Search

Despite recent breakthroughs in deep learning for materials informatics, there exists a disparity between their popularity in academic research and their limited adoption in the industry. A significant contributor to this “interpretability-adoption gap” is the prevalence of black-box models and the lack of built-in methods for model interpretation. While established methods for evaluating model performance exist, an intuitive understanding of the modeling and decision-making processes in models is nonetheless desired in many cases. In this work, we demonstrate several ways of incorporating model interpretability to the structure-agnostic Compositionally Restricted Attention-Based network, CrabNet. We show that CrabNet learns meaningful, material property-specific element representations based solely on the data with no additional supervision. These element representations can then be used to explore element identity, similarity, behavior, and interactions within different chemical environments. Chemical compounds can also be uniquely represented and examined to reveal clear structures and trends within the chemical space. Additionally, visualizations of the attention mechanism can be used in conjunction to further understand the modeling process, identify potential modeling or dataset errors, and hint at further chemical insights leading to a better understanding of the phenomena governing material properties. We feel confident that the interpretability methods introduced in this work for CrabNet will be of keen interest to materials informatics researchers as well as industrial practitioners alike.

Membrane-based separation technologies have the potential to lower the cost of post-combustion CO2 capture from power-plant flue gases through reduced energy and capital costs relative to conventional solvent approaches. Studies have shown promise under controlled conditions, but there is a need for data on performance and reliability under field conditions. Coal-fired power plants in China operate in a dynamic manner, with increases and decreases in output causing changes in flue-gas composition. In this paper, we describe the first field test of a membrane-based post-combustion CO2-capture system connected to a dynamically operating power plant in China. We report the performance of a Membrane Technology Research, Inc. (MTR) PolarisTM membrane-based capture system over a range of plant operating loads ranging from 54% to 84% and conducted an operational stability test over a 168-h period during which the power plant was operating at an average load of 55%, but ramped as high as 79% and as low as 55%. Our results confirm the ability of a membrane capture system to operate effectively over a wide range of host-plant operating conditions, but also identity several issues related to plant integration, system control and resilience in the face of host-plant upsets that require attention as membrane separation systems move towards commercial use.

We experimentally demonstrated V-groove-based self-aligned SiN edge coupler (EC) on a monolithic CMOS-SiPh platform. 520 mW power handling capability were achieved.

In rheumatoid arthritis, the use of Routine Assessment of Patient Index Data 3 (RAPID3) assessments to meet treat-to-target goals is endorsed by the 2021 American College of Rheumatology guidelines. In November 2020, the Baylor Scott & White specialty pharmacy implemented a new service that included more frequent collection of RAPID3 scores and standardized provider communication for patients co-managed by a Baylor Scott & White rheumatology clinic. The objective was to evaluate the impact of this new service on rheumatoid arthritis disease activity. Before the new service started, patients followed a protocol of RAPID3 assessments that occurred every 6 months; once the service began, patients were followed using an algorithm in which patients with higher disease activity were contacted more frequently. Eighty-six percent of patients in the pre-intervention group (n = 7) compared with 100% of patients in the post-intervention group (n = 10) had high to moderate disease activity at baseline. Within a 6-month follow-up period in both groups, the percentage of high to moderate disease activity patients decreased by 30% in the post-intervention group and remained the same in the pre-intervention group. These results support the positive impact increased specialty pharmacy services may have on clinical outcomes; therefore, the continued expansion of these services should be considered.

Critical Need for LDMS Techniques: Laser desorption mass spectrometry (LDMS) enables 2D and/or 3D chemical imaging of organic and inorganic analytes, supporting focused research objectives as well as discovery-based science. Many minerals and (aromatic) organics effectively absorb UV radiation, and photon energies of UV light approximate the first ionization energies of many elements in the Periodic Table. Consequently, UV laser sources, particularly those with controllable output attenuation, serve as effective ionization sources for many materials. In the realm of in situ planetary science, LDMS techniques are valued to detect refractory organic molecules, including prospective biosignatures, and identify the host phases that harbor said compounds. The short pulse widths offered by many solid-state laser systems, such as the 266 nm laser system developed for the Mars Organic Molecule Analyzer (MOMA) onboard the ExoMars rover, have been shown to circumvent the challenges associated with pyrolyzing organic-rich samples in the presence of strong oxidants, like the perchlorates found across the Martian surface [1]. Critical Need for Ultrahigh Mass Resolution: Heritage mass spectrometers, such as that flown on the Sample Analysis at Mars (SAM) investigation onboard the Curiosity rover [2], are limited to mass resolving powers of m/Δm < 1000 (FWHM), leading to uncertainty in the identification of molecular signals. Peaks are often assigned based on known isotope abundances, diagnostic fragmentation patterns, and/or corroborative measurements provided by other payload instruments. Alternatively, hardware additions such as gas chromatographs, resonance lasers sources, and/or collision cells can increase the confidence of molecular assignments, albeit at the cost of additional mass, volume, and power requirements. In contrast, the OrbitrapTM analyzer commercialized by Thermo Scientific [3], adapted for spaceflight by a consortium of French laboratories [4], and miniaturized and ruggedized through a collaboration with the University of Maryland and NASA GSFC [5], enables identification of molecular stoichiometry through mass resolving powers of m/Δm > 100,000 (FWHM) and ppm-level mass accuracies. Unrivaled disambiguation of isobaric interferences and isotopologues across a wide intrascan mass range (e.g., 20 – 600 u) distinguishes the Orbitrap from other high-resolution analyzers. LDMS with an Orbitrap mass analyzer: The analytical and scientific value of integrating a pulsed UV laser source with an Orbitrap mass analyzer for planetary applications was first recognized with the development of the Ion Laser Mass Analyzer (ILMA) for Marco Polo [6], a joint ESA-JAXA sample return mission targeting a primitive Near-Earth Object (NEO). In response to the Pre-Release of an Announcement of Opportunity for a NASA Europa Lander Mission, a laser-enabled Orbitrap was modified to analyze ice residues and seek out physicochemical signs of life in potentially habitable cryogenic environments [5]. More recently, a permutation of the instrument was adapted for operation on the lunar surface, facilitating investigations into the composition of the bulk silicate Moon, dynamics of the lunar interior, space weathering of the surface, rates of exogenous infall, and the alteration of organic materials due to exposure to cosmic rays [7]. Advanced analytical capabilities with a laser-enabled Orbitrap mass spectrometer: To support the evolving objectives of the planetary community, such as the prioritization of an Enceladus Orbilander in the most recent Decadal Survey [8], advanced LDMS techniques are currently being explored to maximize the science return for a spectrum of mission architectures. The characterization of complex organic materials simulating extraterrestrial matter (e.g., Titan-like tholins), and planetary analog samples doped with organic compounds commonly associated with living systems, such as proteinogenic amino acids and the nucleobase uracil found in RNA, demonstrates the detection of prospective biomarkers, classification of host mineralogy, and establishment of geological context/provenance [9-11]. In order to improve spectral reproducibility and enhance limits of detection, a variety of sample plate materials spanning a range of thermal diffusivities, electrical conductivities, and ionization potentials are being tested. Inorganic chemical matrices, such as Si nanoparticles, are being explored to promote ionization efficiency and the preservation of the molecular ion of a compelling suite of organic macromolecules, including short chain peptides [12] and simple, compound, and derived lipids [13]. References: [1] Li et al. (2015) Astrobio 15 (2), 104 – 110. [2] Mahaffy et al. (2012) PSS 170, 401 – 478. [3] Makarov (1999) US Patent 5,886,346. [4] Briois et al. (2016) PSS 131, 33 – 45. [5] Willhite et al. (2021) IEEE Aerospace, 1 – 13. [6] Cottin et al. (2009) 9th Euro Workshop on Astrobio. [7] Willhite et al. (2020) LSSW, LPI Contrib. No. 2241. [8] National Academies (2022) Origins, Worlds and Life. [9] Arevalo Jr. et al. (2018) Rapid Comm 32, 1875 – 1886. [10] Selliez et al. (2019) PSS 170, 42 – 51. [11] Selliez et al. (2020) Rapid Comm 34, e8645. [12] Arevalo Jr. et al. (2022) AbSciCon, 208-03. [13] Hanna et al. (2022) AbSciCon, 130-011.

Age-related cardiac amyloidosis results from deposits of wild-type tranthyretin amyloid (ATTRwt) in cardiac tissue. ATTR may play a role in carpal tunnel syndrome (CTS) and in spinal stenosis (SS), indicating or presaging systemic amyloidosis. We investigated consecutive patients undergoing surgery for SS for ATTR deposition in the resected ligamentum flavum (LF) and concomitant risk of cardiac amyloidosis. Each surgical specimen (LF) was stained with Congo red, and if positive, the amyloid deposits were typed by mass spectrometry. Patients with positive specimens underwent standard of care evaluation with fat pad aspirates, serum and urine protein electrophoresis with immunofixation, free light-chain assay

Introduction: Mutation-targeting and immuno-oncology drugs are revolutionizing the treatment of advanced non-small cell lung cancer (NSCLC). Cost-effectiveness analyses (CEA) of these drugs have be...

Wild-type transthyretin (ATTRwt) amyloid deposits have been found in the ligamentum flavum of patients undergoing surgery for spinal stenosis. The relationship between ATTRwt and ligamentum flavum thickness is unclear. We used pre-operative magnetic resonance imaging (MRI) to analyze ligamentum flavum thickness in lumbar spinal stenosis patients with and without ATTRwt amyloid.We retrospectively identified 178 patients who underwent lumbar spine surgery. Ligamentum flavum thickness of 253 specimens was measured on T2-weighted axial MRI. Amyloid presence was confirmed through Congo red staining of specimens, and ATTRwt was confirmed using mass-spectrometry and gene sequencing.Twenty four of the 178 patients (13.5%) were found to have ATTRwt in the ligamentum flavum. Forty ATTRwt specimens and 213 non-ATTRwt specimens were measured. Mean ligamentum flavum thickness was 4.92 (±1.27) mm in the ATTRwt group and 4.00 (±1.21) mm in the non-ATTRwt group (p 0.01). The ligamentum flavum was thickest at L4-L5, with a thickness of 5.15 (±1.27) mm and 4.23 (±1.29) mm in the ATTRwt and non-ATTRwt group, respectively (p = 0.007). There was a significant difference in ligamentum flavum thickness between ATTRwt and non-ATTRwt case for both patients younger than 70 years (p = 0.016) and those older than 70 years (p = 0.004). ATTRwt patients had greater ligamentum flavum thickness by 0.83 mm (95% confidence interval (CI): 0.41-1.25 mm, p 0.001) when controlled for age and lumbar level.Patients with ATTRwt had thicker ligamentum flavum compared to patients without ATTRwt. Further studies are needed to investigate the pathophysiology of ATTRwt in ligamentum flavum thickening.

BACKGROUND CONTEXT Extension of long posterior cervical fusions into the thoracic spine has been recommended to decrease the risk of kyphosis and spondylolisthesis at the cervico-thoracic junction. In recent years, several studies have addressed this question with conflicting results. We hypothesized that extension of posterior cervical fusions into the upper thoracic spine improved clinical outcomes while decreasing kyphosis. PURPOSE To investigate if it is safe to stop at C7 spinal level during multilevel posterior cervical decompression and fusion surgery. STUDY DESIGN/SETTING Analysis of routinely collected data. PATIENT SAMPLE A total of 264 adult patients undergoing mutlilevel posterior cervical decompression and fusion. OUTCOME MEASURES Estimated blood loss, operative time, length of hospital stay, cervical lordosis, C2-C7 sagittal plumbline, T1 slope, visual analog scale, Oswestry Disability Index and revision rate METHODS Analysis of multicenter radiographic and clinical databases of patients who had undergone a three or more-level posterior cervical fusion for degenerative disease from January 2013 to May 2015, with at least two years of postop data. Patients were divided into three groups: Group I (fusion terminating at C6), Group II (fusion terminating at C7), Group III (fusion extending into the thoracic spine). All radiographic measurements were performed by an independent, experienced clinical researcher. RESULTS A total of 264 patient cases were reviewed and sorted into the three outlined groups, Group I (n=80), Group II (n=88) and Group III (n=96). Demographically, mean age, percentage of nonsmokers and anterior support were greater in Group III than in Groups I and II (p 0.05). Rate of revision was not clinically or statistically significantly different (p>0.05) between Group I (8.8%), Group II (7.95%) and Group III (9.4%). The majority of the revision surgeries occurred between two to five years postop. Groups I and II had greater number of subjacent degeneration/spondylolisthesis compared to Group III (3.8% vs 3.4% vs 1.2%). There were significant improvements in mean clinical outcomes (ie, visual analog scale [VAS] and Oswestry Disability Index [ODI]) at two years postoperative in the three groups, but there were no statistically significant differences between the groups (p>0.05). All groups had similar postop rate of return to work status (group I vs group II vs group III: 11.4 % vs 14.7% vs 12.5%). CONCLUSIONS All groups had similar clinical and radiographic outcomes and comparable revision rates. Higher EBL, OR and LOS in Group III suggest that, absent focal C7-T1 pathology, extension of posterior cervical fusions into the thoracic spine may not be necessary. Extension of posterior cervical fusions into the thoracic spine may be recommended for higher risk patients with limitations to strong C7 bone anchorage. In others, it is safe to stop at C7. FDA DEVICE/DRUG STATUS This abstract does not discuss or include any applicable devices or drugs.

Background Wild-type transthyretin amyloid (ATTRwt) has been noted to deposit in the ligamentum flavum of the spine. Prior studies have focused on ATTRwt in the lumbar region, but studies discussing its presence in other levels of the spine are lacking. We report on the presentation of patients with confirmed amyloid in the cervicothoracic regions and discuss the literature to date. Methods We retrospectively identified patients at a single institution who underwent surgery for spinal stenosis and had pathologic specimens sent for amyloidosis testing with Congo red staining. ATTRwt was confirmed by the presence of transthyretin amyloid by typing and the absence of mutations in the TTR gene sequence. A final study group of patients with ATTRwt and spinal involvement was established (n = 27). Results Of 27 patients with amyloid in the spine, 24 (89%) had amyloid present in the lumbar region, 2 (7%) had amyloid in the cervical region, and 1 (4%) had amyloid in the thoracic region. The median age at which patients in the study underwent surgery was 71 years (interquartile range: 9). Spinal stenosis was the indication for surgery in 26 of 27 (96%) patients. Surgery involved 1 or 2 spinal levels in 24 of 27 (89%) patients. Conclusions ATTRwt amyloid predominantly deposits in the lumbar region, but it can also be present in the cervical and thoracic regions. While the lumbar regions should remain a focus for evaluation of ATTRwt amyloidosis, the cervicothoracic region should not be ignored.

New featurization schemes for describing materials as composition vectors in order to predict their properties using machine learning are common in the field of Materials Informatics. However, little is known about the comparative efficacy of these methods. This work sets out to make clear which featurization methods should be used across various circumstances. Our findings include, surprisingly, that simple fractional and random-noise representations of elements can be as effective as traditional and new descriptors when using large amounts of data. However, in the absence of large datasets or for data that is not fully representative, we show that the integration of domain knowledge offers advantages in predictive ability.

CO2 capture, utilization and storage (CCUS) is recognized as a uniquely important option in global efforts to control anthropogenic greenhouse-gas (GHG) emissions. Despite significant progress globally in advancing the maturity of the various component technologies and their assembly into full-chain demonstrations, a gap remains on the path to widespread deployment in many countries. In this paper, we focus on the importance of business models adapted to the unique technical features and sociopolitical drivers in different regions as a necessary component of commercial scale-up and how lessons might be shared across borders. We identify three archetypes for CCUS development—resource recovery, green growth and low-carbon grids—each with different near-term issues that, if addressed, will enhance the prospect of successful commercial deployment. These archetypes provide a framing mechanism that can help to translate experience in one region or context to other locations by clarifying the most important technical issues and policy requirements. Going forward, the archetype framework also provides guidance on how different regions can converge on the most effective use of CCUS as part of global deep-decarbonization efforts over the long term.

This Methods/Protocols article is intended for materials scientists interested in performing machine learning-centered research. We cover broad guidelines and best practices regarding the obtaining...

Given the dominant share of coal in China’s energy-generation mix and the fact that >50% of the power plants in the country are currently

In this chapter, tensile properties of different grades of post-consumer recycled (PCR) polycarbonate (PC) plastics have been compared with conventional or virgin PC before and after different aging conditions. 50 and 75% recycled PCs showed comparable yield strength (∼57 MPa), maximum tensile strength (∼70 MPa) and maximum strain (∼190–200%) before aging, when compared to virgin PC of same melt flow rate (MFR of ∼10 g/10 min). From the fractography analysis (optical and scanning electron microscopy) of the both virgin and 50% recycled PC, it is evident that the fracture morphologies are very similar and they are indicative of ductile failure. It is observed that with the presence of temperature and humidity (60°C 90% RH) aging, tensile strength starts to drop over time but most importantly both 50% and 75% PCR grades showed similar aging behavior compared to virgin PC (10–13% strength degradation after 500 hours of aging). Reliability modelling showed comparable B10, Weibull Alpha and Weibull Beta values between Virgin PC and PCR grades after different aging conditions. Fractography analysis of fresh and aged 75% PCR also showed ductile features.

Study Design Systematic review and Meta-analysis. Objective This systematic review seeks to compare fusion, reoperation and complication rates, estimated blood loss (EBL), and surgical time between multi-level instrumented fusions with LIVs (lowest instrumented vertebra) in the cervical spine and those that extend into the thoracic spine. Summary of Background Data Several studies address the question of whether to extend a long-segment, posterior cervical fusions, performed for degenerative disease, into the upper thoracic spine. Recommendations for appropriate LIV continue to vary. Methods A comprehensive computerized literature search through multiple electronic databases without date limits up until April 3rd, 2020 using combinations of key search terms and sets of inclusion/exclusion criteria was performed. Results Our comprehensive literature search yielded 3852 studies. Of these, 8 articles consisting of 1162 patients were included in the meta-analysis. In 61.2% of the patients, the fusion did not cross the cervicothoracic junction (CTJ) (cervical LIV, CLV). In the remaining 38.8%, the fusion extended into the upper thoracic spine (thoracic LIV, TLV). Overall, mean patient age was 62.5 years (range: 58.8-66.1 years). Our direct analysis showed that odds of fusion were not statistically different between the CLV and TLV groups (OR: .648, 95% CI: .336-1.252, P = .197). Similarly, odds of reoperation (OR: 0.726, 95% CI: 0.493-1.068, P = .104) and complication rates were similar between the 2 groups (OR: 1.214, 95% CI: 0.0.750-1.965, P = .430). Standardized mean difference (SMD) for the blood loss (SMD: .728, 95% CI: 0.554-.901, P = .000) and operative (SMD: 0.653, 95% CI: .479-.826, P = .000) differed significantly between the 2 groups. The indirect analysis showed similar fusion (Effect Size (ES)TLV: .892, 95% CI: .840-.928 vs ESCLV:0.894, 95% CI:0.849-.926); reoperation rate (ESTLV:0.112, 95% CI: 0.075-.164 vs ESCLV: .125, 95% CI: .071-.211) and complication rates (ESTLV: .108, 95% CI: .074-.154 vs ESCLV:0.081, 95% CI: .040-.156). Conclusions Our meta-analysis showed that fusion, complication, and reoperation rates did not differ significantly between patients in whom multi-level posterior fusions ended in the cervical spine vs those of which was extended into the thoracic spine. The mean blood loss, operative time and length of stay were significantly lower in patients with CLV at C6 or C7, compared to their counterparts. These data suggest that, absent focal, C7-T1 pathology, extension of long, posterior cervical fusions into the thoracic spine may not be necessary.

The fast and affordable development of novel materials is needed in order to enable technological advancements in application areas such as clean energy, healthcare, sustainable transport, and climate-friendly consumption. However, the development of novel materials is not a trivial task. One of the biggest challenges in materials design and discovery is the enormous search space of possible material compositions available, also referred to as the “chemical whitespace”. Faced with the high risk, high reward nature of materials exploration, materials scientists have increasingly moved away from traditional trial and error methods and instead adapted new data-driven methods of materials discovery. The rapid development of data science, machine learning (ML), and deep learning (DL) as well as the influx of high-quality materials property datasets have led to the development of the new field of materials informatics (MI). This new paradigm has drastically changed the way in which materials are understood, predicted, discovered, and designed. Despite promising developments in this relatively young field, there are several open issues that need to be addressed. The lack of guidelines and established procedures to ensure high quality research in MI impedes the pace of further development in this field. Furthermore, the current techniques for representing and modeling chemical compositions are flawed and unsuitable to be used in the search of novel materials. Lastly, the prevalence of black-box DL models without model interpretability limits the trust and adoption of these models in academia and industry. Accordingly, the main aims of this work are (1) to propose a set of best practices and protocols for conducting and reporting MI studies, and (2) to improve the state of the art in materials property predictions by introducing interpretable DL techniques for representing and modeling chemical compounds. In the first work described in this thesis, the fundamental ideas and considerations of using data-driven methods for materials science are introduced. A broad set of guidelines and protocols for ensuring the reliable, reproducible, and comparable reporting of research results in MI studies is established. Common software tools, methodologies, and materials data repositories are presented. Lastly, the full procedure of an ML study including data processing, feature engineering, model training, evaluation and comparison is demonstrated using the prediction of heat capacity for solid inorganic compounds as an example. In the second work, a novel DL model named “Compositionally Restricted Attention-Based network” (CrabNet), based on the Transformer self-attention mechanism, is introduced. CrabNet is benchmarked on 28 materials property datasets and is shown to match or exceed state-of-the-art models in the prediction of inorganic material properties. The benefits of learning element-element interactions within chemical compounds using the self-attention mechanism are discussed. Furthermore, a new way of representing chemical composition which overcomes some of the limitations present in current techniques is developed. Lastly, the opportunities to study model interpretability methods in CrabNet are previewed. Continuing in the third work described in this thesis, the model interpretability of CrabNet is further examined. Intrinsic model interpretability methods are added to CrabNet and used to extract additional information about the model and data representations during the modeling process. The extracted information is processed and visualized into static and interactive figures as well as video animations. The examination of these visualizations and additional information reveals well-known chemical patterns about the elements and compounds, intuitively suggesting that CrabNet is able to learn the element properties, element interactions, and how they together dictate materials properties. Furthermore, the dataset quality as well as the self-attention mechanism are also discussed for their significance towards an improved and interpretable modeling of materials properties. Lastly, the potential benefits of applying interpretable modeling methods in academia and industry are discussed. Overall, the methods, results, and considerations discussed in this dissertation are presented in a way to educate and empower interested materials science researchers to undertake their own materials informatics research. Die schnelle und erschwingliche Entwicklung neuartiger Materialien wird benötigt, um technologische Fortschritte in Anwendungsbereichen wie der sauberen Energie, dem Gesundheitswesen, dem nachhaltiger Verkehr und dem klimafreundlichen Konsum zu ermöglichen. Die Entwicklung neuartiger Materialien ist jedoch keine triviale Aufgabe. Eine der größten Herausforderungen bei Materialdesign und -entdeckung ist der enorme Suchraum möglicher Materialzusammensetzungen, der auch als „chemical whitespace“ bezeichnet wird. Angesichts der risikointensiven aber aussichtsreichen Natur der Materialerforschung ziehen Materialwissenschaftler zunehmend von herkömmlichen Versuchs-und-Irrtums- Methoden weg und adaptieren stattdessen neue datengetriebene Methoden der Materialentdeckung. Die schnelle Entwicklung von Data Science, maschinellem Lernen (ML) und Deep Learning (DL) sowie der Zustrom hochwertiger Materialdatensätze haben zu der Entwicklung des neuen Gebiets der Materialinformatik (MI) geführt. Dieses neue Paradigma hat die Art und Weise mit der Materialien verstanden, vorhergesagt, entdeckt und entworfen werden drastisch verändert. Trotz vielversprechender Entwicklungen in diesem relativ neuen Bereich gibt es mehrere offene Probleme, die behoben werden müssen. Der Mangel an Richtlinien und etablierten Verfahren zur Gewährleistung hochwertiger Forschung in der MI behindert das Tempo der Weiterentwicklung in diesem Bereich. Darüber hinaus sind die derzeitigen Techniken zur Darstellung und Modellierung chemischer Zusammensetzungen fehlerhaft und ungeeignet, um bei der Suche nach neuartigen Materialien verwendet zu werden. Schließlich begrenzt die Prävalenz von Black-Box-DL-Modellen ohne Modellinterpretierbarkeit das Vertrauen und die Akzeptanz dieser Modelle in der Wissenschaft und Industrie. Dementsprechend sind die Hauptziele dieser Arbeit (1) einen Satz bewährter Verfahren und Protokolle zur Durchführung und Berichterstellung von MI-Studien vorzuschlagen und (2) den Stand der Technik in der Vorhersage von Materialeigenschaften durch die Einführung von interpretierbaren DL-Techniken zur Darstellung und Modellierung chemischer Verbindungen zu verbessern. In der ersten in dieser Dissertation beschriebenen Arbeit werden die grundlegenden Ideen und Überlegungen zur Verwendung von datengetriebenen Methoden für die Materialwissenschaft eingeführt. Eine breite Reihe von Richtlinien und Protokollen, um die zuverlässige, reproduzierbare und vergleichbare Berichterstattung von Forschungsergebnissen in MI-Studien zu gewährleisten, wird etabliert. Gängige Software-Tools, Methoden und Repositorien für Materialdaten werden dargestellt. Schließlich wird das vollständige Verfahren einer ML-Studie für die Vorhersage der Wärmekapazität fester anorganischer Verbindungen, einschließlich Datenverarbeitung, Feature Engineering, Modelltraining, -auswertung und -vergleich, als Beispiel gezeigt. In der zweiten Arbeit wird ein neuartiges DL-Modell namens „Compositionally Restricted Attention-Based network“ (CrabNet), basierend auf dem Transformer Self- Attention-Mechanismus, eingeführt. CrabNet wird anhand von 28 Materialdatensätzen evaluiert und kann den Stand der Technik bei der Vorhersage anorganischer Materialeigenschaften erreichen oder übertreffen. Die Vorteile, Wechselwirkungen zwischen Elementen in chemischen Verbindungen unter Verwendung des Self-Attention Mechanismus zu lernen, werden ebenfalls diskutiert. Darüber hinaus wird eine neue Art der Repräsentation für chemische Zusammensetzungen, die einige der in aktuellen Techniken vorhandenen Einschränkungen überwindet, präsentiert. Schließlich wird eine Vorschau der Möglichkeiten Modellinterpretationsmethoden in CrabNet zu untersuchen gezeigt. Fortgesetzt in der dritten Arbeit wird die Modellinterpretierbarkeit von CrabNet weiter untersucht. Intrinsische Modellinterpretationsmethoden werden zu CrabNet hinzugefügt und verwendet, um zusätzliche Informationen über das Modell und die Datenrepräsentationen während des Modellierungsprozesses zu extrahieren. Die extrahierten Informationen werden in statischen und interaktiven Abbildungen sowie Videoanimationen verarbeitet und visualisiert. Die Untersuchung dieser Visualisierungen und der zusätzlichen Informationen ergibt bekannte chemische Muster hinsichtlich der Elemente und deren Verbindungen, die intuitiv darauf hindeuten, dass CrabNet die Elementeigenschaften, -wechselwirkungen und deren Einfluss auf die Materialeigenschaften, lernen kann. Darüber hinaus werden die Qualität des Datensatzes und der Self-Attention-Mechanismus hinsichtlich ihrer Bedeutung für eine verbesserte und interpretierbare Modellierung von Materialeigenschaften diskutiert. Schließlich werden die potenziellen Vorteile der Anwendung interpretierbarer Modellierungsmethoden in der Wissenschaft und der Industrie diskutiert. Insgesamt sind die in dieser Dissertation diskutierten Methoden, Ergebnisse und Erwägungen so dargestellt, dass sie interessierte Materialwissenschaftler dazu ermächtigen sich auf diesem Gebiet weiterzubilden, um ihre eigene Forschung in der Materialinformatik durchzuführen.

For hydrogen to play a meaningful role in a sustainable energy system, all elements of the value chain must scale coherently. Advocates support electrolytic (green) hydrogen or (blue) hydrogen that relies on methane reformation with carbon capture and storage, however, efforts to definitively choose how to deliver this scaling up are premature. For blue hydrogen, methane emissions must be minimised. Best in class supply chain management in combination with high rates of CO2 capture can deliver a low carbon hydrogen product. In the case of electrolytic hydrogen, the carbon intensity of power needs to be very low for this to be a viable alternative to blue hydrogen. Until the electricity grid is deeply decarbonised, there is an opportunity cost associated with using renewable energy to produce hydrogen, as opposed to integrating this with the power system. To have a realistic chance of success, net zero transition pathways need to be formulated in a way that is coherent with socio-political-economic constraints.

Carbon capture, utilization, and storage (CCUS) is a critical technology to realize carbon neutrality target in the Chinese coal-fired power sector, which emitted 3.7 billion tonnes of carbon dioxide in 2017. However, CCUS technology is often viewed as an "alternative technology" option owing to common perceptions of relatively high cost and potential risks. This study indicates that coal power CCUS is likely to be a cost-effective and key technology for helping China reach the ambitious goal of carbon neutrality. This comprehensive, national-scale assessment of CCUS deployment on coal power in China is based on a unique bottom-up approach that includes site selection, coal plant screening, techno-economic analysis, and carbon dioxide source-sink matching. Analysis indicates that, based on 2017 costs and assumptions, more than 70% of coal power plants in this study could be cost-competitive with natural gas-fired power plants, and 22-58% would be cost-competitive with onshore wind generation. These insights suggest that the commercialization of CCUS technology in the coal power sector in China is a viable route toward decarbonizing the economy if a grid price policy similar to that of renewables and natural gas power is applied.

Summary Carbon capture and storage (CCS) is an important option for any “lowest cost” transition to a decarbonized global economy. China, given its large CO2 footprint and strong dependence on coal, is a particularly important player. In this study, we show a pathway for affordable deployment of CO2 capture in the Chinese power sector scaling to greater than 100 million tCO2 per year by 2030 at an avoided cost of less than $37/tCO2. We show that the distribution of costs across a fleet of power plants presents an underappreciated opportunity to reduce costs and that several avenues are available in the existing policy and market regime to further reduce the net avoided cost of capture to as low as $25/tCO2 for some plants. This means the rate of deployment of CCS in China will likely be limited by the maturation of transport and storage, rather than the cost of capture.