Published Online: 22 December 2004
Journal of Mathematical Physics 3, 157 (1962); https://doi.org/10.1063/1.1703774
more...View Affiliations
  • Institute for Advanced Study, Princeton, New Jersey
The distribution function of spacings S between nearest neighbors, in a long series of energy levels with average spacing D, is studied. The statistical properties of S are defined in terms of an ensemble of systems described in a previous paper. For large values of t = (πS/2D), it is shown that the distribution of S can be deduced from the thermodynamical properties of a certain model. The model, which replaces the eigenvalue distribution by a continuous fluid, can be studied by the methods of classical electrostatics, potential theory, and thermodynamics. In this way the distribution function of spacings S is found to be asymptotically
Q(t)=At17/8exp[−14t212t]
for large t. The numerical constant A can in principle not be determined from such a continuum model. Reasons are given for considering the remaining factors in the formula for Q(t) to be reliable.
  • See also: Journal of Mathematical Physics 3, 140
    1. 1. Freeman J. Dyson, J. Math. Phys. 3, 140 (1962). Quoted in what follows as (I). Google ScholarScitation
    2. 2. E. P. Wigner, Gatlinberg Conference on Neutron Physics, Oak Ridge National Laboratory Report ORNL 2309 (1957), p. 59. Google Scholar
    3. 3. N. Rosenzweig and C. E. Porter, Phys. Rev. 120, 1698 (1960). See, also, reference 4. Google ScholarCrossref
    4. 4. C. E. Porter and N. Rosenzweig, Suomalaisen Tiedeakat. Toimituksia A VI, No. 44 (1960); see especially Figs. 19 and 20. Google Scholar
    5. 5. M. L. Mehta, Nuclear Phys. 18, 395 (1960); Google ScholarCrossref
      M. L. Mehta and M. Gaudin, Nuclear Phys. 18, 420 (1960); , Nucl. Phys. , Google ScholarCrossref
      M. Gaudin, Nuclear Phys. 25, 447 (1961)., Nucl. Phys. , Google ScholarCrossref
    6. 6. E. Schrödinger, Statistical Thermodynamics (Cambridge University Press, New York, 1952), pp. 58–62. Google Scholar
    7. 7. E. P. Wigner, Proc. 4th Can. Math. Congress, p. 174 (Toronto, 1959), has in fact used this method to determine the over‐all eigenvalue distribution of the Gaussian ensemble. For the over‐all distribution, in contrast to the distribution of level spacings, a zero‐temperature approximation is sufficient. Wigner was therefore able to derive the “semi‐circle law” for the eigenvalue distribution, using a purely electrostatic model without any thermodynamics. Google Scholar
    8. © 1962 The American Institute of Physics.