Original title: Coherent advantage in the computational expressivity of excitonic networks
Matthew Du, Carlos Floyd, Dipti Jasrasaria, Suriyanarayanan Vaikuntanathan · September 2026
Artificial intelligence uses huge amounts of energy. Because of this, scientists want to use physical matter instead of standard computers to run trainable programs. In this study, researchers looked at networks of light-absorbing molecules called chromophores. They tested these networks as a computational setup where the energy is constantly pumped in and lost. The connections between the molecules act as the input, and the final steady balance of energy acts as the output. The team found that quantum coherence, where energy moves in synchronized waves, allows the network's computing power to grow as the network gets bigger, much like regular artificial neural networks. However, when strong disturbances scramble this coherence, this growth and the overall computing power both get suppressed.
Imagine trying to pass a secret message down a line of people. If each person whispers the message to the next, and they all do it clearly and at the same time, the message gets through perfectly. This is like coherence. But if people start mumbling, or talking over each other, or forgetting parts of the message, the final message gets garbled. Here, the researchers found that the 'whispering' (coherence) is what allows the message (computation) to be passed accurately and to grow in complexity as the line gets longer.
AI uses a lot of electricity. Using natural molecular systems for computation could offer a more energy-efficient alternative. This research shows that a key quantum property, coherence, is essential for these molecular systems to be powerful enough for complex tasks.
The abstract does not specify the exact computational tasks these networks can perform, nor does it quantify the expressivity or the network size achieved. It also doesn't detail the specific methods used to control couplings or measure coherence.
Ask anything about this paper. Answers come from its abstract, and say so when the abstract doesn't cover your question.
Coherent advantage in the computational expressivity of excitonic networks — Matthew Du, Carlos Floyd, Dipti Jasrasaria et al.
Abstract licence: CC0
Explanation generated by Synthesis from the abstract — it is not the original paper. This is a preprint, so it hasn't necessarily been peer reviewed.