| The discovery and development of novel materials in the field of energy are essential to accelerate the transition to a low-carbon economy. Bringing recent technological innovations in automation, robotics and computer science together with current approaches in chemistry , materials synthesis and characterization will act as a catalyst for revolutionizing traditional research and development in both industry and academia. This Perspective provides a vision for an integrated artificial intelligence approach towards autonomous materials discovery , which, in our opinion, will emerge within the next 5 to 10 years. The approach we discuss requires the integration of the following tools, which have already seen substantial development to date: high-throughput virtual screening, automated synthesis planning, automated laboratories and machine learning algorithms. In addition to reducing the time to deployment of new materials by an order of magnitude, this integrated approach is expected to lower the cost associated with the initial discovery. Thus, the price of the final products (for example, solar panels, batteries and electric vehicles) will also decrease. This in turn will enable industries and governments to meet more ambitious targets in terms of reducing greenhouse gas emissions at a faster pace. volume 3 | mAY 2018 | 5 PERSPECTIVES

The observed prevalence of oscillatory signals in the spectroscopy of biological light-harvesting complexes at ambient temperatures has led to a search for mechanisms supporting coherent transport through larger molecules in noisy environments. We demonstrate a generic mechanism supporting long-lasting electronic coherence up to 0.3 ps at a temperature of 277 K. The mechanism relies on two properties of the spectral density: (i) a large dissipative coupling to a continuum of higherfrequency vibrations required for efficient transport and (ii) a small slope of the spectral density at zero frequency.PACS numbers: 03.65. Yz, 78.47.nj, 05.60.Gg Long-lasting oscillatory signals in optically excited light-harvesting complexes (LHCs) point towards the prevalence of a coherent electronic dynamics in molecular networks at physiological temperatures [2][3][4][5]. The experiments probe the electronic dynamics using twodimensional (2d) echo-spectroscopy for a series of delaytimes between the excitation pulse and the probe pulse. The 2d spectroscopy makes studies of the dynamics of dissipative systems possible and has found further applications in mesoscopic systems such as molecular nanotubes [6] and semiconductor devices [7]. Due to its known crystallographic structure and relative simplicity, the Fenna-Matthews-Olson (FMO) complex serves as the prototype system for studying the choreography of the energy transfer from the antenna to the reaction center of a light harvesting complex [8]. In 2d spectra long-lasting beatings are observed, ranging from 1.2 ps at T = 150 K to 0.3 ps at T = 277 K [9]. The interplay of coherent dynamics, which leads to a delocalization of an initial excitation arriving at the FMO network from the antenna, and the coupling to a vibronic environment with slow and fast fluctuations, has lead to studies of environmentally assisted transport in LHCs [10,11].An important open question is whether coherence plays a key-role in the functioning of light-harvesting complexes [8]. The theoretical understanding of the experiments is in its early stages and atomistic simulations based on molecular dynamics have not reached agreement [12,13]. A calculation of 2d echo-spectra based on the molecular dynamics simulation [14] does not show clear coherent oscillations at T = 277 K. One key ingredient for efficient transfer dynamics is the strong coupling to vibronic modes, which induces energy dissipation [10,11,15]. For the FMO complex the thermalization occurs within picoseconds and was observed by Brixner et al. by the decay of diagonal-peak amplitudes to lower energies [16]. It has been proposed that the inclusion of the finite time scale of the reorganization process gives rise to long-lasting coherence in LHCs [17]. While a sluggish bath relaxation leads to prolonged population beatings in the FMO network, calculations of 2d echo-spectra show oscillations of the exciton cross-peaks for only about 1/6th of the experimentally recorded time at T = 150 K [18]. For an even longer bath-relaxation cross-peak osci...

Excitonic models of light-harvesting complexes, where the vibrational degrees of freedom are treated as a bath, are commonly used to describe the motion of the electronic excitation through a molecule. Recent experiments point toward the possibility of memory effects in this process and require one to consider time nonlocal propagation techniques. The hierarchical equations of motion (HEOM) were proposed by Ishizaki and Fleming to describe the site-dependent reorganization dynamics of protein environments ( J. Chem. Phys. 2009 , 130 , 234111 ), which plays a significant role in photosynthetic electronic energy transfer. HEOM are often used as a reference for other approximate methods but have been implemented only for small systems due to their adverse computational scaling with the system size. Here, we show that HEOM are also solvable for larger systems, since the underlying algorithm is ideally suited for the usage of graphics processing units (GPU). The tremendous reduction in computational time due to the GPU allows us to perform a systematic study of the energy-transfer efficiency in the Fenna-Matthews-Olson (FMO) light-harvesting complex at physiological temperature under full consideration of memory effects. We find that approximative methods differ qualitatively and quantitatively from the HEOM results and discuss the importance of finite temperature to achieving high energy-transfer efficiencies.

We report Phoenics, a probabilistic global optimization algorithm identifying the set of conditions of an experimental or computational procedure which satisfies desired targets. Phoenics combines ideas from Bayesian optimization with concepts from Bayesian kernel density estimation. As such, Phoenics allows to tackle typical optimization problems in chemistry for which objective evaluations are limited, due to either budgeted resources or time-consuming evaluations of the conditions, including experimentation or enduring computations. Phoenics proposes new conditions based on all previous observations, avoiding, thus, redundant evaluations to locate the optimal conditions. It enables an efficient parallel search based on intuitive sampling strategies implicitly biasing toward exploration or exploitation of the search space. Our benchmarks indicate that Phoenics is less sensitive to the response surface than already established optimization algorithms. We showcase the applicability of Phoenics on the Oregonator, a complex case-study describing a nonlinear chemical reaction network. Despite the large search space, Phoenics quickly identifies the conditions which yield the desired target dynamic behavior. Overall, we recommend Phoenics for rapid optimization of unknown expensive-to-evaluate objective functions, such as experimentation or long-lasting computations.

Recent experimental observations of time-dependent beatings in the two-dimensional echo-spectra of light-harvesting complexes at ambient temperatures have opened up the question of whether coherence and wave-like behaviour play a significant role in photosynthesis. We carry out a numerical study of the absorption and echo-spectra of the Fenna-Matthews-Olson (FMO) complex in Chlorobium tepidum and analyse the requirements in the theoretical model needed to reproduce beatings in the calculated spectra. The energy transfer in the FMO pigment-protein complex is theoretically described by an exciton Hamiltonian coupled to a phonon bath which accounts for the pigments' electronic and vibrational excitations, respectively. We use the hierarchical equations of motions method to treat the strong couplings in a non-perturbative way. We show that the oscillations in the two-dimensional echo-spectra persist in the presence of thermal noise and static disorder. 3 different information from the population dynamics and needs to be calculated and analysed separately in detail. In principle, techniques such as quantum state and process tomography made out of a sequence of 2D echo-spectra can be used to map out the complete density matrix [13]. In contrast to energy-transfer efficiency studies where an initial excitation enters the complex at specific sites close to the antenna, in 2D echo-spectra the whole complex is simultaneously excited. Also the two-exciton manifold yields prominent contributions to the signal, resulting in negative regions in the 2D echo-spectra.The non-pertubative calculation of 2D echo-spectra presents a considerable computational challenge owing to the presence of two excitons giving rise to excited state absorption and the requirement to consider ensemble averages over differently orientated complexes with slightly varying energy levels. Previous calculations have used Markovian approximations [14,15] or exclude the double-exciton manifold [16]. In addition, the systematic study of beatings in a series of 2D echo-spectra requires an effective means of calculating a huge number of such spectra. So far, no theoretical method has been able to describe the long-lasting beatings in the time-resolved 2D spectra [14,16,17]. One possible explanation for the persistence of long coherence times has been the sluggish absorption of the reorganization energy by the molecule, which requires theoretical descriptions that go beyond the Markovian approximation and the rotating wave approximation [18]. The hierarchical equations of motions (HEOM), first developed by Tanimura and Kubo [19] and subsequently refined in [20][21][22][23], show oscillations in the dynamics of the exciton populations that persist even at temperature T = 300 K [24,25]. The HEOM include the reorganization process in a transparent way and are directly applicable to computations at physiological temperatures. A calculation at temperature 77 K of the 2D echospectra with the HEOM method has recently been performed by Chen et al [17], which does not di...