The simulation of photochemical processes has been a major challenge to theoretical chemistry because of the need to simultaneously describe quantum-mechanical effects of both nuclei and electrons. Photochemical reactions, that is, chemical reactions where the absorption of light carries a molecular system to a different region of its nuclear configuration space, stress not only our understanding of coupled electron-nuclear (or nonadiabatic) dynamics but also the current methods to calculate the (excited) electronic structure of molecules. Numerous advances have been made over the past decade, though, both in nonadiabatic molecular dynamics and electronic-structure theory, and the chemical physics community would argue that excited-state simulations have demonstrated their value in the interpretation of photochemical experiments, ranging from time-resolved spectroscopies (e.g., time-resolved photoelectron or ultrafast electron diffraction) to atmospheric chemistry.

However, one can question whether these simulations have been unambiguously predictive. True predictive capabilities would pave the way to rational design of light-driven molecular systems, with revolutionary implications for renewable solar energy (directly to electricity or fuels), bioimaging, optogenetics, photochemical synthesis, or CO₂ mitigation. Thanks to the development of advanced light sources, new ultrafast diffraction experiments have come online and provide both spatial and temporal resolution on the atomic scale, i.e., they can produce molecular movies of chemical reactions. This provides a novel opportunity – a double-blind test of the accuracy of nonadiabatic molecular dynamics simulations.

Based on intensive discussions during a CECAM workshop in 2023, we arranged for such a test and challenged the community of computational photochemists to predict (i) the photochemistry of cyclobutanone in the gas phase and (ii) its resulting ultrafast electron diffraction signal in advance of an experiment. The prediction challenge was announced in the Summer of 2023, with a deadline for submission set for January 2024. The participants could use any combination of methods of their liking, considering that they should submit a paper with their prediction before the experiment was done, leaving them with six months for their prediction. This prediction challenge resulted in an unexpectedly large number of contributions (sixteen) with a much wider variety of methods for the electronic structure and nuclear dynamics than anticipated (see https://pubs.aip.org/jcp/collection/16531/Prediction-Challenge-Cyclobutanone-Photochemistry).

Yet, the publication of these predictions is only the beginning of the adventure, as the community now needs to analyze and thoroughly discuss these results to agree on their interpretation and lead to the publication of a fact-based roadmap for future developments in the field. The proposed Flagship Workshop will offer this opportunity by using an alternative format, allowing each contributor to this prediction challenge to briefly present their results and the reasoning that motivated the choice of their strategy. Based on a series of micro-presentations, the overarching goal of this alternative-format Workshop is for the community to criticize the results of this prediction challenge in a constructive and inclusive way, leading to an agreement by this consortium on the precise outcome of this challenge and the resulting future directions for the field of nonadiabatic molecular dynamics. These elements would form the main body of a perspective article published after the workshop and conclude the Prediction Challenge.