Multiexpfit includes programs for numerical data processing for
experimental sciences, for the evaluation
of physical processes characterized by exponential or
An experiment is performed as follows. A pump (for example, a short laser pulse) excites some kind of metastable states of a sample (for example, a fluorescent dye). Then the sample decays, emitting the radiation, with a multiexponential law. A time-resolved measure of the emission F (decay) is done (for example, the intensity of fluorescence). Also a time-resolved measure of the pump is performed; we will call this the instrumental response function (IRF).
The task that we want to perform is to express F as the convolution of the IRF with the sum of a given number of decaying exponentials, with time constant τn. The quality of the fit will be evaluated as the mean square difference between F and the convolutions.
The fit parameters are the τn and the amplitudes An. Actually, also other parameters must be cosidered: the time displacement δ between the measurement of F and the IRF, and the background intensity of both F and the IRF. The background intensity of the IRF is evaluated from the starting part of the IRF, before the emission. The background of F is the fit parameter b.
multiexp uses a generic non-linear fit algorithm
for finding the best fitting values for some of the parameters.
By default, only the τn are handled in this way. The values
of δ are discrete, so an exaustive scan of the values is
performed, inside a given range. The An and b values are
evaluated by an analytic formula, since the mean square error
is quadratic in both the An and b. It is possible to define
all these parameters in a way that they are handled by the generic
non-linear fit algorithm of
The IRF can be evaluated by measuring the decay of a short-living exited state. In this case, a correction τR parameter can be added, that represents the (short) decay time of the state. This can be a fit parameter too, though it is preferred to use a known value. (see M. Zuker et al, Delta function convolution method (DFCM) for fluorescence decay experiments, Rev. Sci. Instrum. 56 (1) 1985)
The fit can be performed simulteneously on a set of F. In this case, the parameters τn will be the same for all the F, and the minimization of the mean square error will be performed. The set of F will be divided into sub-sets, each one with a different IRF.
The program is based on
minuit minimizer. A
configuration file must
be provided (see section 3): it defines which file contains the
experimental data. After it has been called with
the configuration file name as argument, it opens a
Operations are described in section 4.
A different way for processing the data is through moment evaluation (
see I. Isenberg, On the theory of fluorescence decay experiments,
J. Chem. Phys. 59 (10) 1973). In this case, the experimental data
are processed in order to evaluate the moments of the distributions. From the moments,
with no fit procedure, the decay times can be extracted by analytical procedures. This
is performed by
A graphic user interface is also provided (multiexpFD)