STUDIA UNIVERSITATIS BABEŞ-BOLYAI, PHYSICA, SPECIAL ISSUE, 2003 THE THEORY OF THE THERMAL-DIFFUSION COLUMN: A CRITICAL REVIEW Ilie HODOR National Institute for R&D of Isotopic and Molecular Technologies P.O.Box 700, 400293 Cluj-Napoca, Romania Abstract. It is critically reviewed the theoretical literature of the thermal-diffusion column: the original and improved Furry-JonesOnsager theories, the forgotten effect, the involvement of the chemical exchange reactions, and the application of separation-of-variables to the linearized problems. A series of weaknesses are pointed out and tentative plans for more rigorous theoretical approach are mentioned. Introduction Thermal-diffusion (TD) consists in the fact that a temperature gradient in a mixture of fluids gives rise to a flow of one component relative to the mixture as a whole. The phenomenon has a small separation effect, which is very small for isotopic mixtures. In 1938 Clusius and Dickel[1] found their thermal-diffusion column, which can multiply considerably the TD elementary separation effect with a countercurrent convection in the column. The mixture is placed between two vertical walls, one hot, and one cold. There are two typical geometries: (a) the plane case, when the two walls are plan-parallel, and (b) the cylindrical case, when the two walls are coaxial cylinders. The horizontal temperature gradient gives rise to a density gradient, which, at its turn, gives rise to the convective countercurrent flow. The geometrical conditions are chosen so that the flow is lamellar. In 1938, Furry, Jones, and Onsager (FJO)[2-4] published their famous theory of the TD column, which has played a central role in numerous studies that have followed. A few years ago I discovered an error in the FJO theory and decided to re-examine the main theoretical TD literature. The purpose of this communication is to show the present state of this undertaking. FJO theory The FJO theory is derived for a binary mixture and assumes the validity of the following equations for the mass flux J 1* due to molecules of species 1 and for the molecular average velocity of all species v respectively, ILIE HODOR J1* [vc1 D( c1 c1c2 ln T )] (1) r ( / r )r (v / r ) (dp / dz ) g (2) 1 where - mass density, c i - mole fraction of species i , D - coefficient of ordinary diffusion, T - absolute temperature, - thermal diffusion constant, p - pressure, - viscosity, g - acceleration of gravity, and (r, z) - cylindrical coordinates. Eq. (2) involves the assumption that v is parallel to z-axis. The main result of the column theory is the transport equation 1* *c1 H *c1 (1 c1 ) ( K c* K d* )( dc1 / dz ) , (3) where 1* stands for mass transport of the species 1 through the column, * stands for total mass transport through the column, and H * , K c* , K d* are some definite integrals. An error in the FJO theory In 1984 I published an overall axiomatic theory of the separation column[5] and, as an application, derived an independent theory of the TD column. My result and that of FJO were similar and I thought, at that time, that they coincide. Kitamoto et al.[6] studied hydrogen isotope separation using thermaldiffusion column for hydrogen gas. When the mixture is binary, say H-D, three molecular species D2-DH-H2 are present and the exchange reaction H2+D2=2HD takes place at the hot wall. For such a system Kitamoto et al.[6] developed a theory with 12 column constants instead of the three constants ( H * , K c* , K d* ) in Eq. (3). From the theory in Ref. 5 it follows that if only two isotopes are present and if the isotopic effects are small, then an equation of the type (3) can be derived, with only three constants, indifferent of how many molecular types and what kind of exchange reactions are involved. Thus, I derived an equation of the type (3)[8] for the Kitamoto et al.’s problem. To verify my result I considered the limit case when deuterium concentration is very small, in which case, my coefficients (H, Kc, Kd) must coincide with those of FJO. After repeated verification I concluded that the two results did not coincide. Then I analyzed the original FJO papers and made an astonishing discovery: the discrepancy was due to an error in the FJO theory. Namely, in the FJO theory, it is constantly assumed that: the mass of a component in a binary mixture is given by the product between the mass of the whole mixture and the molar concentration of that component. This is strictly correct only if m2 / m1 1 , where m1 and m2 (with m2 m1 ) are the molecular masses of the two components. The same error is included in Eq. (1), the corresponding correct equation of the flux being THE THEORY OF THE THERMAL-DIFFUSION COLUMN: A CRITICAL REVIEW J1 m1n[c1v D(c1 c1c2 ln T )] (4) where n is the total molar density. I corrected[9] the FJO theory using the original FJO method of derivation though the more rigorous method from Ref. 5 could be applied. The result was similar to Eq. 3, 1 m1ˆc1 Hc1 (1 c1 ) ( K c K d )( dc1 / dz ) (5) where ˆ is the total molar transport through column. For distinction, the quantities affected of the FJO error are marked with asterisk; that is, the equations for (H, Kd, Kc) differ from those corresponding to ( H * , K c* , K d* ). The correlation between corrected and original theory is given in Ref. 8. The derivation procedure of the FJO theory is not rigorous; it uses a series of intuitive simplifying assumptions. This circumstance has generated discussions in literature and attempts have been made to improve the theory. However, the FJO error was not noticed in an interval of about sixty years; there are two explanations for that: 1. As shown in Ref. 8, the effect produced by FJO error is really large only if the following two conditions are simultaneously fulfilled: i. The TD-column is operated with extraction (not at total reflux); ii. The ratio m2/m1 is sufficiently large (for a 3He-4He mixture, when m2 / m1 1.25 , the error effect can be up to 33%; if m2 / m1 10 the error effect can be huge). 2. Usually, a paper is not analyzed in order to find elementary errors in it, the more so when the authors are eminent (Onsager won Nobel Pries in 1968). (It is to note that I discovered the error by verifying a new theory for which I took FJO theory for granted.) It seems that the two conditions (i and ii) have not been fulfilled in the experimental research. For simplicity, the TD column was generally studied at total reflux (closed column) and not many studies were made with a ratio m2/m1 sufficiently greater then unity. One could assert that the curious FJO-assumption is not an error but merely a simplifying assumption. That cannot be true as this assumption is not necessary in the derivation of the theory and, on the other hand, FJO had in view to apply their theory to helium isotope separation[4] where the ratio m2/m1 is substantially greater then unity. Other comments on TD column theory The comments that follow are based on a series of previous theoretical studies[5,7-11]. 1. Besides of the original[2-4] and corrected[9] FJO theory, a variant derived by Rutherford[12] should be mentioned. ILIE HODOR Rutherford[12] affirmed that “the FJO theory was originally developed for mixtures of heavy isotopes” even if Jones and Furry applied their theory also to helium isotope separation (Ref. 4, pg. 210). He considered that “the theory can be extended to include mixtures of light isotopes” merely by using mass fraction wi instead of mole fractions ci and mass average velocity vm instead of molar average velocity v . Doing this change, Rutherford re-derived the theory using the flux equation J1 w1vm D(w1 w1w2 ln T ) . (6) One can demonstrate by algebraic manipulation that this flux is identical with that given by Eq. (4). That is, Rutherford used a correct flux so that he derived a correct theory, but he did not mention this essential quality as compared with the FJO theory. Both theories, Rutherford’s[12] and mine[9], are similar corrected FJO theories, but strictly speaking they do not coincide: in the first[12] it is supposed that vm, in the second[9] that v is parallel with z-axes; or it was shown[9] that the two conditions generally cannot be fulfilled simultaneously. However, both theories assume that is small, which mathematically means 0 , and for this limit (vm v ) 0 . It follows that the two theories coincide to the limit, or that the difference between them is a term of the second order (proportional to 2 ). Rutherford tested his theory by separation experiments at total reflux with mixtures of 3He and 4He. In spite of the fact that the experimental errors were quite large, he considered that his theory is suitable for mixtures of light isotopes. We know now that, at total reflux, all three theories FJO, ([9]), and ([12]) are equally suitable, the difference between them being only a term of the second order which should be covered by experimental errors. 2. Forgotten effect. In the original[2-4] and corrected[9,12] FJO-theory, it is assumed that the horizontal gradient of density is caused only of temperature gradient. De Groot et al.[13] drew the attention upon a “forgotten effect” which consists in the influence of the concentration gradient upon the density gradient. Numerous attempts[14-19] have been made to derive the phenomenological theory of the TD-column with forgotten effect included but to date there has no rigorous derivation. Prigogine et al.[20] published batch experiments on TD-column of some binary organic solutions in which they obtained bizarre separation results: the difference in concentration between top and bottom started in one direction, slowed, and then went in the opposite direction, giving separations opposite from the original direction. This behavior, which has been called concentration reversal, was attributed by them to the influence of concentration on density, that is to forgotten effect. Concentration reversal was also observed experimentally by Jones and Milberger[21]. THE THEORY OF THE THERMAL-DIFFUSION COLUMN: A CRITICAL REVIEW Using carefully constructed and operated columns, Korchinsky and Emery[22] were unable to obtain concentration reversal under conditions similar to those in which it has been reported in the literature[20,21]. They also solved numerically the basic equations that describe the TD-column, including the forgotten effect. They concluded that: “When density of the liquid in the column varies with concentration, the forgotten effect is important in transient batch operation of columns, increasingly so as the wall spacing decreases, but it has no influence at the steady state.” This last part of the conclusion that the forgotten effect “has no influence at the steady state” cannot be true. Indeed, let us suppose a binary mixture with the property / T 0 . In this case the countercurrent convection is caused only of the concentration gradient. With other words, the multiplication of the elementary separation effect on the TDcolumn is produced only of the forgotten effect. It means that the numerical method used by Korchinsky and Emery, which is not presented in detail in Ref. 22, has weaknesses. Horne and Bearman[23], observing that their results does not agree with the new literature data[22,24], reanalyzed their theory and added a corrective term to their forgotten effect. However, this correction was not sufficient to obtain an agreement with Korchinsky and Emery’s result[22]. The conclusion is that in spite of the fact that many authors have studied the forgotten effect; there is not a satisfactory theoretical treatment of this subject. 3. The thermal diffusion term in Eq. (4) contains the product c1c2 c1 (1 c1 ) , which is not linear. When c1 does not vary much along column, this term can be linearized and the whole mathematical problem became linear in c1. The linearized problem can be solved by separation-ofvariables method. Tsay and Yen[25] used this method but they did not arrived to understand and overcome the encountered special difficulties. However, there are studies on similar systems[7,10,11,26,27] so that linearized TD problems could be correctly solved. 4. Spindel and Taylor[28] studied the concentration of 15N by chemical exchange 14NO2+15NO2=15NO2+14NO in a TD column. No theory was derived so far for such complex processes on a TD column. The chemical exchange in the TD column represents a domain that is insufficiently explored and it is possible to find in it processes of remarkable practical interest. For this reason it would be worth to develop adequate theoretical means. Tentative plans for future theoretical studies Much theoretical work on TD column have been done so far, nevertheless, as shown above, many important problems are not satisfactorily solved. My tentative plans are to complete the TD-column theory by finalization and publication of the following series of theoretical works: ILIE HODOR a. The rigorous theory of the TD column (without forgotten effect; done[8]); b. The rigorous theory of the TD column with forgotten effect included; c. Theory of the TD column with chemical exchange at the wall (done[8]); d. Theory of the TD column with chemical exchange in inner fluid mixture; e. Theory of the TD column with misaligned cylindrical walls (problem studied numerically by Sørensen et al.[29]); f. Second order terms in the theory of the TD-column (the theories a—e refer to the first order terms obtained in the supposition that is small); g. Solving linearized problems of the TD-column by separation-ofvariables method. These problems can be solved using the methods developed in Refs. 5 and 11. REFERENCES [1] [2] [3] [4] [5] [6] K. Cl u si u s a nd G. D ic ke l: Naturforschung, 26, 546 (1938) W . H. F ur r y, R. C. J o ne s, a nd L. O n sa ge r : Phys. Rev. 55, 1083 (1939) W . H. F ur r y a nd R. C. J o ne s : Phys. Rev., 69, 459 (1946) R. C. J o n es a nd W . H. F urr y : Rev. Mod. Phys., 18, 151 (1946) I . Ho d o r : Isotopenpraxis, 20(9), 330 (1984) A. K ita mo to , M. Ot h a, K. Ha se g a wa : In Proceedings of the International Symposium on Isotope Separation and Chemical Exchange Uranium Enrichment, Tokyo, Oct 29 – Nov 1, 1990; Fujii, Y., Ishida, T., Takeuchi, K., Eds.; Bull. Res. Lab. for Nucl. Reactors, Tokyo Institute of Technology, Tokyo 1992, pp. 460-471. [7] I . Ho d o r : ibid., pp. 482-485 [8] I . Ho d o r : Unpublished works [9] I . Ho d o r : Sep. Sci. Technol., 38(5), 1229 (2003) [10] I . Ho d o r : Thesis (Univ. Babeş-Bolyai, Romania, 1971) [11] I . Ho d o r : Studia Universitatis Babeş-Bolyai, Physica, 28, 15 (1983) [12] W . M. R u t her fo r d : J. Chem. Phys., 53, 4319 (1970) [13] S. R. De Gr o o t, W . Ho o ge n str aat e n, a nd C . J . Go rt er : Physica, 9, 923 (1942) [14] S. R. De Gr o o t, C. J . Go rt er, a nd W . Ho o g e n str aat e n : Physica, 10, 81 (1943) [15] S. R. De Gr o o t : “L’Effet Soret,” N. V. Noord Hollandsche Uitgeveres Maatschappij, Amsterdam (1945) [16] E. Vo n Ha ll e : E. Ph.D. thesis, Univ. Tennesse, Knoxville (1959) [17] J . D. B ale sc h wi el er : A.I.Ch.E. J., 7, 34 (1961) [18] F. H. Ho r ne a nd R. J . B ear ma n : J. Chem. Phys., 37, 2842, 2857 (1962) [19] F. H. Ho r ne a nd R. J . B ear ma n : J. Chem. Phys., 46, 4128 (1967) [20] I . P r igo g i ne, L. d e B r o uc k ère, a nd R. A ma nd ,: Physyca 16 577, 851 (1950) [21] A. L. J o ne s a nd E. C. Mi lb er ge r : Ind. Eng. Chem., 45, 2689 (1953) [22] W . J . Ko r c hi n s k y a nd A. H. E me r y J r .: A.I.Ch.E. J. 13 224 (1967) [23] F. H. Ho r ne a nd R. J . B ear ma n : J. Chem. Phys., 49, 2457 (1968) THE THEORY OF THE THERMAL-DIFFUSION COLUMN: A CRITICAL REVIEW [24] J . C. R. T ur n er , B . D. B ut ler, a nd M. J . Sto r y : Trans. Faraday. Soc. 63, 1906 (1967) [25] S-W T s ai a nd H - M Ye n: Sep. Sci. Technol., 21(10), 1142 (1986) [26] R. P . S te i n : Chem. Eng. Prog. Symposium Ser., 61(59), 64, 76 (1965; AIChE J., 12, 1217 (1966); Chem. Eng. Prog. Symposium Ser., 64(82), 219 (1968) [27] R. B ea ls : J. Math. Phys., 22, 954 (1981) [28] W . Sp i nd e l a nd T . I . T a ylo r : J. Chem. Phys., 23, 1318 (1955) [29] J . P . Sø r en se n, M. S. W i ll i s, a nd W . E. S te wa rt : Effects of Column Asymmetry on Thermal Diffusion Separation, Univ. of Wisconsin, Jan. (1974)