Closing the loop: autonomous experiments enabled by machine-learning-based online data analysis in synchrotron beamline environments.

Recently, there has been significant interest in applying machine-learning (ML) techniques to the automated analysis of X-ray scattering experiments, due to the increasing speed and size at which datasets are generated. ML-based analysis presents an important opportunity to establish a closed-loop feedback system, enabling monitoring and real-time decision-making based on online data analysis. In this study, the incorporation of a combined one-dimensional convolutional neural network (CNN) and multilayer perceptron that is trained to extract physical thin-film parameters (thickness, density, roughness) and capable of taking into account prior knowledge is described. ML-based online analysis results are processed in a closed-loop workflow for X-ray reflectometry (XRR), using the growth of organic thin films as an example. Our focus lies on the beamline integration of ML-based online data analysis and closed-loop feedback. Our data demonstrate the accuracy and robustness of ML methods for analyzing XRR curves and Bragg reflections and its autonomous control over a vacuum deposition setup.

Artificial neural networks for quantitative online NMR spectroscopy.

Industry 4.0 is all about interconnectivity, sensor-enhanced process control, and data-driven systems. Process analytical technology (PAT) such as online nuclear magnetic resonance (NMR) spectroscopy is gaining in importance, as it increasingly contributes to automation and digitalization in production. In many cases up to now, however, a classical evaluation of process data and their transformation into knowledge is not possible or not economical due to the insufficiently large datasets available. When developing an automated method applicable in process control, sometimes only the basic data of a limited number of batch tests from typical product and process development campaigns are available. However, these datasets are not large enough for training machine-supported procedures. In this work, to overcome this limitation, a new procedure was developed, which allows physically motivated multiplication of the available reference data in order to obtain a sufficiently large dataset for training machine learning algorithms. The underlying example chemical synthesis was measured and analyzed with both application-relevant low-field NMR and high-field NMR spectroscopy as reference method. Artificial neural networks (ANNs) have the potential to infer valuable process information already from relatively limited input data. However, in order to predict the concentration at complex conditions (many reactants and wide concentration ranges), larger ANNs and, therefore, a larger training dataset are required. We demonstrate that a moderately complex problem with four reactants can be addressed using ANNs in combination with the presented PAT method (low-field NMR) and with the proposed approach to generate meaningful training data. Graphical abstract.

Fiber Optic Train Monitoring with Distributed Acoustic Sensing: Conventional and Neural Network Data Analysis.

Distributed acoustic sensing (DAS) over tens of kilometers of fiber optic cables is well-suited for monitoring extended railway infrastructures. As DAS produces large, noisy datasets, it is important to optimize algorithms for precise tracking of train position, speed, and the number of train cars. The purpose of this study is to compare different data analysis strategies and the resulting parameter uncertainties. We present data of an ICE 4 train of the Deutsche Bahn AG, which was recorded with a commercial DAS system. We localize the train signal in the data either along the temporal or spatial direction, and a similar velocity standard deviation of less than 5 km/h for a train moving at 160 km/h is found for both analysis methods. The data can be further enhanced by peak finding as well as faster and more flexible neural network algorithms. Then, individual noise peaks due to bogie clusters become visible and individual train cars can be counted. From the time between bogie signals, the velocity can also be determined with a lower standard deviation of 0.8 km/h. The analysis methods presented here will help to establish routines for near real-time train tracking and train integrity analysis.

Real-time dynamic strain sensing in optical fibers using artificial neural networks.

We propose to use artificial neural networks (ANNs) for raw measurement data interpolation and signal shift computation and to demonstrate advantages for wavelength-scanning coherent optical time domain reflectometry (WS-COTDR) and dynamic strain distribution measurement along optical fibers. The ANNs are trained with synthetic data to predict signal shifts from wavelength scans. Domain adaptation to measurement data is achieved, and standard correlation algorithms are outperformed. First and foremost, the ANN reduces the data analysis time by more than two orders of magnitude, making it possible for the first time to predict strain in real-time applications using the WS-COTDR approach. Further, strain noise and linearity of the sensor response are improved, resulting in more accurate measurements. ANNs also perform better for low signal-to-noise measurement data, for a reduced length of correlation input (i.e., extended distance range), and for coarser sampling settings (i.e., extended strain scanning range). The general applicability is demonstrated for distributed measurement of ground movement along a dark fiber in a telecom cable. The presented ANN-based techniques can be employed to improve the performance of a wide range of correlation or interpolation problems in fiber sensing data analysis and beyond.

Molecular structure of the substrate-induced thin-film phase of tetracene.

We present a combined experimental and theoretical study to solve the unit-cell and molecular arrangement of the tetracene thin film (TF) phase. TF phases, also known as substrate induced phases (SIPs), are polymorphs that exist at interfaces and decisively impact the functionality of organic thin films, e.g., in a transistor channel, but also change the optical spectra due to the different molecular packing. As SIPs only exist in textured ultrathin films, their structure determination remains challenging compared to bulk materials. Here, we use grazing incidence X-ray diffraction and atomistic simulations to extract the TF unit-cell parameters of tetracene together with the atomic positions within the unit-cell.

Diffusion and nucleation in multilayer growth of PTCDI-C8 studied with in situ X-ray growth oscillations and real-time small angle X-ray scattering.

We study nucleation and multilayer growth of the perylene derivative PTCDI-C8 and find a persistent layer-by-layer growth, transformation of island shapes, and an enhancement of molecular diffusivity in upper monolayers (MLs). These findings result from the evaluation of the ML-dependent island densities, obtained by in situ real-time grazing incidence small angle X-ray scattering measurements and simultaneous X-ray growth oscillations. Complementary ex situ atomic force microscopy snapshots of different growth stages agree quantitatively with both X-ray techniques. The rate and temperature-dependent island density is analyzed using different mean-field nucleation models. Both a diffusion limited aggregation and an attachment limited aggregation model yield in the first two MLs the same critical nucleus size i, similar surface diffusion attempt frequencies in the 1019-1020 s-1 range, and a decrease of the diffusion barrier Ed in the 2nd ML by 140 meV.

Polymorphism in α-sexithiophene crystals: relative stability and transition path.

We present a joint theoretical and experimental study to investigate polymorphism in α-sexithiophene (6T) crystals. By means of density-functional theory calculations, we clarify that the low-temperature phase is favorable over the high-temperature one, with higher relative stability up to 50 meV per molecule. This result is in agreement with our thermal desorption measurements. We also propose a transition path between the high- and low-temperature 6T polymorphs, estimating an upper bound for the energy barrier of about 1 eV per molecule. The analysis of the electronic properties of the investigated 6T crystal structures complements our study.

Miniaturized Bragg-grating couplers for SiN-photonic crystal slabs.

We report on an experimental and theoretical investigation of an integrated Bragg-like grating coupler for near-vertical scattering of light from photonic crystal waveguides with an ultra-small footprint of a few lattice constants only. Using frequency-resolved measurements, we find the directional properties of the scattered radiation and prove that the coupler shows a good performance over the complete photonic bandgap. The results compare well to analytical considerations regarding 1d-scattering phenomena as well as to FDTD simulations.

Thermally driven smoothening of molecular thin films: Structural transitions in n-alkane layers studied in real-time.

We use thermal annealing to improve smoothness and to increase the lateral size of crystalline islands of n-tetratetracontane (TTC, C44H90) films. With in situ x-ray diffraction, we find an optimum temperature range leading to improved texture and crystallinity while avoiding an irreversible phase transition that reduces crystallinity again. We employ real-time optical phase contrast microscopy with sub-nm height resolution to track the diffusion of TTC across monomolecular step edges which causes the unusual smoothing of a molecular thin film during annealing. We show that the lateral island sizes increase by more than one order of magnitude from 0.5 µm to 10 µm. This desirable behavior of 2d-Ostwald ripening and smoothing is in contrast to many other organic molecular films where annealing leads to dewetting, roughening, and a pronounced 3d morphology. We rationalize the smoothing behavior with the highly anisotropic attachment energies and low surface energies for TTC. The results are technically relevant for the use of TTC as passivation layer and as gate dielectric in organic field effect transistors.

Millisecond X-ray reflectometry and neural network analysis: unveiling fast processes in spin coating.

X-ray reflectometry (XRR) is a powerful tool for probing the structural characteristics of nanoscale films and layered structures, which is an important field of nanotechnology and is often used in semiconductor and optics manufacturing. This study introduces a novel approach for conducting quantitative high-resolution millisecond monochromatic XRR measurements. This is an order of magnitude faster than in previously published work. Quick XRR (qXRR) enables real time and in situ monitoring of nanoscale processes such as thin film formation during spin coating. A record qXRR acquisition time of 1.4 ms is demonstrated for a static gold thin film on a silicon sample. As a second example of this novel approach, dynamic in situ measurements are performed during PMMA spin coating onto silicon wafers and fast fitting of XRR curves using machine learning is demonstrated. This investigation primarily focuses on the evolution of film structure and surface morphology, resolving for the first time with qXRR the initial film thinning via mass transport and also shedding light on later thinning via solvent evaporation. This innovative millisecond qXRR technique is of significance for in situ studies of thin film deposition. It addresses the challenge of following intrinsically fast processes, such as thin film growth of high deposition rate or spin coating. Beyond thin film growth processes, millisecond XRR has implications for resolving fast structural changes such as photostriction or diffusion processes.

Neural network analysis of neutron and X-ray reflectivity data incorporating prior knowledge.

Due to the ambiguity related to the lack of phase information, determining the physical parameters of multilayer thin films from measured neutron and X-ray reflectivity curves is, on a fundamental level, an underdetermined inverse problem. This ambiguity poses limitations on standard neural networks, constraining the range and number of considered parameters in previous machine learning solutions. To overcome this challenge, a novel training procedure has been designed which incorporates dynamic prior boundaries for each physical parameter as additional inputs to the neural network. In this manner, the neural network can be trained simultaneously on all well-posed subintervals of a larger parameter space in which the inverse problem is underdetermined. During inference, users can flexibly input their own prior knowledge about the physical system to constrain the neural network prediction to distinct target subintervals in the parameter space. The effectiveness of the method is demonstrated in various scenarios, including multilayer structures with a box model parameterization and a physics-inspired special parameterization of the scattering length density profile for a multilayer structure. In contrast to previous methods, this approach scales favourably when increasing the complexity of the inverse problem, working properly even for a five-layer multilayer model and a periodic multilayer model with up to 17 open parameters.

Formation of carboxy- and amide-terminated alkyl monolayers on Silicon(111) investigated by ATR-FTIR, XPS, and X-ray scattering: construction of photoswitchable surfaces.

We have prepared high-quality, densely packed, self-assembled monolayers (SAMs) of carboxy-terminated alkyl chains on Si(111). The samples were made by thermal grafting of methyl undec-10-enoate under an inert atmosphere and subsequent cleavage of the ester functionality to disclose the carboxylic acid end-group. X-ray photoelectron spectroscopy (XPS) and grazing incidence X-ray diffraction (GIXD) indicate a surface coverage of about 50% of the initially H-terminated sites. In agreement, GIXD implies a rectangular unit mesh of 6.65 and 7.68 Å side lengths, containing two molecules in a regular zigzag-like substitution pattern for the ester- and carboxy-terminated monolayer. Hydrolysis of the remaining H-Si(111) bonds at the surface furnished HO-Si(111) groups according to XPS and attenuated total reflection Fourier-transform infrared spectroscopy (ATR-FTIR) studies. The amide-terminated alkyl SAM on Si(111) assembled in a 2-(6-chloro-1H-benzotriazol-1-yl)-1,1,3,3-tetramethyluronium hexafluorophosphate (HCTU)-mediated one-pot coupling reaction under an inert atmosphere, whereby the active ester forms in situ prior to the reaction with an amino-functionalized photoswitchable fulgimide. ATR-FTIR and XPS studies of the fulgimide samples revealed closely covered amide-terminated SAMs. Reversible photoswitching of the headgroup was read out by applying XPS, ATR-FTIR, and difference absorption spectra in the mid-IR. In XPS, we observed a reversible breathing of the amide/imide C1s and N1s signals of the fulgimide. The results demonstrate the general suitability of HCTU as a reagent for amide couplings to carboxy-terminated alkyl SAMs and the on-chip functionalization toward photoswitchable Si(111) surfaces.

Chirality-Specific Unidirectional Rotation of Molecular Motors on Cu(111).

Molecular motors have chemical properties that enable unidirectional motion, thus breaking microscopic reversibility. They are well studied in solution, but much less is known regarding their behavior on solid surfaces. Here, single motor molecules adsorbed on a Cu(111) surface are excited by voltages pulses from an STM tip, which leads to their rotation around a fixed pivot point. Comparison with calculations shows that this axis results from a chemical bond of a sulfur atom in the chemical structure and a metal atom of the surface. While statistics show approximately equal rotations in both directions, clockwise and anticlockwise, a detailed study reveals that these motions are enantiomer-specific. Hence, the rotation direction of each individual molecule depends on its chirality, which can be determined from STM images. At first glance, these dynamics could be assigned to the activation of the motor molecule, but our results show that this is unlikely as the molecule remains in the same conformation after rotation. Additionally, a control molecule, although it lacks unidirectional rotation in solution, also shows unidirectional rotation for each enantiomer. Hence, it seems that the unidirectional rotation is not specifically related to the motor property of the molecule. The calculated energy barriers for motion show that the propeller-like motor activity requires higher energy than the simple rotation of the molecule as a rigid object, which is therefore preferred.

Faster and lower-dose X-ray reflectivity measurements enabled by physics-informed modeling and artificial intelligence co-refinement.

An approach is presented for analysis of real-time X-ray reflectivity (XRR) process data not just as a function of the magnitude of the reciprocal-space vector q, as is commonly done, but as a function of both q and time. The real-space structures extracted from the XRR curves are restricted to be solutions of a physics-informed growth model and use state-of-the-art convolutional neural networks (CNNs) and differential evolution fitting to co-refine multiple time-dependent XRR curves R(q, t) of a thin film growth experiment. Thereby it becomes possible to correctly analyze XRR data with a fidelity corresponding to standard fits of individual XRR curves, even if they are sparsely sampled, with a sevenfold reduction of XRR data points, or if the data are noisy due to a 200-fold reduction in counting times. The approach of using a CNN analysis and of including prior information through a kinetic model is not limited to growth studies but can be easily extended to other kinetic X-ray or neutron reflectivity data to enable faster measurements with less beam damage.

Band-offset engineering in organic/inorganic semiconductor hybrid structures.

Control over the electronic structure of organic/inorganic semiconductor interfaces is required to realize hybrid structures with tailored opto-electronic properties. An approach towards this goal is demonstrated for a layered hybrid system composed of p-sexiphenyl (6P) and ZnO. The molecular orientation can be switched from "upright-standing" to "flat-lying" by tuning the molecule-substrate interactions through aggregation on different crystal faces. The morphology change has profound consequences on the offsets between the molecular frontier energy levels and the semiconductor band edges. The combination of ZnO surface dipole modification through molecule adsorption and the orientation-dependence of the ionization energy of molecular layers shift these offsets by 0.7 eV. The implications for optimizing hybrid structures with regard to exciton and charge transfer are discussed.

Fast fitting of reflectivity data of growing thin films using neural networks.

X-ray reflectivity (XRR) is a powerful and popular scattering technique that can give valuable insight into the growth behavior of thin films. This study shows how a simple artificial neural network model can be used to determine the thickness, roughness and density of thin films of different organic semiconductors [diindenoperylene, copper(II) phthalocyanine and α-sexithiophene] on silica from their XRR data with millisecond computation time and with minimal user input or a priori knowledge. For a large experimental data set of 372 XRR curves, it is shown that a simple fully connected model can provide good results with a mean absolute percentage error of 8-18% when compared with the results obtained by a genetic least mean squares fit using the classical Parratt formalism. Furthermore, current drawbacks and prospects for improvement are discussed.

Thin film growth studies using time-resolved x-ray scattering.

Thin-film growth is important for novel functional materials and new generations of devices. The non-equilibrium growth physics involved is very challenging, because the energy landscape for atomic scale processes is determined by many parameters, such as the diffusion and Ehrlich-Schwoebel barriers. We review the in situ real-time techniques of x-ray diffraction (XRD), x-ray growth oscillations and diffuse x-ray scattering (GISAXS) for the determination of structure and morphology on length scales from Å to µm. We give examples of time resolved growth experiments mainly from molecular thin film growth, but also highlight growth of inorganic materials using molecular beam epitaxy (MBE) and electrochemical deposition from liquids. We discuss how scaling parameters of rate equation models and fundamental energy barriers in kinetic Monte Carlo methods can be determined from fits of the real-time x-ray data.

Direct Photoalignment and Optical Patterning of Molecular Thin Films.

A novel strategy for direct photoalignment of molecular materials using optothermal re-orientation is introduced. Photoalignment for molecular materials such as the organic semiconductor tetracene is shown, without relying on additional photoreactive dopants or alignment layers. Patterning and polarized light emission, e.g., for polarized organic light emitting diodes is demonstrated.

Cooperative Switching in Nanofibers of Azobenzene Oligomers.

Next-generation molecular devices and machines demand the integration of molecular switches into hierarchical assemblies to amplify the response of the system from the molecular level to the meso- or macro-scale. Here, we demonstrate that multi-azobenzene oligomers can assemble to form robust supramolecular nanofibers in which they can be switched repeatedly between the E- and Z-configuration. While in isolated oligomers the azobenzene units undergo reversible photoisomerization independently, in the nanofibers they are coupled via intermolecular interactions and switch cooperatively as evidenced by unusual thermal and kinetic behavior. We find that the photoisomerization rate from the Z-isomer to the E-isomer depends on the fraction of Z-azobenzene in the nanofibers, and is increased by more than a factor of 4 in Z-rich fibers when compared to E-rich fibers. This demonstrates the great potential of coupling individual photochromic units for increasing their quantum efficiency in the solid state with potential relevance for actuation and sensing.

Lattice matching as the determining factor for molecular tilt and multilayer growth mode of the nanographene hexa-peri-hexabenzocoronene.

The microstructure, morphology, and growth dynamics of hexa-peri-hexabenzocoronene (HBC, C42H18) thin films deposited on inert substrates of similar surface energies are studied with particular emphasis on the influence of substrate symmetry and substrate-molecule lattice matching on the resulting films of this material. By combining atomic force microscopy (AFM) with X-ray diffraction (XRD), X-ray absorption spectroscopy (NEXAFS), and in situ X-ray reflectivity (XRR) measurements, it is shown that HBC forms polycrystalline films on SiO2, where molecules are oriented in an upright fashion and adopt the known bulk structure. Remarkably, HBC films deposited on highly oriented pyrolytic graphite (HOPG) exhibit a new, substrate-induced polymorph, where all molecules adopt a recumbent orientation with planar π-stacking. Formation of this new phase, however, depends critically on the coherence of the underlying graphite lattice since HBC grown on defective HOPG reveals the same orientation and phase as on SiO2. These results therefore demonstrate that the resulting film structure and morphology are not solely governed by the adsorption energy but also by the presence or absence of symmetry- and lattice-matching between the substrate and admolecules. Moreover, it highlights that weakly interacting substrates of high quality and coherence can be useful to induce new polymorphs with distinctly different molecular arrangements than the bulk structure.