layer. This question has already been discussed elsewhere\(^5\).

DENSITIES OF THE LAYER

The question of the maximum densities of the layer and their determination by observation of critical penetration was examined in detail by other speakers. The results of individual measurements are very diverse. A large amount of statistical material is needed in order to show the usual course of the variation both daily and seasonally.

I have tried to obtain, in a short time, a large amount of material by a method which, being perhaps not sufficiently accurate, quickly gives a large number of results that can be statistically analyzed. The method is based on observations of local transmitting stations and on determining, by the method of angles, whether the signals of a given station disappear or not.

The phenomenon of disappearance (skip) is well known. For a given distance between transmitter and receiver, and assuming a certain height of the layer, it is possible to calculate the electron density which is just sufficient to turn the ray toward the Earth at the receiver. If at the moment of observation the station is at the skip distance, i.e., receives little energy reflected from the layer, the density must be less than this critical quantity; beyond the skip, the density is greater than this quantity.

The results of these observations showed that the ionization density does not remain constant over the course of the day. In 80% of cases the density proved to be greater than that required for disappearance of the signal, and in 20% less. Although this method is based on assumptions that may give rise to doubt, it may be noted with satisfaction that the results are monitored by a more direct method of determining limiting frequencies for a normally directed pulse. By this method results were obtained for the winter, spring, equinox, and summer of 1932; they relate to the daily variation of the density, as well as to its seasonal variation. The ratio of the observed maximum density of a summer noon to the maximum density of a winter noon is equal to \(1.9 : 1\).

Next, the maximum value of the ionization density was calculated under the assumption, made by Chapman, that the ionization is produced by monochromatic ionizing radiation arriving directly from the Sun, and that there exists a constant number of recombinations.

If it is assumed that the main factor is recombination, then the mean recombination coefficient can be determined

by the disappearance of the right and left rays in the experiments and observations made in January and February of this year.

The values obtained differ somewhat from the values given earlier.

Undoubtedly the density does not always fall uniformly at night; many cases of sudden increase have been observed, but I am inclined to regard as correct the fall at night after the ultraviolet radiation has ceased, along with the sudden increases caused by solar eclipses of charged corpuscles. The mean recombination coefficient, determined above and equal to \(1.3\)–\(1.5 \cdot 10^{-10}\), is probably somewhat lower than it should be.

Theoretical densities according to Chapman’s theory were calculated by Millington’s numerical method. The data obtained by us were confirmed by many observations and measurements of short-wave transmission over long distances.

These results have some bearing on the question of the manner in which the electron density decreases during the night, when the ionizing agent has been removed.

There are two hypotheses: first, that the number of electrons decreases owing to their recombination with positive ions; second, that it decreases owing to the attachment of electrons to neutral ions. Chapman showed that the ratio of the summer maximum of ion density to the winter maximum, according to the first theory, should be \(1.84\) for latitude \(50^\circ\), and according to the latter theory \((1.84)^2 = 3.4\). The observations agree with the first theory. Further agreement of the ion-density diagrams with the interpretation of the results of short-wave transmission over long distances confirms the assumption that recombination is the chief factor determining the decrease of the electron density at night.

In this part Dr. Ratcliffe will be as brief as possible, since I fear that, for lack of time, only those persons whom I would call professional ionospheric physicists will be able to take part in the discussion, which would be extremely undesirable. Appleton dealt chiefly with the fine structure of the ionosphere in vertical section; I should like to say a few words about the structure in horizontal section. Appleton dealt with the layered structure; I should like to dwell on the “patchy” structure.

With a view to investigating the effect of thunderstorms on the ionosphere, my colleagues and I at Slough made a comparison of data relating to ionospheric conditions with detailed observations of thunderstorms. Appleton and Naismith have already reported the results of investigations carried out at Slough, but it seems to me of interest to point to one or two examples of the study of local thunderstorms in connection with the instantaneous state of the local ionosphere.

The first quite definite observation was made by Lutkin in October 1932, while he was observing at Abinger. If, during a thunderstorm with thunder audible at the observing site, after the limiting frequency had been reached at which no reflection was observed even from the upper region, reflection from the lower region again appeared at an even higher frequency, this means that there was a temporary state in which the ionization density in the lower region increased to approximately \(8 \cdot 10^5\) electrons (per cubic centimeter), whereas the density in the region \(F\) amounted to from \(3\) to \(4 \cdot 10^5\) electrons (per cubic centimeter). The statistical work reported by Appleton and Naismith appeared very soon after this observation; a second example of a comparatively short-lived change in the ionospheric density took place in April of this year.

A continuous record of reflections during the second half of the day at a frequency of 6 megacycles per second revealed several periods

continuing for only 3–5 sec, during which waves of this frequency were reflected from the \(E\)-region. The critical frequency at noon was 3.1 megacycles per second; if it is assumed that this frequency determines the normal ionization for the given day, then these brief reflections indicate instantaneous increases of ionization at least threefold. These increases were so short-lived that, for lack of the necessary time, it was impossible to operate the hand-held camera prepared for this purpose.

This phenomenon is observed fairly often in summer. To illustrate other numerical measurements of the rate of increase of ionization under these anomalous, probably thunderstorm, conditions, the following examples may be cited: 1) an increase in the ratio of three to one, which took place within one hour in the May observations; 2) an increase in the ratio of two to one within half an hour, when thunderstorms were definitely observed at a distance of 25 km; 3) a series of observations of the results of the work of the current month of June. In this series it was established, by recording atmospherics and reflections of radio signals, that immediately before the occurrence of a thunderstorm at a distance of 50 km from the recording station there was observed an anomalous increase in ionization density, reaching values of \(8.5 \cdot 10^{5}\) and \(12 \cdot 10^{5}\) electrons (per cubic centimeter of measurement at two different frequencies).

The very method of removing these anomalous densities from the zenith of the receiving station must be carefully considered. Wilson made a very valuable observation: the secondary free electrons are held in a single homogeneous group; on this basis we may suppose that these may be ionic clouds arising in a similar way, and simultaneous observations by stations situated not very far from one another may establish the course of dispersion or the change of these anomalous concentrations.

This view of the action of thunderstorms on the ionosphere contains an indication of the necessity for a much more precise study of the relations between magnetic storms and thunderstorm activity. Therefore one should expect anomalous conductivity in these anomalous gaps, produced by as yet unidentified terrestrial sources, and reflected in magnetic observations.

It seems to me risky to have a large number of explanations available for the ion content in the ionosphere. Apparently ultraviolet light alone is already a sufficient source for all the electrons of the ionosphere; and in the tropics and in other thunderstorms we have, to an equal extent, a sufficient explanation of the entire electron content.

Allow me to mention, on the basis of the introductory part of Appleton’s report, one unfinished, only just begun, investigation of the 27-day period of recurrence of thunderstorms. We applied Bartels’s method of representation on four verticals, one beneath another, of 27-day periods of recurrence of various phenomena. We plotted the number of atmospherics received at Slough, under specially defined conditions, and the intensity of these atmospherics. The data, as I have already said, are not final, since neither the integral intensity nor even the number of atmospherics during a given period is the sole indication of the influence of a thunderstorm in the sphere of a given radius. This influence makes it possible to connect changes in the intensity of received atmospherics with changes in the coefficient of effective reflection, which in turn is due to the very change in the ion content that we are discussing. But this probable secondary effect of the possible 27-day period in the ion content should not, in my opinion, diminish the significance of the data presented.

... based on records from a comparatively insensitive instrument, i.e., such an instrument which, in all probability, does not take account of more distant sources undergoing a strong change in the reflection coefficient; nevertheless, the material presented gives sufficient grounds for further study of the 27-day tendency toward recurrence of thunderstorm activity.

Further study would require a more detailed investigation of the distribution of thunderstorms; such an investigation, it seems to me, may be better carried out by some other method than one depending on radiotelegraphic localization of the sources of atmospherics. But this method must first be further developed in relation to the interpretation of the observed intensities of atmospherics.

Returning to the questions of fine vertical structure discussed by Appleton, I should like to say that, in my opinion, in region \(E\) there exists not just a single sharply defined maximum of density. In many cases of short-period increase in density we see that the increase takes place (very often in succession) at two or three well-defined equivalent levels, each of which differs from its neighbor by \(10\)—\(15\) km. Thus, on June 20, 1932, between 8 and 10 hours Greenwich mean time, the equivalent height of reflection for pulses with a frequency of 4 megacycles per second was 110 km. But with the rapid changes in ionization density of which I spoke above, there were observed: a) 10 primary reflections (without secondary ones) from a height of 97 km, b) 4 primary reflections, accompanied by 7 secondary ones, from a height of 112 km, c) 7 primary reflections, accompanied by one secondary reflection, from a height of 122 km. The secondary reflections give a mean height of 117 and 127 km. The maximum deviation from these mean numbers was 3 km, which represents the order of accuracy of the measurements. Although a partial explanation might be found in the conditions of the magnetic-ion distribution, the conditions do not favor such an explanation, which would require a change from complete absorption of one component to complete absorption of the other, and vice versa, at repeated five-minute intervals. To test this rather improbable explanation, no corresponding arrangement has yet been made, although the construction of one would meet with no obstacles. To my satisfaction I believe that these data indicate an additional fine structure in the vertical section.

Ratcliffe. During the last twelve months in Cambridge, with the aid of an instrument designed by White, automatic records have been obtained of radio signals reflected from the ionosphere. In these records the effective height is plotted as a function of the time of day, while the wavelength remains constant; in addition, recording of the polarization of the descending waves has been prepared. A preliminary examination of these records has revealed several questions of interest for the present discussion.

The most outstanding feature, preliminarily noted by several workers, is the nocturnal increase of ionization in region \(E\). It usually occurs without a corresponding increase of ionization in region \(F\) and was observed even in those cases when ionization in layer \(F\) is abnormally weak. There is a direct correlation between the case of increased ionization in layer \(E\) and a disturbance of magnetic conditions. Anomalous magnetic behavior happens to be associated also with unusually low ionization in region \(F\). The presence of intense ionization in region \(E\) for an unusually long time after sunset is connected with the thunderstorm phenomenon.

The records abound in indications of the existence of an intermediate region. In the winter months this region is often more intensely

is manifested by day than the region \(F\). Thus, in November and March, reflections appear (\(\lambda = 150\ \text{m}\)) an hour or two before sunrise, before reflection from the region \(E\) arises, and after sunset between the disappearance of reflection from the region \(E\) and the appearance of reflection from the layer \(F\). In December, January, and February it is quite usual for the intermediate ionization region to remain more intense than the region \(E\) throughout the day, so that (\(\lambda = 150\ \text{m}\)) reflection is observed throughout the day from the intermediate region, with occasional isolated reflections from the region \(E\). The effective heights recorded for the region \(E\) are unusually constant; they do not deviate from the mean value (for a wavelength of 150 m the height is \(105\ \text{km}\)) by more than 5%, which only slightly exceeds the accuracy of measurement. The effective heights of reception for the intermediate region are more varied; they range from 120 to 180 km at different hours of the day (\(\lambda = 150\ \text{m}\)).

The effective height of the region \(E\) at night is equal to the height of the region \(E\) measured by day (\(105\ \text{km}\)).

As a result, reflection from the level \(E\) (\(105\ \text{km}\)) in winter is a quite general phenomenon at night, whereas by day reflection occurs from a higher intermediate region (approximately \(140\ \text{km}\)).

One case was recorded when, in December 1932, instead of the expected daytime reflection from the intermediate region, throughout the whole day there was reflection from the region \(F\), which indicated a weakening of ionization in the intermediate region. A weakening of the \(F\)-ionization was also observed on the preceding night; hence one may suppose with a known probability that one and the same source of ionization forms both the \(F\) region and the intermediate region. In the case described, the nocturnal \(E\)-echo manifested itself very intensely, which makes it necessary to suppose a different origin for the ionization of the region \(E\).

The night and day regions \(E\), observed at a height of \(105\ \text{km}\) during the winter months, evidently have common causes, whereas the considerably greater daytime absorption of the echo can be attributed to the widespread ionization of the intermediate region at lower heights, i.e. below \(105\ \text{km}\). Polarization measurements indicate that in the regions \(F\), intermediate, and \(E\) there appear free electrons, as also in the absorption region, at any rate up to the limit where the frequency of collisions of electrons with molecules is approximately \(5 \cdot 10^{6}\) per sec.

The diurnal variation of the ionization density of the region \(F\) appears more complicated. Rukop and Paul discovered an “evening concentration” of ionization approximately between 20 and 23 hours in August 1932; we checked these data and found that in May 1933 it was observed between 19 and 22 hours. There is reason to suppose that there exists an increase of ionization around midnight, which at the critical wavelength often causes the appearance of an echo during the night hours. The receptions carried out by us did not confirm Elmas’s hypothesis that this returning echo is determined by the presence of a still higher layer (“layer II” in his terminology).

There are indications of a weakening of the ionization of the region \(E\) as compared with the preceding year, which cannot be said of the region \(F\). This fact once more confirms the assumption made above concerning the difference in the cause of ionization of the region \(F\) and the region \(E\).

Finally, in the course of the whole discussion we dealt with equivalent heights, defined as the heights calculated on the basis of the assumption that the pulse propagates with the speed of light. To establish a correlation between our data on the ionosphere and our geophysical knowledge, it is neces-

It is necessary to determine the actual heights at which the reflection of radio waves takes place, since all our observations and calculations concerning the temperature, density, structure, etc., of the upper layers of the atmosphere relate to effective heights. In this connection it is important to note the absence of any data on the relation of equivalent heights to actual ones.

It is well known that the phenomenon of so-called radio echoes was observed tens of seconds, and sometimes even minutes, after the sending of these signals. To the present time two types of explanations of this fact have been given. According to the first, waves are reflected by a cloud of charged particles emanating from the Sun and collected into a cloud of corresponding form by the magnetic field of the Earth. Does this explanation withstand criticism? First of all, it compels us to assume an accumulation of from \(10^5\) to \(10^6\) particles in one cubic centimeter of the cloud. If they were all of the same sign of charge, their mutual repulsion would be strong enough to prevent the formation of the cloud. If, on the other hand, they were distributed in equal quantities of opposite charge, then a calculation proceeding from the assumption of the existence of the aforesaid cloud ceases to be valid. Independently of this objection, the theory of the cloud is vulnerable to the following considerations. If radio signals really had to traverse distances measured in millions of kilometers, corresponding to the time of return of the echo, then they would weaken to inaudibility. This argument is usually parried by the assumption that the cloud of particles assumes a form suitable for focusing radio waves back onto the Earth, by virtue of which the inverse-square law loses its force. In reply to this we point out that such a reverse-focusing action can take place only if the curved surface of the cloud of particles is constructed with an accuracy corresponding to the wavelength of the given signals. It is scarcely possible to assert that these particles, flying from the Sun at the most varied velocities, are arranged along a curved surface with a radius measured in millions of kilometers and constructed with an accuracy of several meters. On the other hand, let us recall that the time of return of the echo oscillates over the course of several minutes within limits of many seconds. Then the protective theory of the cloud would have to be set aside, since the surface of the cloud moves over the course of several minutes by many millions of kilometers, while, of course, preserving all the time its sharply, to an accuracy of several meters, defined surface. This objection is so substantial that the explanation of the echo phenomenon by means of an electron cloud must be decisively rejected.

Thus, if one does not believe in the activity of certain inhabitants of the Moon sending us their own signals, one must turn to the second explanation: according to it, the velocity of the signals is sometimes abnormally distorted in the ionosphere owing to a special distribution of ion densities, causing anomalies in dispersion sufficient to diminish the speed of propagation of the wave over a time measured in seconds. This theory has its shortcomings, but no other explanation has been created up to the present time. Be that as it may, in any case it is beyond doubt that delays can occur and actually do occur. It would be reasonable to beware of the assumption that analogous, but infinitely smaller (on the order of thousandths of a second) anomalies in dispersion are the cause of delays observed in ordinary measurements of the height of various ionized layers. If this were so, then it is quite probable that in fact the two principal layers are not separated from one another by a distance of about 100 km, as this is accepted on the basis of some tacit agreement, but are in close contact and represent only more or less sharp oscillations of the gradient of ionic densit-

particular. Questions of this kind must be carefully studied and clarified before time and effort are spent on establishing a connection between the existence of these layers and any other physical phenomena.

Of course, it is quite correct that the height of these layers is always determined as the equivalent height, and that no one has ever regarded it as the actual height. However, under conditions of discussion such as the present one, this fact can easily be blurred, and a tendency arises to identify equivalent heights with actual heights. It is very probable that, if one constantly keeps in mind the possibility of such an error, some of the difficulties and apparent contradictions encountered in attempts to explain these phenomena would be alleviated, and perhaps even entirely removed.

References

1. Septer, Met. Ztshr. 43, 229, 1926.
2. Henderson, Canad. J. Res. 8, 1, 1933.
3. Plendl, Z. N. T. 10, 75–94, 1933.
4. Chapman, Proc. Roy. Soc. A 132, 353, 1931.
5. J. Inst. Elect. Eng. London 71, 405, 1932.