VOL.

46, 1960

GEOPHYSICS: H. CRAIG

1221

for on the basis of induced chromosome interchanges. In addition, it was found that diethyl sulfate treatments induced a high frequency of mutations, and further investigation revealed that the spectrum of mutation types appeared to be different for the two mutagenic agents being compared. The distinct lack of association of chromosome structural aberrations with mutations for a highly effective mutagenic agent has broad fundamental as well as practical implications.

The writers wish to acknowledge that the selection of diethyl sulfate as a mutagen for study was made as a result of a suggestion from Dr. Mogens Westergaard, Copenhagen, Denmark. Correspondence with Dr. James MacKey was helpful in developing experimental techniques. Discussions with Drs. R. F. Foster and Kermit Groves contributed much to the design of the experiments. We are indebted also to Miss Edith Froese-Gertzen, Mrs. Sally Wilbur, and Mrs. Diann Robbers who assisted in preparation of certain treatments or in cytological analyses. The helpful suggestions of many others at W. S. U. are also gratefully acknowledged.
* This research was supported in part by funds from the U. S. Atomic Energy Commission Contract AT-(45-1)-353, in part under U.S. Public Health Service Grant A-2184, by the Washington State Department of Agriculture, and conducted under Washington Agricultural Experiment Stations, Projects 4068 and 1435. t Contribution from the Departments of Agronomy and Agricultural Chemistry, Washington State University. Scientific paper no. 1973, Washington Agricultural Experiment Stations. ' von Wettstein, D., Arbeitsgemeinschaft fur Forschung des Landes Nordrhein-Westfalen (Koln und Opladen: Westdeutscher Verlag) Heft 73, 7-60. 2 Rapoport, A., Dok. Vsesoyuz. Akad. Sel'sko-Khoz Nauk V. I. Lenina, 12, 12 (1947).

3Heslot, H., and R. Ferrary, Ann. Inst. Nat. Agron., 44, 3 (1958). 4Ehrenberg, L., in Proceedings of the Symposium on Chemical Mutagens (held at Gatersleben, Germany, August 1959, in press).

6 Stadler, L. J., Missouri Agr. Expt. Station, Res. Bull. 204, 29 (1933). 6 Gustafsson, A., Lunds Univ. Arsskr. N. F. Avd. 2, Bd. 36, Nr. 11, 1-40 (1940).
7Gaul, H., in Proceedings of the Sixth Congress of the Societa Italiana di Genetica Agraria (held at Forli, Italy, October 4-6, 1959, in press). 8 Steel, R. G. D., and J. H. Torrie, Principles and Procedures of Statistics (New York: McGraw-Hill, 1960), p. 456-7. 9 Ehrenberg, L., et al., in Conference on Chromosomes (Zwolle, the Netherlands: W. E. J. Tjeenk Willink, 1956), pp. 131-159. 10 Westergaard, M., Experientia, 13, 224-234 (1956). '1 Loveless, A., and S. Howarth, Nature, 184, 1780-1782 (1959).

THE THERMODYNAMICS OF SEA WA TER BY HARMON CRAIG

DEPARTMENT OF EARTH SCIENCES AND SCRIPPS INSTITUTION OF OCEANOGRAPHY, UNIVERSITY OF CALIFORNIA, LA JOLLA

Communicated by Walter H. Munk, July 15, 1960

In a recent volume dealing with the physics and chemistry of sea water,' Fofonoff2 presents a systematic application of the equations of thermodynamics to an ocean water system. Few subjects are more fundamental to the study of the sea and more neglected in application, reflecting the inherent difficulties in treating a

1222

GEOPHYSICS: H. CRAIG

PROC. N. A. S.

multicomponent system in natural surroundings and in treating nonequilibrium systems. Confronted with these obstacles, it seems only reasonable to hope that oceanography, in contrast to meteorology which presents much simpler problems, might refrain from attempting to establish a unique thermodynamic jargon of its own. Especially in dealing with natural systems, a clear and precise formulation according to established theory is necessary to prevent isolation from scientists in other fields, and indeed to insure the correct development of the subject. Fofonoff's treatment seems to me to be erroneous in several respects, and to be misleading in the development of terminology which obscures the formal simplicity of general usage. In what follows, I discuss in particular two aspects of Fofonoff's paper: the ocean as a thermodynamic system in a gravitational field, and his treatment of the chemical potential in dealing with the components of sea water. Gravitational Field Treatment.-Fofonoff points out that the ocean is a complex thermodynamic system in a gravitational field, consisting of infinitely many phases in vertical extent. To describe the changes in internal energy of a general sea water system, he states the "First Law" as (1) dU= TdS-PdV+Z(, +4)dmi
an equation which he quotes from Guggenheim.3 (I use Guggenheim's terminology, rather than Fofonoff's, throughout this paper, except that mi is the mass of component i and , is the "specific" chemical potential referred to mass units instead of moles.) 0 is the gravitational potential gh. Integrating (1) in the usual way gives: (2) U = TS-PV Z + mii + mk where m = mi. (Fofonoff uses intensive form from this point on, but this does not affect the argument.) Differentiating (2) and subtracting (1), Fofonoff derives what he proposes as the "Gibbs-Duhem relation" for sea water, (3) SdT- VdP + Z midL + md4, = 0 and states that the partial derivative of specific "internal" entropy with respect to salinity can be calculated from this relation when the chemical potential of water is known. Fofonoff's relation (3) is, however, incorrect because equation (1) actually applies only to a specific phase, i.e., to matter at a fixed gravitational potential. Thus in the differentiation of (2), the term md4 = 0, and this term will not appear in (3), as indeed it should not. Gibbs, on the other hand. in his discussion of the gravitational field,4 treats an arbitrary element of mass within the system; here there is an additional term md4 in (1) as applied to the mass element. This term will also occur on differentiating (2), so that it cancels in the subtraction. In either case we obtain the correct Gibbs-Duhem equation: SdT - VdP + E midyi = 0 which is therefore invariant in the gravitational field. Gibbs himself stated that this relationship (which he derived without considering the gravitational field) is true for any continuous variations of phase, which is the reason that the equilibrium conditions in a gravitational field are overdetermined.'

VOL. 46, 1960

GEOPHYSICS: H. CRAIG

1223

Equation (2) is always true for an arbitrary mass element and therefore is also true in the special case of fixed 4 treated by Guggenheim. The difficulty only occurs when we try to integrate Guggenheim and differentiate Gibbs. To describe a general sea water system at equilibrium, as Fofonoff wishes to do, the correct Gibbs-Duhem equation must be integrated over all mass elements (Gibbs) or all phases (Guggenheim), under the equilibrium constraint dT = 0 (neglecting the insignificant relativistic gradient). This simply gives the condition for hydrostatic equilibrium, as was shown by Gibbs. Chemical Potential.-Considering sea water as a system of water and "salinity," Fofonoff defines:
A
=

As

AW

where ,Uw = chemical potential of water (mass units); IA, = chemical potential of dissolved salts ("salinity"), assuming the components of salinity occur in fixed proportions; and = "chemical potential of sea water." (Fofonoff actually calls all these functions "partial chemical potentials" in contrast to an ordinary chemical potential which he attributes to a pure phase. This is equivalent to partial Gibbs free energy, a term which has no meaningful definition.) On page ix of the volume referred to,' we find that the Committee on Physical Properties considers the "chemical potential of sea water" to be an important property which should be measured and computed. It is therefore somewhat discouraging to note that the M defined above is not by any possible definition a chemical potential. If we consider salinity to be a pure salt of fixed composition, then this function is:

/h9

VJXSPT

a6s,

"

Xs

where x, is the mass fraction of salinity and g is the specific Gibbs free energy per unit mass of sea water. Gibbs used specific chemical potentials in terms of mass units, but he did not treat the thermochemistry of solutions. eofonoff does not discuss the concentration dependence of his potentials in terms of the mass fractions he uses, but the greatest difficulty in actually using such functions is that the chemical potential of a component of an ideal solution is no longer linear in the logarithm of concentration. Thus in Fofonoff's units, we have for the chemical potentials of water and "salinity":
JA

= JA+ R',fT In x.- RwT In -x 1- ..) + M"]

L ~Ms

Msj

As= As* + RT In x, -RT In x,(1

M-

+M

where M, and M, are the molecular weights of water and "salinity," and R_

1224

GEOPHYSICS: H. CRAIG

PROC. N. A. S.

specific gas constant, = R/M1. These expressions involving mass functions are to be contrasted with the simple ideal solution dependence in terms of molal functions and mole fractions:
= (i.)? + RT In (x) where the parentheses signify molal chemical potentials and mole fractions. The use of Fofonoff's function ji puts standard equations in a form difficult to use. For example, Fofonoff's equation (12) for the freezing point depression in sea water in equilibrium with ice (assuming the ideal solution law) is:
[SW
-

Sice](x

X= -Xs

)P

where si is specific entropy of phase i and his ji is as defined above. One may ask what the right hand side of this equation is, in terms of usable functions. In Fofonoff's units, the equation as it must be used becomes:
[SW
x) + x8] = SiceI(aZ) [1-i (1c-

-R4T

This equation and the previous one are to be compared with the simple expression given in molal units and mole fractions:

[(S)

(s) - ice] I

t)=
_
.,

(T1

Fofonoff comments that he uses specific chemical potentials in mass units in order to use salinity as a state variable. It will be noted, however, that both here and above the molecular weight of "salinity" enters his equations whenever one puts them into the form in which they are actually used. Salinity can be used as a variable in any type of units, depending on the problem. However, in considering the properties of sea water, the function p is only an apparent convenience which actually obscures the simplicity of most of the thermodynamic equations, regardless of the type of units used for salinity. In reality, of course, ;,u cannot be defined without treating salinity as NaCl, for if the proportions of dissolved salts are not constant, the Gibbs function changes even though x, may be constant. Since G is also a function of components not included in salinity, such as dissolved gases, combined CO2, and dissolved organic matter, the concept of ins as a summation involving fixed multipliers is not very useful. One either treats the ocean directly as a two-component system or one doesn't, but in either case the molecular functions are always the more useful when even the simplest chemistry is involved. The "chemical potential of sea water" could be correctly defined by treating a system whose components are sea water (of a fixed composition), pure water, and any other components desired, using a three or more component system to treat the effects of any particular species. For example we could take standard Copenhagen sea water as a component of sea water and compute or measure y for this component in simple systems. It is doubtful whether much would be gained compared to

VOL. 46, 1960

MATHEMATICS: WALSH AND MOTZKIN

1225

treating the ocean simply as water and sodium chloride. Still, the nomenclature should be preserved for an actual component of the system in case such a treatment becomes useful.
I Physical and Chemical Properties of Sea Water, National Academy of Sciences-National Research Council (Washington, D. C.), Pub. 600 (1959). 2 Fofonoff, N. P., "Interpretation of Oceanographic Measurements-Thermodynamics," ibid., 38-66. 3 Guggenheim, E. A., Thermodynamics, (North-Holland Publishing Company, 1957), p. 404. 4Gibbs, J. W., Collected Works, (Yale University Press, 1928), Vol. 1, p. 146.

BEST APPROXIMATORS WITHIN A LINEAR FAMILY ON AN INTERVAL* BY J. L. WALSH AND T. S. MOTZKIN

HARVARD UNIVERSITY AND UNIVERSITY OF CALIFORNIA

Communicated July 19, 1960

It is our purpose in the present note to investigate the behavior of functions p(x) of best approximation to a given function f(x), where p(x) is supposed to belong to a linear family P and where deviation is measured on E:O < x < 1 by a generalized norm - p(x)ll = fE r(lf(x) - p(x)I)w(x)dx

11f(x)

involving a transformer r(t) and a weight function w(x). This norm was recently studied by the authors' in the special case T(t) = tp, p > 0, w(x) = 1, and is now to be studied more generally (but not including the case p < 1). For positive r(t) we may write r(t) = ta(i),-and our problem becomes that of variable-power approximation, an aspect that we do not stress for nonconstant a(t). We use r(t) directly to define our norm, in contrast to Orlicz who introduces a function comparable with our r(t) but defines quite differently a norm in the classical sense. We considered in reference' and consider particularly here properties of oscillation of f(x) - p(x) on E and identical vanishing of f(x) - p(x) on a subset of E of positive measure, extensions of classical results due to D. Jackson. Our previous results derived necessary conditions for best approximators, whereas here we treat also sufficient conditions. Under suitable conditions we prove the existence of a weight function so that a given p(x) shall be a best approximator. Our methods involve a certain algebraic inequality akin to orthogonality (inequality [1] below), of a type which was recently used by one of us2 in an invited address. The authors have previously investigated approximation on a finite set with the pth power norm, but postpone such consideration of the r-norm to another occasion. 1. We consider henceforth a deviation as defined by a generalized norm

lg(x)JJ =fE r(g(x)I)w(x)dx,

E:O <

< 1

where -r(t) is defined for all t, 0 < t < a, where r'(t) exists and is continuous for such t, and where w(x) is continuous on E and positive a.e. We study for a given function