The excited-state chemistry of DNA bases is directly relevant to understanding UV-induced photodamage, which is a major cause of skin cancers. Thus, there has been considerable interest in the excited-state chemistry of DNA bases in a variety of environments. Studies have been carried out in isolation, in solution, and in fully assembled DNA strands. The chemistry of electronically excited states, particularly in the case of DNA bases, has often been described in terms of a single minimal energy conical intersection (MECI) and a single mechanistic pathway involving a minimal energy path (MEP). However, the energetics and geometries of isolated MECIs can be insufficient to explain photochemical mechanisms because the locus of conical intersections (CIs) forms a high-dimensional seam. Furthermore, MEPs can be of little relevance because CIs are often reached before equilibration can be achieved. Although considerable insight can be gained from an understanding of the geometry and topography of these seams in the neighborhood of an MECI, dynamics calculations are required to fully elucidate photochemical mechanisms. Here, we show that the DNA base cytosine is an example where multiple, chemically distinct, CIs and multiple corresponding pathways are simultaneously operative in excited-state relaxation. In this work, we focus on the excited-state behavior of isolated cytosine, which has been the subject of much controversy. Establishing the mechanism of deactivation of the isolated excited DNA bases is a critical stepping stone to complete modelling in more complex biologically relevant environments. The results we present here are consistent with the rapid quenching and trapping pathways observed in solution, thus suggesting that the basic mechanism is largely preserved in complex environments. The first femtosecond spectroscopic experiments on isolated cytosine showed multiexponential decays, with lifetimes of 820 fs and 3.2 ps. Subsequent resonant ionization experiments reported lifetimes of 160 fs and 1.86 ps. Many different MECIs and relaxation pathways have been proposed to explain this ultrafast internal conversion, and the focus to date has been on determining which of these pathways is the dominant pathway. In fact, as shown below, many of the previously proposed pathways are operative in parallel and the excited-state chemistry of cytosine resists simple description in terms of a single reaction path. The electronic structure of cytosine in the FC region is well established, for example, by multireference perturbation theory (CASPT2). The lowest singlet excited states are of pp*, nOp*, and nNp* character, [12] where nO and nN are nonbonding orbitals centered on oxygen and nitrogen atoms, respectively (see Figure 1). The optically bright pp* state is the lowest of these excited states (S1). Although there is agreement on the existence of several chemically distinct S1/S0 MECIs that are potentially relevant to cytosine photochemistry, considerable controversy remains concerning the photochemical mechanism. There are three low-lying S1/S0 MECIs, where S1 has nOp*, nNp*, or pp* character, [13–16] as well as a much higher-lying three-state S2/S1/S0 MECI. [17] Minimum-energy pathways have been determined to each of these MECIs, and there has been considerable effort to determine which of these pathways best represents the photochemical mechanism. The lowest energy of all the candidate intersections is the pp* MECI, which leads to the suggestion that the dominant pathway should involve quenching through this MECI. The geometric similarity of the pp* S1/S0 MECI in cytosine 20] and the S1/S0 MECI in uracil and thymine [3,21] prompted Merchan and co-workers to propose a unified mechanistic model for all the pyrimidine bases. As acknowledged in some of the work aimed at determining MECIs and MEPs, only molecular dynamics can definitively elucidate the interplay of these MECIs in the photochemical mechanism of cytosine. Using ab initio multiple spawning (AIMS) dynamics, we show that many of the previously described pathways occur in parallel within 1 ps. The AIMS method solves the electronic and nuclear Schrodinger equations simultaneously, allowing full flexibility in the description of the potential energy surfaces (PESs) and proper treatment of quantum mechanical nonadiabatic effects, that is, “surface crossing.” Our results are verified by direct comparison to femtosecond time-resolved photoelectron spectroscopy (TRPES) experiments and benchmark multi-state CASPT2 (MSPT2). We used the AIMS method to simulate the gas-phase singlet excited-state dynamics of cytosine following photoexcitation to the lowest bright state (S1). All nuclear coordinates are included in the simulation and the electronic structure is determined “on the fly” using the state-averaged complete active space self-consistent field (SA-CASSCF) method for the electronic structure. An adaptive basis set of frozen Gaussian trajectory basis functions (TBFs) is used to describe the nuclear dynamics. During regions of high non-adiabatic coupling between electronic states, additional basis functions are “spawned” to describe population transfer between electronic [a] Dr. H. R. Hudock, Prof. T. J. Mart nez Department of Chemistry, The Beckman Institute, and Center for Biophysics and Computational Biology University of Illinois at Urbana-Champaign 600 S. Mathews, Urbana, IL 61801 (USA) Fax: (+1)217-244-3186 E-mail : tjm@spawn.scs.uiuc.edu Supporting information for this article is available on the WWW under http://dx.doi.org/10.1002/cphc.200800649.