fewer than 96 new organic compounds have been isolated in the course of this work.

Compound formation in binary organic mixtures is therefore fundamentally dependent upon diversity in character (i. e., differences in positive or negative natures of the constituent groups) of the two components.

Compound Formation and Ionization. Before proceeding further, it was necessary to confirm a point which the above results had indicated as highly probable-namely, that the order of electroaffinity of different radicals in these organic mixtures was substantially the same as in aqueous solutions. This was satisfactorily established by careful conductivity determinations upon various systems of the types examined above. In a fixed ketone or ester, for example, different acids HX were found to give more highly conducting solutions the more electronegative the radical X. In a fixed acid HX, conversely, different ketones R.CO.R1 or esters R.COOR1 gave solutions of better conductivity the more electropositive the radicals R and R'. Different acids HY in a fixed acid HX, finally, gave more highly conducting solutions the greater the difference in the electroaffinity of the radicals X and Y.

An important generalization may therefore be deduced: in a conducting solution, compound formation between the components and ionization proceed in parallel. Where compound formation is very slight, ionization is inappreciable; as compound formation increases in extent, ionization becomes evident; where compound formation is very marked, ionization is extensive.

The validity of this generalization for aqueous solutions may now be examined. Acids and bases obviously furnish a stringent test, since ionization here varies very considerably. Examination of the available data shows that strong acids invariably give stable hydrates, while among the myriads of weak acids listed in Beilstein not a single example of hydrate formation is indicated. Quantitative confirmation of the parallelism between hydrate formation and ionization for acids in aqueous solution was obtained from freezing-point depression determinations with a series. of acids of gradually increasing strength. With bases the data, though not so extensive, showed similar agreement.

A Modified Ionization Theory. On the basis of the above results, a modification of the current ionization theory has been proposed, under the assumption that "ionization is preceded by combination between solvent and solute and is, indeed, a consequence of such combination." According to this view, ionization in solutions is due not to solute alone nor to solvent alone, but to combination between the two to form unstable complexes. The various factors affecting the formation and stability of such complexes have been critically discussed elsewhere. It has been shown that, while in simple molecules the two radicals are in general attracted to each other so strongly that no dissociation is observable, yet

3

in complex "addition compounds" the attractive forces between the radicals are so diminished that disintegration of the complex into oppositely charged ions may readily occur. Union with another molecule promotes ionization.

The theoretical arguments in favor of this assumption and the experimental evidence in its support cannot be presented in detail here. The immediate consequence of the modified ionization theory, as connected with the anomaly of strong electrolytes, may however be indicated.

Instead of a simple equilibrium such as KCl — K+ + Cl− for a uniunivalent electrolyte in aqueous solution, we now have to consider a complicated series of equilibria between many different molecular and ionic types, such as (KCl),; (KCl),, (H2O)y; [(KCl),, Cl]; [(H2O),, Cl]−, etc. To some of these equilibria the law of mass action alone may apply, to others the Milner-Ghosh conception of electrostatic equilibrium may need to be superadded. Little wonder, therefore, that the problem has baffled all attempts at explanation under the hypothesis RX R+ + X-. If we persist in postulating the reaction as inconceivably simple, we can hardly expect to make progress, even in the field of dilute solutions, where many of the equilibria may be neglected.

It might seem that the complexity in the equilibria necessarily introduced by the consideration of so many molecular types could, after all, only make the problem still more intricate and elusive. The first effect, nevertheless, is in the opposite direction. Many subsidiary assumptions which have been attached to our theory of solutions in order to conceal the failure of the simple equation RX — R+ + X- may now be shown to be unnecessary. For example, the hypothesis of the "catalytic activity of the undissociated molecule," which purports to explain why the speed of reactions such as ester catalysis is not exactly proportional to hydrogen ion concentration, may be discarded in favor of a view which recognizes several types of "hydrogen ion" (e. g., H+, [H(H2O)]+, [H(R.COOR1)]+) each possessing a different catalytic activity. In the same way abnormal results in electromotive force measurements, which have been interpreted as demonstrating a variable activity of the hydrogen ion with concentration, may now be alternatively referred to variations in the ratios of the different types of "hydrogen ion" with concentration. There is no more reason, for instance, why the ion [H(H2O),]+ should show the same potential difference as the ion H+ with respect to gaseous hydrogen than there is for the ion [Cu(NH3),]++ to exhibit the same potential difference as the ion Cu++ with respect to metallic copper. The extension of the theory of complex ions to simple aqueous solutions is evidently one of the first steps to be taken in the further development of the views here presented. Detailed discussion must be postponed to a later article.

y

Salts in Aqueous Solution.-The correlation of compound formation and ionization in this important field offers difficulties. Nearly all uni

univalent salts are practically equally dissociated in aqueous solution, yet hydration values so far as we can judge from present data are widely divergent. It is a very striking fact, however, that with mercuric salts (the one series in which ionization varies considerably) all highly ionized salts yield hydrates (e. g., Hg(NO3)2, SH2O; Hg(ClO4)2, 6H2O; HgF2,2H2O) while all slightly ionized are non-hydrated.

It appeared that the problem might be taken up most profitably in two stages-by the examination of systems of the type RX-HX (acid salts) and of the type RX-ROH (basic salts). By comparison of the results in these fields with those already obtained for systems of the types: HX-H2O and ROH-H2O it was hoped that generalizations for the more complex type RX-H2O might be successfully formulated. The results obtained from a detailed investigation of two series of acid salts (formates and sulfates) have justified this procedure.

Acid Salts.-Compound formation in the system acid-salt is found to be directly dependent upon differences in electroaffinity, as before. With salts of highly electropositive metals (e. g., K, Na) extensive compound formation is obtained. As the metal approaches hydrogen in the electromotive series, compound formation decreases rapidly (with Ba, Ca, Mg) and finally becomes negligible (with Ni, Fe, Cu). With highly electronegative metals, however (Hg, Ag), compound formation once more becomes appreciable. Ionization again proceeds in parallel with compound formation, the best conducting solutions in pure formic or sulfuric acid being given by the salts of K and Na.

Solubility. A new and very significant relationship which appears in these systems is that of solubility, which also parallels compound formation. The sulfates of K and Na are exceedingly soluble in pure H2SO4, those of Ba, Ca and Mg are decreasingly soluble, those of Ni, Fe, Cu practically insoluble. With electronegative metals the solubility again becomes appreciable (with Hg) and finally extensive (with Ag). The solubilities of the formates in formic acid follow exactly the same course.

That the rule is valid also in aqueous solutions may be demonstrated by applying it to systems of the type ROH-HOH. The hydroxides of the alkali metals are very soluble in water, those of the alkaline earths decreasingly soluble, those of Ni, Fe, Cu practically insoluble. Silver again shows an increase, limited however by precipitation of the more stable oxide.

In systems of the type RX-HOH, finally, the same rules hold if R is not very different in character from H, or if X is not very different in character from OH. Thus the salts of weak bases (e. g., ferric salts) or of weak acids (e. g., fluorides) show regular gradations of solubility in water, which substantially follow the generalizations given above. For salts of a strong acid with a strong base, however, where R and X both differ widely in

character from H and OH, no such simple behavior is to be expected, nor is it found Rules for such systems have still to be discovered.

The reason for the above dependence of solubility upon compound formation becomes evident when we examine the freezing-point depression curve of a system such as Li2SO4-H1⁄2SO1. The absolute freezing-point of the salt is so high that, unless compound formation took place, solubility in the acid at ordinary temperatures would, according to the Schröder-Le Chatelier equation, necessarily be negligible. The removal of simple solute molecules by combination with solvent, however, results in an increased solubility, and the extent of such increase depends, naturally, upon the extent of compound formation. The further development of this topic and the presentation of quantitative results will be given in forthcoming articles in the Journal of the American Chemical Society.

*Contribution No. 350.

1 Kendall and Booge, J. Amer. Chem. Soc., 39, 1917 (2323).

2 Carpenter, Ph.D. Dissertation, Chicago, 1915; Gibbons, J. Amer. Chem. Soc., 37, 1915 (149); Booge, Ph.D. Dissertation, Columbia, 1916; Gross, Ph.D. Dissertation, Columbia, 1919; Landon, Ph.D. Dissertation, Columbia, 1920; Adler, Ph.D. Dissertation, Columbia, 1920; Davidson, Ph.D. Dissertation, Columbia, 1920.

3 Kendall and Booge, loc. cit.; Gross, loc. cit.

Roozeboom, Heterogene Gleichgewichte, 2, 1904 (270–84).

THE PELTER EFFECT

BY EDWIN H. HALL

JEFFERSON PHYSICAL LABORATORY, HARVARD UNIVERSITY

Read before the Academy, November 16, 1920

When a current of electricity goes from metal a, in which the freeelectron fraction of the conductivity is (kk) a, to the metal ẞ, in which the corresponding fraction is (k÷k), ionization or reassociation must take place at the junction, according as (k, k) is greater or less than (kk) The resulting change of condition is the same whether we assume the ionization or reassociation to take place after or before the crossing of the interface between a and B. We shall use each of these hypotheses in turn, beginning with the first.

The gain of energy, and so the heat energy absorbed, by the unit quantity of electricity, 10 coulombs, in the form of electrons, in crossing the boundary from a to ẞ is

II

=

-(-4). (P. - P.), + (?) (P. - P.) + (P2- P.)

B

B

a

+

λβ, (1)

where (P-Pa) means the Pƒ of ẞ minus the P of a, etc., the nomenclature according in general with that of my preceding papers.

From my paper in these PROCEEDINGS for April, 1918, we get, by integration of equation (5) there given,

Integrating equation (6) of the same paper, we get

(2)

(3)

(4)

If we had used at the beginning the hypothesis that the ionization occurred in a, just before the crossing, we should have obtained the expres

The total change of state being the same in the process which gives equation (5) as in that which gives equation (4), the total heat absorbed must be the same in both cases—that is, the II of (5) is the same as the II of (4). Subtracting (5) from (4) we get

Equation (8) has a familiar form, and in fact it might have been written down at once without circumlocation as soon as the fact was recognized, as it was in my paper of April, 1918, that the free electrons in the one metal are in equilibrium with those in the other, there being no circulatory motion of the electrons at the junction if the metals are at one temperature. Making use of equation (7) we can change (4) to the form

It is now in order to see how well values of II calculated by means of