6 August 1962

Dear Drs. Davidson and Finland:

I am pleased to learn about the plans of recognizing in a suitable way the retirement of Dr. William B. Castle, perhaps together with the celebration of the Centennial of the Boston City Hospital and the fortieth anniversary of the founding of the Thorndike Memorial Laboratory.

During the couple of decades when I have been interested in the scientific basis of medicine, I have been pleased to hear, over and over again, about the contributions that Dr. Castle has made, over a period of 40 years, to the scientific understanding of the human body and of the nature of disease.

An example of the significance of Castle's early work is given by the statement on page 233 of the book "Nobel: The Man and His Prizes." In the discussion of the award of the Nobel Prize to Whipple, Minot, and Murphy for their discoveries concerning liver therapy against anemias, the authors of the book, which is edited by the Nobel Foundation, state "On the basis of the observation that in cases of pernicious anaemia there is usually a cessation of the gastric juice secretion, W. B. Castle performed experiments that improved our understanding of the nature of this illness. In 1928 he discovered that meat digested with normal gastric juice was almost as effective against pernicious anaemia as liver. By itself, however, meat did not have this effect, nor after having been digested with the aid of an artificial gastric juice (pepsin plus hydrochloric acid). Consequently, normal gastric juice seems to contain an 'intrinsic' factor which together with some element in the meat (the 'extrinsic factor') forms the active substance, which is stored, among other places, in the liver. . . ."

I have in past years learned much through conversations with Castle, and I am especially indebted to him for having, through his interest in the disease sickle-cell anemia, and his enthusiasm about the possibilities of developing the science of medicine, led me to formulate the idea that the disease is connected with the gene-controlled manufacture by the patients of an abnormal form of adult human hemoglobin, and encouraged me to carry out experimental studies in this field. I am glad to be able to express my appreciation to him, and to say that I cherish his friendship.

Cordially yours,

[Linus Pauling]

6 August 1962

Dear Peter,

I have been thinking that another paper ought to be written on metallic valence. Because of its relation to the paper on valence and atomic size that you and I wrote together, published in 1956 in Acta Crystallographica, I am writing to ask if you would be interested in joining me in writing the new paper. This would require that you check the literature and look for arguments to supplement those that I have thought of and also examine my arguments.

About a month ago I wrote to Professor G. V. Raynor of the Department of Physical Metallurgy, University of Birmingham, mentioning his paper on the effective valences of transition metals in solid solution in gold and silver alloys (B. Henderson and G. V. Raynor, Proc. Roy. Soc. A 267, 313 (1962)), and asking if he had considered using my theory of metals in the discussion of his interesting and valuable experimental results. He has not yet answered the letter. I think that when he does it is likely that he will say that he doesn't like the theory.

I have decided that we (or I) ought to publish a paper in which his results and other information about pertinent alloys are discussed on the basis of the system of metallic valences.

In this paper and another one, the apparent metallic valences of transition metals in solid solution, B. Cockayne and G. V. Raynor, Proc. Roy. Soc. A 261, 175 (1961), he and his coworkers have discussed alloys with the hexagonal close packed structure, called the zeta alloys. In various other papers to which reference is made in these two papers, he and others have pointed out that the axial ratio is a function of composition, which they express in terms of the electron-atom ratio calculated with the old valences, one for copper, two for zinc, and so on, and that for some alloys, such as copper-gallium-germanium, it is essentially a function of this electron-atom ratio alone. In the two last papers they have determined the axial ratios for alloys such as copper-germanium with added iron, cobalt, or nickel. By assuming that the axial ratio should be determined by the electron-atom ratio, they have evaluated the effective valence of the transition metals. The values of the effective valence are about 0.5 + 0.5 for nickel, 0.4 + 0.6 for cobalt, 0.5 + 0.5 for iron in copper and 0.8 + 0.2 for iron in gold or silver, and about 1.5 + 0.3 for manganese.

If, instead of these old valences, the metallic valences are used, 5.56 for copper, and so on, the valences calculated in the same way for the transition metals are 6.56 minus the values given above.

Accordingly the valences for iron, cobalt, and nickel come out about 6, as, in fact, they are assigned in the system of metallic valences. For manganese, which behaves oddly anyway, the value about 5 is found.

On the other hand, the values reported by Raynor on the old scheme do not make much sense. The value 0.5 for nickel is close to the value 0.6 that is calculated from the magnetic moment assuming that the magnetic moment represents a hole in the d shell, but the values for iron and cobalt do not correspond to those calculated by the same assumption, and these small valences for the transition metals seem to me not to make much sense anyway.

A simple statement that could be made in the paper on this subject is that the new system (1938 on) of metallic valences can be used to discuss the axial ratio in relation to electron-atom ratio in very nearly the same way as the old system, but the values of the valences are different.

Some of the properties of the alloys are discussed in terms of the Brillouin zones. This has been done in particular by J. Goodenough, Phys. Rev. 89, 282 (1953). I have not yet read Goodenough's paper, but only the comment on it by Raynor and Massalski, Acta Metallurgical 3, 480 (1955). You remember that in the paper on the ratio of valence electrons to atoms in metals and intermetallic compounds by Fred Ewing and me published in Reviews of Modern Physics 20, 112 (1948) we found that some of the important Brillouin polyhedra are just filled when the metallic valences are used to calculate number of electrons per unit cell - better correlation, I think, than with the old valences and the smaller polyhedra. We did not discuss the hexagonal close packed structure, and one of the things that needs to be done is to check the planes that give strong reflections and to make similar calculations for the Brillouin polyhedron containing about 5 electrons per atom, which is the value for the zeta phase given by the metallic valences. Perhaps a better discussion of the changes in slope of lattice constants and axial ratio with change in electron-atom ratio could be given with the new valences than with the old ones.

A possible way of differentiating between the old valences and the new ones is by discussion of alloys containing aluminum (or some other metal in the third, second, or first column of the periodic table). With the old valences, there is no reason why the aluminum alloys should be different from those of silican, gallium, germanium, and so on, with copper, zinc, gold, and so on. But with the new valences a difference would be expected.

The metallic valence for germanium is 2.5, and I judge for tin also, in the zeta alloys. I think that this could be checked by consideration of the interatomic distances, as in our 1956 paper. In that paper we say that silican in copper has valence 4.6, but the phase is the cubic close packed phase, rather than the hexagonal close packed phase. The axial ratio of the zeta phase for copper-silican, as given by Foley and Raynor, Trans. Faraday Soc. 57, 51 (1961), seem to be normal, corresponding to metallic valence on our scheme of 2.5 for silican.

I think that the alloys containing aluminum might be used as a test. If the metallic valence 1.56 were used for aluminum, the same consistency would be found with the new metallic valences as with the old valences (aluminum 3, copper 1, and so on). But there is no reason to use metallic 1.5 for aluminum, because all three valence electrons can be used in forming bonds, and the fourth orbital can serve as the metallic orbital. Accordingly the aluminum alloys should give electron-atom ratios on the metallic-valence system that do not fit into the same scheme as the other alloys, and a different discussion should be applied to aluminum alloys for the metallic valence system than for the old valence system.

I have only checked on one zeta phase involving aluminum. It is the silver-aluminum zeta phase, discussed by T. B. Massalski and B. Cockayne, Acta Metallurgica 7, 762 (1959). Calculated on the old valence scheme, the zeta phase runs from electron-atom ratio 1.52 to 1.77. This is rather different from the usual range, about 1.3 to 1.5 (silver-tin runs from 1.36 to 1.60). The range, of course, may not be a good means of determining electron-atom ratio corresponding to stability, because the limits involve also the structure of the adjacent phases; nevertheless the range of composition may as a first approximation be taken to indicate the stability of the phase.

If aluminum is given valence 3 and silver valence 5.5, this range corresponds to the electron-atom ratio from 4.84 to 4.54. These values are somewhat lower than those found for the zeta phase in general, which for silver-tin is from 4.94 to 5.20, and for other alloys also about 5.1 + 0.1.

However, electron transfer has to be considered in the silver-aluminum alloys, whereas for most of the other alloys it does not have to be considered. The silver-aluminum alloys involve a hyperelectronic atom, silver, and a hypoelectronic atom, aluminum. We might expect that 0.28 electrons could be transferred from silver to aluminum, still leaving aluminum with 0.72 metallic orbital, and increasing the valence of both silver and aluminum by 0.28, to 3.28 for aluminum and 5.84 for silver. The electron-atom ratio then becomes 4.82 to 5.12 - that is, rather close to the standard range.

It may be that the amount of electron transfer would change somewhat with composition. I think that a study of the interatomic distances in the silver-aluminum zeta phase, as a function of concentration, might well give information about the amount of electron transfer.

Possibly there are some magnesium, beryllium, or lithium alloys that have zeta phases, and that could be examined.

I think that some very interesting results might well be got out of examination of available information about the zeta phases. Please let me know if you would like to collaborate in this work. If you would like to, you might check the aluminum-silver alloys with respect to interatomic distance and apparent effective valence of silver and aluminum, assuming that both silver and aluminum increase by the same change in valence from their standard values 5.56 for silver and 3.00 for aluminum.

Do you have a copy of the Review of Modern Physics article and of the electron-transfer article? If not, let me know and I shall send them to you.

Much love from

[Linus Pauling]