No Access Submitted: 12 December 1969 Published Online: 08 September 2003
J. Chem. Phys. 52, 3869 (1970);
more...View Affiliations
  • Theoretical Chemistry Institute and Chemistry Department, University of Wisconsin, Madison, Wisconsin 53706
View Contributors
  • Robert J. LeRoy
  • Richard B. Bernstein
An expression is derived which relates the distribution of vibrational levels near the dissociation limit D of a given diatomic species to the nature of the long‐range interatomic potential, in the region where the latter may be approximated by D − Cn / Rn. Fitting experimental energies directly to this relationship yields values of D, n, and Cn. This procedure requires a knowledge of the relative energies and relative vibrational numbering for at least four rotationless levels lying near the dissociation limit. However, it requires no information on the rotational constants or on the number and energies of the deeply bound levels. D can be evaluated with a much smaller uncertainty than heretofore obtainable from Birge–Sponer extrapolations. The formula predicts the energies of all vibrational levels lying above the highest one measured, with uncertainties no larger than that of the binding energy of the highest level. The validity of the method is tested with model potentials, and its usefulness is demonstrated by application to the precise data of Douglas, Mo/ller, and Stoicheff for the B 3Π0u+ state of Cl2.
  1. 1. R. T. Birge and H. Sponer, Phys. Rev. 28, 259 (1926). Google ScholarCrossref
  2. 2. R. T. Birge, Trans. Faraday Soc. 25, 707 (1929). Google ScholarCrossref
  3. 3. For a recent review see A. G. Gaydon, Dissociation Energies (Chapman and Hall Ltd., London, 1968), 3rd. ed. Google Scholar
  4. 4. G. Herzberg, Molecular Structure and Molecular Spectra: I. Spectra of Diatomic Molecules (D. Van Nostrand Co., Inc., Toronto, 1950), 2nd ed. Google Scholar
  5. 5.Note that while ΔG(υ) is Herzberg’s ΔGυ, ΔG(υ+12)≡dE(υ+12)/dυ is not identical to the “observable” vibrational level spacing
    ΔGυ+1/2 = υυ+1ΔG(υ)dυ
    (see p. 98 in Ref. 4).
  6. 6. (a) H. Harrison and R. B. Bernstein, J. Chem. Phys. 38, 2135 (1963); Google ScholarScitation
    (b) H. Harrison and R. B. Bernstein, Erratum 47, 1884 (1967)., J. Chem. Phys. , Google ScholarScitation
  7. 7. See the review by E. A. Mason and L. Monchick, Advan. Chem. Phys. 12, 329 (1967). Google Scholar
  8. 8. The parameters of the LJ (12, 6) potential were chosen to allow for 24 bound states. In the notation of Ref. 6, this corresponds to Bz = 2μDeRe2/ℏ2 = 10 000, where De is the well depth, and Re the position of the potential minimum. Eigenvalues were calculated numerically and are accurate to 10−7 De. This was done using a slightly modified form of the Cooley‐Cashion program: J. W. Cooley, Math. Computation 15, 363 (1961); Google Scholar
    J. K. Cashion, J. Chem. Phys. 39, 1872 (1963). Google ScholarScitation
  9. 9. I. S. Gradshteyn and I. M. Ryzhik, Table of Integrals, Series and Products (Academic Press Inc., New York, 1965), Sec. 3.251, p. 295. Google Scholar
  10. 10. M. Abramowitz and I. Stegun, Natl. Bur. Std. (U.S.), Appl. Math. Ser. 55 (1964); Google Scholar
    also Handbook of Mathematical Functions (Dover Publications, Inc., New York, 1965). Google Scholar
  11. 11. This is because a direct fit of experimental data to Eq. (4) requires a prior numerical smoothing of the data to obtain accurate values of the derivatives dE(υ)/dυ. Google Scholar
  12. 12. It is interesting to note that for n = 4 (ion‐induced dipole forces) Eq. (6) is simply a quartic in υ, and for n = 6 (induced dipole‐induced dipole, London dispersion forces) it is cubic. Google Scholar
  13. 13. By comparing Eq. (1) for E(υ) = D and E(υ) at a slightly smaller υ, W. C. Stwalley (private communication, 1969) independently obtained a result for n = 6 which, upon generalization for any n>2, may be cast into the useful form of Eq. (6). However, his factor equivalent to the present Kn is slightly less general, and his approach (unlike the present one) cannot be applied to cases with n⩽2. Google Scholar
    While seeking a “natural” analytic expression to describe the vibrational spectrum of H2, C. L. Beckel [J. Chem. Phys. 39, 90 (1963)] proposed empirical formulas somewhat similar in form to Eq. (6). Google ScholarScitation
  14. 14. P. M. Morse and H. Feshbach, Methods of Theoretical Physics (McGraw‐Hill Book Co., New York, 1953), Vol. 2, Sec. 12.3. Google Scholar
  15. 15. For pure inverse‐power potentials with n>2, there are a finite number of levels within any finite neighborhood of the dissociation limit, but there are an infinite number of discrete levels below it, extending down to infinite binding energy. For potentials with n<2, there exists a lowest level bound by a finite energy, while there are an infinite number of levels within any finite neighborhood of D. For n = 2, the levels extend down to infinite binding energy, and there are an infinite number of levels in any finite neighborhood of D. Google Scholar
  16. 16.Using the Langer WKB modification [i.e., replacing J(J+1) by (J+12)2] would require replacing Eq. (3a) by
    V(R) = D−(Cn/Rn)+(ℏ2/2μ)(1/4R2)
    . For n = 2 this just means that C2 in Eq. (7) becomes [C2−(ℏ2/8μ)], but for n≠2, the integral arising from Eq. (2) is no longer analytically soluble. However, for realistic systems the Langer correction is fortunately very small.
    R. E. Langer, Phys. Rev. 51, 669 (1937). Google ScholarCrossref
  17. 17. Within the context of the present approach, potentials with exponential long‐range tails (such as the Morse potential) correspond qualitatively to inverse‐power potentials with very large n. The purely attractive exponential potential has both a discrete lowest level and a finite number of bound states within any finite neighborhood of D. Google Scholar
  18. 18. A linear B–S plot for levels near the dissociation limit of a potential will be considered as an indication that the potential in the given region is effectively exponential in form. Google Scholar
  19. 19. Care should be taken to avoid confusion between the well depth De and D, the position of the dissociation limit. Google Scholar
  20. 20. See the discussion of intermolecular forces in (a) J. O. Hirschfelder, C. F. Curtiss, and R. B. Bird, Molecular Theory of Gases and Liquids (John Wiley & Sons, Inc., New York, 1964). Google Scholar
    (b) J. O. Hirschfelder and W. J. Meath, Advan. Chem. Phys. 12, 3 (1967). Google Scholar
  21. 21. Note that in the case where some of the dominant terms in Eq. (13) are repulsive (i.e., their Cm<0), some of these weighting factors will have differing signs, and the resulting value of n may then lie outside the range of the m’s of the contributing terms. If the lowest inverse‐power term is repulsive while the higher power terms are attractive, this gives rise to a potential maximum at large R. This appears to be the case for the 3Π0g+ state of I2; R. J. LeRoy, J. Chem. Phys. 52, 2678 (1970). Google ScholarScitation
  22. 22. In this context a potential is “well behaved” if it has no potential maximum and no nonadiabatic perturbation. Google Scholar
  23. 23. Of course, both errors approach zero for levels approaching D. Google Scholar
  24. 24. (a) See, e.g., the discussion by J. K. Cashion, J. Chem. Phys. 48, 94 (1968); see also Appendix A. Google ScholarScitation
    (b) A. S. Dickinson (private communication, 1968). , Google Scholar
  25. 25. See Appendix B for a summary of the theoretical ñ values for a wide variety of cases. Google Scholar
  26. 26. Nonlinear least‐squares regression computer programs for fitting arbitrary analytic functions are available at most computing centers. The present calculations used the University of Wisconsin Computing Center subroutine GASAUS for such fits. Google Scholar
  27. 27. Primes denote differentiation with respect to υ; e.g., E′(υ) = dE(υ)/dυ. Google Scholar
  28. 28. Parameter values obtained from Eqs. (15) and (16) should, in principle, be just as reliable as those obtained from Eq. (6). However, the former approach requires a prior smoothing of the data to obtain accurate values of the derivatives E′(υ) and E″(υ),27 and in practice this introduces some error. Experience has shown that while trial parameter values from Eqs. (15) and (16) are satisfactory, they are measurably improved by four‐parameter fittings to Eq. (6).26, Google Scholar
  29. 29. In all of the results presented, an initial fit of the data to Eqs. (15) and (16) yielded trial parameter values which were used to initiate the general nonlinear fit to Eq. (6).26,30, Google Scholar
  30. 30. R. J. LeRoy and R. B. Bernstein, (a) Wisc. Theoret. Chem. Inst. Tech. Rept. WIS‐TCI‐369, 1970. The report contains in an appendix FORTRAN listings of the programs used for carrying out fits to Eq. (6), (15), and (16). Google Scholar
    (b) See also, “Dissociation Energies of Diatomic Molecules from Vibrational Spacings of Higher Levels: Applications to the Halogens,” Chem. Phys. Letters (to be published). Google Scholar
  31. 31.This is mainly because of the problem of averaging the estimates of D and Kn obtained at different values of υ, to yield a mutually consistent set of parameters. It is interesting that analogous to Eq. (17)
    n = [4[E″(υ)]2/E′(υ)E‴(υ)]−2
    , but because of the above problem, this expression is less reliable than is Eq. (15).
  32. 32. Since the derivatives are obtained from the highest 11 energies only, they cannot be accurate at the end points, so only the 9 points shown on Fig. 2 are reliable. Google Scholar
  33. 33. Since the input data (level energies) are never completely error free, a given fit should always utilize at least one level more than the number of free parameters being fitted. If there is significant experimental uncertainty in the energies (e.g., more than a few percent of the level spacings), a redundancy of more than one level may be required to yield meaningful values of the parameters. Google Scholar
  34. 34. In the application of this method to the B 3Π0u+ state of I2,30 the experimental uncertainty introduces considerable imprecision into the four‐parameter fits, so that n could not be directly determined within required accuracy of better than ±1. Google Scholar
  35. 35. D. E. Stogryn and J. O. Hirschfelder, J. Chem. Phys. 31, 1531 (1959). These authors derived an analytic expression [their Eq. (89)] for the exact first‐order WKB value of υD (which omits the effect of the Langer correction16). A more exact value of the numerical constant in their Eq. (92) is 1.6826. Google ScholarScitation, ISI
  36. 36. A. E. Douglas, Chr. Kn. Møller, and B. P. Stoicheff, Can. J. Phys. 41, 1174 (1963). Google ScholarCrossref
  37. 37. The experimental data for this system are for the most common isotope 35,35Cl2; all energies are expressed relative to the υ″ = 0, J″ = 0 level of its ground electronic state. Google Scholar
  38. 38. T. Y. Chang, Mol. Phys. 13, 487 (1967); see also the discussion in Appendix B. Google ScholarCrossref
  39. 39. M. A. Byrne, W. G. Richards, and J. A. Horsley, Mol. Phys. 12, 273 (1967). Google ScholarCrossref
  40. 40. In choosing these values it is assumed that the “hook” at the end of the n = 5 curves in Fig. 5 is significant, illustrating the decrease of the error term for levels farther into the asymptotic (n = ñ) region. The indicated uncertainties (including the error bars in Figs. 5 and 6) correspond to a statistical confidence limit of 95%. Google Scholar
  41. 41. It has been shown by J. K. Knipp [Phys. Rev. 53, 734 (1938)] Google ScholarCrossref
    that C5 coefficients may be expressed as a product of an angular factor and [〈rA2〉〈rB2〉], the product of the expectation values for the square of the electron radii in the unfilled valence shells on interacting atoms A and B. Knipp presented values of the angular factors and approximate expectation values for a few systems, and T. Y. Chang [Rev. Mod. Phys. 39, 911 (1967)] extended these results considerably. , Google ScholarCrossref
    Recently C. F. Fischer [Can. J. Phys. 46, 2336 (1968)] has reported Hartree‐Fock values of 〈r2 for all shells of atoms from He to Rn. , Google ScholarCrossref
  42. 42. The erratic nature of the curve in Fig. 6 is due to the influence of small errors in the experimental energies on the fitted values of the parameters; the corresponding values of n, Cn, and υD show similar behavior. Including more levels in each fit dampens these oscillations. Google Scholar
  43. 43. Holding D fixed dampens the “noise” due to experimental uncertainty,42 yielding a more reliable segmented potential. Google Scholar
  44. 44. J. A. C. Todd, W. G. Richards, and M. A. Byrne, Trans. Faraday Soc. 63, 2081 (1967). Google ScholarCrossref
  45. 45. For a related discussion of the quasibound states, see A. S. Dickinson and R. B. Bernstein, “Some Properties of Bound and Quasibound States for Various Interatomic Potential Functions,” Mol. Phys. (to be published). Google Scholar
  46. 46. While the present method is expected to give values of Cn which are slightly small (see Sec. II.C), there is reason to suspect that the theoretical C5 value used for comparison41 may be somewhat too large. M. T. Marron (private communication, 1969) points out that Fischer’s41 values of 〈r2 are based on Hartree‐Fock wavefunctions which do not have correct asymptotic tails and that correcting for this may decrease 〈r2〉, and hence the theoretical C2. Google Scholar
  47. 47. For a few systems, such as isotopic hydrogen and most hydrides, the inverse‐power long‐range forces are relatively weak, so that the B–S plot shows negative or zero curvature even for the very highest levels. Google Scholar
  48. 48. R. B. Bernstein, Phys. Rev. Letters 16, 385 (1966). Google ScholarCrossref
  49. 49. J. A. Horsley and W. G. Richards, J. Chim. Phys. 66, 41 (1969). Google ScholarCrossref
  50. 50. See, for example, H. Pauly and J. P. Toennies, in Atomic and Electron Physics: Atomic Interactions, Part A, L. Marton, B. Bederson, and W. L. Fite, Eds. (Academic Press Inc., New York, 1968), Vol. 7, Chap. 3.1, p. 227. Google Scholar
  51. 51. Although all of the cases thus far considered correspond to ñ = 5 or 6, the present method should be even more successful for systems with smaller ñ (e.g., ñ = 4, for molecules which dissociate to ion+neutral) because of the relatively higher density of levels near D. Google Scholar
  52. 52. The present work utilized the corrected tables reported in Ref. 6b. These are available as Document No. 9499 in the ADI Auxiliary Publications Project, Photoduplication Service, Library of Congress, Washington, D.C. 20540. Google Scholar
  53. 53. Comparison of the φ values6,52 for θ = 0 and θ = 10−4 shows that this introduces negligible error. Google Scholar
  54. 54. This was done by piecewise fitting of third‐order polynomials in φ. Despite the rather large gaps between the tabulated points for large φ, this is expected to be fairly accurate since the eigenvalue distribution for the highest levels of an R−6‐tailed potential is expected to be cubic in υ (i.e., in φ).12, Google Scholar
  55. 55. Although the exact υD is infinite for the pure R−6 attractive potential, there are a finite number of levels within any finite interval about D.15 Hence the quantities D−υ) and Curve A in Fig. 8 are significant in the semiclassical (WKB) approximation. Google Scholar
  56. 56. G. W. King and J. H. Van Vleck, Phys. Rev. 55, 1165 (1939). Google ScholarCrossref
  57. 57. H. Margenau, Rev. Mod. Phys. 11, 1 (1939). Google ScholarCrossref
  58. 58. This conclusion is partly based on Chang’s conclusion41 that for the 0g+ states of O2 and Cu2, these effects do not dominate the interaction until R>60 a.u. Google Scholar
  59. 59. This case is, however, relatively uncommon; Hirschfelder and Meath20b point out that only an excited H atom can have a permanent dipole moment. Google Scholar
  1. © 1970 American Institute of Physics.