The single-crystal-to-single-crystal (scsc) synthesis of a 2D polymer based on photochemically triggered [2 + 2]-cycloaddition is reported. Both monomer and polymer single crystals are analyzed by X-ray diffraction, which is the first case of a scsc two-dimensional polymerization based on this cycloaddition and the third ever case for a scsc synthesis of a 2D polymer. The product crystals at quantitative conversion are wet-exfoliated under mild conditions and afford countless features that are single and double layers as judged by their AFM heights of h ≈ 1.2 ± 0.5 and 2.2 ± 0.5 nm, respectively. The X-ray-structure-based molecular weight of the 2D polymers and their degree of polymerization per μm are M = 360 MDa and P = 464 900, respectively. The sheet size is on the order of 5 × 5 μm.

Two-dimensional single-crystal-to-single-crystal polymerization and depolymerization are described in detail. The results are based on in-house and synchrotron X-ray diffraction experiments conducted on several samples at 100 K and room temperature. The reactions are associated with considerable molecular motions of all components (monomer, template and incorporated solvent molecules), which can be as large as 1 Å . Continuous polymerization leads to a gradual gap opening between the emerging two-dimensional polymer layers, which allows for increased mobility of the solvent molecules. The positional flexibility of both the solvents and the weakly bound templates buffers the local strain induced by polymerization through a complex chain of movements. As a consequence, the accumulated global strain remains small enough to essentially preserve the single-crystalline state in the course of a complete polymerization/depolymerization cycle. The unit-cell parameters evolve in an unusual way. The a and c axes of the trigonal lattice slightly increase during polymerization, even though van der Waals interactions are replaced by shorter covalent bonds and the involved molecules shrink. However, the c axis experiences a significant drop of more than 1 Å during the first depolymerization step. Progressive depolymerization expands the c axis again, but it does not quite reach the value of the fresh crystal. These effects can be explained by local strain formation and compensation mechanisms and by annealing effects during heat-induced depolymerization. An interesting side effect of the polymerization is the reorientation of incorporated solvent molecules, which give the crystal a tunable dipole moment. Of particular importance for the understanding of two-dimensional polymers is the evolution of the connectivity between molecules during polymerization and depolymerization. Combining reaction kinetics with structural information, such as the polymerization-induced displacement of reactive sites, allowed for the development of a propagation model, in which both polymerization and depolymerization proceed in a self-impeding fashion. This model is supported by Monte Carlo simulations. research papers J. Appl. Cryst. (2018). 51, 481-497 Gregor Hofer et al. Two-dimensional polymerization in the single crystal 483 research papers J. Appl. Cryst. (2018). 51, 481-497 Gregor Hofer et al. Two-dimensional polymerization in the single crystal 485 research papers J. Appl. Cryst. (2018). 51, 481-497 Gregor Hofer et al. Two-dimensional polymerization in the single crystal 495 Figure 14(a) Graphical representation of the average anthracene-anthracene distance between M S and M T . (b) Change of the average anthracene-anthracene distance between M S and M T upon polymerization.

We provide a critical analysis of what and what not to expect when elucidating the mechanism of 2D polymerization in crystals by single crystal X-ray diffraction. Analysing Bragg and diffuse scattering helps the understanding and provides guidelines for future developments.

2D polymer sheets with six positively charged pyrylium groups at each pore edge in a stacked single crystal can be transformed into a 2D polymer with six pyridines per pore by exposure to gaseous ammonia. This reaction furnishes still a crystalline material with tunable protonation degree at regular nano‐sized pores promising as separation membrane. The exfoliation is compared for both 2D polymers with the latter being superior. Its liquid phase exfoliation yields nanosheet dispersions, which can be size‐selected using centrifugation cascades. Monolayer contents of ≈30 % are achieved with ≈130 nm sized sheets in mg quantities, corresponding to tens of trillions of monolayers. Quantification of nanosheet sizes, layer number and mass shows that this exfoliation is comparable to graphite. Thus, we expect that recent advances in exfoliation of graphite or inorganic crystals (e.g. scale‐up, printing etc.) can be directly applied to this 2D polymer as well as to covalent organic frameworks.

Speech Emotion Recognition (SER) is an important and challenging task for human-computer interaction. In the literature deep learning architectures have been shown to yield state-ofthe-art performance on this task when the model is trained and evaluated on the same corpus. However, prior work has indicated that such systems often yield poor performance on unseen data. To improve the generalisation capabilities of emotion recognition systems one possible approach is cross-corpus training, which consists of training the model on an aggregation of different corpora. In this paper we present an analysis of the generalisation capability of deep learning models using crosscorpus training with six different speech emotion corpora. We evaluate the models on an unseen corpus and analyse the learned representations using the t-SNE algorithm, showing that architectures based on recurrent neural networks are prone to overfit the corpora present in the training set, while architectures based on convolutional neural networks (CNNs) show better generalisation capabilities. These findings indicate that (1) cross-corpus training is a promising approach for improving generalisation and (2) CNNs should be the architecture of choice for this approach.