|
A new approach to fitting a statistical model to time-resolved laser-induced fluorescence spectroscopy (TRLFS) could reveal hidden details and remove background noise, according to the German team developing the technique. Andre´ Steinborn and Boris Flach, of Dresden University of Technology, Steffen Taut of the University's Radiation Safety Group, and Vinzenz Brendler and Gerhard Geipel of the Institute of Radiochemistry, in Dresden, Germany, explain that TRLFS spectra are obtained by counting the photons emitted by a material at defined time intervals. The researchers say that any photon can be described by emission time and wavelength, which are observable parameters and so generate a spectrum. Additionally, the same photon is also described by component and peak affiliation, which are hidden from conventional techniques. TRLFS is a powerful technique often used for samples containing more than one component of interest and where each will provide at least one peak. For instance, it is used as an analytical method in actinide chemistry, for the direct speciation of luminescent metal ions such as protoactinium(IV) and uranium(VI), americium(III), and curium(III). The sample is excited using a laser and the luminescence emitted is channelled using fibre optics or lenses into a spectrograph to produce a luminescence spectrum, which can be recorded to an intensified charge coupled device (ICCD) type camera. The intensifier not only amplifies the signal but also acts as a "shutter" for the camera allowing exposure times from one millisecond down to 100 nanoseconds to be obtained. However, standard approaches to analysing such TRLFS spectra do not address the problem/opportunity of the decomposition of the fluorescence signal. This means that it is difficult, say the researchers, to evaluate the plausibility of the fitted component and peak models by means of their spectra. It is an understanding of these various attribute values for emitted photons could allow researchers to draw a probability density distribution for the material. As such, solving the problem of extracting meaning from these attributes distils to a statistical problem with incomplete data. Such problems are usually handled using expectation-maximization (EM) algorithms. The team has now devised a particular EM algorithm that can essentially fill the gaps by uncovering the hidden data. The team's initial tests demonstrate that this algorithm has the predicted advantages over the well-known least-squares analysis method. The main advantage being that the model can reveal details within the spectra, its components and peaks, as well as the hidden attributes revealed by the analysis. It also allows the spectroscopist to account for the longer fluorescence lifetimes in a mixture of components and so separate the signal from the background noise resulting in a better estimation of the component's lifetime. "The simultaneous detection of temporal and spectral model parameters provides a mutually consistent description of TRLFS spectra," the researchers say. The team concludes that their approach allows them to simultaneously obtain parameter sets for actinide mixtures that are internally consistent and reveal details of the fluorescent components within the sample, including fluorescence lifetime, peak maxima for the single-component spectra due to each species, and the associated weights of each spectrum. They concede that several simplifications were made for a comprehensive mathematical proof of the principle, but point out that similar, more sophisticated algorithms were used for experiments with real spectra. As shown in the appendix to their paper these algorithms account for realistic experimental conditions, involving variable delay times, skipping certain wavelength ranges as well as background noise. Reference:
Related links:
|
|
