We reexamine the time-delay formalism of Wigner, Eisenbud and Smith, which was developed to analyze both elastic and inelastic resonances. An error in the paper of Smith has propagated through the literature. We correct this error and show how the results of Eisenbud and Smith are related. We also comment on some recent time-delay studies, based on Smith's erroneous interpretation of the Eisenbud result.
I. THE RESULT OF WIGNER AND EISENBUDIn 1948, Eisenbud [1] used a simple wave-packet approach to show that, for an elastic resonance, the time-delay (∆t) in a collision process was related to the energy-derivative of the phase shiftthis result, apart from a factor of two [2], also appeared in a paper by Wigner [3], which used causality to place a limit on dδ/dE. Less well known is the Eisenbud result for scattering into several final states [1,4],In the case of elestic scattering, where S = e 2iδ , this gives the result in Eq.(1). Note that ∆t, in the multi-channel case, becomes a matrix. The matrix S being symmetric and unitary was diagonalized and one element was assumed to be resonant. The resulting matrix ∆t was again found to have the form of Eq. (1) for each entry, however, with the phase shift replaced by the resonant eigenphase. Unfortunately, the same notation was used for both the phase shift and the eigenphase, and (as we will see) this may have caused some confusion.
II. THE RESULT OF SMITHSmith [5] derived the time-delay matrix based on the flux passing through some interaction region of radius R. His main result was for the average lifetime of a metastable state due to a collision beginning in the i th channel. The result wasSmith further claimed that his result and the result of Eisenbud could be connected, using the following representation for Eisenbud's resultand a relation for the average over all outgoing channelsgiving Eq. (3), as claimed. The problem with this argument lies in Eq. (4), which is not equivalent to Eisenbud's result given in Eq. (2) above.