ichf_logo instytut

ZB 24 - Zjawiska złożone, procesy nierównowagowe i przetwarzanie informacji przy wykorzystaniu reakcji chemicznych (J. Gorecki)


news foto

Konrad Giżyński

Konrad Gizynski (PhD degree in chemistry, 2015) works as an assistant professor at the Department of Complex Systems and Chemical Processing of Information at the Institute of Physical Chemistry of the Polish Academy of Sciences in Warsaw. His research interests cover experimental and numerical studies on self-organization phenomena, computer simulations of non-equilibrium systems and unconventional computing.



Information processing in reaction-diffusion media.

Chemical reactions are responsible for information processing in living organisms. It is believed that the basic properties of biological computing activity are reflected by a reaction-diffusion medium.We illustrate it discussing information processing with Belousov-Zhabotinsky (BZ) reaction and its photosensitive variant. The computational universality of information processing is demonstrated. Constructions of the simplest signal processing devices are described for

different methods of information coding. The function preformed by a particular device is determined by the geometrical structure of oscillatory (or excitable) and non-excitable regions of the medium. In living organisms brain is created as self-grown structure of interacting nonlinear elements and reaches its functionality as the result of learning. We study if such scenario can be repeated for generation of chemical information processing devices. The recent studies have shown that lipid-covered droplets containing solution of reagents of BZ reaction are mobile. Droplet structures can be spontaneously formed at a specific nonequilibrium constrains, for example forced by a fluid flow. Applications of droplet structures for classification tasks are considered. We present how to introduce information to a droplet structure, track the information flow inside and optimize the evolution to achieve the maximum reliability. The perspectives of such approach are investigated.


Figure 1. Image transformations observed during the time evolution of oscillating medium with Ru-catalyzed BZ reaction. (a) - initially projected image, (b-f) a sequence of snapshots showing image processing during a typical time evolution. Initial phase differences are sharp. They decrease in the following cycles as the result of reagent diffusion smoothing out contours of subsequent images.


Figure 2. Snapshots from the experimental realization of the distance sensor [1]. The upper figure shows the source: 1-mm-thick silver wire placed 2 mm away from the sensor; in the lower one the source was 12 mm away. The firing numbers are given next to the corresponding channels.


Figure 3. Artificial chemical neuron constructed with structured excitable medium. (a) - the geometry of excitable (dark) and non-excitable areas (white); (b) - the excitation of neuron body by 3 arriving stimuli [2].The medium parameters have been selected such that the neuron is excited by any 3 stimuli, but no combination of 2 stimuli can trigger it.


Figure 4. Microfluidic platform for generating uniform droplets containing BZ medium [3]. (a) Layout of the microfluidic chip for generation of a double emulsion with BZ-solution as the encapsulated phase and oil as the shell phase. The two different components BZ(A: sulfuric acid, malonic acid, ruthenium complex and potassium bromide) and BZ(B: sodium bromate and bathoferroin) mix on chip which prevents precipitation of gas bubbles. The sequence of BZ droplets gets encapsulated at the first T-junction, gets carried downstream at the second T-junction and relaxes towards final structure in the long and wide outlet chamber. (b) An example of a 9-droplet structure generated in such microfluidic reactor.


Figure 5. CANCER dataset classifier based on oscillating BZ droplets [4]. (a) The trajectory of the fitness values at the best individual droplet (the thick, solid line) and the average fitness for the population (the thin, solid line). (b) Illumination pattern for a 25 droplet network evolved as a classifier of cancer cells. Circles represent droplets in a network and the brightness of blue color is proportional to the initial illumination time. The conversion of time to blue color is shown on vertical bar. If no blue color is visible then the droplet was active from the beginning of the experiment whereas high intensity corresponds to illumination time close to the total simulation time. The amount of mutual information contained in each droplet is marked with a red color in form of a pie chart where the sector size is normalized to the maximal value of mutual information that can be obtained from employed inputs. The output droplet is marked with a wide, black border and the numbers of the input droplets correspond to the predictor number in the dataset.



1. Gorecki J, Gorecka JN, Yoshikawa K, Igarashi Y, Nagahara H. 2005. Sensing the distance to a

source of periodic oscillations in a nonlinear chemical medium with the output information

coded in frequency of excitation pulses. Phys. Rev. E 72, 046201.

2. Gorecka JN, Gorecki J. 2006. Multiargument logical operations performed with excitable

chemical medium J. Chem.Phys. 124, 084101.

3. Guzowski J, Gizynski K, Gorecki J, Garstecki P. 2016. Microfluidic platform for reproducible self-assembly of chemically communicating droplet networks with predesigned number and type of the communicating compartments. Lab Chip 16, 764-772

4. Gizynski K, Gruenert G, Dittrich P, Gorecki J. 2016. Evolutionary design of classifiers made of droplets containing a nonlinear chemical medium. Evol. Comput. (in press). 



Evolutionary design of classifiers made of droplets containing a nonlinear chemical medium
Author(s): K. Gizynski, G. Gruenert, P. Dittrich, J. Gorecki
Journal: Evolutionary Computation
Pages: In press
Year: 2016
URL: http://www.mitpressjournals.org/doi/abs/10.1162/EVCO_a_00197#.WFeNh4grJhE
DOI: 10.1162/EVCO_a_00197

Microfluidic platform for reproducible self-assembly of chemically communicating droplet networks with predesigned number and type of the communicating compartments
Author(s): J. Guzowski, K. Gizynski, J. Gorecki, P. Garstecki
Journal: Lab on Chip
Volume: 16
Pages: 764-772
Year: 2016
URL: http://pubs.rsc.org/en/Content/ArticleLanding/2016/LC/C5LC01526J#!divAbstract
DOI: 10.1039/C5LC01526J

Understanding Computing Droplet Networks by Following Information Flow
Author(s): G. Gruenert, K. Gizynski, G. Escuela, B. Ibrahim, J. Gorecki, P. Dittrich
Journal: International Journal of Neural Systems
Volume: 25
Pages: 1450032
Year: 2015
URL: http://www.worldscientific.com/doi/pdf/10.1142/S0129065714500324
DOI: 10.1142/S0129065714500324

Chemical computing with reaction–diffusion processes
Author(s): J. Gorecki, K. Gizynski, J. Guzowski, J. N. Gorecka, P. Garstecki, G. Gruenert, P. Dittrich
Journal: Proceedings of the Royal Society A
Volume: 373

Year: 2015
URL: http://rsta.royalsocietypublishing.org/content/373/2046/20140219
DOI: 10.1098/rsta.2014.0219

Droplets with information processing ability
Author(s): J. Szymanski, J. Gorecka, Y. Igarashi, K. Gizynski, J. Gorecki, K. Zauner, M. De Planque
Journal: International Journal of Unconventional Computing
Volume: 7
Pages: 185 - 200
Year: 2011
URL: http://www.oldcitypublishing.com/IJUC/IJUCabstracts/IJUC7.3abstracts/IJUCv7n3p185-200Szymanski.html