Group leader: Géza Groma


Group website:


Group members






senior research associate




research associate




scientific administrator




Fluorescence spectroscopy of femtosecond time resolution

Recently in our lab we constructed a measuring system suitable for the detection of fluorescence kinetics in the 100 fs – 10 ns time and the complete visible spectral range. On this apparatus presently we are focusing on the study of the coenzymes FAD and NADH – substantial components of redox systems – in different microenvironments. We have found that in aqueous solution the kinetics of FAD fluorescence (Fig 1) can be characterized by several time constants: one of them, falling in the ns range, can be associated to the open conformation of the molecule, while three components in the ps range correspond to closed formations. The Hofmeister effect – controlling the ratio of molecules in the different conformational states – is well observable on the kinetics, in accordance with the models based on molecule dynamics calculations. The setup also makes possible the observation of Stokes-shift, characterized by a lifetime of ~100 fs, and the study of the related solvation dynamics. In contrast to the above parameters, in the case of FAD bound covalently to the flavocytochrome c sulfide dehydrogenase enzyme, we observed only a single sub-ps component, indicating that the coenzyme is in an extremely strong interaction with the protein microenvironment. For coenzyme NADH the equilibrium of the open and closed conformations is also traceable, however the Stokes-shift has three components, related to an unusually complex process of vibrational relaxation.


Figure 1. Fluorescence kinetics of FAD in time and wavelength domain

Analysis of complex first-order reaction kinetic systems

The above-mentioned reactions, taking place in the excited state can be characterized by first-order kinetics, as well as, for example, the light-induced ground-state reactions of retinal proteins. The standard way of the analysis of such complex system reactions is based on the method of nonlinear regression with exponential terms, the application of which faces to many difficulties. The major problem is that the solution of the fitting task is an ill-posed problem: a small noise on the experimental data generates high uncertainties on the predicted parameters, also including the number of the necessary parameters. Applying simulated data based on a very complex model of the bacteriorhodopsin photocycle we elaborated a new method for handling this problem. The essential part of this procedure is that instead of discrete exponential components, the solution is characterized by a distribution over the continuous space of time-constants, and the instability of the regression is handled by adding two regularizing terms. The weight of the first regularizing term () controls the number of the peaks of the distribution, and hence the complexity of the selected model, while the second one () determines the width of them. Applying the techniques of modern statistics (cross-validation, Bayesian optimization) we developed a machine learning method, which is able to automatically determine the proper value of these hyperparameters, exclusively from the simulated experimental data and the level of noise corrupting them (Fig 2). This method also performs very well on the analysis of the experimentally detected fluorescence kinetics.


Figure 2. Selection of the hyperparameters at the minimum of cross-validation error

Development of a multifunctional femtobiological workstation

Made possible by a grant acquired recently, and based on the unique infrastructure of the ELI-ALPS laser center of Szeged, collaborating with the groups of BRC and University of Pécs we are in the process of developing a permanent workstation with the purpose of studying ultrafast processes on biological systems utilizing light energy (light-harvesting complexes and photosynthetic reaction centers, retinal proteins, plant photoreceptors). The measuring techniques to be applied are founded partially on the expansion of the most modern technology of femtosecond multidimensional electronic spectroscopy, and partially on the improvement of the methods we developed for the detection of low-intensity light-induced coherent THz radiation. After designing the workstation, in the present first phase of implementation we work on the conversion of the infrared beam of the HR laser of ELI into the visible range, maintaining the very short pulse-length (~6 fs) and the corresponding wide spectral range. On concluding with the complete implementation, the workstation will be publicly available.