A molecular computing approach to solving optimization problems via programmable microdroplet arrays

Si Yue Guo, Pascal Friederich, Yudong Cao, Tony C. Wu, Christopher J. Forman, Douglas Mendoza, Matthias Degroote, Andrew Cavell, Veronica Krasecki, Riley J. Hickman, Abhishek Sharma, Leroy Cronin*, Nathan Gianneschi, Randall H. Goldsmith, Alán Aspuru-Guzik

*Corresponding author for this work

Research output: Contribution to journalReview articlepeer-review

Abstract

The search for novel forms of computing to the dominant von Neumann model-based approach is important as it will enable different classes of problems to be solved. Molecular computers are a promising alternative to semiconductor-based computers given their potential biocompatibility and cost advantages. The vast space of chemical reactions makes molecules a tunable, scalable, and energy-efficient computational vehicle. In molecular computers, memory and processing units can be combined into a single, inherently parallelized device. Here, we present a microdroplet array molecular computer to solve combinatorial optimization problems by employing an Ising Hamiltonian to map problems heuristically to droplet-droplet interactions. The droplets represent binary digits and problems are encoded in intra- and inter-droplet reactions. We propose two implementations: first, a hybrid classical-molecular computer that enforces inter-droplet constraints in a classical computer and, second, a purely molecular computer where the problem is entirely pre-programmed in the nearest-neighbor droplet reactions. Among the alternatives to conventional silicon-based computers, chemical reactions have the potential for high information density and massive parallelization while remaining accessible and cost-effective. Previous work in this field has focused on using chemistry and biology to emulate circuit components, such as logic gates, or to explore a combinatorial space in parallel. These approaches can be limited in their scalability, ease of programming, and applicability to a wide range of computational problems. The molecular computer presented here attempts to address these challenges by drawing inspiration from the Ising model for magnetism. Ideal for tackling combinatorial optimization, the device is an array of microdroplets subjected to pre-programmed droplet-droplet interactions that encode a given problem. This molecular computer offers the opportunity to use chemistry to overcome barriers in classical computers, such as high energy consumption, the von Neumann bottleneck, and the combinatorial explosion of computational problems.

Original languageEnglish (US)
Pages (from-to)1107-1124
Number of pages18
JournalMatter
Volume4
Issue number4
DOIs
StatePublished - Apr 7 2021

Keywords

  • combinatorial optimization
  • hybrid classical-molecular computing
  • Ising machine
  • molecular computing
  • Monte Carlo simulation

ASJC Scopus subject areas

  • Materials Science(all)

Fingerprint

Dive into the research topics of 'A molecular computing approach to solving optimization problems via programmable microdroplet arrays'. Together they form a unique fingerprint.

Cite this