Micro-scaled chemical oscillator networks - towards chemical computing

The need for diversification and optimization of information processing systems is increasing. Current computers based on micro-electronic chips struggle with the demand of energy-efficient processing of continuously increasing amounts of data. Unconventional computers, such as chemistry-based devices, also known as a ‘chemical computers’(CCs), are a potential answer. Chemical computing can be based on information processing by a non-linear oscillatory chemical reaction network (CRN). Other than current computation architectures, a CC can process information by exploiting the enormous parallelism of collective networks even in the presence of large amounts of noise. Likewise, it provides energy efficiency as the processing unit and memory reside in the same space. Also, many CRNs do not require any external energy for their operation [1].
Towards the development of a CC, we built a diffusion driven network for information processing. The network comprises a set of micro-scaled chemical reactors (MCRs) connected by nanofluidic channels. The MCRs contain reagents of a non-linear oscillatory CRN, the Belousov–Zhabotinsky (BZ) reaction. This reaction is based on the oxidation of an organic compound by bromic acid (HBrO3) [2] mediated by a transition-metal catalyst in acidic aqueous solution. During oscillations, the catalyst switches between two oxidation states that offer several ways of readout (color, fluorescence, redox potential). In our case, the catalyst of the BZ reaction is a ruthenium bipyridyl complex, Ru(bpy)32+, which enables visualizing oscillations by fluorescence microscopy. The Ru complex fluoresces only in the Ru²⁺ state and is dark in the Ru³⁺ state.
Previous efforts regarding the use of BZ reaction for information processing have been described: at macro-scale, a language-recognizing Turing machine based on the addition of aliquots of BZ reagents in a ‘one pot reactor’ was shown [3]. Another example features a programmable chemical processor in a reaction array of cells fluidically connected [4]. For higher-speed communication among reactors, micro-scaled systems were proposed [5]. However, in these cases, complex architectures of reactors presenting directional diffusive coupling of MCRs were not explored.
To harvest the energy efficiency and nanoscale dimensions of molecules, chemical computing units, we will have to be scaled to dimensions in the micrometer range or below. Considering that, as a first step, we successfully fabricated miniaturized, complex architectures of BZ-MCR networks in silicon by photolithography, as presented in Figure 1. The advantages of silicon-based fabrication of MCRs are the use of a chemically inert material and the scalability and freedom of design inherent with silicon based micro-fabrication. On our chips, the MCRs’ diameters range from 5 to 30 μm, their depth is 20 μm. In Figure 2, the proof-of-concept of communication in a MCR network is presented. We visualize the branching of a BZ reaction wave in a couple of 22.5μm diameter MCRs. When the wave arrives at the branching point (figure 2-d), it is transmitted to both directions (figure 2-e). In Figure 3, one example of the synchronization of different MCR sizes is presented. Two equally sized MCRs (reactor 10 and 12) are both connected to a MCR with a 4x larger diameter (reactor 11). The coupling mediates synchronization of the slower oscillations in the large reactors with the faster oscillation in the smaller reactors. After a few oscillation cycles, reactor 11 synchronizes to twice the period of reactors 10 and 12 (figure 3-d).
In conclusion, we demonstrated uniform device behavior in large arrays of silicon based MCRs. Connecting channels mediate a coupling between compartments which leads to synchronization, an important ingredient for data processing. These steps are important first results towards scalable chemical computing architectures based on simple molecules.
We thank European Union’s Horizon 2020 - Marie Skłodowska-Curie grant agreement No 812868.

[1] E. Bergh et al. Advances in Unconventional Computing. Springer, Cham (2017) 677-709.
[2] R.J. Field et al. Journal of the American Chemical Society 94.25 (1972) 8649-8664.
[3] M. Dueñas-Díez, and J. Pérez-Mercader, iScience 19 (2019) 514-526.
[4] J.M. Parrilla-Gutierrez et al. Nature communications 11.1 (2020) 1442.
[5] I.L. Mallphanov and V. K. Vanag. Russian Chemical Reviews 90.10 (2021) 1263.