MODERN PHOTOCHEMISTRY*

E. V. Shpolsky, Moscow

The study of photochemical processes has made significant progress in recent years. We have advanced further in understanding the mechanism of the chemical action of light than in all other areas of chemical kinetics. Photochemistry owes these successes to the modern development of the theory of atomic phenomena, in particular the theory of atomic and molecular spectra. The study of photochemical processes apart from the spectral properties of molecules is at present impossible.

In what follows we shall give a brief survey of the present state of photochemistry, dwelling exclusively on the mechanism of the primary photochemical process, i.e., the process occurring immediately after the absorption of light, and limiting ourselves to the consideration of reactions in the gas phase and in solutions. We shall leave aside the photochemistry of crystals, which is of outstanding importance for understanding the nature of the photographic process, since its exposition would require consideration of a whole series of peculiar facts and would excessively broaden the scope of the present article1.

I. Photochemical Law of Equivalence

The first major impulse toward understanding the nature of photochemical processes was given by the formulation of the so-called Einstein photochemical law of equivalence[^2].

According to the original formulation of this law, the number of molecules that have undergone a photochemical transformation must be equal to the number of absorbed light quanta, or, in other words, the quantum yield \(\left(= \frac{\text{number of transformed molecules}}{\text{number of absorbed quanta}}\right)\) for all photochemical processes must be equal to unity. For at least ten years this law served as the “apple of discord” for photochemists of all countries. In fact, determinations of the quantum yield for numerous reactions showed that, although in a number of cases this law is obeyed, in others it is violated, and moreover in both directions and in the most pronounced manner. If one looks at the tables

quantum yields³, it is immediately apparent that, alongside values close to unity, one very often encounters values many times greater than unity (up to \(10^5\)), and values expressed by very small fractions (a few thousandths and even ten-thousandths). The final result of the heated controversy concerning this law was reached only in 1925 at the photochemical conference organized by the Faraday Society⁴. At the present time we may state with complete certainty that the photochemical law of equivalence, in its original primitive formulation, of course has no place and cannot have one. But if it is expediently modified and it is asserted that the number of primary photochemical acts is equal to the number of absorbed light quanta, then in this form the law raises no objections, since it is a simple consequence of the firmly established quantum nature of light absorption. For the modern photochemist, the magnitude of the quantum yield is the guiding thread in unraveling the nature of a photochemical process, since, along with kinetic data, this quantity immediately enables us to decide whether a given reaction is simple or chain-like, and in what direction the interpretation of the process must proceed. We shall not dwell further here on this law, since discussion of its fundamental bases is no longer necessary because they are generally known, while consideration of particular values of the quantum yield is meaningful only in connection with the examination of specific reactions. In the further exposition we shall repeatedly return to this law on various occasions.

II. Photochemical reactions of excited atoms

Turning to the consideration of the mechanism of elementary photochemical processes, we shall first of all dwell on the simplest case of reactions between excited atoms and molecules. Although such reactions are relatively rare, their consideration is of interest, since it makes it possible to establish certain propositions important for the subsequent analysis of more complex cases.

A classical example of photochemical reactions of the indicated type is provided by reactions of excited mercury atoms. It is known that when mercury vapor is illuminated with light of wavelength \(\lambda = 2537\) Å, the mercury vapor fluoresces intensely, emitting the same wavelength, 2537 Å. This is the so-called resonance fluorescence, whose occurrence is explained by the fact that the magnitude of the quantum corresponding to 2537 Å (\(4.9\) electron volts \(= 112.5\) kg cal/mol) is exactly equal to the separation between the normal (spectroscopic symbol \({}^{1}S_0\)) and the first excited (symbol \({}^{3}P_1\)) level of the mercury atom. Therefore mercury atoms that have absorbed quanta \(h\nu_{2537}\), after the mean residence time in the excited state \(\tau = 10^{-8}\) sec. has elapsed, can return only to the ground state, emitting light of the same wavelength \(\lambda = 2537\) Å. If, however,

if a foreign gas is mixed with mercury vapor, the fluorescence is partly (sometimes very substantially) quenched. Especially intense quenching is observed when oxygen and hydrogen are admixed; in both cases a photochemical reaction is observed simultaneously: in the first case mercury oxide, \(\mathrm{HgO}\), is formed; in the second, free hydrogen atoms appear and, in addition, mercury hydride, \(\mathrm{HgH}\). Let us dwell somewhat on the first case. The processes occurring here may be represented schematically by the equations

\[ \mathrm{Hg}+h\nu \to \mathrm{Hg}^\ast, \tag{1} \]

\[ \mathrm{Hg}^\ast \to \mathrm{Hg}+h\nu, \tag{2} \]

\[ \mathrm{Hg}^\ast+\mathrm{O}_2 \to \mathrm{HgO}+\mathrm{O}, \tag{3} \]

where the asterisk denotes an excited mercury atom possessing a large excess of energy, \(112.5\ \mathrm{kg/cal}\) per mole. Equations (2) and (3) show that the excited mercury atom may meet a twofold fate. If, during the time interval \(\tau = 10^{-8}\) sec, it does not undergo a collision with an oxygen molecule, then the absorbed quantum of light will be emitted as a quantum of the same kind—resonance fluorescence will occur. If, however, during the same interval of time it undergoes a collision, then its excess energy will be used for the chemical process, and the fluorescence is quenched. Thus quenching of fluorescence is a sign of the occurrence of a photochemical reaction. It can easily be shown\(^{5,6}\) that the quantum yield of the reaction in this case is given by the formula

\[ \gamma=\frac{\tau Zp}{1+\tau Zp}, \tag{1} \]

where \(Z\) is the number of collisions at a pressure of \(1\ \mathrm{mm}\), and \(p\) is the pressure in \(\mathrm{mm}\)*. We see from this that the quantum yield depends substantially on the gas pressure. Only under the condition \(\tau Zp \gg 1\), i.e., at sufficiently high pressure, does the quantity \(\gamma\) reach the value 1; at low pressures \(\gamma\) is certainly less than 1. Likewise, at constant pressure, \(\gamma\) depends substantially on the number of collisions \(Z\), which is directly related to temperature. Thus, the characteristic signs of reactions of excited atoms are a small value of the quantum yield and a strong dependence of this value on temperature and pressure.

In formula (1), besides \(Z\) and \(p\), there also enters \(\tau\)—the residence time in the excited state: as \(\tau\) increases, the rate of the reaction increases, and consequently so does the quantum yield. A significant increase in \(\tau\) can be achieved if the excited atom is transferred into a metastable state, in which it (the excited atom) can remain for time intervals considerably exceeding \(10^{-8}\) sec. Thus, for example, in the mercury atom

* It is assumed that the reaction occurs at every collision of an excited mercury atom with an oxygen molecule.

(Fig. 1), close to the excited state under consideration, there lies the level \(^{3}P_{0}\) (at a distance of \(0.2\ \mathrm{V}\) from it), the transition from which to the normal level \(^{1}S_{0}\) is “forbidden.” Thus, an atom that has entered this state cannot spontaneously return to the normal state with the emission of a light quantum. It will remain in the metastable state, retaining its store of energy, until a collision occurs in which this metastable atom can give up its excess energy, or until, as a result of a collision, the atom receives the additional energy needed in order to return to the level \(^{3}P_{1}\), from which spontaneous return to the normal level \(^{1}S_{0}\) with emission of the excess energy is possible. If one takes into account that thereby \(\tau\) in the formula is very considerably increased (up to \(10^{-2}\) sec.), it is clear that in the presence of metastable atoms the reaction rate must increase very strongly.

Fig. 1.

It turns out that if nitrogen or water vapor is added to mercury vapor, the excited mercury atoms \(\mathrm{Hg}(^{3}P_{1})\) will, with high probability, transfer to these foreign molecules an energy of \(0.2\ \mathrm{V}\), passing into the metastable state. And indeed, Wood and Gaviola\(^7\) established that the addition of \(\mathrm{N}_{2}\) or water vapor to a mixture

\[ \mathrm{Hg} + \mathrm{O}_{2} \]

has a strong accelerating effect on the reaction of photochemical oxidation of mercury.

In connection with this it is interesting to note that Mecke and Childs\(^8\) recently found a metastable state in the oxygen molecule (spectroscopic symbol \(^{1}\Sigma\)), the lifetime of which was estimated at \(7\) sec. Oxygen molecules in this state possess an excess energy of \(37\ \mathrm{kg\ cal/mole}\) and are therefore capable of producing considerable chemical effects. Recently Kautsky\(^9\) attributed to these metastable oxygen molecules the principal role in the processes of sensitized photooxidation, photosynthesis of carbohydrates, etc.

In conclusion, one should note one more extremely characteristic feature of the interactions of excited atoms. Experiment shows that in a collision the probability of energy transfer from an excited atom to an unexcited atom or molecule increases many times if the colliding particle has an energy level close in magnitude to the energy being transferred. Franck formulated this peculiar resonance condition as follows: the probability of energy transfer will be the greater, the smaller the fraction of the transferred energy that passes into the kinetic energy of translational motion of the colliding partners. This resonance in energy transfer can be demonstrated most clearly in the case of the “sensitized fluorescence” of sodium vapor in the presence of mercury vapor.\(^ {10}\) If a mixture of Na and Hg vapors is illuminated with light of the resonance line of mercury \((\lambda = 2537\ \text{Å})\), then one observes

sensitized fluorescence of sodium vapor is observed: the resonance line of Hg is not absorbed by pure sodium vapor; therefore, upon illumination with \(\lambda = 2537\) Å, pure sodium vapor does not fluoresce, but in the presence of Hg vapor fluorescence arises, caused by the transfer of energy in collisions from excited Hg atoms \(({}^{2}P_1)\) to Na atoms. It turns out, moreover, that the greatest intensity in the light of such sensitized fluorescence is possessed not by the resonance \(D\)-line of sodium, \(\lambda = 5890\) Å, but by the doublet \(\lambda = 4223/4420\) Å \((2\,{}^{2}P_{\frac{3}{2},\,\frac{1}{2}} — 7\,{}^{2}S_{\frac{1}{2}})\). On the other hand, the excitation energy of Na in the state \(7\,{}^{2}S_{\frac{1}{2}}\), equal to \(4.880\) eV, almost coincides with the excitation energy of the state \({}^{3}P_1\) for Hg \((4.86\ \mathrm{eV})\). If, however, a certain amount of nitrogen is also added to the mixture of Na and Hg vapors, then in a collision of an excited Hg\(({}^{3}P_1)\) atom with an \(\mathrm{N}_2\) molecule, the former with high probability* gives up a small excess of energy \({}^{3}P_1 — {}^{3}P_0\) to the nitrogen molecules and passes into the metastable state \({}^{3}P_0\). Under these conditions, the greatest intensity in the spectrum of sensitized fluorescence of Na is acquired by a new pair of lines, \(2752/2748\) Å, which is caused by the transitions \(2\,{}^{2}P_{\frac{3}{2},\,\frac{2}{1}} — 5\,{}^{2}S_{\frac{1}{2}}\); but the energy of the Na state \((5\,{}^{2}S_{\frac{1}{2}})\), equal to \(4.68\) eV, is very close to the \({}^{3}P_0\) level of the mercury atom \((4.64\ \mathrm{V})^{11}\).

The physical mechanism underlying these resonance phenomena in energy transfer can be understood only on the basis of wave mechanics \(^{12}\). Consideration of the questions involved here would lead us too far afield. Let us note only that the phenomena of quantum-mechanical resonance undoubtedly play a large role in the processes of photochemical sensitization.

III. Photochemical decomposition of diatomic molecules

1. Let us now turn to consideration of the most important elementary photochemical process—the photochemical dissociation of a molecule—and begin with the simplest case: the dissociation of a diatomic molecule. Photochemical decomposition occurring immediately after absorption of a light quantum was first discovered by Terenin \(^{13}\) for the case of vapors of alkali-halide salts. Somewhat earlier, Franck \(^{14}\) gave a completely clear picture of the physical mechanism of this decomposition. As is known, the spectra of molecules differ from atomic spectra by a significantly greater complexity and abundance of lines. This is explained by the fact that in a molecule, in addition to the purely electronic energy levels, the only possible ones in the case of atoms, there are also levels of vibrational and rota-

* This ease of transfer is also based on resonance between the difference of the \({}^{3}P_1 — {}^{3}P_0\) levels and the vibrational levels of the \(\mathrm{N}_2\) molecule.

tional energy. Let \(A\) and \(B\) (Fig. 2) be two electronic energy levels. In the case of an atom, the presence of these levels determines the only possibility of the transition \(A \to B\) in the case of absorption and \(B \to A\) in the case of emission. In a molecule, however, above each of these levels there lies a series of levels \(A_1, A_2 \ldots, B_1, B_2 \ldots\), determined by the possibility of vibrations of the atoms of the molecule relative to one another, and in the intervals between the levels \(A\) and \(A_1\), \(A_1\) and \(A_2 \ldots, B\) and \(B_1\), etc., there lies still another series of levels (not shown in Fig. 2), corresponding to various states of rotation of the molecule. Since transitions may occur between all these levels, accompanied by absorption (or emission) of light, it is clear that instead of a single line (corresponding to the transition \(A \to B\)) we obtain a large number of lines, grouped into bands, very often characterized by the presence of a sharp edge or head. The position of these heads corresponds to the sequence of vibrational levels. Tracing the course of the vibrational levels\(^{15}\) shows that these levels, both in the lower normal state and in the upper, excited state, gradually approach one another up to complete coalescence. Beyond this coalescence lies a continuous region of non-quantized energy values.

Fig. 2.

Such a course of the vibrational levels corresponds to the gradual weakening of the bonds in the molecule as its store of vibrational energy increases, so that ultimately the bond between the atoms is destroyed and the molecule dissociates into its constituent atoms. This dissociation is precisely what corresponds to the point of coalescence of the levels. With a further increase of vibrational energy, the excess above the energy necessary to raise the system to the point of coalescence is converted into the kinetic energy of the separated atoms. And since the latter is not quantized, i.e., may have any values, the region lying above the point of coalescence has a continuous character—no levels can be distinguished in it.

The gradual convergence of the vibrational levels in the spectrum of the molecule corresponds to a gradual convergence of the heads of the bands up to their complete coalescence. Beyond the point of coalescence, on the side of the shorter waves, there must lie a continuous spectrum corresponding to transitions into the hatched region of Fig. 2.

Series of bands of this kind, ending at a point of coalescence with a continuous spectrum adjoining it, are in fact observed. This whole picture can be followed with particular clarity in the spectrum of molecular iodine\(^{16}\). From what has been said it is clear that, by determining the position of the point of coalescence, one can calculate the dissociation energy of the molecule. Thus, for iodine molecules the point of coalescence is observed at \(\lambda = 4\,89{,}3\ \text{Å}\). To this wavelength there corresponds a quantum energy of \(2{,}468\ \mathrm{eV}\);

upon absorption of such a quantum the \(J_2\) molecule dissociates into two atoms, of which one is in the normal state and the other in an excited (metastable) state with an excitation energy of \(0.937\ \mathrm{eV}\). Consequently, the energy going into the dissociation process itself will be \(2.488 - 0.937 = 1.531\ \mathrm{eV}\), or \(35\ kg\ cal/mole\). Thus a new and very precise method is opened for determining the dissociation energy of molecules—a quantity so difficult to measure by ordinary physicochemical methods.*

From the photochemical point of view, it is especially interesting that in the region of continuous absorption “optical” dissociation must occur. The fact that this dissociation actually takes place for the case under consideration, molecular iodine, has been demonstrated by the most diverse methods: by the disappearance of fluorescence upon illumination with wavelengths from the continuum region \((\lambda < 4995)^{18}\), by the decrease in pressure resulting from adsorption by the walls of the vessel of the free iodine atoms produced \(^{19}\), by the change in the thermal conductivity of illuminated iodine vapor \(^{20}\), and, finally, directly spectroscopically, by the appearance of absorption lines of iodine atoms \(^{21}\).

Fig. 3.

The physical mechanism of this dissociation becomes especially clear from consideration of the curves of potential energy of the normal and excited molecules. The form of these curves, representing the dependence of the potential energy of the atoms of a molecule on their distance, is shown in Fig. 3. When the atoms approach one another, their potential energy decreases to a certain minimum corresponding to the distance between nuclei at equilibrium \(r_0\); upon further approach the potential energy rises steeply owing to the appearance of repulsive forces. Obviously, the ordinate equal to the magnitude of the arrow \(D\) is equal to the work that must be expended in order to remove the atoms from the equilibrium position to an infinite distance, i.e. to the dissociation energy. Curve \(I\) corresponds to the normal molecule, curve \(II\) to the excited one. If electronic excitation is accompanied by a weakening of the bonds in the molecule (which happens very often), then curve \(II\) must be shifted to the right relative to curve \(I\) (the internuclear distance at equilibrium increases), and, in addition, the dissociation energy \(D'\) will be less than \(D\). If now, as a result of electronic excitation, a transition occurs from curve \(I\) to curve \(II\), then, as Franck first showed \(^{6}\), this transition must occur in such a way that at the first moment the internuclear distance does not change (Franck–Condon principle).

* Cf. the tables of dissociation energies determined spectroscopically, compiled by Gerda Sponer \(^{17}\).

Indeed, since electronic excitation is an instantaneous process, at the moment when the excitation has already ended, the heavy nuclei still occupy their former positions. As a consequence of this, as is easy to see from the figure, even when in the normal state we have a non-oscillating molecule (or, more precisely, a molecule with zero oscillations), which is the case in the majority of instances when the gas temperature is not too high, upon transition to curve II the nuclei acquire an excess of potential energy and begin to oscillate strongly. If, moreover, it turns out that the point \(A'\), into which the vertical arrow \(AA'\) falls, lies above the horizontal asymptote of the upper curve II, then during the first half-period of oscillation (\(10^{-13}\) sec) the molecule will be torn into parts, since its potential energy at point \(A\) will be greater than the energy of dissociation in the excited state.

Fig. 4.

1. Another case, presenting features of fundamental interest, is encountered when the absorption spectrum of a gas has no discrete part at all, consisting of separate lines, but is entirely continuous. Such a case is observed, for example, in the hydrogen halide molecules HJ, HBr, HCl\({}^{22}\). In this case the potential curve of the excited molecule corresponds entirely to an unstable state. The existence of such states was first theoretically substantiated by London\({}^{23}\) in constructing the quantum-mechanical theory of the homeopolar bond. In the simplest case of the interaction of two hydrogen atoms it turns out that in only \(25\%\) of all possible cases does the interaction of hydrogen atoms proceed along the “stable” potential curve I, possessing a minimum that corresponds to the equilibrium position; in \(75\%\) of cases the atoms repel one another at all distances. In all these cases the potential curve rises monotonically upward as the distance decreases (curve II in Fig. 5), and at no distance is a stable molecule formed. If now the absorption spectrum of the molecule is continuous throughout, this means that the transition to the excited state is always accompanied by dissociation. The mechanism of such a process is depicted—

Fig. 5.

appears to be Fig. 5, which, after what has been said, needs no special explanation.

3. Let us consider, finally, the case of decomposition in the so-called predissociation spectrum. The phenomenon of predissociation, discovered by V. Henri^24, is outwardly manifested in the fact that in some molecular spectra, when the bands are followed toward shorter waves, beginning with a certain wavelength, sometimes suddenly and sometimes gradually, the fine structure of the bands disappears. The bands as such continue to exist and do not merge into one continuous absorption region, but the individual lines of which they consist, and which are due to the rotation of the molecule, become broadened and indistinguishable. Spectra of this kind are observed very often, and moreover both for simple and for complex molecules (NH$_3$, NO$_2$, CS$_2$, benzene, formaldehyde, and many others).

The correct explanation of the origin of predissociation was first indicated by Born and Franck^25. Briefly stated, it amounts to the following: the disappearance of the fine structure of the bands is caused by dissociation during a time comparable with the period of rotation of the molecule (10$^{-12}$–10$^{-10}$ sec.). Since Rayleigh we know that an ideally monochromatic and, consequently, ideally fine spectral line is an abstraction. Such a line could arise only if the oscillations exciting it continued infinitely long and, consequently, the wave train were infinitely long. In reality the oscillations continue for a limited time, which means, in the language of quantum theory, that the residence time in the excited state $\tau$ has a finite value. Therefore a real spectral line always has a finite width, which is further increased by broadening due to the Doppler effect. The natural width of a spectral line $\Delta\nu$ is related to the time $\tau$ during which the oscillations occur (or, what is the same thing, to the residence time in the excited state) by the relation

\[ \tau \cdot \Delta\nu = 1. \]

For a normal value of $\tau$ (10$^{-8}$ sec.) the width $\Delta\nu$ amounts to several hundredths of an ångström. If, however, $\tau$ decreases, then $\Delta\nu$ increases correspondingly. In the case of predissociation the width $\Delta\nu$ is such that it exceeds the distance between the separate rotational lines of the band. Hence one can estimate the order of magnitude of $\tau$ as 10$^{-12}$–10$^{-10}$ sec., which corresponds to the period of rotation of the molecule. That the decomposition of the molecule actually occurs after the lapse of this interval of time is attested by numerous experiments of Henri himself and of others^27.

We shall consider the mechanism of this decomposition with the aid of a scheme of energy levels and potential curves. The occurrence of decomposition here can be followed with the aid of the scheme in Fig. 6. Let us assume that, alongside the excited state $B$, into which the molecule passes directly upon excitation, there exists also a second state $B'$, situated in such a way that the discrete level of the state $B$ cor—

occurs at the same energy a continuous region of the state \(B'\). Then, for a molecule in state \(B\), there arises a definite probability of transition to a state of equal energy belonging to the system \(B'\). But since at the height of level \(B\) there already lies a continuous region of the system \(B'\), this transition ends in the dissociation of the molecule. It turns out, however, that the transition from the discrete energy level shown in Fig. 7 to the continuous absorption region can occur only when an additional condition is satisfied, namely that

Fig. 6.                Fig. 7.

the potential curves corresponding to the states \(B\) and \(B'\) must intersect. Let us consider two typical cases. Let \(n\) be the potential curve of the normal state; \(\alpha\) and \(\alpha'\) are the curves of the excited states. A transition from the curve \(n\) leads directly to the curve \(\alpha\), since, according to the Franck–Condon principle, this transition must occur without a change in the internuclear distance. If this transition, as, for example, the transition \(AB\), leads to a point of the curve \(\alpha\) situated below the point of intersection of the curves \(\alpha\) and \(\alpha'\), then nothing unusual occurs. The resulting absorption bands have a normal structure. If, however, the transition leads to the point \(D\), lying below the horizontal asymptote of the curve \(\alpha\), but above the intersection of the curves \(\alpha\) and \(\alpha'\), then, during vibrations in the “potential well” \(DBEF\), the molecule will each time pass through the point of intersection of the curves \(\alpha\) and \(\alpha'\), and there will thereby arise a probability of transition from the curve \(\alpha\) to the curve \(\alpha'\). As a result, after several vibrations along the curve \(\alpha\), the molecule will pass to the curve \(\alpha'\). But once such a transition has occurred, it will inevitably end in dissociation, since the potential energy of the molecule will be greater than the dissociation energy corresponding to the curve \(\alpha'\). Thus, so long as transitions from the curve \(n\) lead to that same part of the curve

of \(\alpha'\), which lies below the level \(EG\) corresponding to the intersection of the curves, the bands have a normal form. As soon as a transition arises to a point lying at the level \(n\) or above it, it ends in dissociation, and the band acquires a diffuse form.

Another characteristic case is shown in Fig. 8. Here the second curve \(\alpha'\) is an unstable repulsion curve. The difference from the preceding case is as follows. Suppose the molecule is at the level \(H\) of the curve \(\alpha\). At the same height as this level lies the point \(K\) of the potential curve \(\alpha'\), upon reaching which the molecule must undergo decomposition. But the points \(H\) and \(K\) are separated by the potential barrier \(HEK\), and therefore it is necessary to raise the molecule to the level of the top of this barrier, \(E\), in order that it may roll down along the curve \(\alpha'\). However, from quantum mechanics we know that between two regions of equal energy separated by a potential barrier, “tunnel” transitions under the barrier are possible, although of low probability. Moreover, the closer to the top of the barrier the molecule has been raised, the greater the probability of a tunnel transition. Thus we may expect that, in the case when the potential curves have the character described, diffuseness does not appear at once, but increases gradually, as the probability of transition to the curve \(\alpha'\) increases.

Fig. 8.

1. We saw above that iodine molecules, when illuminated by light from the region of the continuous absorption spectrum \((\lambda < 4989\ \text{Å})\), undergo spontaneous dissociation in a time interval comparable with the period of vibration. At wavelengths greater than \(5000\ \text{Å}\), the absorption spectrum of iodine is discrete and decomposition is not observed, although in each elementary act the molecule can absorb quanta exceeding its dissociation energy. It turns out, however, that if a foreign gas, for example argon, is mixed with iodine vapor, dissociation is observed both in the continuous and in the discrete parts of the spectrum \((\lambda > 5000\ \text{Å})\) \(^{28}\). Obviously, in this case dissociation in the discrete part of the spectrum is the result of a collision experienced by the excited molecule during the time with a foreign molecule. The cause of this forced decomposition is the phenomenon of so-called induced predissociation, which plays a large role in photochemical processes. Kronig \(^{29}\) showed that, for predissociation to occur, in addition to the conditions considered above, a number of other conditions must be fulfilled. Thus, for example, conservation of total angular momentum is necessary

amount of motion, the observance of a whole series of symmetry properties, etc. If any of these conditions is not fulfilled, then, although the coincidence of a discrete level of one system with the continuous region of another system will take place, predissociation will not occur, and the molecule will preserve its integrity. If, however, factors are present that make possible a violation of the selection rules, then the transition to the unstable potential curve becomes possible, and absorption ends in decomposition. Iodine has precisely a repulsive potential curve (Fig. 9), transition to which under normal conditions is forbidden \(^{30}\). It is sufficient, however, to place the illuminated vapors in a magnetic field in order to create conditions that remove the prohibition of the transition. And indeed, it has long been known that in a magnetic field the fluorescence of iodine vapor is quenched. This quenching is explained by the occurrence in the magnetic field of a “forbidden” transition to the repulsive curve, as a result of which the excitation energy is not returned in the form of fluorescence light, but is expended on dissociation of the molecule. The same role of a factor permitting, under normal conditions, a forbidden transition to the repulsive curve may also be played by collisions with foreign molecules.

Fig. 9.

In order to clarify the role of collisions, Kondrat’ev and Polyak \(^{31}\) studied the influence of gases with different physical and chemical properties; namely, nitrogen was taken as an inert gas, oxygen as a paramagnetic gas, and hydrogen chloride as a polar gas. It turned out that the order of magnitude of the inducing action of all three gases is the same. From this it may be concluded that, for induced predissociation to occur, only a collision with some foreign particle is necessary, while the individual properties of this particle play no role. This confirms the supposition expressed by Terner, according to which the reason for the prohibition of the transition to an unstable potential curve is the violation, in this transition, of the conservation law for the total angular momentum. The role of the foreign particle in such a case is reduced to its taking upon itself the excess angular momentum, so that for the system iodine molecule + foreign particle the law of conservation of angular momentum will be fulfilled. We thus see that spontaneous photochemical

decay in a continuous spectrum always occurs, whereas induced decay may also occur in a discrete absorption spectrum.

IV. Photochemistry of Complex Molecules

1. Analysis of the primary photochemical process in the case of complex molecules is, of course, associated with great difficulties. However, in this field too, considerable successes have recently been achieved.

As in the case of diatomic molecules, the guiding thread for understanding the primary photochemical act is furnished above all by absorption spectra. It has long been known[^32] that certain groups of atoms reveal themselves in a definite way in the spectra of various complex molecules of which they form a constituent part. Thus, for example, the carbonyl group \(> C = O\) gives a characteristic absorption band with a maximum at \(2800\ \text{Å}\); the benzene ring or the phenyl radical, \(\mathrm{C_6H_5}\), shows absorption in the region from \(3000\) to \(2300\ \text{Å}\), and so on. Such groups are called “chromophoric,” although this word must be put in quotation marks because the “color” often turns out to be “ultraviolet.” On the other hand, it is known that the introduction, alongside the chromophoric group, of other groups into the molecule often leads to a shift of the absorption region of the chromophoric group. In addition, the absorption spectrum of chromophoric groups in complex molecules is usually so diffuse that it cannot yet serve as an unambiguous indication of the presence of one group or another. In this respect, significant help is provided by the study of the frequencies of the molecule’s normal vibrations. Here the vibration frequencies in the excited state can be found from absorption spectra, if the latter preserve their banded structure. However, especially abundant material is available for the vibration frequencies in the normal state of the electron shell, since these latter vibrations can with great convenience be found by means of the Raman effect[^33].

For example, the absorption spectrum of formaldehyde \(\mathrm{H_2CO}\) has a discrete character and reveals two vibration frequencies: \(\omega_1' = 1180\) and \(\omega_2' = 830\ \mathrm{cm^{-1}}\). The Raman effect makes it possible to find in the same formaldehyde the frequencies \(\omega_1'' = 1768\) and \(\omega_2'' = 1039\ \mathrm{cm^{-1}}\). Thus the frequencies detected by the absorption spectrum are noticeably lower than the frequencies appearing in the Raman effect. This is explained by the fact that the latter makes it possible to find frequencies in the ground state, whereas the absorption spectrum gives frequencies in the state with an excited electron shell, when the bonds have already been weakened and therefore the frequencies are naturally smaller. Indeed, direct determination of the vibration frequencies in the ground state for a gaseous* formaldehyde molecule, carried out with the aid of the fluorescence spectrum of \(\mathrm{H_2CO}\), gave \(\omega_1'' = 1713\) and \(\omega_2'' = 1039\ \mathrm{cm^{-1}}\). The first frequency \((1713\ \mathrm{cm^{-1}})\) undoubtedly belongs to the carbonyl-

* This and the following examples are borrowed from the work of A. N. Terenin[^34].

group, since it appears in all compounds containing this group. The lower frequency, \(\omega_2' = 1023\ \mathrm{cm}^{-1}\) and \(\omega_2' = 830\), according to Herzberg, is due to transverse vibrations of the H atoms in the molecule

\[ \begin{array}{c} \mathrm{H}\\[-2mm] \ \backslash\\[-2mm] \mathrm{H}/ \end{array} \mathrm{C}=\mathrm{O}. \]

Another very characteristic example is provided by the benzaldehyde molecule

\[ \begin{array}{c} \mathrm{H}\\[-2mm] \ \backslash\\[-2mm] \mathrm{C_6H_5}/ \end{array} \mathrm{CO}, \]

which differs from formaldehyde by the introduction of a phenyl group \(\mathrm{C_6H_5}\). This molecule gives a banded absorption spectrum, located in the region from 2700 to 2400 Å and showing frequencies 945 and \(197\ \mathrm{cm}^{-1}\), characteristic of the benzene ring. It is noteworthy that the presence of the carbonyl group has no effect on the absorption spectrum, since in the region \(\lambda = 2800\) Å no absorption is observed. Likewise, the absorption spectrum of acetophenone

\[ \begin{array}{c} \mathrm{CH_3}\\[-2mm] \ \backslash\\[-2mm] \mathrm{C_6H_5}/ \end{array} \mathrm{CO} \]

in its position and in the frequencies it exhibits is also characteristic of the phenyl group and does not reveal the presence of the carbonyl. The last two examples are especially interesting from the point of view that they show that if a molecule contains two chromophoric groups, absorption may take place at one of them, while the other remains unaffected.

1. Turning to the consideration of various types of photochemical reactions in complex molecules, we shall first of all dwell on cases in which the elementary act reduces to excitation of the molecule. The most interesting example of such reactions may be the so-called transformation of stereoisomers occurring under the action of light. Thus, trans-stilbene is converted into cis-stilbene

\[ \begin{array}{ccc} \mathrm{C_6H_5{-}C{-}H} & & \mathrm{H{-}C{-}C_6H_5}\\ \| & +\,h\nu \ \longrightarrow & \|\\ \mathrm{H{-}C{-}C_6H_5} & & \mathrm{H{-}C{-}C_6H_5}\\ \text{trans-stilbene} & & \text{cis-stilbene} \end{array} \]

Maleic acid is converted into fumaric acid:

\[ \begin{array}{cc} \mathrm{H{-}C{-}COOH} & \mathrm{HOOC{-}C{-}H}\\ \| & \|\\ \mathrm{H{-}C{-}COOH} & \mathrm{H{-}C{-}COOH}\\ \text{maleic acid} & \text{fumaric acid} \end{array} \]

Analysis of the absorption spectra of these substances showed that the chromophoric character here belongs to the ethylene double bond \(\mathrm{C{=}C}\): namely, absorption of light occurs on the electrons forming this bond. But in that case the transformation reactions written above receive a very simple explanation. Absorption of light breaks one of the two bonds \(\mathrm{C{=}C}\), as a result of which rotation is restored

free rotation about the remaining bond, and the molecule, returning to the normal state, may pass either into one or into the other stereoisomeric form. In the case of the transformation of maleic acid into fumaric acid, the probabilities of transition from the excited state into one form or the other differ little from one another, which explains the small quantum yield of this transformation; on the contrary, in the transformation of trans-stilbene into cis-stilbene the quantum yield is of the order of unity, which indicates a sharply expressed preference for transition from the excited state into the form of cis-stilbene. It is evident that the two forms (trans-cis) are energetically separated by a potential barrier, for the overcoming of which the molecule must be raised to a higher energy level. But upon returning from the excited state to the normal state of the cis-isomer, the excess energy must be given off either in the form of fluorescence light or expended by collisions of the second kind. In fact, the transformation of maleic acid into fumaric acid is not accompanied by radiation; consequently, the excess energy is given off here by a collision of the second kind; on the contrary, trans-stilbene fluoresces upon illumination. Since the quantum yield in the latter case is equal to unity, the emission of light must occur simultaneously with the act of transformation.

An experimental proof of the hypothesis set forth for the case of the transformation of trans-stilbene into cis-stilbene may be served by the fact that the quantum yield for wavelengths corresponding to the absorption by the electron of the ethylenic bond is equal to unity.

1. We shall now turn to the processes of decomposition of complex molecules, occurring in a single elementary act in a time interval of the order of a vibrational period, i.e. \(10^{-13}\)—\(10^{-14}\) sec. The absorption spectrum in the presence of decomposition in such a time interval must be continuous. However, the converse assertion in the case of complex molecules is, generally speaking, not true: here the absorption spectrum very often has a continuous character, but by no means always is the cause of this diffuseness dissociation occurring upon absorption. The presence of many parts in a molecule and the mutual influence of their electrostatic fields often lead to a complete blurring of the spectrum (as a result of the Stark effect) without any dissociation.

Thus, from the appearance of the spectrum alone, in the case of complex molecules we cannot judge whether dissociation occurs or not.

For detecting decomposition and identifying its products, one has to apply other methods. In order, however, to “catch” the products of direct decomposition, it is necessary to use methods that make it possible to discover these products within a very short time after the completion of the act of decomposition, since the resulting free atoms or radicals may have time to recombine or to enter into reaction with other molecules. The most refined method, as always, is the spectroscopic method. If the magnitude of the absorbed quantum is sufficient not only to destroy the molecule, but also to transfer

if the decomposition products are in an excited state, then these products can be identified from the spectrum of the fluorescence that arises. In the case of complex molecules consisting of H, C, and O atoms, the quanta required for this are so large that the corresponding wavelengths lie in the Schumann ultraviolet (wavelengths of the order of 1000 Å). A. N. Terenin^35 succeeded in overcoming the experimental difficulties connected with work in this region of the spectrum, where even air already gives strong absorption, and in studying the decomposition of a number of molecules—namely, the molecules of water, methyl and ethyl alcohol, formic and acetic acid, CH₃CN and NH₃. In all these cases it was possible to detect the emission spectra of the radicals OH, CN, and NH₂, whence it follows that, upon decomposition in the Schumann region of the spectrum, the aforementioned radicals are obtained in an excited state.

Another method, applied by Terenin for the rapid capture of decomposition products, is of a more chemical character. It is based on the use of Paneth’s method for detecting free alkyl radicals (CH₃, C₂H₅, etc.) and consists in the following. Onto the walls of the vessel (or tube) in which the reaction takes place, a mirror of various pure metals (bismuth, tellurium, lead) is deposited; when the radicals interact with the metal, volatile organometallic compounds are formed, and the mirror disappears. By following the disappearance of the metal of the mirror by measuring its transparency with the aid of a photoelement, Terenin converted this method from a qualitative one into a quantitative one. A peculiar modification of this method was proposed by Mortensen and Leighton. By depositing, instead of lead, its radioactive isotope (radium D) as the mirror and following the transfer of radioactivity from the mirror to the trap, it was possible to increase the sensitivity of the method considerably. In this way it is possible to observe a stationary concentration of radicals corresponding to a partial pressure of \(10^{-8}\) mm.

With the aid of the methods described, Terenin studied the decomposition of the organometallic compounds Hg(CH₃)₂ and Pb(C₂H₅)₄, and Mortensen and Leighton studied the decomposition of Pb(CH₃)₄. In all cases it was found that the decomposition is indeed accompanied by the splitting off of free radicals. It should be noted that the “half-life” of free radicals is of the order of magnitude of one thousandth of a second.

In all the cases considered, the decomposition occurs at a definite bond and is, in essence, completely analogous to the dissociation of diatomic molecules in the true continuum.

1. Let us now turn to the consideration of the decomposition of complex molecules in a time interval of the order of the period of rotation of the molecule (\(10^{-10}\)—\(10^{-12}\) sec). Such decomposition takes place as a result of predissociation (see III, §§ 3, 4), which is observed in the spectra of complex molecules rather often (for example, in phosgene, thiophosgene, acetone, formaldehyde, acrolein, etc.). In the case of complex molecules the mechanism of this decomposition process in the predissociation spectrum, which reduces to a “switching over” of energy from a stable exc-

from the excited state to an unstable state of the same energy reveals certain very characteristic and interesting features (Franck, Sponer, and Teller)\(^{36}\). In order to clarify these features, let us return to the picture of potential curves considered in connection with the predissociation of diatomic molecules. As we have seen, predissociation of diatomic molecules occurs in the case of an intersection of the potential curve of the excited state \(a\) with another potential curve \(a'\), corresponding to a less stable or unstable state. In such a case, as was explained, there arises a probability of transition from the stable potential curve to the unstable one—a transition that ends in dissociation. Owing to the considerable shortening of the lifetime of the excited molecule (by \(10^3\)—\(10^4\) times), the rotational lines of the bands broaden to the point of merging, and the band assumes a diffuse appearance.

In the case of polyatomic molecules the process described is considerably more complicated. The chief complication consists in the fact that even when the vibrational energy is sufficient to bring the nuclei into a position in which predissociation is possible, a very long interval of time may elapse before this position is actually reached.

Let us consider the reason for this “delay.” In the case of diatomic molecules the energy state of the molecule is determined by only one parameter. For polyatomic molecules the relative arrangement of the nuclei is, generally speaking, determined by several parameters. Let us consider the simplest case and suppose that we are dealing with a molecule whose state is determined by two parameters.* Accordingly, the potential curves \(a\) and \(a'\) are replaced here by potential surfaces \(A\) and \(A'\), and the point of intersection of the curves is replaced by a line of intersection. If the state of our polyatomic molecule is known, being specified by the values of two parameters, then this state will be represented on the potential surface by some “representative” point, the two coordinates of which are determined by the given values of the parameters. While vibrations (changes of the parameters) are taking place in the molecule, the “representative point” will describe some curve on the potential surface. It is obvious that only at the moment when, moving along this curve, the “representative point” falls on the line of intersection of the potential surfaces \(A\) and \(A'\), will the nuclei in the molecule be brought into a state in which predissociation is possible. But before this moment, from the beginning of the excitation of the molecule, a very considerable interval of time may elapse. In fact,

* In reality, even in the case of a triatomic molecule it is necessary to know three parameters. However, even the simplified model with two parameters makes it possible conveniently to trace all the characteristic features of the phenomenon. The complications connected with an increase in the number of parameters only considerably enhance those features which are revealed in consideration of the simplest model.

let us examine in somewhat greater detail the various possible cases of motion of the “representative point”:

a) The representative point performs a one-dimensional motion on the potential surface. This means that only one of the two “normal vibrations” is excited. If, in this one-dimensional motion, the representative point falls on the line of intersection of the potential surfaces \(A\) and \(A'\), then the conditions will be realized under which predissociation can occur. It is clear that in this case the process of predissociation proceeds in exactly the same way as in diatomic molecules.

b) Both normal vibrations are excited. In such a case the representative point traces, on the potential surface, generally speaking, a very complicated Lissajous curve. In such a case it is clear that, before the conditions necessary for the occurrence of predissociation are created, i.e. before the representative point passes at the proper place through the line of intersection of the potential surfaces, a considerable interval of time may elapse.

c) The representative point performs a one-dimensional motion, but its trajectory does not at all pass through the line of intersection of the potential surfaces. In this case predissociation does not occur at all. If, however, one takes into account that the forces under whose influence the vibrations occur differ from quasi-elastic ones, so that the vibrations are anharmonic, then, as is known from the theory of vibrations, normal vibrations cannot in general be regarded as unrelated: if one has arisen, then after a certain interval of time the other will necessarily be excited as well. But in such a case the one-dimensional trajectory of the representative point will, in the course of time, turn into a Lissajous figure, in moving along which the representative point will finally reach the line of intersection. Thus predissociation will in the end occur; however, the interval of time necessary for all the nuclei to occupy a position in which dissociation is possible will increase so much that the residence time in the excited state before dissociation may reach the normal value \(\tau \cong 10^{-8}\) sec.

As the final result of the consideration we have carried out, we can state that in the case of polyatomic molecules the time after which the process of predissociation is completed generally increases by \(10^2\)—\(10^3\) times. But, owing to the already known relation between the “frequency width” \(\Delta\nu\) and the residence time in the excited state, \(\tau\cdot\Delta\nu=1\), when \(\tau\) increases, \(\Delta\nu\) decreases, and if \(\tau\) reaches the usual value \(10^{-8}\) sec, the lines again acquire normal sharpness. In such a case the spectrum will retain its discrete structure, despite the fact that absorption will end in dissociation.

Norrish \(^{37}\) has indicated a number of interesting cases confirming the considerations set forth by Franck, Sponer, and Teller. If dissociation occurs in the discrete part of the spectrum, then the occurrence of decomposition can be established either by detecting the decomposition products or—much more subtly—by establishing the fact

of the disappearance of fluorescence. Indeed, in formaldehyde dissociation is observed at wavelengths from 3300 to 2000 Å, but the diffuseness of the bands can be detected only for wavelengths shorter than 2700 Å; on the other hand, fluorescence already ceases at 3600 Å, i.e., at wavelengths even somewhat longer than those at which dissociation becomes noticeable. The latter circumstance is undoubtedly explained by the considerably greater sensitivity of the method of detecting the quenching of fluorescence in comparison with other physicochemical methods.

A very interesting result was found in the case of acetone ³⁸ and NO₂³⁹. Here it turned out that there is a region, approximately at 300 Å, where both fluorescence and predissociation are observed simultaneously. This circumstance especially clearly confirms the considerations of Franck, Sponer, and Teller set forth above, for if fluorescence and dissociation have the same probability, then this means that the residence time in the excited state is the same in both cases, i.e., of the order of \(10^{-8}\) sec.

A peculiar and interesting type of decomposition is exhibited by aldehydes and ketones, as well as by certain organic acids, in the gaseous state. In all aldehydes

\[ \begin{matrix} R\\ H \end{matrix} \!\!>\!CO, \]

where \(R\) is some radical, and in ketones

\[ \begin{matrix} R_1\\ R_2 \end{matrix} \!\!>\!CO \]

there is a carbonyl group

\[ >C=O, \]

which is here the “chromophore.” Indeed, all these compounds have the band characteristic of the carbonyl group, a continuous absorption band extending from 3500 to 2000 Å with a maximum at 2800—2900 Å⁴⁰; this band appears both in the gaseous and in the liquid state, and also in solutions. The decomposition of aldehydes and ketones has the character of predissociation. Thus, for example, in the case of formaldehyde and acetaldehyde, in addition to the indicated continuous band, there are bands which at first show a rotational structure, which at a certain wavelength (for formaldehyde at \(\lambda = 2750\) Å, for acetaldehyde at \(\lambda < 3000\) Å) disappears, and at the same time photochemical dissociation also begins. In this case the decomposition of all the simple aldehydes investigated proceeds according to one and the same scheme

\[ \begin{matrix} R\\ H \end{matrix} \!\!>\!CO + h\nu = RH + CO. \]

For example, in formaldehyde decomposition into hydrogen and carbon monoxide is observed⁴¹

\[ \begin{matrix} H\\ H \end{matrix} \!\!>\!CO + h\nu = H_2 + CO, \]

in acetaldehyde the reaction leads to the formation of methane and carbon monoxide \(^ {42}\)

\[ \begin{matrix} \mathrm{CH_3}\\[-2pt] \mathrm{H} \end{matrix} \!\!>\mathrm{CO}+h\nu=\mathrm{CH_4}+\mathrm{CO}, \]

benzaldehyde decomposes into benzene and carbon monoxide \(^ {43}\)

\[ \begin{matrix} \mathrm{H}\\[-2pt] \mathrm{C_6H_5} \end{matrix} \!\!>\mathrm{CO}+h\nu=\mathrm{C_6H_6}+\mathrm{CO}. \]

It is remarkable that in all cases decomposition and recombination proceed in one elementary act. This is evident from the following fact. If the first stage of decomposition consisted in the splitting off of hydrogen or of a hydrocarbon radical and in their subsequent reaction with each other or with the remainder of the molecule, then among the reaction products there would necessarily have to be molecular hydrogen and molecules that are products of recombination of radicals with one another (for example, \(\mathrm{C_2H_6}\) in the case of acetaldehyde). But precisely these products are not observed, and the decomposition proceeds almost \(100\%\) according to the indicated schemes. It follows from this that the whole decomposition is completed within one excited molecule, and free atoms or radicals do not arise.

A different picture is obtained in the case of ketones. Here everything indicates that the first stage of decomposition consists precisely in the splitting off of a free radical. This is especially clearly seen in the example of mixed ketones. Here the reaction products represent an almost equimolecular mixture of hydrocarbons according to the following scheme:

\[ \begin{matrix} \mathrm{R_1}\\[-2pt] \mathrm{R_2} \end{matrix} \!\!>\mathrm{CO}+h\nu=\frac{1}{3}\left(\mathrm{R_1R_1}+\mathrm{R_1R_2}+\mathrm{R_2R_2}\right)+\mathrm{CO}. \]

For example, for the case of methyl ethyl ketone one obtains

\[ \begin{matrix} \mathrm{CH_3}\\[-2pt] \mathrm{C_2H_5} \end{matrix} \!\!>\mathrm{CO}+h\nu=\frac{1}{3}\left(\mathrm{C_2H_6}+\mathrm{C_3H_8}+\mathrm{C_4H_{10}}\right)+\mathrm{CO}. \]

The difference in the mechanism of decomposition of aldehydes and ketones was shown with special clarity by Pearson, who detected free radicals upon illumination of simple ketones, but was unable to do so with aldehydes.

The reason for the difference in the behavior of aldehydes and ketones has not yet been finally established. One may, however, put forward a probable supposition. For this purpose we shall first formulate the experimentally established fact as follows. The elementary photochemical process in aldehydes consists in a rearrangement in which hydrogen changes its position, attaching itself to the radical, whereas in ketones the elementary photochemical process consists in the splitting off of a radical. Since,

however, the hydrogen and the radical in the aldehyde have saturated valences; according to quantum mechanics^44, on approaching one another they repel. Therefore the rearrangement requires a certain activation energy to overcome the potential barrier separating the two parts of the molecule. On the other hand, in the presence of a potential barrier there is always a certain probability of penetration beneath the barrier (“tunnel effect”). Terenin^34 expressed the supposition that the possibility of rearrangement in the case of aldehydes is due precisely to the small magnitude of the mass of the hydrogen atom, since this circumstance increases the probability of penetration beneath the barrier.

1. The photochemical behavior of organic acids in the gaseous state has as yet been insufficiently investigated. It is known that all acids have practically identical absorption (a continuous band with a maximum at 2040 Å), irrespective of the number of carbon atoms entering into the composition of the acid^45,46. This absorption, evidently, must be ascribed to the COOH group, which is present in all acids.

On the basis of the available data one may think that the elementary photochemical act in acids is analogous to that in aldehydes. Thus, in formic acid the following process is observed^47

\[ \begin{array}{c} \mathrm{H}\\[-0.2em] \diagdown\\[-0.2em] \mathrm{HO} \end{array} \mathrm{CO}+h\nu=\mathrm{H_2O}+\mathrm{CO}, \]

in acetic acid^48

\[ \begin{array}{c} \mathrm{CH_3}\\[-0.2em] \diagdown\\[-0.2em] \mathrm{HO} \end{array} \mathrm{CO}+h\nu=\mathrm{CH_4}+\mathrm{CO_2}. \]

\[ \hspace{7.5em}(50\%)\qquad(50\%) \]

However, in the case of acetic acid, in addition to the process indicated, another process is also observed, in which \(\mathrm{C_2H_6}\), \(\mathrm{CO_2}\), \(\mathrm{CO}\), and \(\mathrm{H_2O}\) are obtained. This process, however, undoubtedly takes place on double molecules of acetic acid, since it is observed only where these double molecules are present (in the gaseous state and in solution in hexane), and is not observed where there are no double molecules (in aqueous solution; see below). It is characteristic that molecular hydrogen is not formed in this case,* which indicates that the process, as in the case of aldehydes, takes place within the molecule itself and is not connected with the splitting off of free radicals.

1. Let us now turn to the consideration of yet another peculiar phenomenon, observed in the case of complex molecules and possible precisely because of the presence of a large number of atoms and various bonds. To this end we shall note first of all that, in a polyatomic molecule, processes of stabilization are possible, in which the energy absorbed by some chromophoric group,

* In fact, the formation of hydrogen is sometimes observed in very small amounts. It is explained, however, by secondary causes.

may also not be used to break the bonds in this group, but may be transferred to other atoms or groups. This may explain, for example, the fact that the quantum yield in the decomposition of acetone in the region of the continuous absorption spectrum is not equal to unity, but amounts to only 0.4. Indeed, if the process of absorption of light by a diatomic molecule occurs just at the moment of its collision with another particle, then the excess energy can be transferred to this colliding partner, and dissociation will not occur. In a gas consisting of diatomic molecules, such absorption processes at the moment of collision are rare. A polyatomic molecule has within itself a permanent reservoir, in the form of the various groups composing it, to which excess energy can be transferred. Therefore, in a polyatomic molecule stabilization processes are much more probable than in a diatomic one.

Considerably more interesting, however, is the following fact. If a polyatomic molecule absorbs light, then, as we have seen, this usually occurs at some definite place in the molecule—where the absorbing bond of the chromophoric group is located. In this case the remaining bonds, situated in other parts of the molecule, may remain completely unaffected at the moment of absorption. It turns out, however, that the absorbed energy is capable of producing fluctuations within the molecule, as a result of which its action may be detected not at the place where it was absorbed, but in an entirely different place. Let us give several examples. Acetophenone

\[ \begin{array}{c} \mathrm{CH_3}\\[-2pt] \diagdown\\[-2pt] \mathrm{C_6H_5}\!\diagup\ \mathrm{CO} \end{array} \]

and benzaldehyde

\[ \begin{array}{c} \mathrm{H}\\[-2pt] \diagdown\\[-2pt] \mathrm{C_6H_5}\!\diagup\ \mathrm{CO} \end{array} \]

contain two chromophoric groups: the carbonyl \(>\mathrm{C}=\mathrm{O}\) and the phenyl radical \(\mathrm{C_6H_5}\) (benzene ring). On comparing the absorption and fluorescence of these molecules it turns out \(^{49}\) that absorption in both substances takes place at the phenyl group (absorption maximum at \(2600\ \text{\AA}\), presence of frequencies at \(900\ \text{cm}^{-1}\), characteristic of the benzene nucleus), whereas the emission spectrum (fluorescence) is characteristic of the carbonyl group (presence of a frequency at \(1750\ \text{cm}^{-1}\), characteristic of the group \(>\mathrm{C}=\mathrm{O}\)).

From the photochemical point of view, the facts discovered by Norrish are still more interesting. \(^{37}\) It proved that in the decomposition of aldehydes and ketones with a long chain, cleavage occurs at such a place in the molecule as is considerably removed from the place where absorption of the quantum takes place. Absorption, as always, occurs in the carbonyl group, whereas cleavage occurs in the long chain attached to the car-

bonyl group. For example, methyl butyl ketone decomposes according to the scheme

\[ \begin{array}{c} \mathrm{CH_3} \\ \mathrm{CH_3CH_2(CH_2CH_2)} \ \underset{\uparrow\uparrow}{\overset{\downarrow}{>}}\mathrm{C=O} + h\nu = \begin{array}{c} \mathrm{CH_3}\\[-2mm] \mathrm{CH_3} \end{array} \mathrm{>CO} + \mathrm{CH_3CH:CH_2}. \end{array} \]

Here the arrow \(\downarrow\) indicates the place where the quantum is absorbed, and the double arrow \(\uparrow\uparrow\) the place where decomposition occurs. In exactly the same way, in the case of valeric aldehyde the principal decomposition proceeds according to the scheme

\[ \mathrm{CH_3CH_2CH_2CH_2} \underset{\uparrow\uparrow}{\overset{\downarrow}{\backslash}} \mathrm{C=O} + h\nu \rightarrow \begin{array}{c} \mathrm{CH_3}\\[-1mm] \mathrm{H} \end{array} \mathrm{>CO} + \mathrm{CH_3CH:CH_2}. \]

The amount of energy absorbed is \(90\text{—}90\ \mathrm{kg\ cal/mol}\), whereas the energy expended in carrying out the decomposition that occurs lies between 25 and \(65\ \mathrm{kg/cal}\). Such an amount of energy cannot be transmitted by means of vibrations of parts of the molecule (the energy of vibrations is of the order of tenths of a large calorie per mole). We thus arrive at the conclusion that within a complex molecule there are possible processes of energy transfer analogous to impacts of the second kind—peculiar internal impacts of the second kind. As is known, these impacts are a quantum-mechanical effect of resonance character, and it is the task of theoretical physics to give an explanation of this interesting phenomenon.

V. Photochemical reactions in solutions

1. Photochemical reactions in solutions are of the greatest interest to the chemist. For the physicist, however, who is called upon to interpret the mechanism of elementary processes, the liquid state presents the greatest difficulties. In fact, to this day we know least of all about the nature of liquids, and if formerly we transferred to liquids everything that we know about gases, now we are inclined to regard them as closer to solids than to gases[^50]. Despite all these well-known difficulties, in recent years we have advanced fairly far in understanding photochemical processes in solutions.

Unfortunately, in this case absorption spectra tell us still less than in the case of complex molecules. Indeed, the absorption spectra of solutions, especially aqueous ones, are, with rare exceptions, continuous. However, the diffuseness of the spectrum in the present case is by no means an unambiguous indication of a dissociation taking place. In fact, a molecule in solution is surrounded by a huge number of solvent molecules and, one may say, is constantly in a state of collision. Thus the broadening of spectra in solutions is caused predominantly by two reasons: a) the influence of the electric fields of the surrounding molecules (the Stark effect), and b) the shortening of the lifetime of the excited state

as a consequence of collisions with the surrounding molecules. These collisions, in turn, may lead (a) to dissociation and (b) to the dissipation of the excitation energy and its degradation into heat. We thus see that dissociation, photochemical decomposition, is only one of a whole series of causes determining the diffuseness of the spectrum. It follows from this that, in the case of solutions, the occurrence of dissociation can be established exclusively by the appearance of decomposition products; spectroscopic methods here do not give an unambiguous answer.

1. The most characteristic feature of reactions in solutions is the small value of the quantum yield observed in such reactions. The reason for this can be understood if we take into account precisely the circumstance that the molecules of the dissolved substance may be regarded as being constantly in a state of collision with the surrounding molecules of the solvent. We have already mentioned above that even in gases absorption at the moment of collision can lead to stabilization of the absorbing molecule, since the excess energy acquired by the latter may be transferred to the colliding partner and, in particular, may pass into the kinetic energy of translational motion of the colliding partners. But in a gas the probability of such absorption at the moment of collision, or of a collision in the interval of time occurring between the moment of absorption and the act of decomposition, is negligibly small ($\sim 10^{-4}$ at atmospheric pressure), since the lifetime of a molecule absorbing in the true continuum is small ($\tau = 10^{-13}$ sec). In a liquid, however, the dissolved molecule is constantly within the sphere of influence of the surrounding molecules; therefore in a solution there are constantly present conditions favorable for stabilization. This leads to the probability of stabilization becoming very considerable, despite the fact that the probability of conversion of excitation energy into translational energy is, generally speaking, very small.

But even if the absorption of the dissolved molecule does end in its decomposition, there is a high probability that the results of this decomposition will be reduced to zero by subsequent recombination of the products. Indeed, in solution there is such “crowding” that the decomposition products, over a considerable interval of time, cannot move far apart and remain close to one another. As a consequence, the possibility arises for them of repeated collisions, during which recombination may occur. The number of such collisions may reach several hundred before the decomposition products have separated to a sufficient distance (if, of course, recombination does not occur earlier).51

An equally important role is played by the possibility of secondary reactions of the decomposition products with solvent molecules. Indeed, there are no reasons that would prevent the atoms or radicals arising in a photochemical process from entering into reaction with the solvent. If this reaction proves to be reversible, then after illumination the initial state will be restored,

and the energy is simply dissipated. If, however, the reaction is irreversible, then it will yield certain additional products, the possibility of whose appearance must be taken into account.

All the reasons analyzed lead to a decrease in the quantum yield even in a continuous spectrum, i.e., where in gases we obtain a quantum yield equal to unity (or close to unity).

Let us note, however, that in those cases where decomposition occurs in the discrete part of the spectrum, for example as a result of induced predissociation, the probability of predissociation itself in solution will be incomparably greater than in the gas, since in solution the subsequent collision for the molecule that has absorbed is assured. However, this increased probability of decomposition is completely offset by the factors considered, which lead to a decrease in the final yield.

1. In studying reactions in solutions, especially in aqueous ones, another distinctive factor comes to the fore—electrolytic dissociation. We shall therefore consider separately photochemical processes involving undissociated molecules and ions. The mechanism of the process in the two cases is, of course, entirely different.

Let us begin with reactions involving undissociated molecules. The absorption spectra of such molecules in solution are continuous throughout and therefore differ greatly from the corresponding gas spectra. But the behavior of the absorption coefficient, i.e., the absorption curve itself, is in both cases almost identical, and sometimes completely identical*. For example, in solutions of \(J_2\) in chloroform and carbon disulfide the absorption maximum lies between 5000 and 6000 Å, whereas in gaseous iodine it lies at 5000 Å. Likewise, for other halides \(Br_2\) and \(Cl_2\), the absorption spectra in solutions are close to the gaseous spectra. For \(H_2O_2\), the absorption spectra in the gaseous state and in aqueous solutions are completely identical. The same is observed for complex organic molecules: the absorption spectra of aldehydes in solutions, except in those cases where solvation or hydration occurs, are identical with the spectra of the same substances in the gaseous state, which indicates the identity of the chromophoric groups in the two cases. Similarly, organic acids in solutions give, as in gases, an absorption spectrum common to all acids and belonging to the chromophoric group COOH.

The closeness, or identity, of the absorption curve in solutions and in gases indicates the sameness of the elementary photochemical processes in the two cases. Thus, the various reactions in which the absorbing component is the molecule \(J_2\) are undoubtedly due to the photochemical dissociation of \(J_2\) and to the appearance of free iodine atoms in the solution. Particularly detailed work on this

* For literature concerning absorption spectra in solutions, see Ley\(^{52}\).

from the point of view of study is the photoreaction between \(J_2\) and oxalic-acid salts \(^{53}\).

Reactions in which the molecule \(Cl_2\) participates are also connected with the preliminary photochemical dissociation of this molecule and the appearance of free Cl atoms. These reactions, however, are always complicated by the interaction between the chlorine atoms formed and the solvent. These include the reactions of chlorine water \(^{54}\) and the chlorination of tribromomethane. Unfortunately, precisely these simpler cases from the photochemical point of view have been studied extremely insufficiently, so that the available experimental data provide little material for discussion.

1. The photochemical behavior of aliphatic carboxylic acids has been studied more fully \(^{25}\). What is most interesting here is that these acids in aqueous solution exhibit the same elementary process as in gases, namely: they decompose in one elementary act, i.e. without prior splitting off of radicals, into carbon dioxide and a paraffin hydrocarbon. Thus, acetic acid gives methane and \(CO_2\)

\[ \begin{array}{c} CH_3\\[-2pt] \phantom{CH_3}\backslash\\[-2pt] HO \end{array} \!\!\!>CO + h\nu \to CH_4 + CO_2, \]

propionic acid gives ethane and \(CO_2\),*

\[ \begin{array}{c} CH_3CH_2\\[-2pt] \phantom{CH_3CH_2}\backslash\\[-2pt] HO \end{array} \!\!\!>CO + h\nu \to C_2H_6 + CO_2. \]

A more complex picture is observed with butyric acid. Here, in addition to the reaction proceeding according to the same scheme as in acetic and propionic acids, namely:

\[ \begin{array}{c} CH_3CH_2CH_2\\[-2pt] \phantom{CH_3CH_2CH_2}\backslash\\[-2pt] HO \end{array} \!\!\!>CO + h\nu \to C_3H_8 + CO_2, \]

the formation of ethylene is observed, probably by virtue of the reaction

\[ \begin{array}{c} CH_3CH_2CH_2\\[-2pt] \phantom{CH_3CH_2CH_2}\backslash\\[-2pt] HO \end{array} \!\!\!>CO + h\nu \to CH_2{=}CH_2 + CH_3COOH. \]

Here we are dealing with a case entirely analogous to that found by Norrish in the decomposition of aldehydes and ketones with a long chain: absorption of light occurs in the \(COOH\) group, while the reaction occurs in a part of the molecule remote from this group.

Another very interesting fact, characteristic specifically of reactions proceeding in solution, is the following. If one compares the number of acid molecules decomposed with the amount of it found in the form of gaseous decomposition products, it turns out that the first quantity is always considerably greater than the second. This means that, along with gaseous products, soluble products are also obtained

soluble. Indeed, in one of the experiments 55 millimoles of acetic acid decomposed, while 38 millimoles were found in the form of \(\mathrm{CH_4 + CO_2}\); in addition, however, 13 millimoles were found in the solution in the form of formic acid. This last circumstance—the formation of formic acid—can be explained most simply and naturally by the participation of water in the reaction

\[ \mathrm{CH_3COOH,\ H_2O} + h\nu \rightarrow \mathrm{CH_3OH} + \mathrm{HCOOH}. \]

Exactly the same results were found also for other acids—propionic, butyric, succinic: in all cases the amount of acid decomposed proved to be greater than the amount corresponding to the gaseous products, and in all cases the formation of formic acid was observed.

1. Let us now turn to photochemical processes involving ions. The nature of these processes was first established using the example of the halide ions \(\mathrm{J^-}\), \(\mathrm{Br^-}\), \(\mathrm{Cl^-}\). In studying the absorption spectra of aqueous solutions of various halide salts\({}^{54}\), it was shown that these spectra in the region \(2500\text{—}1800\ \text{\AA}\) are determined exclusively by the anions and do not depend on the nature of the cation. These spectra are continuous; moreover, in iodine and bromine two maxima are observed, separated from one another, in frequency units, by distances of \(\sim 8000\ \mathrm{cm^{-1}}\) for iodides and \(3000\ \mathrm{cm^{-1}}\) for bromides. If one finds the energy values \((h\nu)\) corresponding to these separations, then for the \(\mathrm{J^-}\) ions one obtains \(0.9\ \mathrm{eV}\), and for \(\mathrm{Br^-}\), \(0.3\ \mathrm{eV}\). These values are characteristic of iodine and bromine atoms. Namely: \(0.9\ \mathrm{eV}\) is nothing other than the separation between the normal level of iodine \(\left({}^{2}P_{3/2}\right)\) and the first excited (metastable) level of the same atom (spectroscopic symbol \({}^{2}P_{1/2}\)). Similarly, for bromine \(0.3\ \mathrm{eV}\) is the separation between the same levels of the neutral atom. On this basis Franck and Scheibe\({}^{55}\) gave the following interpretation of the nature of the process responsible for the appearance of these spectra. The elementary process occurring upon absorption of a quantum, for example by an iodine ion \(\mathrm{J^-}\), according to Franck and Scheibe, consists in the detachment of an electron from this ion; the remaining neutral iodine atom \(\mathrm{J}\) may then be either in the normal or in the excited state. In either case, different energies must be expended for the detachment process, namely: let the magnitude of the quantum required for the first process be \(h\nu_1\); then

\[ \mathrm{J^-} + h\nu_1 = \mathrm{J}\left({}^{2}P_{\frac{3}{2}}\right) + e^-; \]

let the magnitude of the quantum required for the second process be \(h\nu_2\),

\[ \mathrm{J^-} + h\nu_2 = \mathrm{J}\left({}^{2}P_{\frac{1}{2}}\right) + e^-. \]

Then the energy difference will be

\[ h\nu_1-h\nu_2 \equiv h\Delta\nu = {}^2P_{\frac{3}{2}}-{}^2P_{\frac{1}{2}} . \]

This quantity, \({}^2P_{\frac{3}{2}}-{}^2P_{\frac{1}{2}}\), as is known from the analysis of the iodine spectrum, is precisely equal to 0.9 V, or, in frequency units, \(\Delta\nu = 8000\ \mathrm{cm}^{-1}\). In an analogous way, for the bromine atom as well, the energy difference of the first levels is precisely equal to 0.3 V.

Since the energy relations in both cases are determined by how strongly the electron is bound to the bromine and iodine atoms, these spectra were called electron-affinity spectra. However, the attempt to determine from these spectra the quantity of electron affinity, very important for atomic phenomena, did not give satisfactory results. Therefore the very interpretation of the spectra in later work by Franck and Haber\(^{56}\) was somewhat modified by the following refinement. An ion dissolved in water cannot be regarded as isolated. It is surrounded by a shell of water molecules, i.e. it is hydrated. Therefore the elementary process, according to Franck and Haber, occurs in the complex \(\mathrm{J^-}, \mathrm{H_2O}\) and consists in the detachment of an electron from the iodine atom and its attachment to the hydrogen of a neighboring water molecule. As a result one obtains

\[ \mathrm{J^-},\ \mathrm{H_2O}+h\nu \to \mathrm{J}+\mathrm{H}+\mathrm{OH^-}. \tag{1} \]

Thus the final result of the elementary photochemical process must consist in the formation of free iodine and hydrogen atoms and in an increase of the alkalinity of the solution. It should be noted, however, that we may expect in advance that the quantum yield of this process will be very small. Indeed, although process (1) proceeds with a quantum yield equal to unity, the products arising in this process in the solution cannot separate to large distances, but undergo repeated collisions (see above). Therefore already at the next moment, with great probability, a charge-exchange reaction must take place,

\[ \mathrm{J}+\mathrm{H}\to \mathrm{J^-}+\mathrm{H^+}, \tag{2} \]

restoring the initial state. Only a small part of the atoms formed remains in the free state and, recombining, gives molecules

\[ 2\mathrm{J}\to \mathrm{J_2}\quad \text{and}\quad 2\mathrm{H}\to \mathrm{H_2}. \]

The available experimental material confirms this scheme. Thus, according to Warburg and Rump\(^{57}\), aqueous solutions of hydrogen iodide, when illuminated with ultraviolet light, decompose with liberation of free iodine. The quantum yield of this reaction is very small and only in very concentrated solutions (7.5 mole/l) reaches the value 2, normal for the decomposition of hydrogen iodide. But

In this latter case the primary processes (1) and (2) are complicated by the secondary ones

\[ \mathrm{H}+\mathrm{HJ}\to \mathrm{H}_2+\mathrm{J} \]

\[ \mathrm{H}+\mathrm{J}^{-}(\mathrm{H}_2\mathrm{O})\to \mathrm{H}_2+\mathrm{J}+(\mathrm{OH})^{-}. \]

The photochemical reactions of solutions of alkali-halide salts were studied by Butkov,^58 who showed that, in solutions of iodides and bromides, iodine and bromine are liberated and the alkalinity of the solutions increases. At the same time the region of photochemical sensitivity of the solutions coincides with the absorption spectrum of the corresponding ions. The quantum yield in all cases proved to be very small, since the reverse reactions

\[ \mathrm{J}+\mathrm{H}\to \mathrm{J}^{-}+\mathrm{H}^{+}, \]

\[ \mathrm{Br}+\mathrm{H}^{+}\to \mathrm{Br}^{-}+\mathrm{H}^{+} \]

proceed at rates close to the rate of the direct reaction. In chlorides, the formation of free chlorine was not observed at all, evidently because the reverse reaction

\[ \mathrm{Cl}+\mathrm{H}\to \mathrm{Cl}^{-}+\mathrm{H}^{+} \]

has a probability equal to unity.

1. Let us now consider a more complicated case, but one of special interest for the interpretation of photochemical reactions in solutions: the case of an elementary process in the spectrum of electron affinity. The reaction in question is that on sulfite ions, \(\mathrm{SO}_3^{--}\). The reaction of oxidation of sodium sulfite by oxygen is one of the most interesting and thoroughly studied photochemical reactions in solutions.^59 This reaction has a typical chain character, since the number of sulfite molecules undergoing oxidation as a result of one absorbed quantum is \(5\cdot 10^{4}\); in this same reaction Bäckström first demonstrated the identity of chains in photochemical and dark processes.

The absorption spectrum of sodium sulfite solutions was subjected to a special investigation,^60 and it turned out that at a definite acidity of the solution (\(\mathrm{pH}\geq 7\)) this spectrum belongs wholly to sulfite ions. This spectrum is interpreted by Franck and Haber as a spectrum of electron affinity; the elementary process occurring here, (3), in accordance with what has been said above, may be represented as follows:

\[ \mathrm{SO}_3^{--},\ \mathrm{H}_2\mathrm{O}+h\nu\to \mathrm{SO}_3^{-}+\mathrm{H}+\mathrm{OH}^{-}. \tag{3} \]

The energy relations in such a process can be established from the following considerations. The energy of the quantum is expended on detaching one electron from the ion \(\mathrm{SO}_3^{--}\) (obviously, the work required for this detachment is equal to the electron affinity of the ion \(\mathrm{SO}_3^{-}\), \(E_{\mathrm{SO}_3}\)) and on decomposing the water molecule into H and OH (the dissociation energy \(D_{\mathrm{H_2O}}\)). But, in addition, in this process there is liberated the energy of attachment of the electron to OH (the electron affinity ...

side of \(E_{\mathrm{OH}}\)), and, finally, there arises still a certain potential energy of the decomposition products \(X\). This potential energy is due to the possible repulsion of the decomposition products if they are charged with like charges, and also to the potential energy of the surrounding water molecules, which tend to arrange themselves in an ordered manner around the ions that arise (hydration energy). Since, however, the elementary process is connected with electronic excitation, which occurs instantaneously, whereas the enumerated sources of potential energy are connected with the displacement of heavy nuclei, which do not have time to change their position during the time in which the electronic process is completed, the corresponding energy figures only as potential energy and cannot be used in the process itself. It is easy to see that these considerations constitute a modification and extension of the Franck–Condon principle considered above for the case of elementary processes in solutions.

On the basis of what has been said, the energy balance of the reaction may be represented by the equation

\[ h\nu = D_{\mathrm{H_2O}} + (E_{\mathrm{SO_3}} - E_{\mathrm{OH}}) + X . \tag{4} \]

The presence in this equation of the term \(X\), which is difficult to estimate, makes the process itself unsuitable for an exact determination of the electron affinity. Thus Franck’s original hope, which saw in this process a possibility for the direct determination of the magnitude of the electron-affinity energy, was not justified. However, from the photochemical point of view it will be justified if it turns out that the products to which this process leads are in fact obtained.

Let us consider the question from this point of view. The primary act, as equation (3) shows, leads to the formation of \(\mathrm{SO_3^-}\), \(\mathrm{H}\), and \(\mathrm{OH^-}\). However, already at the next moment, owing to the equilibrium between the appearing ion \(\mathrm{SO_3^-}\) and the present ions \(\mathrm{H^+}\), undissociated molecules \(\mathrm{HSO_3}\) must be formed. The radical \(\mathrm{HSO_3}\), unknown in the free state, Franck and Haber call monothionic acid, and they ascribe to this radical a large role in the development of the further chain process. Thus we may say that for each quantum absorbed in the elementary process the following products arise: one atom \(\mathrm{H}\), one molecule \(\mathrm{HSO_3}\), and two ions \(\mathrm{OH^-}\). It is not difficult to convince oneself, however, that the final quantum yield of the reaction (in the absence of oxygen) must be very small. Indeed, after the primary process (1) and the formation of the molecule of monothionic acid there follows a series of secondary processes which completely restore the initial state. First, between \(\mathrm{H}\) and \(\mathrm{HSO_3}\), which remain for a long time in proximity, there occurs a charge-exchange reaction

\[ \mathrm{H} + \mathrm{HSO_3} \to \mathrm{H^+} + \mathrm{HSO_3^-}. \tag{5} \]

This reaction has the character of the usual analytical reactions of charge transfer. It proceeds practically without activation energy, and therefore its probability is very high.

Following reaction (5), the process

\[ \mathrm{H}^{+}+\mathrm{HSO}_{3}^{-}+2\mathrm{OH}^{-}\to \mathrm{SO}_{3}^{--}+2\mathrm{H}_{2}\mathrm{O}, \]

is possible, completely restoring the initial state.

Part of the products of the primary photochemical act nevertheless escapes reverse reactions and undergoes further transformations. Namely, first, reactions of the atoms and radicals formed are possible with the participation of water molecules (disproportionation)

\[ \mathrm{H}+\mathrm{HOH}+\mathrm{HSO}_{3}\to \mathrm{H}_{2}\mathrm{O}+\mathrm{H}_{2}\mathrm{SO}_{3}, \]

\[ \mathrm{H}+\mathrm{HOH}+\mathrm{HSO}_{3}\to \mathrm{H}_{2}+\mathrm{H}_{2}\mathrm{SO}_{4}, \]

second, H atoms and \(\mathrm{HSO}_{3}\) radicals may undergo dimerization, giving stable products

\[ 2\mathrm{H}=\mathrm{H}_{2}; \qquad 2\mathrm{HSO}_{3}=\mathrm{H}_{2}\mathrm{S}_{2}\mathrm{O}_{6}. \]

Thus, upon irradiation of sulfite in the absence of oxygen, we should expect the formation of the following products: hydrogen, dithionic acid \(\mathrm{H}_{2}\mathrm{S}_{2}\mathrm{O}_{6}\), and sulfuric acid. Special experiments carried out to test this theory made it possible to find all these products and thus confirmed the theory of Franck and Haber. In the presence of oxygen the products formed undergo further oxidation, and a chain arises. The mechanism of development of this chain was also indicated by Haber and Uonsborough-Jones\(^{62}\) and was somewhat modified by Bäckström\(^{62}\). However, since these questions are no longer connected with primary photochemical processes, we shall not dwell on them here, referring those interested to the original literature or to the author’s review cited above.

1. To the examples considered above one should add one more interesting and instructive case of photochemical reactions on ions of organic acids. The spectra of the ions of acetic, formic, and succinic acids are very similar to one another. Farkas and Uonsborough-Jones interpret these spectra as the electron-affinity spectra of the anions. In this case the elementary process must proceed according to the scheme

\[ \mathrm{A}^{-},\ \mathrm{H}_{2}\mathrm{O}+h\nu \to \mathrm{A}+\mathrm{H}+\mathrm{OH}^{-}, \]

where \(\mathrm{A}^{-}\) is the symbol of some anion. Owing to the reverse reaction

\[ \mathrm{A}+\mathrm{H}\to \mathrm{A}^{-}+\mathrm{H}^{+}, \]

which restores the initial state, the final quantum yield should be very small.

Experimental investigation led to the following results. The acetate ion, as the principal products of photochemical decomposition, gives carbon dioxide and methane. The quantum yield, referred to \(\mathrm{CH}_{4}\), as in the case of sulfite, is very small. In the be—

In the more favorable case (\(\mathrm{pH}\simeq 8.0\)) it is equal to 0.065, i.e. in 93.5% of cases the products of the primary photochemical process enter into a reaction restoring the initial state. The formation of carbon dioxide and methane according to Farkas and Wansbrough-Jones is the result of the subsequent reaction of the ion \(\mathrm{CH_3COO^-}\) at the moment of its formation in the course of recharge. This process is accompanied by the liberation of a large quantity of heat, which in some cases proves sufficient for a reaction with water to take place according to the scheme

\[ \mathrm{CH_3COO^-}\ \longrightarrow\ \boxed{\begin{matrix} \mathrm{CH_3}\\ \mathrm{H} \end{matrix}} \ +\ \boxed{\begin{matrix} \mathrm{COO^-}\\ \mathrm{OH} \end{matrix}} \]

The circumstance that the latter reaction does not occur every time, but only in a small number of cases, is explained, in the opinion of Farkas and Wansbrough-Jones, by the rapid dissipation of the excess energy: the newly formed ion \(\mathrm{CH_3COO^-}\) is immediately hydrated and becomes capable of giving up its excess energy to several water molecules bound to it.

The radical \(\mathrm{CH_3COO}\), arising in the primary photochemical process, in an insignificant number of cases avoids recharge and continues to exist for a certain interval of time, entering into side reactions. In the case of sulfite, one such reaction was the formation of dithionic acid as a result of the dimerization of monothionous acid, and the formation of sulfuric acid as a result of a reaction involving a water molecule. To the first reaction in the case under consideration there corresponds the formation of succinic acid

\[ 2\mathrm{CH_3COO}\rightarrow \mathrm{COOH}-\mathrm{CH_2}-\mathrm{CH_2}-\mathrm{COOH}, \]

and to the second—the formation of peracetic acid. The formation of succinic acid proved possible to detect after 28 hours’ illumination of a \(0.1\,n\) solution of sodium acetate \(+\) a \(0.1\,n\) solution of \(\mathrm{NaOH}\); the formation of the peracid, however, was not detected, evidently because from the thermochemical point of view the conditions for obtaining it are considerably less favorable than in the case of the formation of sulfuric acid from \(\mathrm{HSO_3}\).

The phenomena observed upon illumination of the ions of formic and succinic acids are analogous to those considered. The only difference is that in the case of formate, alongside \(\mathrm{CO_2}\), evolution of \(\mathrm{CO}\) is also observed. Farkas and Wansbrough-Jones attribute the origin of \(\mathrm{CO}\) to the reaction

\[ \mathrm{HCOO^-},\ \mathrm{H_2O}\rightarrow \mathrm{H_2O}+\mathrm{CO}+\mathrm{OH^-}. \]

Upon dimerization of the radical \(\mathrm{HCOO}\) arising in the primary act, oxalic acid should evidently be obtained, and this is indeed observed: upon illumination of formate for 4.5 hours, 0.0068 moles of oxalic acid were detected.

Summing up everything that has been said so far, we see that elementary photochemical processes on ions may be considered fully established in those cases when the absorption spectrum of the ion can be interpreted as the spectrum of electron affinity.

LITERATURE

A. Monographs

Bonhoeffer and Harteck. Fundamentals of photochemistry, ONTI, 1935.
A. N. Terenin. Photochemistry of salt vapors, GTTI, 1934.
V. N. Kondrat’ev. The structure of molecules and elementary chemical processes, GTTI, 1934.
V. N. Kondrat’ev. Photochemistry, GTTI, 1933.
V. N. Kondrat’ev and Elyashevich. Energy exchange, GTTI, 1933.
Mitchell a. Zemansky, Resonance Radiation and excited Atoms. Cambridge 1934.

B. Literature cited in the text

1. For a review of the photochemistry of crystals see the article by M. V. Savost’yanova, Uspekhi fizich. nauk 11, 451, 1931.
2. A. Einstein, Ann. d. Phys., 17, 132, 1905; 37, 832, 1912; J. Stark, Phys. Z., 9, 398, 1912.
3. Cf., for example, K. Bonhoeffer and P. Harteck. Fundamentals of photochemistry, pp. 9–18.
4. Photochemical Processes. Trans. Farad. Soc., 21, 1925; see also the German translation Z. physikal. Chem., 120, 1926.
5. O. Stern u. M. Volmer, Physik. Z., 20, 186, 1919.
6. I. M. Frank, Acta Physicochimica URSS, 1, 833, 1935.
7. R. Wood a. Gaviola, Phil. Mag., 6, 271, 1928.
8. Childs u. Mecke, Z. Physik, 68, 344, 1931.
9. Kautsky, de Bruijn, R. Neuwirth, Ber., 66, 1588, 1932.
10. Beutler u. Josephy, Z. Physik, 53, 747, 1929.
11. Cf., for example, W. Grotrian, Graphische Darstellung der Spektren, II. Teil.
12. H. Kallmann u. F. London, Z. physik. Chem., B 2, 207, 1929.
13. A. Terenin, Z. physik, 37, 120, 1926.
14. J. Franck, Trans. Farad. Soc., 21, 3, 1925.
15. See, for example, Weizel, Bandenspektren.
16. R. Mecke, Ann. d. Phys., 71, 104, 1923.
17. Landolt-Börnstein, Physikalisch-chemische Tabellen, II. Ergänzungsband.
18. Dymond, Z. Physik, 34, 553, 1925.
19. Bonhoeffer u. Farkas, Z. Physikal. Chem., 132, 255, 1928.
20. Senftleben, Ann. d. Phys. (5) 2, 847, 1929.
21. Turner, Phys. Rev., 27, 396, 1826.
22. Bonhoeffer u. Steiner, Z. physik. Chem., 122, 287, 1926.
23. F. London, Z. Physik.
24. V. Henri, Structure des Molecules, Paris Hermann, 1926.
25. M. Born u. J. Franck, Z. Physik 31, 411, 1925.
26. L. De Broglie, Einführung in die Quantenmechanik, Lpz. 1929.
27. V. Henri, Leipziger Vorträge, 1931.
28. L. Turner, Phys. Rev., 41, 627, 1932.
29. R. de L. Kronig, Z. Physik 50, 347, 1928.
30. I. H. VanVleck, Phys. Rev., 1930.
31. Kondratjew u. Polak, Sow. Phys., 4, 770. 1933.

1. Cf., for example, H. Leu, Lichtabsorption und chemische Konstitution (Handb. d. Physik von Geiger u. Scheel, B. XXI).
2. K. W. F. Kohlrausch, Der Smekal-Raman-Effekt.
3. A. N. Terenin, Acta Physicochimica URSS, 3, 181, 1935.
4. A. Terenin and H. Heuminn, Nature, 134, 255, 1934.
5. J. Franck, H. Sponer u. E. Teller, Z. physik. Chem., B. 18, 88, 1932.
6. R. G. N. Norrish, Acta physicochimica URSS, 3, 171, 1935.
7. Norrish, Crone a. Saltmotsh, J. Chem. Soc. (Lond.) 1934, 1456.
8. Norrisch, J. Chem. Soc. (Lond.) 1927, 761; 1929, 1158; 1604, 1611.
9. Cf. Bonhoeffer and Harteck, Foundations of Photochemistry, p. 160; the literature is also indicated there.
10. Kirkbride a. Norrisch, Trans. Farad. Soc., 27, 404, 1931.
11. V. Henri, Leipziger Vorträge, 1931, 132.
12. V. Henri, l. c. p. 134.
13. F. London, Sommerfeld Festschrift.
14. V. Henri, Photochimie, p. 90.
15. Leu u. Arends, Z. physik. Chem., B. 4, 234, 1929.
16. E. Gorin a. H. Taylor, J. Am. Chem. Soc., 56, 2042, 1935.
17. L. Farkas u. O. H. Wansbrough-Jones, Z. physik. Chem., B. 18, 124, 1932.
18. N. Prileshajewa, Acta physicochimica URSS, 3, 195, 1935.
19. Bernal and Fowler, J. chem. Phys.
20. Ya. K. Syrkin, Journal of Physical Chemistry
    Acta physicochimica URSS
21. H. Leu, Handb. d. Physik
22. Berthoud et Bellenot, J. d. Chim. phys., 21, 308, 1934; see the analysis of this reaction in the article by E. Shpolsky, Journal of Physical Chemistry, 2, 468, 1931.
23. Allmand, Cunliffe a. Madison, J. chem. Soc. (Lond.) 127, 822, 1925.
24. G. Scheibe, Z. Elektrochem., 34, 497, 1928; Z. physik. Chem. B. 5, 355, 1929, Fromherz u. Menschik, Z. physik. Chem. 7, 439, 1930. J. Franck u. G. Scheibe, Z. physik. Chem. Haber-Band, 22, 1928.
25. J. Franck u. F. Haber, Sitz. Ber. Preuss. Akad. Wissensch. Phys.-Math. Klasse, 1931, XIII.
26. E. Warburg u. Rump, Z. Physik, 47, 305, 1928.
27. K. Butkow, Z. Physik, 62, 71, 1930.
28. E. Shpolsky, Journal of Physical Chemistry, 2, 468, 1931.
29. Albu u. Goldfinger, Z. physik. Chem. (B) 16, 338, 1932; references to later works are also given there.
30. E. Shpolsky, Journal of Physical Chemistry, 6, issue 2—3, 1935.
31. H. L. Bäckström, Z. Physik. Chem. (B) 25, 122, 1934.
32. H. Leu u. B. Arends, Z. physik. Chem. (B), 17, 177, 1932.

1. * The article is an expanded presentation of a lecture delivered on December 12, 1935, at the “N. D. Zelinsky University of Physicochemistry” at Moscow University. ↩