The transition towards a society based on 100% renewable energy requires massive deployment of photovoltaics of 30-70 TW until 2050. This requires huge amounts of resources, while their limited availability is already becoming apparent. A major lever to reduce resource consumption is to increase the solar cell efficiency. As best single junction solar cells approach their fundamental limits, higher efficiency can only be reached with so-called tandem solar cells, made of two or more subcells. All tandem technologies so far are based on relatively thick absorber layers, reducing resource demand compared to single junction devices by efficiency increase. There, light trapping strategies are used to maximize absorptance close to the band gap of the materials and improve efficiency by few percent relative. However, by applying advanced light trapping techniques such as nanophotonic metasurfac-es, ultrathin single junction devices with a 5-10-fold decrease in semiconductor material were realized. To reduce resource demand further, the concept of ultrathin solar cells must be extended to tandem devices. This introduces severe challenges, as not only absorption needs to be maximized within the active part, but a spectrally dependent light guiding strategy is required. Metasurfaces have shown the ability to manipulate light e.g. spectrally dependent; however, they have never been implemented into tandem solar cells. Thus, the overarching goal of PHASE is to generate a deep physical understanding of metasurfaces for ultrathin tandem solar cells and to develop process flows to implement nanopho-tonic structures into such devices with efficiencies above 30%. This will proof that the chosen tech-nology pathway can support the urgently needed energy transition. More specific, the goal of PHASE is to realize tandem solar cells, where the resource demanding semiconductor part is 10 times thinner (and thus needs 10 times less semiconductor material) than similar existing devices.

The abusive use of antibiotics has led to multidrug-resistant bacteria and the acute threat of a post-antibiotic era. However, apart from resisters, there is a subgroup of bacteria called persisters that surviveby recalcitrance to antibiotic treatment. Persisters are not resistant to antibiotics but simply survive by metabolic shutdown. Upon withdrawal of antibiotics, these persisters resuscitate and regenerate the colony. They are heavily involved in failure of antibiotic treatment and the development of chronic infections. Bacterial persistence is controlled by the stringent response, which itself is mediated by hyperphosphorylated nucleotides, known as the magic spot (MS) nucleotides or (p)ppGpp. The importance of the stringent response, its omnipresence in the domain of bacteria, its connection to persister formation and tolerance to (antibiotic) stress, and its absence in mammals has led to significant research in microbiology. However, until recently these activities have not been paralleled by the development of chemical biology approaches. The current proposal aims to fill this gap by research into (1) synthetic methodology targeting the magic spot nucleotides and their analogs, (2) tools to identify target proteins of (p)ppGpp, and more generally (p)ppNpp (3) analytical approaches to extract, resolve, and quantify (p)ppGpp, (4) strategies to control the stringent response and persister formation with light (5) inhibitors of the stringent response. These new tools will enable a detailed understanding of the stringent response and thus ultimately help in the design of new antibiotics effective against persisters. The goal is to develop methods to force bacteria into the persistent state or inversely wake them up by using light and small molecules. Forcing bacteria out of persistence and blocking their entry into this state in combination with antibiotic treatment is a highly promising strategy to avoid the development of chronic bacterial infections.

Why are contemporary readers so fascinated by retellings of the Iliad or Beowulf? And why have retellings of premodern – ancient and medieval – texts not been taken seriously as a literary practice to date? In DERIVATE, I investigate the striking surge of retellings in contemporary English literature by setting the contemporary texts in a productive dialogue with practices of writing in the Middle Ages. The medieval period offers a perfect point of departure for theorizing retellings as medieval literature was inherently derivative and medieval authors had developed a system for being inventive within a system of derivations. Taking issue with the privileging of that which is new in literary history, I propose a new paradigm for literary history: literary history as a history of derivations. Starting from the premise that retelling is a transhistorical concept, the project sheds new light on the processes of reception that find their expression in the current interest in and relevance of premodern material. The project triangulates (classical) reception studies, medieval literary studies as well as literary theory, especially postmodernist theory, and scrutinizes the practice of retelling premodern (ancient and medieval) texts in contemporary English literature. DERIVATE thus develops a theory of retelling based on the intense engagement and critical comparison with medieval practices of retelling in order to map the wider cultural, historical, and literary contexts and implications of the current trend in retellings of classical and medieval texts (audiences, canon, literary market) as a springboard for developing a literary history of derivations.

Describing the set of rational solutions to polynomial equations is the oldest field in mathematics and one of the fundamental goals in number theory. In arithmetic geometry such solutions are studied using geometric tools: rational solutions to equations correspond to rational points on the corresponding geometric object. Fano varieties are among the simplest geometric objects, but still far from fully understood. This makes them a great class to test conjectures and develop new techniques. It is generally believed that Fano varieties, if they have a rational point, should have many, and they should be well-distributed. For curves (dimension 1), this is the case. Fano varieties in dimension 2 are del Pezzo surfaces, and already here there are many open questions. These surfaces have been a very active area of research in the last 50 years. Del Pezzo surfaces are classified by their degree, an integer between 1 and 9. The lower the degree, the more complex these surfaces become, and especially del Pezzo surfaces of degree 1 are notoriously difficult. Current results on the rational points on these surfaces make use of ad-hoc constructions, and a general geometric approach is missing. This forms a sharp contrast with del Pezzo surfaces of higher degree, and leaves a big gap in the understanding of rational points on Fano varieties. This project proposes to create a systematic approach to construct rational points on del Pezzo surfaces of degree 1, and use this to prove several different results on ‘abundant’ rational points for new families of surfaces. This will lead to answering big open questions (unirationality, Hilbert property and weak weak approximation for del Pezzo surfaces of degree 1), and providing evidence towards a long-standing conjecture on rational points on rationally connected varieties. Recent developments on the construction of rational points and low-genus curves on del Pezzo surfaces make this the perfect time to tackle the proposed objectives.

Deep neural networks (DNNs) have led to dramatic improvements of the state-of-the-art for many important classification problems, such as object recognition from images or speech recognition from audio data. However, DNNs are also notoriously dependent on the tuning of their hyperparameters. Since their manual tuning is time-consuming and requires expert knowledge, recent years have seen the rise of Bayesian optimization methods for automating this task. While these methods have had substantial successes, their treatment of DNN performance as a black box poses fundamental limitations, allowing manual tuning to be more effective for large and computationally expensive data sets: humans can (1) exploit prior knowledge and extrapolate performance from data subsets, (2) monitor the DNN's internal weight optimization by stochastic gradient descent over time, and (3) reactively change hyperparameters at runtime. We therefore propose to model DNN performance beyond a blackbox level and to use these models to develop for the first time: 1. Next-generation Bayesian optimization methods that exploit data-driven priors to optimize performance orders of magnitude faster than currently possible; 2. Graybox Bayesian optimization methods that have access to -- and exploit -- performance and state information of algorithm runs over time; and 3. Hyperparameter control strategies that learn across different datasets to adapt hyperparameters reactively to the characteristics of any given situation. DNNs play into our project in two ways. First, in all our methods we will use (Bayesian) DNNs to model and exploit the large amounts of performance data we will collect on various datasets. Second, our application goal is to optimize and control DNN hyperparameters far better than human experts and to obtain: 4. Computationally inexpensive auto-tuned deep neural networks, even for large datasets, enabling the widespread use of deep learning by non-experts.