“…As we have seen, this assumption is incorrect, since the acceptor and donor counts are distributed according to the binomial law (eq 10). In some cases (ε ≪ 1), a binomial law of parameters (ε, N) is well-approximated by a Poisson law of identical mean εN, but in general, this is not the case, as can be seen from the different variances of the two distributions (20) where Π(n|μ) is the Poisson distribution of mean μ (21) If one ignores this limitation, it is possible to proceed formally and note that, for large enough mean values, a Poisson distribution is well-approximated by a γ distribution γ(ν, λ), which is in a sense its extension to noninteger values (22) Using a known property of independent γ distributions A and D of means a and d, the random variable r = A/A + D can be shown to be well-approximated by a β distribution (23) Using a = εS and d = (1 − ε)S and the properties of β distributions, we obtain that the random variable r has a mean and standard deviation approximately equal to (24) This latter value is the estimate of the upper bound to the PRH width proposed by Dahan The next step toward a better understanding of the contribution of shot noise to the PRH was made by Gopich and Szabo, who first presented a complete analysis of the theoretical shape of the PRH using a fixed time bin approach and simple models of excitation profiles. 29 In particular, they obtained theoretical expressions for the standard deviation of the PRH in diverse cases (including two-state model molecules).…”