SPONTANEOUS FISSION OF URANIUM¹)

K. A. Petrzhak and G. N. Flerov, Leningrad

1. INTRODUCTION

N. Bohr and Wheeler¹ pointed out the possibility of spontaneous fission of uranium with a half-life of the order of \(10^{22}\) years. The calculation was carried out for the principal isotope \(^{238}_{92}\mathrm{U}\) by the formula for the penetration of a particle through a potential barrier. The width of the barrier was taken equal to the radius of a fragment, which is an assumption whose validity can be checked only experimentally.

The question of the possibility of spontaneous fission of uranium was studied experimentally by Libby², who proceeded from the assumption that, in the process of spontaneous decay, neutrons should be emitted, as in the case of fission of uranium under the action of neutrons. Libby attempted to detect the neutrons produced in the process of spontaneous fission of uranium nuclei by means of a \(\mathrm{BF}_3\) counter sensitive to slow neutrons. On the basis of the negative result of the experiment it was possible to establish a lower limit for the half-life of uranium \((T > 10^{14}\) years).

2. METHOD OF WORK

In the experiments described below we used the method first proposed by Frisch³ for recording processes of nuclear fission. An ionization chamber with plates coated with a layer of uranium oxide is connected to a linear amplifier adjusted in such a way that the \(\alpha\)-particles emitted from uranium are not registered by the system; the pulses from fragments, which greatly exceed in magnitude the pulses from \(\alpha\)-particles, unlock the output thyratron and are counted by a mechanical relay. The thickness of the working layer of uranium oxide and the area of the plates of the ionization chambers ordinarily used (2 plates with a diameter of 30 mm) made it possible to establish approximately the same lower limit for the half-life of spontaneous fission as in Libby’s experiments. In order to increase the sensitivity of this method, it was necessary to increase the working surface of the uranium oxide. For this purpose an ionization chamber was specially constructed in the form of a multilayer flat condenser with a total area of 15 pla-

¹) Report at the Conference on the Atomic Nucleus, 1940; see p. 241 of this issue. Published in ZhETF, 10, 1013, 1940.

stins of \(1000\ \mathrm{cm}^2\) (Fig. 1). The plates, placed \(3\ \mathrm{mm}\) apart, were coated with a layer of uranium oxide \(10\)—\(20\ \mathrm{mg}/\mathrm{cm}^2\). The collecting potential was \(360\ \mathrm{V}\). To reduce the microphonic effect, the ionization chamber, together with the first tube of the amplifier, was mounted on a double damper consisting of lead plates on rubber cushions.

When testing this chamber for recording fragments, it became clear that the special features of its design also imposed special requirements on the amplifier. The large number of pulses from \(\alpha\)-particles required a considerably higher resolving power of the amplifier than when working with chambers of the usual type.

Fig. 1

Fig. 2

The superposition of pulses from individual \(\alpha\)-particles, with the usual resolving power of the amplifier, could produce pulses comparable in magnitude with those obtained from fragments. In order to increase the resolving power of the system, the grid of the first tube was connected to ground through a resistance of \(10^5\ \Omega\). With the chamber’s own capacitance \(\sim 150\ \mu\mu\mathrm{F}\) and a leakage resistance of \(10^5\ \Omega\), the charging time of the capacitance was \(10^{-5}\ \mathrm{sec}\). The pulse at the anode of the first tube had the form shown in Fig. 2. The ordinary pulse shape without leakage resistance is shown by the dashed line in the same figure. A further increase in the resolving power of the amplifier was achieved by reducing the coupling capacitance between the first and second tubes to \(10\ \mu\mu\mathrm{F}\) (instead of the usual \(100\)—\(1000\ \mu\mu\mathrm{F}\)). Installing this capacitance shifted the frequency pass band toward higher frequencies. This sharpened the pulse still more, since only the high-frequency components from the Fourier decomposition of the voltage pulse were passed by the amplifier. Narrowing the amplifier’s frequency pass band also gave a number of other advantages; namely, it made it possible to reduce to a minimum the low-frequency flicker effect and the microphonic noise of the ionization chamber. The optimum value of the coupling capacitance between the first and second tubes was selected experimentally by inserting a capacitance box between the tubes and choosing the capacitance at which the ratio between the fragment pulse and the amplifier noise was maximal.

The increased capacitance of the ionization chamber and the process of sharpening the pulse by including a leakage resistance greatly reduced

voltage amplitude from the particles, so that the pulses from the α-particles were of the same order as the Johnson and shot noises of the first tube. The pulses from the fragments exceeded the amplifier background; in order to make this excess more noticeable, the last tube of the 6-Ф-5 amplifier was shifted by the grid bias into the nonlinear region, after which the pulses from the fragments exceeded by 4–5 times the background due to the superposition of pulses from α-particles on the voltage fluctuations caused by Johnson and shot noise. The total amplification factor was \(\sim 10^7\).

The pulses from the output of the amplifier were fed to a cathode oscillograph for visual observation and to a counting-tube circuit of the usual type for recording the number of pulses by means of a mechanical relay.

When operating a linear amplifier in the nonlinear region, special requirements are imposed on the constancy of the amplification factor. In our experiments, over 5–10 hours of operation the amplification factor changed by no more than 1–2%. The constancy of the amplification factor was monitored by means of a calibration device. The anode and filament circuits were supplied from storage batteries with constant buffer recharging.

3. RESULTS OF MEASUREMENTS AND CONTROL EXPERIMENTS

The method developed made it possible easily to observe the processes of uranium fission under the action of neutrons. The sensitivity of the chamber, as was to be expected, was 30–40 times greater than the sensitivity of chambers of ordinary construction. The amplifier with an ionization chamber adjusted for the registration of fragments gave approximately one pulse per minute from \(1\ \mathrm{mCu}\) \((\mathrm{Rn} + \mathrm{Be})\) of fast neutrons when the source was brought close to the chamber.

In the very first experiments with the amplifier adjusted for counting fragments, it was possible to observe spontaneous (in the absence of a neutron source) pulses on the relay and on the oscillograph. The number of these pulses was small (6 in 1 hour), and it is therefore quite understandable that this phenomenon could not be observed with chambers of the usual type. Although in their form the observed pulses were very similar to pulses from fragments, it was necessary to carry out a series of control experiments in order to ascertain the reality of the existence of uranium fission under these conditions. The origin of the observed pulses could be explained by:

1) reception of external oscillations by the amplifier;
2) superposition of pulses from individual α-particles;
3) the presence of regions of gas amplification in individual regions of the ionization chamber;
4) random discharges on the surface of uranium oxide.

Special experiments showed that none of the above-mentioned causes can serve to explain the effect. A chamber with plates without uranium oxide did not give a single pulse in 5 hours. This showed that the observed spontaneous pulses were обуслов—

are due to the presence of uranium oxide on the plates of the chamber, and not to the reception of external vibrations.

To check the second assumption about the origin of these pulses, a chamber was assembled in which, on the plates, instead of uranium oxide, thorium oxide with an addition of Po was deposited in such an amount that the total ionization current caused by $\alpha$-particles was twice as great as the ionization current in the uranium chamber. During 10 hours of operation only three relay kicks were observed. In these experiments there was no complete equivalence of the operating conditions of the chamber because of the difference in the surfaces of uranium oxide and thorium oxide. Therefore another control experiment was carried out, in which thorium emanation was introduced into the uranium ionization chamber in such an amount that the products of its decay gave an ionization current, caused by $\alpha$-particles, twice as large as that from the $\alpha$-particles of the uranium itself. In this experiment no increase of the effect was observed, from which it follows that accidental coincidences of $\alpha$-particles were excluded by the resolving power of the amplifier.

From this same experiment it follows that there were no regions of gas amplification in the chamber. The active deposit should have been distributed uniformly over the chamber plates. Therefore, in the event that such regions of gas amplification existed, the introduction of the active deposit, increasing the number of ionizing particles in these regions, should have led to an increase in the number of pulses being studied. Raising the voltage on the chamber from 360 to 600 V did not give any noticeable increase of the effect, which also excludes the possibility of gas amplification. Covering the uranium oxide with bronze foil 1 $\mu$ thick, ensuring good conductivity of the surface, led to a decrease both in the magnitude and in the number of spontaneous pulses; but the number and magnitude of pulses from uranium fragments caused by neutrons from an (Rn + Be) ampoule were decreased in the same ratio. Thus, none of the four secondary causes advanced to explain the effect is in fact applicable. It should also be noted that, in the course of the experiments, the measurements were made in three different chambers, always with the same result. This excludes the possibility of explaining the observed spontaneous pulses by defects in the assembly of the chamber.

However, in order to establish definitively that the spontaneous pulses are obtained as a result of the fission of uranium nuclei, the distribution curve of the pulses by magnitude was taken. The dependence of the number of spontaneous pulses on the bias on the grid of the first tube of the circuit controlling the mechanical relay was measured. The curves are shown in Fig. 3. Along the ordinate axis is plotted the number of counts of the mechanical relay; along the abscissa axis, the negative bias in volts on the tube grid. Curve 1 was obtained in the absence of the neutron source and represents the counting of $\alpha$-particles against the background of Johnson noise. Curve 2 gives the same dependence with the chamber additionally loaded with the active deposit Th Em. With the neutron source, curve 3 was taken, giving the distribution of fragments by magnitude under conditions of nonlinear operation of the apparatus. The character of this curve coinci-

gives with the character of the distribution of spontaneous pulses by magnitude. The figure shows the number of counted spontaneous pulses in 1 hour for three different biases of the counting circuit. All these control experiments were carried out with an ionization chamber having an active surface of \(\sim 1000\ \mathrm{cm}^2\). Subsequently, an even larger chamber was constructed, with a total surface of 15 plates of \(6000\ \mathrm{cm}^2\). To increase the mobility of the ions, the chamber was filled with dried argon. The maximum number of spontaneous kicks that

Fig. 3

Fig. 4

could be observed with the aid of this chamber increased to 25–30 kicks per hour. The distribution curve of these spontaneous pulses by magnitude, in the same coordinates, is presented in Fig. 4. In the same figure, on a changed scale, the dependence of the number of spontaneous pulses on the bias on the grid of the counting-circuit tube is shown. On the basis of all these experiments it may be concluded that the spontaneous pulses observed by us are indeed caused by fragments from the fission of uranium.

4. DISCUSSION OF THE RESULTS

The established effect may be explained by the action on uranium of cosmic neutrons.

The experimental data at present available, both on the number and on the energies of cosmic neutrons at the earth’s surface, compel us to reject this hypothesis. In order that the observed effect could be ascribed to cosmic neutrons, it would be necessary for a flux of 5 neutrons per \(1\ \mathrm{cm}^2\) per 1 sec to pass through. This number is much greater than the upper limit for the number of both cosmic and terrestrial neutrons present in the atmosphere. Moreover, the absence of a noticeable effect in the chamber with \(\mathrm{ThO}_2\) (3 kicks in 10 hours) convinces us that it is impossible to ascribe the observed effect to cosmic neutrons. From the works of Petrzhak and Flerov\(^4\) and of Nikitinskaya and Flerov\(^5\) it follows that the thresholds for the fission of uranium and thorium are close and lie near 1 MeV. The effective cross section for the fission of thorium is 5 times smaller than the effective cross section for the fission of uranium. Consequently, cosmic neutrons, not

causing an effect in the thorium chamber, cannot explain the effect observed in the uranium chamber.

The fission of uranium could have been attributed to the action of cosmic electrons and mesotrons. There is no indication of any such mechanism of interaction between electrons and heavy nuclei. But even if such an interaction is assumed possible, the requirements on the effective interaction cross section, \(\sigma_{\mathrm{вз}} \sim 10^{-23}\ \mathrm{cm}^2\), make this assumption improbable.

Hoffmann ionization bursts could have produced similarly large pulses. The number of Hoffmann ionization bursts in a large chamber filled with gas at high pressure is smaller than the effect observed by us. Moreover, in the uranium chamber an ionization burst would not have produced sufficiently large ionization, since the electrons would have lost only a small fraction of their energy in the gas of the chamber. Experimentally this is confirmed by the absence of an effect in the chamber with plates without uranium.

One could have attributed the observed effect to neutrons arising from nitrogen and impurities in uranium oxide under the action of the \(\alpha\)-particles of uranium itself. The observed 6 recoils per hour correspond to \(1/20\) mCu of neutron activity of an \((\mathrm{Rn}+\mathrm{Be})\) source. To obtain such a number of neutrons, roughly 100 times more \(\alpha\)-particles would be needed than are emitted by uranium. Moreover, on the basis of Libby’s data, this hypothesis too can be refuted. Measurements with a \(\mathrm{BF}_3\) counter showed that \(1.5\ \mathrm{kg}\) of uranyl nitrate yields a number of neutrons not exceeding the number of neutrons emitted by a \(1/10\) mCu \((\mathrm{Rn}+\mathrm{Be})\) source. Since in our experiments the total amount of uranium oxide on the plates was \(15\ \mathrm{g}\), the neutrons emitted by uranium oxide could explain only \(1/50\) of the observed effect.

We are inclined to think that the effect observed by us should be attributed to fragments produced as a result of the spontaneous fission of uranium.

To clarify the question of the possibility of spontaneous fission of the nearest decay products of uranium, which are in an excited state after \(\beta\)-transition, we carried out special experiments with a layer of \(\mathrm{U}_3\mathrm{O}_8\) deposited on the chamber plates, enriched in \(\mathrm{UX}_1\) by a factor of 12 compared with its equilibrium content in U, and observed no increase in the effect. Therefore spontaneous fission should be attributed to one of the unexcited isotopes of U, with half-lives obtained from an estimate of our results:

\[ \mathrm{U}^{238} — 10^{16} \sim 10^{17}\ \text{years}, \]

\[ \mathrm{U}^{235} — 10^{14} \sim 10^{15}\ \text{years}, \]

\[ \mathrm{U}^{234} — 10^{12} \sim 10^{13}\ \text{years}. \]

We express our sincere gratitude for the direction of the work to Prof. I. V. Kurchatov, who outlined all the principal control experiments and took the most direct part in discussing the results of the investigations.

REFERENCES

1. N. Bohr and A. Wheeler, Phys. Rev., 56, 426, 1939.
2. W. F. Libby, Phys. Rev., 55, 1269, 1939.
3. Frisch, Nature, 143, 276, 1939.
4. K. A. Petrzhak and G. N. Flerov (in press, ZhETF).
5. T. I. Nikitinskaya and G. N. Flerov (in press, ZhETF).

ADDENDUM

Further experiments on the study of the phenomenon of spontaneous fission of heavy nuclei were carried out with the aid of the apparatus described above, with a somewhat modified method of recording fragments. Voltage pulses from the Wynn-Williams amplifier were fed to a power amplifier, at the output of which a ballistic galvanometer was connected. From the deflection of the galvanometer it was possible to judge the ionization produced by a fragment in the chamber.

Despite the nonlinearity of the system—ionization chamber–amplifier—this method could be used to compare the distributions, by pulse magnitude, of fragments produced as a result of irradiation of uranium by (Rn + Be) neutrons and of pulses appearing spontaneously. The coincidence of both curves in Fig. 5 is not only additional evidence of the fragmentary nature of the spontaneous pulses, but at the same time indicates that there is not too great a difference in a number of characteristics of the fragments produced in both types of fission—the fragment masses, energies, and effective ranges.

Fig. 5. ●—number of galvanometer deflections of a given magnitude in the presence of an (Rn + Be) source; ○—number of galvanometer deflections of the same magnitude in the absence of a source. \(N\)—number of deflections.

A series of experiments was carried out with an ionization chamber on whose plates thorium oxide was deposited instead of uranium oxide. On the basis of the negative result of the experiments and an estimate of the sensitivity of the method, it was possible to determine only the lower limit for the half-life of thorium by fission. The half-life proved to be greater than \(2 \cdot 10^{18}\) years. These results, as well as data on the abundance of the uranium isotopes \(U^{238}\), \(U^{235}\), \(U^{234}\) and experimentally obtained information on the degree of instability of nuclei as a function of their \(Z\) and \(A\), make it possible to put forward the hypothesis that one of the light isotopes of uranium, apparently \(U^{234}\), undergoes spontaneous fission.

For a more exact determination of the half-life of uranium by spontaneous fission, a series of experiments was carried out with layers of uranium oxide of various thicknesses. It turned out that for a layer

with uranium oxide with a surface density of \(1.4\ \mathrm{mg/cm^2}\), it could be assumed that all cases of fission of uranium nuclei were being recorded. The half-life of uranium, calculated for all atoms, proved to be \((4 \pm 1)\cdot 10^{16}\) years.

An estimate of the effective range of the spontaneously produced fragments from experiments with uranium-oxide layers of different thicknesses gave a value of the order of \(5\ \mathrm{mm}\) of air, which is in good agreement with the measurement of the effective range of the fragments under the conditions of a large ionization chamber.

A number of experiments were carried out by us at a depth of \(50\ \mathrm{m}\) underground at one of the stations of the Moscow Metro. These experiments were performed in order to establish definitively the impossibility of attributing the effect of spontaneous fission observed by us to one of the components of cosmic rays. In the underground experiments we obtained results analogous to those previously obtained in Leningrad at sea level. The intensity of all known components of cosmic rays should have decreased at this depth by no less than a factor of 40. Consequently, the absence of any noticeable change in the number of spontaneous decay events at this depth once again indicates the impossibility of explaining the effect by cosmic rays.