id
stringlengths 24
24
| idx
int64 0
402
| paragraph
stringlengths 106
17.2k
|
|---|---|---|
67ac29c881d2151a02169df6 | 7 | To illustrate how specialized organs help prepare biological materials for their rapid on demand solidification, we detail the mussel byssus example . Byssal fibers cure after release, triggered by the higher pH of saltwater compared to that of the gland . This fiber anchors mussels to the shore and withstands the crushing waves with its outstanding toughness and underwater adhesion properties . Within the mussel's foot, plaque gland cells secrete plaque vesicles and metal storage particles (Fe/V) into longitudinal ducts . Cilia transport these concentrated droplets to distal depressions, where the material forms reversible metal-containing coordination complexes with catechol groups of dihydroxyphenylalanine (DOPA)-containing vesicles . After secretion from the foot, the fiber is held at its stem by billions of cilia and detaches upon neurochemical stimulation . While synthetic and biotechnological methods aim to mimic this hierarchical material , self-microfabrication offers a promising alternative for replicating complex material processing. |
67ac29c881d2151a02169df6 | 8 | In this sections, we explore the spider silk gland to illustrate how an organ tightly regulates material self-assembly 159 . Its physiology 160 , as the silkworm's 161 , have been extensively studied, making both excellent candidates for selfmicrofabrication. Spider silk is a prime example of a sustainable biological material with high mechanical performance and industrial applications 128 . Tougher than Kevlar 162 and not dependent on fossil fuels, spider silk has been extensively studied in various species 127 for its potential uses, notably in bulletproof vests or composites 128 . Replicating the silk's intricate microstructure synthetically has posed a challenge for decades, notably due to its complex processing conditions (Fig. ). Chemical microenvironments (ChemµEnv), such as the pH and salt concentrations, guide the self-assembly of biological materials through thermodynamics principles . Spider silk glands use specialized cell types to synthesize and secrete precursor proteins (spidroins) into the extracellular space. Other cells subsequently regulate the extracellular ChemµEnv 165 , driving the proper folding of proteins and material assembly . For these reasons, the spider silk gland example is detailed to showcase the role of multicellular tissues in hierarchical material production. However, an in-depth understanding of spider biology is not required in the following sections. In a spider, up to seven types of glands can produce silk with different properties, such as the tough dragline from the major ampullate (MA) glands or sticky fibers from the aggregate glands. Furthermore, environmental adaptations across 50,000 spider species have diversified their silk properties , including hydrophobic motifs in underwater spiders . We focus on Larinioides sclopetarius and Nephila golden orb weaver MA silk glands, due to the large literature available on these species, the high mechanical properties 127 and versatile applications of their silk 128 . Despite significant progress in spider silk research, no current technique can replicate the gland's complex material processing, encouraging efforts to develop spider silk gland organoids in the future. When trying to reproduce a biological material, approaches of various scales should be considered. At the organism scale, farming silkworm colonies industrially is efficient 169 but impractical for spiders, owing to their cannibalistic nature, as revealed by early attempts in the late 18 th century 170 . At the molecular scale, fully synthetic approaches can mimic specific aspects, such as spider silk discoveries improving nylon's properties 171 . Recombinant expression of silk protein (spidroin) has been developed in various hosts to better replicate the material composition . Many challenges associated with that approach have been addressed in the past decade 175 , including preventing aggregation in highly concentrated spidroin solutions . However, the yield and scalability continue to be significant bottlenecks 177 . Furthermore, the native silk dope composition is more complex than a few recombinant spidroins: including peptides, lipids, glycoproteins that serve functions such as lubrication, antibacterial and more 178 . While the synthesis function of silk glands can be simplified synthetically (e.g. recombinant silk solutions), the process of transforming the liquid phase into a solid fiber has not been fully replicated 179 . Wet spinning, dry spinning, and electrospinning 180 can produce tough threads but often induce aggregation instead of proper protein folding, reducing the resulting properties . Biomimetic aqueous spinning using microfluidics is the most promising synthetic approach yet that may control the ChemµEnv 181 . However, it does not match the in vivo multicellular regulation of the spinning process where MA glands regulate the pH, salts, water content and shear forces . To address the spinning challenge at a tissue scale, silkworm glands have been genetically engineered to secrete spidroin . Despite not fully mimicking the spider glands, transgenic silkworms present a promising strategy for largescale 184 , high-performance silk production 185 . It further highlights the importance of a multicellular organ for hierarchical material processing. These examples at various scales can be generalized to other biological materials and organs. |
67ac29c881d2151a02169df6 | 9 | This section details how the spider silk MA gland regulates material synthesis and assembly at the tissue, cellular, and molecular scales. At the tissular scale, the MA gland, located in the spider's abdomen, is organized in three regions: the tail for spidroin secretion, the ampulla for storage and coatings, and the duct for solid material formation 186 (Fig. ). The tail is a long, widening tubular epithelium on the anterior side of the organ (zone A). It synthesizes and secretes spidroin vesicles in the lumen 187 . The ampulla stores the liquid silk dope at high concentrations , enabling rapid, on demand, spinning . The ampulla is itself segmented in two regions, zone B and C, secreting each different coatings layers 187 . These middle and outer layers wrap the core silk dope produced in the tail, creating a concentric meso-structure. The third part of the organ, the duct, can be seen as an extruder and is specialized in converting the silk dope from liquid phase to solid fibers. To do so, it regulates the ChemµEnv in the lumen, which guides spidroins' proper protein folding . Pulling the hanging thread at the posterior side, exiting the spigot, initiates the process. Pulling the fiber manually, namely forced silking, also induces spinning. Tensile forces propagate via intermolecular interactions, from the spigot to the ampulla, initiating the flow . A ratchet and pump mechanism is hypothesized to restart spinning if the fiber breaks within the gland . The solidification of silk solution into fiber occurs as it flows from the ampulla and through the duct. The tapering funnel and duct induce shear forces 190 and align the molecules 191 , improving the resulting material properties . The funnel's thick cuticle connects the flexible duct to the more stable ampulla 194 , and is hypothesized to prevent the dope from excessive shearing when the spider moves around . The thicker cuticle also helps to oppose the induced forces from the flow. The duct regulates ChemµEnvs, establishing gradients of salts, pH and water content along its length 195 . As it flows through these conditions, the dope self-assembles in a hierarchical material . To complete the conversion to solid fiber and reduce water losses, terminal tubules on the posterior side pump remaining water 196 and the spigot's lips retain the aqueous phase inside . This example shows how a multisectional organ can perform coordinated functions for material production. |
67ac29c881d2151a02169df6 | 10 | Cellular scale: specialized functions acting in a collective At the cellular scale, specialized cell types have been localized and characterized by single-cell sequencing . The core of silk fibers 159 is secreted by three columnar epithelial cell types localized in the tail (zone A). The middle and outer layer coatings are secreted by two cell types in zone B and C of the ampulla (Fig. ). The columnar epithelia are filled with spidroin vesicles, secreted to the lumen from their apical side 187 . Each cell type produces a different combination of spidroins, used to define their identity 159 , and realizes post-translational modifications 197 impacting their self-assembly 198 . Another cell type is scarcely present in both the tail and the ampulla, and does not seem to be directly involved spidroin expression 159 ; we therefore hypothesize its role in supporting its surrounding cells or for tissue homeostasis. The duct consists of a thickening epithelium along its length, as revealed by histology, electron microscopy and single-cell sequencing, which has identified two duct cell types . A cuticular intima separates the cells from the lumen but enables the diffusion of molecules with its pore-like structure 199 . The duct cells regulate the lumen's ChemµEnv, notably the pH with high carbonic anhydrase expression levels 165 ; they are filled with vesicles and have dense apical microvilli . At the posterior side, flask-shaped cells contain secretory granules and may add an additional coating layer 189 . Water is continuously pumped out of the lumen during spinning, notably with terminal tubule cells . The specialized cell functions in spider silk glands exemplify the complexity of material-producing organs, challenging their biomimicry using conventional recombinant approaches. |
67ac29c881d2151a02169df6 | 11 | Molecular scale: regulation of self-assembly At the molecular scale, the spidroins are high molecular weight proteins with repetitive domains, driving their self-assembly in nanometer-scale structures 200 (Fig. ). The silk dope is itself highly complex, made of many components 178 . Each cell types secretes different composition and quantities of spidroins 159 , tuning the material properties at each layer 127 . The silk dope also contains a heterogeneous group of non-spidroin proteins with low molecular weight, named spider-silk constituting elements (SpiCE) . Peptides, lipids 202 , glycoproteins , and more 204 have also been reported in the silk dope. The hypothesized functions of the main components are: mechanical properties for spidroins; plasticity and support for SpiCE proteins; water balance, lubrication, and pliancy for glycoprotein coatings; and antibacterial protection and social signaling for outer-layer lipids . In most MA spidroins, crystalline regions are formed by polyA and polyGA domains, while glycine rich domains give an amorphous phase . The ratio and size of these domains affects the stiffness and extensibility of the fiber 207 . Spidroins are stored in the ampulla at high concentrations, up to 50% w/v 208 , higher than most artificial spinning . Spidroins self-assembly 178 can be described according to three models: micellar , liquid crystalline 209 , and liquid-liquid phase separation 210 (LLPS). Terminus domains (NT and CT) play a major role in this structure: NT dimerizes spidroins and transfers tensile forces, while CT contributes to the rapid, on demand polymerization of spidroins . Shear forces in the duct and drawdown effects further align spidroin nanoassemblies, stabilized in β-sheets structures . Once again, the spider silk gland showcases the intrinsic complexity of biological material composition and processing in vivo. To mimic silk self-assembly with recombinant expression and artificial spinning, the conditions should be greatly simplified. It is unclear if such reduction in complexity will be capable of producing the proper microstructure and versatile properties of biological materials, such as spider silk. Self-microfabrication, using tissue engineering, offers a complementary approach, aiming to grow material-producing organoids . Insights from spider development , at the single-cell resolution , are encouraging contributions toward this goal. |
67ac29c881d2151a02169df6 | 12 | We have seen how organs can regulate complex material selfassembly. This section explores how to grow such multicellular tissues, notably from stem cells. Biological systems consist of numerous interacting components organized across scales, including molecules, cells, organs, and entire organisms . At the macroscopic scale, top-down approaches extract materials directly from organisms, such as tree trunks, cotton, or silkworm cocoons. At the microscopic scale, bottom-up manufacturing produces recombinant proteins (e.g. bio-based polymers ) or guides material production (e.g. concrete biohealing 217 ). |
67ac29c881d2151a02169df6 | 13 | Despite ongoing advancements , biofabrication approaches often depend on monoclonal colonies 219 . These systems may lack the intracellular and intercellular complexity required by specialized cell types and organs, as exemplified by the spider silk gland example. Bacterial consortia partially address this limitation, notably in biomineralization , but solutions remain to be found for material-producing multicellular organs , which are not covered by current engineering approaches . This leaves a research gap at intermediate scales, between top-down and bottom-up biofabrication strategies. A rational approach to biofabrication may bridge this gap, by directing resources exclusively toward a necessary and sufficient organ, without requiring an entire organism. For instance, snake venom gland organoids produce toxins for anti-venom research without the need to develop other organs, such as the brain . This approach could also enable new functionalities and benefits; i.e. removing the risk of accidental bites from traditional snake breeding 225 . To rationalize biofabrication even further, tissues can themselves be edited and improved , such as multicellular engineered living systems (M-CELS) . We propose to define such rational approach as selfmicrofabrication: the process of fabricating structures at microscopic scale through self-organization. Conventional microfabrication in the semiconductor industry already creates desired patterns , but barely uses self-organization 229 . Selfmicrofabrication also builds microstructures guided by DNAencoded patterns, and achieves an unprecedented level of control over complex material self-assembly 230 . This property differs from passive thermodynamic processes 231 , such as phase transitions in metals, with its common use of energy-dependent molecular chaperones or nanoreactors 232 . Thus, selfmicrofabrication occurs at multiple scales: self-organizing both production units (organogenesis) and materials (molecular selfassembly). This article explores these principles, exemplified by organoids producing materials (Fig. ). |
67ac29c881d2151a02169df6 | 14 | Regulation by the microenvironment Self-organization processes are heavily influenced by surrounding conditions. Cells and organs regulate ChemµEnvs in microscopic compartments , enabling precise spatiotemporal control over material assembly (Fig. ). Synthetic approaches mimic these conditions and explore beyond natural processes 233 . Nevertheless, cellular collectives tune ChemµEnvs at microscopic scales and higher complexity than synthetics, and thus achieve hierarchical structures that are difficult to mimic 187 . The formation of such biological tissues, or organogenesis, is a complex orchestration process . Previous reviews have explored the principles of organoid self-organization . We focus on how environmental conditions influence their development 236 . To introduce a distinction from chemical microenvironments (directing the self-assembly of materials), cellular microenvironments (CellµEnv) describe the conditions regulating cell self-organization 237 (Fig. ). CellµEnvs use various conditions, communication pathways, diffusing biochemicals 238 , extracellular matrix (ECM), mechanical or bioelectric cues 241 , to name a few. Stem cell niches illustrate CellµEnvs, where surrounding signals either maintain stemness or induce cell differentiation 242 . They guide embryonic development, adult homeostasis, regeneration, and more 243 . For instance aged cells can be rejuvenated when placed in a biologically "younger" environment 244 , such as blood or skeletal cells 245 . When the blood circulations of a young and an old mouse are connected through parabiosis, systemic rejuvenation of the older mouse's tissues is observed . Key factors expressed by certain cells can also reprogram a tissue's identity, notably the thickening of skin upon volar fibroblast transplants . In many cases, only a few factors are necessary and sufficient to guide cell behavior and organoid development. For instance, only four Yamanaka factors are sufficient to reprogram differentiated cells into iPSCs 248 . As a second examples, four key signaling factors guide Lgr5-positive stem cells to grow intestinal organoid , not mentioning a few more essential parameters such as the ECM , culture medium and small molecules 250 . Tissue engineering leverages this ability of complex biological systems to arise from relatively simpler external inputs. By leveraging both tissue self-organization and organ-regulated material self-assembly, self-microfabrication aims to create high-value products, such as spider silk-based body armor 251 . |
67ac29c881d2151a02169df6 | 15 | Proof of concept: self-organized production units Self-microfabrication has yet to be fully established. While many examples exist to demonstrate parts of the approach , the hair follicle organoid is the most comprehensive example yet 2 . This organoid self-organizes from human pluripotent stem cells (hPSCs) to produce a hierarchical material: the multilayered hair shaft. It illustrates how relatively simpler engineered conditions can give rise to a highly complex organ producing material (Fig. ). The hair complements the spider silk example by demonstrating a material produced within cells, rather than externally. We describe the hair follicle organoid development in detail to showcase the complex autonomous processes occurring during organogenesis. Hair organoid in vitro growth begins with the aggregation of hPSCs, recapitulating normal stages of embryonic development 254 by differentiating into various cell lineages 255 (Fig. ). A spheroid forms and creates an inner cavity 256 . Mesenchymal cells migrate outwards and selforganize in the dermis while the inner surface ectoderm gives rise to the epidermis 2 . Hair placodes appear, develop into hair germs and bulge out in hair pegs to form the dermal papilla (DP) and bulge 256 . Hair-bearing follicles eventually grow out, each containing tissue-specific stem cells that can be maintained in vitro for half a year . Within each follicle, a hierarchical material is produced -the hair shaft -implying a complex multicellular regulation of the keratinization process 257 . Hair follicle organoids replicate cell types found in vivo 258 and mimic the 3D organization of the real organ, such as DP, melanocytes, and surrounding connective tissue 259 . Even the follicles' patterning is reproduced, with regular intervals between them 260 . This example demonstrates that complex production units can self-organize from stem cells and assemble microstructured materials. |
67ac29c881d2151a02169df6 | 16 | Beyond their development, organoids have integrated mechanisms to self-maintain their function through their lifespan. The hair-bearing organoid homeostasis mechanisms are described to illustrate this unique property of selfmicrofabrication production units. In vivo, hair follicles contain bulge stem cells that differentiate and eventually replenish the hair matrix cells, which will form the solid hair shaft 261 (Fig. ). This stem cell homeostasis strategy shares common characteristics with other adult tissue homeostasis 262 . A small pool of cells remains quiescent, maintaining their stemness via microenvironmental signals from their niche, including bone morphogenetic protein (BMP), transforming growth factor beta (TGFβ) and Wnt inhibition . The anagen phase activates bulge stem cells 264 , which migrate downward through various CellµEnvs guiding their differentiation 265 . They become transit amplifying cells for rapid proliferation, and then progenitors before reaching the DP and giving rise to matrix cells 266 . At the dermal side of the hair follicle bulb, DP cells regulate signaling and nutrient supply, influencing the surrounding CellµEnv 267 (Fig. ). DP cells interact with matrix cells using diffusing signals through the basement membrane and regulate their differentiation . Matrix cells proliferate rapidly and undergo a keratinization process while migrating away from the DP signals . They increase the expression of type I and II keratin proteins, which assemble in heterodimers, then tetramers and finally align headto-tail in intermediate filaments . Keratohyalin granules begin to form, increasing keratin's concentration and storing it within the cells in preparation for their solidification. Inside the granules, filaggrin and trichohyalin proteins help bundle filaments into microfibrils while transglutaminase stabilizes them by cross-linking cysteines into disulfide bonds 271 , which contributes to the mechanical stiffening and hardening of the hair 272 . During cell migration, keratinocytes' nuclei and organelles slowly degrade, halting metabolic activity. The cells are progressively transformed and compacted together in a multilayer solid material composed of microstructured keratin fibers . The hair-bearing organoid is currently used as a model in development, wound repair, and hair loss research and creates opportunities beyond medical applications . A keratinization organoid model offers a promising foundation for studying other keratin-based biological materials found in vertebrates (e.g. nails, horns, scales, shells, beaks, feathers, and hooves 273 ) and invertebrates (e.g. marine organisms 274 ). The hair organoid produces a hair shaft with all seven concentric keratinized cell layers found in vivo 2 (Fig. ). None of the material processing steps, including the keratinization or crosslinking, have been directly engineered nor regulated as it is in artificial systems. The complex organogenesis and maintenance of this materialproducing organoid require minimal inputs and leverages existing self-organization programs 260 . To the best of our knowledge, no synthetic system can autonomously regulate the formation of such hierarchical material at this scale, nor its production unit, highlighting the innovative potential of selfmicrofabrication. However, developing organoids in new species raises fundamental challenges 275 , developed in the next section. |
67ac29c881d2151a02169df6 | 17 | Exploring non-model species Most biological materials of interest are typically found in species less studied than model organisms (Table ). However, growing an organoid relies on preliminary achievements such as cell culture conditions, stem cell identification and a certain understanding of growth factors . This poses a central challenge to material-producing organoids, introduced here and further addressed with emerging solutions in following sections. Intraspecies variability already illustrates the difficulty of generalizing findings between individuals, notably in personalized medicine 277 . This issue is further complicated by interspecies variability, notably with greater phylogenic distances . Collegial efforts are necessary to unravel new cell identities 279 , signaling cues 280 , gene regulatory networks 281 and differentiation pathways 282 , to cite a few. The development of the first organoid relied on decades of research on dissociationreaggregation experiments, embryoid bodies, 3D culture systems and iPSCs . In comparison, recent developments of organoid protocols in new tissues have accelerated rapidly in the past fifteen years 275 . For instance, hair follicle organoids have advanced due to progress in tissue regeneration , growth factor discoveries 286 , niche characterization 287 , embryonic cell aggregation , stem cell differentiation media 290 , and patterned hydrogel scaffold development 291 . |
67ac29c881d2151a02169df6 | 18 | The "missing body problem": cellular microenvironment When trying to grow a tissue outside its native environment, a basic question is: what is lacking for its survival? In a nutshell, the rest of the body is missing. In practice, only a finite number of cues are sufficient to guide a tissue's development and maintenance. For instance, the self-renewal and proliferation of many stem cells is regulated by their niche 292 , which can be replicated in vitro . Characterizing a new tissue's CellµEnv is challenging due to its complexity and potential differences from established systems 294 . Defining these dynamic environments involves the understanding of signals from surrounding cells 295 , diffusing molecules gradients 296,297 , ECM 298 , mechanical 299 and bioelectric cues 300 . To enable self-microfabrication, addressing the "missing body problem" by maintaining functional tissues ex vivo is a crucial initial proof of concept that the organ may eventually produce material autonomously, which has been achieved with spider silk explants. |
67ac29c881d2151a02169df6 | 19 | The technology readiness level of self-microfabrication is between 2 and 3 301 when considering the hair follicle organoids as proof-of-concept. Scaling up organoids for medicine has revealed significant challenges, but potential solutions can be leveraged for material-producing organoids 302,303 (Table ). Unlike medicine, self-microfabrication faces fewer regulatory constraints but emphasizes cost-efficiency 304 . Focusing solely on costs at this stage would be shortsighted, as it overlooks future innovations, automation and economies of scale. Much is to learn from the cultivated meat sector that has even stricter cost requirements 307 . Biological materials may balance the cost by focusing on high value products 308 that exhibit high mechanical performance . Addressing production rates and biological constraints Growing biological materials is generally slower than producing their synthetic counterparts. For instance, Nephila spider MA silk is produced 30 times slower than Kevlar, i.e. 5 µm diameter at 10 cm/s for silk 332 and 10 µm diameter at 80 cm/s 333 . However, spiders use liquid silk preliminary stored in the ampulla. Accounting for protein synthesis rates would likely reveal an even smaller production rate than synthetic polymers, as observed in other species 334 . Although essential, this comparison is somewhat inequitable since industrial polymer synthesis has already gone through decades of optimization. Synthetic biology can also accelerate cell cycle and optimize the growth and metabolism of tissues 335 . Indeed, modeling growth rates and gene regulation 336 highlight potential improvements 337 and trade-offs 338 . To compensate growth limitations, material-producing organoids may produce in parallel to increase the throughput. Organ production units are self-organized and compact, potentially enabling their large proliferation in number, facilitating their parallelization . Since biological materials aim to improve sustainability compared to fossil-based solutions 177 , the environmental impact of biotechnologies should be considered 339 . Large scale selfmicrofabrication may also eventually compete for arable land to produce biomass, as already observed for biofuels . Potential solutions include polycultures, yield optimization 341 and circular economies with the valorization of wastes 342 . Challenges in developing material-producing organoids are mostly addressed in Section 6. Beforehand, we present a theoretical framework for self-microfabrication to examine key concepts. |
67ac29c881d2151a02169df6 | 20 | Extensive studies have detailed the formation of organs ; here, we focus on the plasticity of developmental trajectories. Given the challenges in developing material-producing tissues, it is important to note that organoids can often be cultivated using multiple methods, providing alternatives if one approach fails. For instance, intestinal organoids can be derived either from adult stem cells (ASC) , or from embryonic pluripotent stem cells (PSC) and iPSC 345 ; each using different developmental trajectories. The organoid's maturity is affected by its cellular origin 234 , but can be partly compensated by longterm cultures 346 and co-culture interactions 314 . Organoids can still be grown in tissues lacking ASCs in vivo, such as the heart or brain , by leveraging cellular plasticity 349 . Many culture methods have been developed but the field needs standardized protocols and unified representations 321 . For instance, signaling pathways, cell differentiation trajectories 350 and other biological processes can be formalized into workflows with step-by-step descriptions . We suggest that similarly, a systematic definition of organoid developmental trajectories in standardized workflows would help future discoveries. Such workflow description is generally useful for computational approaches and may eventually contribute to developmental trajectory inference, as it is currently the case with signaling pathways . Organoid protocols only use a few signaling factors to trigger stem cell self-organization 353 . Developmental workflows should emphasize key induction events, such as differentiation signals, symmetry breaking, or pattern formation . The triggers for these events can be transcription factors, chemical gradients, or bioelectric fields 355 . Specific to materialproducing organs, workflows can also describe biomolecule synthesis and assembly. |
67ac29c881d2151a02169df6 | 21 | Once an organ is formed, it is maintained by multiple mechanisms that may be exploited for self-microfabrication. Under normal physiological conditions, also called steady-state homeostasis, tissues use specific maintenance strategies to function throughout the lifetime of an organism 356 (Table ). Other strategies may be activated following a perturbation, such as regeneration in injuries, dysfunctions, pathologies or aging . This diversity of available tissue maintenance strategies, often within a single organ, can be exploited to grow organoids of interest. For instance, many organoids rely on ASC-mediated tissue renewal , such as the hair bulge, basal keratinocytes or intestinal crypts 356 . Other organs are maintained by progenitor cells with restricted self-renewal, or terminally differentiated cells that revert to proliferation 405 . Others leverage dedifferentiation or transdifferentiation, such as hepatocytes upon chronic injuries . A few tissues have little to no cell turnover, such as most post-mitotic neurons or the inner ear hair cells 390 , and may rely on increased intracellular repair mechanisms to maintain their function 391 . Other non-proliferative strategies, such as cell compensations for metabolism, size or shape help maintain the organ function 407 ; for instance, physical exercise induces skeletal muscle cell hypertrophy 394 . Some organs use more than one tissue maintenance strategy, the liver being a prime example. Hepatocytes use cell proliferation during steady-state homeostasis. Following the surgical removal of 30% of the liver, hepatocytes activate compensatory hypertrophy, and following other injuries like toxin exposure, they can transdifferentiate or recruit progenitor cells 370 . Switching from one mechanism to the next can be either preprogrammed within cells, or triggered by external signals, such as apoptosis 405 , phagocytosis, ECM remodeling, and inflammation. Surrounding cells can also signal what tissue maintenance strategy to use, such as macrophages 408 , adipocytes 409 , or microbes 410 . These signals can be induced in culture and engineered to enable tissue growth in non-regenerative organs . Thus, self-microfabrication is conceivable even in non-regenerative organs. |
67ac29c881d2151a02169df6 | 22 | We have seen how multiple developmental trajectories and tissue homeostasis strategies enable organoid growth. These mechanisms are complex, interconnected, and operate at multiple biological scales, making them challenging to apprehend. This section examines how to simplify and better understand essential mechanisms of self-microfabrication. Waddington introduced simple representations of complex biological systems, named epigenetic landscapes 413 . We propose to extend the idea into hierarchical landscapes, describing organ formation across biological scales (Fig. ). |
67ac29c881d2151a02169df6 | 23 | For instance, at the cellular scale, cell types are viewed as balls rolling down a hilly landscape, where valleys are differentiation trajectories, from stem cells at the top to lower specialized cells at the bottom. These simple diagrams allow researchers to depict and share diverse phenomena, such as cell-cell communication (Fig. ). Each layer may be built by integrating experimental data from various sources, such as gene sequences, cell's gene expression profiles or tissues' spatial omics (Fig. ). Since tissues are made of cells, themselves composed of biomolecules, we suggest that this formalism may allow a "vertical" flow of information across scales (Fig. ). Similarly, "horizontal" information transfer may be envisioned within a layer (Fig. ). The purpose is to offer a simple mental image of complex organs and eventually generalize their principles to new material-producing organoids. As diagrams may be abstract, Figure provides a more detailed representation of hierarchical landscapes. At the tissular scale, valleys represent the multiple developmental routes discussed in the previous section (Fig. ). The width of a valley, or canalization, indicates how robust development is to perturbations. For example, the "Picasso frog" can self-correct significant embryonic defects 414 . Each location represents an organ state, composed of multiple cell types structured together and linking this tissular scale to the next. |
67ac29c881d2151a02169df6 | 24 | The cellular scale showcases basins of attraction, where a region canalizes cells to an attractor. ASC are metastable attractors, differentiating in multiple specialized cell types 415 (Fig. ). Landscapes are useful to find genes guiding differentiation 416 or reprogramming factors . At the molecular scale, a free-energy landscape models biomolecules conformation and self-assembly 418 , it supports discoveries 419 and enables microstructure engineering 420 (Fig. ). This representation should not be seen as static, some hills and valleys can dynamically change, influenced by GRNs and surrounding microenvironments 421 . For instance, cell reprograming factors may "invert the slope" during dedifferentiation 417 (Fig. ). Such hierarchical landscapes offer a simple and generalizable representation of complex selfmicrofabrication systems. Modeling complex tissues Beyond a conceptual tool, an epigenetic landscape enables computational modeling 415 , using General Systems Theory 422 . For instance, at the cellular scale a cell type's unique gene expression profile can be viewed as a point in multidimensional space, with each coordinate corresponding the expression level of each particular gene 423 (Fig. ). It is commonly used in single-cell sequencing, with the gene expression matrix containing all information and dimensionality reduction techniques enabling clustering and other data processing . Different biological processes can be modeled depending on how the coordinates, or parameters, are defined. For instance, at the molecular scale, using chromatin modifications as variables for the landscape give insights about epigenetics . If the topography is built using DNA sequences, it may represent GRNs , protein conformations 428 , chromosome interaction 429 or phylogenic relations . Similarities between these modalities may help to interconnect landscapes (Fig. ). Landscapes have been defined at many biological scales 431 , including the tissular layer describing developmental trajectories during morphogenesis . We suggest that interconnecting them can allow the flow of information across scales (Fig. ). As cells are easier to characterize than organs, tissular landscapes may be inferred from more accessible cell measurements, such as single-cell transcriptomics used in cellular landscapes , even in non-model species 434 . To support the idea of vertical interconnections, we propose a theoretical demonstration of its feasibility. Consider a space X representing a molecular landscape with biomolecules x, Y as the cellular landscape, and Z tissular. P(x) is the probability of a molecule x to be in a particular state, the same applies to P(y) for a cell, and P(z) for a tissue. Using experimental data, the density distributions of P(x), P(y) and P(z) can be computed (e.g. gaussian mixture models, normalization flows, variational autoencoders or other methods 435 ). Similarly, conditional probabilities can be derived from empirical data, such as P(y|z), the probability of a cell type y being in a tissue z, and P(z|y) that a tissue z contains a given cell y. To carry the contribution from one scale to the next, we need to connect them hierarchically (X→Y→Z). As an example, we focus solely on two scales (Y→Z). Equation expresses P(z), the probability of a tissue state, as a marginal distribution derived from P(y,z), the joint probability of a cell state y and a tissue state z: |
67ac29c881d2151a02169df6 | 25 | using only a few observations of cells within a tissue P(z|y), along with the more readily available probability P(y) of a cell being in a specific state. This development illustrates one way to bridge landscapes across biological scales. While the proposed framework provides a theoretical basis for bridging scales, more applied approaches are available 436 (e.g. coarse grained 437 , agent-based 438 , cell-cell communication models , or virtual cells 440 ). In silico representations aim to complement experimental evidence and facilitate the understanding, prediction, simulation, and optimization of biological tissues over time 226 . In the meantime, valuable insights can be gained from the models themselves. For instance, how cellular collectives, or decentralized multi-agent systems, are programmed to reach specific outcomes 441 . Improved modeling, prediction and simulation approaches could accelerate the development of new organoids. |
67ac29c881d2151a02169df6 | 26 | Since CellµEnvs control tissue behaviors, we can guide development by engineering these microenvironments. This paragraph discusses key aspects: characterization, modeling, understanding the effects of CellµEnvs, and engineering. Characterization techniques have improved in throughput and sensitivity in the past decade, reaching single-cell resolution with spatial omics 442 , measurements of key regulators 443 , functional perturbations using CRISPR 444 , and mechanical tests . Moreover, versatile data-driven approaches help uncover cellular communication pathways and their downstream effects 447 . |
67ac29c881d2151a02169df6 | 27 | To model spatiotemporal CellµEnvs, biosignaling fields can be developed ; using scalar values (e.g. temperature, diffusing chemicals or electric potentials), vectors (e.g. velocities 448 ), and tensors (e.g. mechanical stresses ). As each cells emits and receives signals, their individual contribution can be added or subtracted to the field. For instance, such "sources" and "sinks" can be localized in 3D and signal patterning devices . 3D cultures are influenced by their scaffold's composition 469 , porosity, topography 310 , functionalization, degradation and release of signals . CellµEnv engineering use techniques such as coated beads, soft lithography 471 , gradients 472 , bioprinting 473 , optogenetics and magnetic fields 475 . To grow larger tissues, bioreactors and microfluidic channels help oxygen and nutrients supply, as well as waste and dead cells removal 477 . Finally, mechanical signals can be tuned by materials stiffness and viscoelasticity or using pressure chambers and shear forces . Using these techniques, tissue engineering can control the biosignaling field to regulate organoid long-term cultures 479 . Self-microfabrication benefits from continuous improvements in CellµEnvs characterization, modeling and engineering to support future material-producing organoid development. |
67ac29c881d2151a02169df6 | 28 | Exploring non-model species Many biological materials of interest come from less-studied species compared to model organisms (Table ). The development of new model species is supported by versatile research tools 480 , notably single-cell multi-omics . Adult stem cells are identified across kingdoms and for many species, cells can be dedifferentiated into iPSC 486 owing to greater insights into reprogramming factors 487 since Takahashi and Yamanaka's discovery in 2006 248 . Similarly, the achievement of organoid development by has reached many new tissues and organisms (Table ). Expanding organoids to lesser-studied species, notably for material production, is challenging due to the difficulty to transferring biological knowledge from one species to another . Computational approaches help to generalize information across species, such as the inference of gene annotations, sets, pathways, GRN 492 , phenotypes or cell types. Many algorithms often use conserved molecular components (e.g. orthologs) to find similarities between organisms 493 . Similarly, cross-species knowledge transfer can be viewed as "horizontal" connections between the landscapes of different organisms (Fig. ). As hierarchical landscapes can also link biological scales (Fig. ), we suggest their use to infer cell types 494 and organogenesis mechanisms from known species to non-model organisms (Fig. ). Transferring landscape topography across species could and guide experiments by highlighting regions of interest, likely to contain cell types or molecular components, such as promoter, growth factors or "missing links" in signaling pathways. Such cross-species knowledge transfers could help accelerate the current expansion of organoid technologies to non-model species that produce biological materials. |
67ac29c881d2151a02169df6 | 29 | The "missing body problem": cellular microenvironment When an organ is removed, or grown outside its native body, it requires specific signals to support its survival. Developing material-producing organoids encourages the study of CellµEnvs in non-model species, which can lead to unexpected breakthroughs . Key signals from the "missing body" can already be studied, independently from other selfmicrofabrication challenges, using transplants 496 and ex vivo explant cultures 497 . Before growing material-producing organs, explants serve as effective in vitro models to examine their physiology and characterize essential CellµEnv factors for survival 498 . To demonstrate this principle, we have developed a spider silk gland explant that spins fibers in vitro. The MA glands are extracted and cultured in a dish, preserving all structures intact, i.e. tail, ampulla, duct, spigot and the 2 µm diameter threads. Pulling the hanging fibers (the spinning trigger) produces new native spider silk. This explant first demonstrates that spider silk gland may produce material autonomously from the rest of the body and serves as in vitro model for this organ. Other material producing tissues could be studied with explants. As an example, developing a byssusproducing mussel foot explant could leverage existing culture media , growth factors , cell lines 503 and in vitro expression protocols . Other approaches can be considered to circumvent the "missing body problem". A material may be produced in a different organ with similar characteristics (e.g. ChemµEnv). For instance, growth factors and gene editing tools are available in silkworms but not in spiders. Thus, silkworms glands can be edited to produce a spider silk analogue 183 . Although the silk glands of spiders and silkworms have different phylogenies and embryogenic origins , their convergent evolution demonstrates material-processing similarities , enabling transgenic silkworms to produce exogenous materials . Edited spidroin, combined with a silkworm silk promoter, can be transfected in Bombyx mori's genome and sustained through generations, aiming to produce spider-silkworm hybrid fibers at industrial scales 183 . We call this approach "environmental transfer", for producing an exogenous material in an already similar organ. Rather than adapting a material to match a similar tissue, the opposite could be considered by editing an organ to produce the desired material. We refer to this as "environment morphism". This principle of changing the CellµEnv to affect a tissue's physiology is used in transplants 506 , tissue rejuvenation 244 and humanized mice 507 . A tissue can even change its anatomy following the addition of a single cell type, such as thickening skins following volar fibroblast transplantation 247 . Similarly, we suggest that self-microfabrication could eventually affect an organ's physiology to improve material processing. |
67ac29c881d2151a02169df6 | 30 | Although natural selection led to optimal biological systems, synthetic approaches can go beyond 222 . Biological machinery provides key insights to improve biological materials production, though it also has limitations that can eventually be overcome. For instance, spider silk properties can be improved with highly hydrophobic domains, but cells are incapable of synthesizing them as their increased hydrophobicity prevents the engineered proteins from passing the endoplasmic reticulum membrane 547 . Synthetic routes can reach normal inaccessible regions of the molecular epigenetic landscapes and create new optimal materials 548-551 (Fig. ). |
67ac29c881d2151a02169df6 | 31 | The recyclability of biological materials may also expose them to degradation by other organisms and environmental conditions 572 , eventually requiring protection layers to meet industrial standards. Interestingly, spiders create a natural coating on their silk 573 , limiting protease digestion 574 . Aquatic species even keep their silk dry with hydrophobic motifs 575 , inspiring self-assembled composite materials. Biological systems not only grow materials but also composite structures and living machinery. This opens new avenues for self-organized devices, such as insulin-producing organoids transplants for pancreatic applications 576 or high specific surface electrodes for batteries 577 . Taking inspiration from the electrocytes in electric eels, high density batteries 578,579 may eventually be self-organized using an organoid approach. Devices could also be grown by synthesizing multiple materials at different locations simultaneously. For instance, existing anatomies may be edited to express exogenous materials in specific tissues using cell-specific promoters . Biohybrid solar cells 582 may eventually be grown to harvest electricity from photosynthesis . Synthetic biology can also improve the carbon capture in algae 585 and could be applied to trees 586 for bioenergy and carbons storage 587 . These advancements raise ethical questions explored in the next section 588 . |
67ac29c881d2151a02169df6 | 32 | Material-producing organoids can not only produce better materials, but also improve manufacturing techniques. Bringing fabrication to microscopic scales has enabled pattern-driven material microstructures 589 and miniaturized devices, revolutionizing everyday lives 590 . For instance, photolithography has supported the semiconductor industry, telecommunication, modern computing 590 , and brought microelectromechanical systems (MEMS) 591 , sensors 592 and implants 593 . However, both manufacturing costs 594 and production volumes 595 are constrained by available technologies. In contrast, nature achieves complex selfmicrofabrication in large sizes and at low costs 596 through selforganization (Fig. ). Engineering material microstructure in bulk materials can enhance their mechanical properties, as demonstrated in hierarchical materials 597 . For instance, while chemically similar to chalk 598 , bivalve shells are up to 100 times tougher (3.3 to 9 |
67ac29c881d2151a02169df6 | 33 | MPa . m 1/2 of fracture toughness), thanks to their microstructure . At the nanoscale, CaCO3 crystal grains form microscopic tiles 598 , blocking fracture propagation and enhancing material properties 599 . Organic bridges measuring 1-5 nm highlight the competitive feature size of self-organized biological materials 599 compared to photolithography's limits 600 . Self-microfabrication has the potential to improve existing achievements, and introduce innovative manufacturing methods 601 , such as pattern-driven materials self-assembly. Applications of this approach 602 promise biological materials combining sustainability with high mechanical performance 592 . |
67ac29c881d2151a02169df6 | 34 | Polymers have shaped modern products, with copolymers diversifying applications by increasing the complexity of their components 603 . However, copolymers are typically limited to 2-5 monomers with simple assembly patterns, such as random and alternating blocks, or graft. Biology achieves higher complexity by freely assembling twenty monomers, along with noncanonical amino acids 604 , post-translational modifications and multimerization 605 . This combinatorial genetic code generates all proteins and biological materials, of the molecular landscape 606 (Fig. ). Biomimicry first explores naturally selected optima, or peaks on a molecular landscape, and applies them to engineer superior materials 222 using existing biological machinery and synthetic approaches 602 . Spatiotemporal control over chemical reactions and materials processing conditions is precisely regulated in biological systems, using nanoscale bioreactors, including membrane-less organelles 232 . More than materials, the biological machinery itself is selforganized from the bottom-up, maintained through homeostasis and self-repaired 229 . Using existing developmental trajectories, complex production units can emerge from lower scales of organization 215 . For instance, only a few genes in Drosophila initiate the complex development of a limb, while altering solely the bioelectric field in flatworms affects their whole body plan 241 . A single seed encodes the blueprint for complex plant development. Similarly, self-microfabrication seeks to grow intricate systems from minimal inputs , blending high-tech development with low-tech implementation ('high-dev lowtech'). By engineering the conditions of emergence , living machines self-organize and produce hierarchical biological materials 2 differently than conventional machines 608 . |
67ac29c881d2151a02169df6 | 35 | The complexity objection A key concern of self-microfabrication is its complexity, notably its need for developmental knowledge in non-model species. Some complex problems require equally complex solutions and may justify why material-producing organoids challenge the "Keep It Simple, Stupid" principle. Even with a simple example, Ashby's law of requisite variety demonstrates that a certain level of complexity is required to regulate a variety of disturbances 609 . Interest in biological materials stems from their capacity to collectively optimize multiple criteria that simpler synthetic methods struggle to solve ; including recyclability, raw material abundance, high mechanical performance, use of aqueous solvents, non-harmful chemicals and room temperature synthesis . Organoids could even simplify the manufacturing of complex systems by leveraging emergence and self-organization 607 , as discussed in the previous section with the high-dev low-tech approach. |
67ac29c881d2151a02169df6 | 36 | Most ethical concerns regarding organoids pertain to human medical applications and are not relevant for material production 610 . Concerns about consciousness do not apply to material-producing organoids lacking a central nervous system, but remain valid for brain-computer hybrids with advanced cognitive functions 554 . Animal experimentation should follow the 3R rule ("replace", "reduce", "refine") and selfmicrofabrication could eventually reduce the need for animals in some conventional farming, as intended for cultured meat 611 . |
67ac29c881d2151a02169df6 | 37 | We explore the idea of growing biological organs for material production. Resource scarcity and global warming present challenges in balancing sustainability with the demand for highperformance materials. Biological materials bridge this gap with versatile properties, available resources and biodegradability. We describe the spider silk gland to illustrate how organs regulate the self-assembly of hierarchical materials, giving rise to their properties. Similar organ production units can be cultured in vitro and self-organized from pluripotent stem cells, such as hair-bearing organoids. Growing tissues for material production, helps to approach organoid challenges from a new angle, notably the generalization of knowledge acquired from well-studied species to non-model organisms, such as arachnids. We propose a theoretical framework to help transfer information across biological scales and species. Finally, we discuss research opportunities to overcome biological limitations and produce optimal materials, using synthetic biology and biohybrids. |
67ac29c881d2151a02169df6 | 38 | Organoids can expand beyond medicine toward new applications, notably improving material microfabrication. Viewing organoids as living machines, and developmental biology as a manufacturing process, opens new avenues for self-microfabrication of next generation materials, such as spider silk. Self-organized organ production units, which can generate complex biological materials, have the potential to revolutionize the microfabrication industry by enabling the scalable production of large microstructured materials with unique properties. The prospect of achieving lower marginal costs for producing sustainable, high-value-added products creates opportunities for multidisciplinary collaborations among material scientists, tissue engineers, and developmental biologists, paving the way for innovative research trajectories and applications. |
67ac29c881d2151a02169df6 | 39 | Material produced by living organisms, such as proteins or polysaccharides, often referencing hierarchically microstructured materials like spider silk. Self-microfabrication Self-organized production units regulate the self-assembly of microscale structures from the bottom-up. Material-producing organoids such as the hair-bearing organoid 2 , are great examples. This self-organized organoid regulates the selfassembly of the hair shaft. |
67ac29c881d2151a02169df6 | 40 | The immediate surroundings of a cell, influencing its behavior, like growth factors, other diffusing biosignals, gradients, ECM composition, surrounding cells, mechanical stresses, bioelectric fields, etc. Biosignaling fields Field representation a CellµEnv's subset (e.g. scalar, vector, matrix or tensor in one or more dimensions). Homeostasis Tendency of a system toward a state of equilibrium, notably during the physiological or pathological maintenance of an organ. |
67ac29c881d2151a02169df6 | 41 | A cell that can self-renew and differentiate into other cell types. Potency Potency refers to a stem cell's ability to differentiate into different cell types. The more potent a cell is, the more types of cells it can become. Quiescence A dormant, non-dividing state of cells. Senescence Permanent cell growth arrest due to aging or damage. HOX genes HOX genes are a group of genes that determine the body plan and the identity of structures during embryonic development Simplicity A system with few components. Complexity A system with many interconnected parts. Emergence Complex patterns arising at a higher order (or scale) from simpler interactions from a lower order. |
67ac29c881d2151a02169df6 | 42 | Th epigenetic landscape metaphor is extended to multiple scales of biology, stacking them up, including molecular, cellular and tissular scales (Fig. 5-6); they may be referred to as "episystemic landscapes". At the tissular level the same concept can represent not cell differentiation trajectories but developmental routes and organogenesis. A location on such "epi-tissular landscape" represents a particular tissue, typically composed of multiple cell types, connecting the tissular layer to the cellular layer and so on (Fig. ). |
67ac29c881d2151a02169df6 | 43 | Concept combining a high-tech, or cutting-edge, technological research and development, with a simple, low-tech implementation. Low-tech implies simpler technologies requiring minimal tools or techniques. Self-microfabrication requires a complex development (e.g. studying organogenesis) but may eventually offer a simpler implementation by leveraging self-organization (e.g. planting a seed). Single-cell and multi-omics techniques Techniques integrating multiple types of molecular data at the single-cell resolution (such as genomics, transcriptomics, proteomics, and metabolomics). |
64831e0c4f8b1884b70f34ad | 0 | Energetic materials (EMs) are a class of compounds which, upon initiation, for example, by mechanical impact or heat, release a large amount of stored chemical energy. The development of new energetic materials is critical as EMs are a vital component of many industrial processes (e.g. mining and aero-space industries) as well as their many uses in the defence sector (e.g. propellants and explosives). However, the inherent dangers with synthesizing, handling and testing EMs presents a clear opportunity to develop computational methods to predict the structure and properties of new materials in advance of their synthesis, thus minimizing unnecessary experimentation. |
64831e0c4f8b1884b70f34ad | 1 | Inherent molecular properties are clearly important in the design of EMs, but the mutual arrangement of molecules in the material dictates how energy is transferred, and in turn also dictates important properties such as impact sensitivity and detonation velocity: this is evident from property variation between polymorphs of some EMs. Therefore, computational approaches for guiding the development of EMs should include the prediction of how EM molecules are arranged in the solid state. |
64831e0c4f8b1884b70f34ad | 2 | Crystal structure prediction (CSP) aims to predict all possible polymorphs for a compound given only its molecular structure. This is usually approached through a computational exploration of the lattice energy surface to identify all low energy local minima, which correspond to putative stable crystal packings of the molecule being studied. The structures associated with each local energy minimum are ranked according to their predicted energy, with the assumption that the most likely observable crystal structure corresponds to the lowest energy predicted structure -the global energy minimum. The two greatest challenges in CSP are the high dimensionality of the relevant energy surface, which creates a challenge in locating all possible crystal structures, and the small energies that typically separate competing predicted crystal structures, meaning that high quality models are required for evaluating the relative energies of predicted crystal structures. |
64831e0c4f8b1884b70f34ad | 3 | CSP can be applied prior to synthesis; hypothetical molecules can be assessed via CSP to decide whether they are worth synthesizing. For known molecules, CSP can help anticipate polymorphism, beyond what has been experimentally observed to-date. Thus, a CSP landscape is a powerful tool for anticipating the crystal structures of as-yet unsynthesized molecules and predicting possible polymorphic forms of known molecules. The field has seen rapid development in methods used for structural exploration and energy ranking, such that CSP has been succesfully applied in a range of applications, such as polymorph screening of active pharmaceutical ingredients, discovery of photocatalysts and porous materials. The present work aims to assess the performance of CSP for molecular organic EMs. Previous studies have demonstrated successful CSP methodologies for individual systems; for example, Bier et al. demonstrated a successful CSP methdology using a genetic algoritm for structure generation combined with solid state density functional theory (DFT) energy ranking which reproduced the known crystal structures of 2,4,6-trinitrobenzene-1,3,5-triamine (TATB) and 2,4,6-trinitrobenzene-1,3-diamine (DATB). Here, we validate the performance of a quasi-random structure search combined with an anisotropic atom-atom force field, as well as DFT reranking, for CSP on a broader set of ten EMs to assess the general applicability of CSP in the area of EMs. We also evaluate the sensitivity of force field predictions to the parameter set and electronic structure calculation from which electrostatics are derived. |
64831e0c4f8b1884b70f34ad | 4 | Ten known EMs, all with experimentally determined crystal structures, were chosen for validation of CSP methods. The test set of EMs studied is shown in Figure . Some of the validation molecules are known to exhibit polymorphism; the number of known polymorphs for each molecule is shown in parentheses in Figure . CSP was performed using the Global Lattice Energy Explorer (GLEE) package, which uses a low-discrepancy, quasi-random sampling of crystal packing variables to explore the lattice energy surface, followed by rigid-molecule lattice energy minimization using an intermolecular force field (described below). Molecular geometries of each molecule were optimized using DFT (PBE0/aug-cc-pVDZ, using Gaussian 36 ) starting from the conformations in their observed crystal structures. The test set contains mostly rigid molecules, so a rigidmolecule approach was taken for generating predicted crystal structures using the gas phase optimized minimum of each EM. Low energy predicted crystal structures were re-optimized using solid state DFT to account for molecular distortion due to intermolecular interactions in the crystal structures. We analyze the impact of DFT re-optimization on the quality of the predictions. |
64831e0c4f8b1884b70f34ad | 5 | HMX was the only EM in the set with more than one conformer present within the known polymorphs. The conformations differ in the geometry of the eight-membered ring and are sometimes referred to as chair (β, ϵ polymorphs) and chair-chair (α, δ) in discussions of HMX polymorphism. For HMX, CSP was performed on the chair and chair-chair conformers, both of which were optimized as described above. In the force field evaluation studies, these CSPs were treated separately due to the large conformational energy difference which exists between the two conformers of HMX. For the force field results, we compare total relative energies, calculated from the sum of intermolecular energies and relative conformational energies. |
64831e0c4f8b1884b70f34ad | 6 | DMACRYS was used with an anisotropic atom-atom force field energy model for all lattice energy minimizations. To test the sensitivity of CSP results to the force field, CSP was performed using two commonly used interatomic potentials (FIT 39 and W99rev 40 ), combined with four combinations of generation methods for the electrostatic model: B3LYP with 6-31G**, 6-311G**, 6-31G** with the polarizable continuum model (PCM ) and 6-311G** with the PCM applied. The PCM is used to mimic the in-crystal environment of a molecule. This provides a efficient, but approximate model of polarization contributions to the overall lattice energy. Distributed Multipole Analysis (DMA) has been used to provide a description of the intermolecular electrostatic interactions. Atomic multipoles, up to hexadecapole, were calculated using the Gaussian Distributed Multipole Analysis (GDMA) program, which analyses the charge density calculated for the gas phase molecular geometry to produce the multipoles. Each EM in the test set was examined with the following combinations of methods for the electrostatic model: HF, 49 MP2, 50 PBE, 51,52 PBE0 53 and B3LYP all with augcc-pVDZ and aug-cc-pVTZ. The aforementioned results with B3LYP and 6-31G** and 6-311G** are also included. |
64831e0c4f8b1884b70f34ad | 7 | Crystal structures were generated with one molecule in the asymmetric unit cell (Z' = 1) for each combination of EM, force field and electrostatic model. These structures were produced by sampling in the 10 most frequently observed space groups for organic molecular crystals (P 1, P 2 1 , C2, Cc, P 2 1 /c, C2/c, P 2 1 2 1 2 1 , P ca2 1 , P na2 1 , P bca). |
64831e0c4f8b1884b70f34ad | 8 | To analyse the success of each CSP, the predicted structures were compared to experimental structures using the COMPACK algorithm. COMPACK searches were carried out by comparing intermolecular atom-atom distances within a cluster of 30 molecules with 30% distance tolerances and 30 • angle tolerances and a root mean squared deviation in atomic positions (RMSD 30 ) was calculated for the overlay of predicted and experimental structures. |
64831e0c4f8b1884b70f34ad | 9 | For all 10 EMs, the CSP set of crystal structures generated with the PBE0/aug-cc-pVTZ multipole model was further optimized using plane-wave-based periodic DFT as implemented in the VASP package. For each EM, all structures on the respective CSP landscape within 10 kJ mol -1 of the global energy minimum were re-optimized. For ATZ, the energy window was increased to within 12 kJ mol -1 of the global energy minimum due to finding the experimental match just below 12 kJ mol -1 . This re-optimization was performed in a three-step procedure that has been found to improve the convergence rate of periodic DFT optimizations for crystal structures. The first step involves optimizing only the atomic positions with the unit cell fixed, the second step optimizes both atomic positions and unit-cell parameters, and the third step is a final single-point calculation which updates the plane wave basis set to the newly relaxed lattice parameters to provide an accurate final energy. |
64831e0c4f8b1884b70f34ad | 10 | All VASP calculations were performed using the PBE exchange correlation function with the GD3BJ dispersion correction. The projector augmented wave method was used for all VASP calculations with the standard supplied pseudopotentials. Following this optimization, a more stringent removal of duplicates was carried out by using the COMPACK algorithm for all structures within 10 kJ mol -1 of the global energy minimum for each CSP, irrespective of the space group of each structure. |
64831e0c4f8b1884b70f34ad | 11 | CSP is relatively under-explored for EMs. Because of this, we present an evaluation of the performance of different commonly used force fields and electrostatic models for CSP. The force field we use combines an empirically parameterized repulsion-dispersion model with an electrostatic model using atomic multipoles derived from a calculated electron density. |
64831e0c4f8b1884b70f34ad | 12 | Certain structures are very sensitive to the electrostatic model used for CSP. We have found that structures with high nitrogen content can be more sensitive to the electrostatic model than those without. Our initial CSPs on NTO, ATZ and TATB produced landscapes of varying quality using a typical electrostatic model; the RMSD 30 values (RMSD of atomic positions in a finite cluster of 30 molecules taken from the two crystal structures being compared) of the matches to the experimental structures covered a range from 0.186 Å to 1.004 Å. (W99rev), was also examined. |
64831e0c4f8b1884b70f34ad | 13 | FIT and W99rev are two force fields that are commonly used in modelling molecular organic crystals. Both of these force fields are parameterized against structural and energetic information from known crystal structures, and were developed to function in tandem with anisotropic atomic multipole electrostatics. Using atomic multipoles has clear benefits over using isotropic point changes: they result in a more accurate electrostatic model due to the difference in the quality of the representation of electrostatic features such as π -electron density and lone pairs, which leads to improved performance of structure prediction. The revised version of W99 has been parameterized for use with multipole electrostatics derived from a specific set of combinations of DFT functionals and basis sets: B3LYP/6-31G** and B3LYP/6-311G**. Both of these electrostatic models also have a variant that has been derived from the charge density calculated using a Polarizable Continuum Model (PCM). Applying PCM to the electrostatic model can be used as an approximate treatment of molecular polarization in crystal structure modelling. As discussed, our initial CSPs on ATZ and NTO proved unsatisfactory; the closest predicted match to the experimental structure for both EMs possessed a high relative energy and large geometric deviations. Thus, our testing for the choices of force field and electrostatic model involved performing CSPs on the 10 molecules shown in Figure for Z' = 1 structures. Results for eight out of the ten molecules are summarized in Figure ; ATZ and α-FOX7 are not included in this initial evaluation of force fields because we found that flexibility of the NH 2 groups has an important impact on the results for these crystal structures. |
64831e0c4f8b1884b70f34ad | 14 | Full CSP data tables for these box plots can be found in the supplementary information. Figure shows an example CSP result for 2-methyl-5-nitramino-2H-tetrazole (MNT), where a predicted structure matching the experimentally determined structure is identified as the second lowest energy structure (Figure ) and the structure reproduces the experimentally determined crystal structure accurately (RMSD 30 = 0.190 Å, Figure ). To evaluate the force fields, we examined how well they reproduced the experimentally determined crystal structures by measuring the RMSD 30 between the experimental structure and its match within each CSP set, and the relative energy of the match to the experimental structure. The best performing force field and electrostatic model combination is that which minimizes the RMSD 30 , which measures structural deviations between predicted and experimental crystal structures, and ranks the matches to the observed crystal structures well (low energies relative to the global energy minimum). No choice of force field emerges from this study as performing better than the others by both measures. The best force field at ranking energies of the observed crystal structures (W99rev + B3LYP/6-31G**(PCM)) produces the largest geometric deviations, whereas that producing the smallest range of RMSD 30 (W99rev + B3LYP/6-31G**) gives the worst overall performance on energy ranking. Without a clear optimum from these results, we proceed with the FIT force field, which produces a small range in RMSD 30 and low median relative energies in these tests, and performs best in previous benchmarks against measured sublimation enthalpies for a variety of small organic crystal systems. |
64831e0c4f8b1884b70f34ad | 15 | We also examined the sensitivity of CSP results to the charge density used to derive the electrostatic model by comparing the CSP results obtained from atomic multipoles generated by HF, MP2 and the DFT functionals PBE, PBE0 and B3LYP with the Dunning basis sets: aug-cc-pVDZ and aug-cc-pVTZ. 6-31G** and 6-311G** basis sets are present with B3LYP since they were used in the initial force field evaluation presented in the previous section. . For each electrostatic model, 11 total comparisons were performed between predicted crystal structures and experimentally known polymorphs. ATZ and α-FOX-7 are not included in these results due to the flexible NH 2 groups not being well modelled by the rigid-molecule calculations (as detailed in the SI). β-FOX-7 was included as the NH 2 conformation within the experimental structure matched the conformation within the DFT optimized geometry used in CSP. HMX has two conformers (referred to as chair and chair-chair), which were treated as separate CSPs in these summary results due to the large conformational energy difference. (a) the distribution of relative energies (the energy difference between the predicted match and the global energy minimum) for all 13 predicted matches to their respective experimentally known polymorph. (b) the distribution of RMSD 30 between for all 13 predicted matches to their respective experimentally known polymorph. |
64831e0c4f8b1884b70f34ad | 16 | The electrostatic models were evaluated analogously to the force fields, by comparing RMSD 30 and relative energies for predicted crystal structures that match the experimental structures across the molecules in the test set. HF performs surprisingly well on geometries, which might be due to cancellation of errors (HF exaggeration of charge separation could make up for the lack of explicit polarization in the force field method), but gives among the largest ranges for relative energies. The higher cost of MP2 calculations does not lead to better results than DFT-based electrostatic models. The results are similar among the DFT methods tested, showing that the average performance of CSP is not particularly sensitive to the functional and basis set from which the electrostatic model is derived. |
64831e0c4f8b1884b70f34ad | 17 | For comparison with the force field-based results, we also summarize the results of periodic DFT re-optimizations of the low energy crystal structures. To generate the DFT-based CSP results, all force field predicted structures within 10 kJ mol -1 of the global energy minimum for each of the ten EMs were re-optimized. For ATZ the energy window was increased to 12 kJ mol -1 as the experimental match was found at 11.46 kJ mol -1 ; ATZ is an outlier in our overall results, which we discuss further below. For HMX, due to the large molecular energy difference between chair and chair-chair conformers, we performed peroidic DFT on the lowest 10 kJ mol -1 for each conformer. FOX-7, HMX and NTO all possess an experimental polymorph with Z' > 1. Thus, to examine where these would occur in energy if CSP had been performed with higher Z', the experimental structures were independently optimized using the force field (FIT with PBE0/aug-cc-pVTZ multipoles) and periodic DFT and their numerical ranking and relative energies, after each stage of optimization, are shown in Table . Note that for these polymorphs, the numerical ranking and relative energies are lower bounds on what they would be if CSP searches had included high Z'. |
64831e0c4f8b1884b70f34ad | 18 | Table : CSP results for the landscapes of the 10 EMs shown in Figure produced using the FIT force field and PBE0/aug-cc-pVTZ atomic multipoles. The numerical ranking is the energy-ranked position of the match to the experimental structure within the list of CSP structures. Relative energy is the energy gap between each EM's predicted global energy minimum and the CSP match to the experimental structure. RMSD 30 represents the quality of our predicted matched structure when compared to the experimental structure: a low value corresponds to a higher quality match. RMSD 30 values were obtained via a COMPACK comparison using 30% distance tolerances and 30 • angle tolerances. For HMX, the two conformers possess a large intramolecular energy difference of 12.84 kJ mol - and thus for α-HMX and δ-HMX we report the numerical ranking and relative energy within the CSP of the individual conformer as a number in parentheses for the force field columns (DFT accounts for this intramolecular energy difference and thus does not the issue does not apply here). γ-FOX- |
64831e0c4f8b1884b70f34ad | 19 | The rigid-molecule force field-based CSP performs well at reproducing many of the known crystal structures with low RMSD and good energetic ranking, although the results are more variable than when similar methods are applied for non-EM molecular crystals. This demonstrates the power of a force field based model as a simple and fast method which can produce accurate predictions for crystal structures. Geometrically, the known crystal structures are reproduced well, with small RMSD in atomic positions. Only ATZ, DNIT, HBT and β-NTO have RMSD 30 values above 0.5 Å. |
64831e0c4f8b1884b70f34ad | 20 | Energetically, the rigid-molecule force field CSP results are mixed. A main assumption of CSP is that observed crystal structures will correspond to low energy structures; a perfect prediction corresponds to the known structures matching the global minimum (rank = 1) structure, or the lowest set of structures for polymorphic molecules. We observe excellent rankings for several of the EMs (ABT, HBT, HNB, MNT, TATB and the β and ϵ polymorphs of HMX, which adopt the lower energy chair conformer). However, several of the matches to known crystal structures are energetically ranked poorly at the force field level, which we attribute to the rigid-molecule approximation more than the quality of the intermolecular force field itself. |
64831e0c4f8b1884b70f34ad | 21 | The relatively poor energetic rankings for ATZ and FOX-7 are likely related to the flexibility of the NH 2 groups causing mismatches between our gas-phase optimized geometry and the molecular geometry within the experimental crystal structures. For HMX, the α and δ polymorphs are ranked poorly. Both of these polymorphs contain the chair-chair HMX conformer and although α-HMX corresponds to the lowest energy predicted structure from the chair-chair conformation, it is far above the global minimum in total energy. The HMX results show that the mixed force field intermolecular energy + DFT conformational energy model does not work well in this case. |
64831e0c4f8b1884b70f34ad | 22 | The poor results for NTO are not fully understood. Although the choice of force field and electrostatic model was not found to influence the average CSP performance across all EMs, we observed that the results for individual molecules could be sensitive to the choice of force field. For example, CSP using the W99rev force field with B3LYP/6-311G** provided a match to the experimental β-NTO crystal structure with an RMSD 30 of 0.281 Å (see SI for more detail) and a relative energy of 1.98 kJ mol -1 . |
64831e0c4f8b1884b70f34ad | 23 | Of the EMs that performed poorly in the rigid-molecule force field CSP, results for NTO are improved significantly, ranking the α and β polymorphs as 3rd and 2nd lowest energy structures, ranking of the FOX-7 polymorphs is improved such that the known polymorphs are 2nd, 3rd and 32nd lowest energy structures, and HMX predictions now place the four known polymorphs as 1st, 2nd, 3rd and 9th ranked in energy. The HMX results demonstrate that the high molecular energy of the chair-chair conformation can be compensated by improved interactions in the solid state. The energy range of all of the polymorphic EMs falls well within the range normally seen for organic molecular polymorphs. Only three out of the 16 experimentally observed crystal structures being ranked outside the three lowest energy predictions with γ-FOX-7 and ATZ being worst ranked, as 32nd and 113th ranked structures by energy after DFT re-optimization. |
64831e0c4f8b1884b70f34ad | 24 | The only EMs with RMSD 30 values above 0.4 Å after DFT re-optimization are ATZ and β-NTO. These higher RMSD 30 matches are still plausibly close matches by visual inspection, as displayed in Figure and Figure respectively. For comparison, the quantitatively much closer agreement for ABT is also depicted in Figure . |
64831e0c4f8b1884b70f34ad | 25 | Although solid state DFT re-optimization should allow the hydrogens to find their energetically most favourable orientations within each crystal structure, we repeated CSP starting with a molecular geometry constrained to be planar. These constraints made minimal difference to the final CSP predictions (see SI Table for more details); the known structure remained high in energy on the resulting CSP landscape. |
64831e0c4f8b1884b70f34ad | 26 | To test whether the CSP structure generation had failed to locate the local energy minimum nearest to the observed crystal structure, we performed lattice energy minimization (using FIT multipoles) on the experimentally determined crystal structure after replacing the molecular geometry by the DFT gas phase optimized molecule. The obtained structure was virtually identical to the predicted structures obtained from CSP. This result confirmed that the structure found during CSP is the nearest local energy minimum to the observed crystal structure. The same test was performed with β-NTO, with the same result. For both molecules, further sampling during CSP structure generation would not find a better match to the observed crystal structure. |
64831e0c4f8b1884b70f34ad | 27 | The structure report for ATZ had some uncertainty in the hydrogen position defining the tautomer, as the structure was determined from powder diffraction data. Our initial CSP assumed the 1H-tautomer, as shown in Figure . We ran CSP with the 2H-tautomer of ATZ (see SI for further details) to determine if the alternative tautomer would lead to a lower energy or closer geometric match to the experimental solution; the known structure forms hydrogen bonds between the N1 and N2 positions in the ring, so the 2H-tautomer could plausibly form the same crystal packing with the H atom shifted across the hydrogen bond. CSP with the 2H-tautomer does produce matches to the packing of the known crystal structure. However, after DFT re-optimization (PBE-GD3BJ), the energies of these structures were found to be more than 11.5 kJ mol -1 above the matches found with the 1H-tautomer (see Figure ). Thus, including the alternative tautomer of ATZ does not improve the CSP results for this molecule. |
64831e0c4f8b1884b70f34ad | 28 | As a final test, we tested whether the poor energy ranking of the ATZ crystal structure is a limitation of the GGA functional used in DFT re-optimization of structures. 11 CSP structures, including the global energy minimum, of ATZ were selected from the DFT reoptimized set, evenly sampling the low energy range. Single-point PBE0 calculations were performed with the GD3BJ dispersion correction to assess the extent of re-ranking by changing to a hybrid DFT functional. The reranking was found to be up to a few kJ mol -1 (see Figure ), which lowered the relative energy of the match to the experimental structure to 5.65 kJ mol -1 (among this subset of structures). PBE0 energy calculations did not significantly affect the relative energies of structures with the 2H-tautomer compared to the 1H tautomer. These results hint that the outlier results for ATZ could be due to limitations of GGA functionals, which are known to suffer from self-interaction error. However, it is plausible that the reported experimental crystal structure of ATZ is in fact a meta-stable polymorph and the true thermodynamically stable polymorph is present within our predicted landscape. |
64831e0c4f8b1884b70f34ad | 29 | We have evaluated crystal structure prediction by global exploration of the lattice energy surface on a set of ten highly energetic organic molecules. The computational methodol-ogy applied a rigid-molecule approach using an empirical force field with atomic multipole electrostatics, followed by re-optimization of the lowest energy predicted structures using dispersion-corrected solid state DFT. The results demonstrate that this is a reliable approach for structure prediction of this class of energetic materials. |
64831e0c4f8b1884b70f34ad | 30 | The rigid-molecule, force field-based CSP approach provides useful results. Matches within the CSP results are found for all of the known Z'=1 crystal structures and, for four of the ten molecules, the predicted global energy minimum corresponds to the known crystal structure, or one of the known polymorphs. Many of the other experimentally observed crystal structures correspond to predicted structures with low numerical ranks in the energyordered sets of predictions. Therefore, this approach can act as a first stage in CSP, which provides a small set of structures for re-optimization at a higher level of theory. |
64831e0c4f8b1884b70f34ad | 31 | Solid state DFT re-optimization of CSP structures produces excellent CSP results; of the 13 target experimentally-determined Z'=1 crystal structures of the ten molecules, 10 are reproduced with excellent quality, within the 3 lowest energy predicted structures, and with RMSD 30 to the experimental structure below 0.4 Å. Two of the other structures showed either slightly higher energy (δ-HMX) or worse geometric agreement with the experimentallydetermined structure (β-NTO), and the known crystal structure of ATZ remained poorly predicted after DFT. We believe that the result for ATZ is due, at least in part, to limitations of the GGA DFT functional. |
64831e0c4f8b1884b70f34ad | 32 | We envision that CSP could be carried out in future to guide screening efforts for polymorphs of energetic molecules. The reliability of structure prediction methods also provides a route to the anticipation of materials properties in advance of molecular synthesis, which could be achieved by combining CSP with methods for evaluating, for example, impact sensitivities. |
64c0f1de658ec5f7e5324bbc | 0 | Electrochemical biosensors have the potential to transform biomarker analysis by replacing bulky laboratory equipment with portable devices. This transition would allow routine clinical tests to be performed in primary care settings, facilitating early disease detection and timely intervention. Over the past decade, activities have shifted towards fully-integrated microsystems, with increasing research efforts dedicated to developing low-cost, disposable devices made from accessible materials, such as plastic or paper. Unfortunately, the fabrication of the reference electrode in these low-cost electrochemical cells remains a significant and unmet challenge. A typical electrochemical detector includes a reference electrode that acts as a reference potential for the working electrode. To ensure measurement stability, the reference electrode is usually made by a standard material with a known redox potential, such as silver/silver chloride or the saturated calomel electrode. However, these materials are often incompatible with common microfabrication techniques or demand complex fabrication processes, thus hindering their integration into low-cost disposable cartridges. Circumventing the overreliance on reference electrodes would enable researchers to design electrochemical sensors using a far broader array of substrates, and thus pave the way towards simpler and more accessible diagnostic technologies. A popular approach is to use an uncoated electrode material, such as carbon or gold, as the reference electrode (i.e. pseudo-reference electrodes). However, because the electric potential of pseudo-reference electrodes strongly depends on their chemical environment, this approach fails to deliver a consistent reference potential in continuous flow detection assays. As a result, the integration of dedicated silver/silver chloride electrodes remains the norm in flow injection analysis (FIA), hindering the fabrication of low-cost disposable devices. FIA is a widely used analytical technique that enables reliable and high-throughput testing of multiple samples by injecting them into a continuous flow; its rapid and automated nature makes it particularly attractive for clinical chemistry. In the realm of point-of-need testing, paper-based FIA devices have attracted considerable attention due to low fabrication costs and their passive, pump-free operation. Because of the difficulty of fabricating electrodes in cellulose paper, early attempts to incorporate electrochemical sensing into paper-based FIA relied on the use of external electrodes in contact with the fluidic path. Such a strategy essentially undermines the advantage of paper as a single-use substrate, since the external electrodes are reused multiple times, raising the risk of crosscontamination. To address this shortcoming, we recently developed laser-pyrolysis of cellulose as a versatile technology for fabricating electrodes fully embedded into paper. Despite their advantages, the electrochemical sensors fabricated in this manner still relied on a three-electrode configuration, with a laser-induced graphenized (LIG) pseudo reference, resulting in signal aberration in continuous-flow assays. The adoption of two-electrode, or reference-free, electrochemical sensors, has the potential to address many of the challenges associated with the use of pseudo-references in FIA. Moving to two-electrode systems would substantially simplify device fabrication and readout hardware by eliminating the need for a reference potential altogether. Unfortunately, due to their unconventional nature and our limited insight of their workings, two-electrode systems remain underexplored and are often used empirically without rational design. Accordingly, understanding the inherent limitations of these systems is essential to harnessing their full potential and ensuring widespread adoption. Here, using paper-based laser-pyrolyzed devices as a model system, we demonstrate that two-electrode detection can provide excellent performance in FIA cartridges for clinical analysis. As proof of concept, we designed a cartridge to perform urinary creatinine analysis, an essential diagnostic assay in need of miniaturization and automation. Urinary creatinine serves as a key biomarker for kidney function, as its production is steady and proportional to a healthy individual's muscle mass. It is also widely used as a normalizing factor in urine assays, such as the albumin-to-creatinine ratio, or as an internal control for sample collection adequacy. The gold-standard method for quantifying urine creatinine remains the Jaffe reaction, which is a colorimetric reaction between creatinine and picric acid developed in 1886 (Fig. ). However, this assay has limitations, e.g. cross-reaction with glucose, and is not suitable for miniaturization and high-throughput analysis due to the need of bulky and expensive optics for signal readout. A number of electrochemical approaches for creatinine analysis have been proposed, including those based on iron ion mediators, picric acid detection, or direct detection catalyzed by copper nanoparticles. Unfortunately, these approaches suffer from slow electron transfer, large redox potential or demanding tedious electrode modification, respectively. Thus, they are unsuitable for use in low-cost disposable devices. |
64c0f1de658ec5f7e5324bbc | 1 | When searching for a more suitable detection system, we discovered that reactions between ferricyanide complexes and creatinine, which were discovered in the 19 th century (Fig. ), remained relatively unexplored within electrochemical systems, despite the attractive features of ferricyanide (fast, reversible, wide electrode compatibility) for low-cost systems (Fig. ). Accordingly, we investigated the ferricyanide-mediated electrochemical detection of urinary creatinine and its implementation in paper-based FIA with two-electrode detectors. We first explored the working range and limitations of the two-electrode detectors, and then optimized their sensing performance for urinary creatinine. Finally, we demonstrated high-throughput testing of clinical urine samples using disposable cartridges and a low-cost reader. |
64c0f1de658ec5f7e5324bbc | 2 | To develop our FIA assays we utilized two different device designs. Both comprise a paper-based electrofluidic layer with an embedded graphenic two-electrode detector made up of working and counter electrodes (WE and CE) (Fig. ). The simplest device, henceforth referred to as the "static analyzer", allows for a single measurement of one sample, introduced through a wide pipetting window (Fig. ). |
64c0f1de658ec5f7e5324bbc | 3 | We used this device to examine the fundamental properties of two-electrode detection systems and to optimize the electrochemical detection of the ferricyanide-mediated reaction with creatinine. We additionally designed a flow injection device for analyzing multiple clinical samples in a continuous flow format. This device relies on a flow of carrier buffer from a reservoir, through the electrofluidic channel before reaching a large-capacity absorbent pad (Fig. ). In this configuration, samples can be sequentially injected upstream of the detector, offering high throughput and continuous detection of multiple samples. |
64c0f1de658ec5f7e5324bbc | 4 | The theory and design principles for electrochemical sensors have been primarily developed for threeelectrode configurations and thus were not entirely applicable to our proposed system, Thus, before focusing on our target analytes, we aimed to gain some insight into two-electrode electrochemical detection, including its working range and limitations. To this end, we first evaluated the optimal voltage for the detection of ferrocyanide by measuring sampled-current voltammograms in the presence or absence of the redox probes (Fig. ). This allowed us to extract the faradic current-overpotential response by subtracting non-faradic contributions (Fig. ). The response resembles that of a characteristic Nernstian system, revealing fast and reversible electron transfer kinetics and diffusion control. In subsequent measurements, we chose an applied voltage of 0.35 V corresponding to approximately 90% of the maximum faradic current attainable, with a small degree of non-faradic contribution (~25%). Compared to a typical three-electrode format, the applied voltage in the twoelectrode system is larger because it represents the voltage difference between the WE and CE, rather than that between the WE and the reference electrode (Fig. ). It can be seen that when the ferrocyanide ([Fe II (CN)6] 4-) concentration varies relative to ferricyanide ([Fe III (CN)6] 3-), whilst maintaining a total concentration of 250 mM, two distinct regimes in the current responses exist (Fig. ). Up to a ferrocyanide concentration of 125 mM, the oxidation signal is linearly proportional to its concentration and consistent with the Cottrell equation for a conventional cell (𝑖 = 𝑛𝐹𝐴𝐶 0 𝐷 0.5 𝑡 -0.5 𝜋 -0.5 , where i and t are current and time, respectively, n the number of electrons transferred, A the surface area of the WE and C0 the analyte concentration with diffusion coefficient D). Beyond 125 mM, a sudden current drop is observed, owing to the deficiency of ferricyanide available at the CE for reduction, and thus unable to absorb the electrons generated at the anode. That said, if we consider the complete chronoamperometric curves (Figs. 3d, S2 and S3), responses appear to deviate from the classical Cottrell equation, namely i ∝ t -0.5 . Indeed, there exists a noticeable delay in response at high ferrocyanide concentrations. Further modelling efforts are required to fully explain this observation. These results demonstrate the importance of considering the linear regime of twoelectrode detectors, as it will ultimately determine the upper detection limit for a given application. |
64c0f1de658ec5f7e5324bbc | 5 | Despite a long history, the reaction of creatinine with ferricyanide remains relatively unexplored, and mechanistic studies and parametric optimization are lacking. This knowledge gap exists because the reaction is difficult to incorporate into colorimetric analysis since ferrocyanide requires further reaction steps to generate a measurable color change, e.g., with Fe 3+ to form Prussian blue (Fig. ). By removing this second step, the two-electrode detector format presented here offers an attractive platform for studying the kinetics and mechanisms of this reaction. |
64c0f1de658ec5f7e5324bbc | 6 | Extreme pH values resulted in the undesired formation of Prussian blue (Fig. ). At pH = 5, we fitted the current responses using i = imax (1-exp(-kobs•t)), where the fitting parameters of imax and kobs correspond to the maximum current and observed reaction rate constant, respectively, and found that the fitted kobs values are nearly independent of creatinine concentration, indicating pseudo-first order kinetics (Fig. ). The observed reaction rates increased with temperature following the Arrhenius equation (Figs. and), with an activation energy, Ea, of 10.9±0.4 kJ mol -1 and a pre-exponential parameter, A, of 26.0±1.4 s -1 . Using these values, we created a reaction yield map as a function of time and temperature (Fig. ). For example, at 65°C, one can reach a reaction yield of >95% after 25 minutes. We then carried out mechanistic analysis of the reaction, starting with an assessment of the role of ferricyanide in the rate-limiting step (Figs. and). The observed reaction rates were linearly proportional to the initial ferricyanide concentration (the first-order contribution), indicating that the rate-limiting step involves a bimolecular reaction between creatinine and an excess of ferricyanide (Fig. ). We also investigated the impact of creatinine concentration (Fig. ). The response reveals a linear dependence up to 27.5 mM, beyond which the current suddenly drops due to CE limitations. By correlating the sampled current values in the linear region with the ferrocyanide oxidation signals calibrated in Fig. , we estimate that on average 3.80±0.11 ferricyanide molecules reacted with one creatinine molecule (Fig. ). In other words, the total reaction corresponds to the oxidation of two chemical bonds per creatinine molecule, leading to the reduction of four ferricyanide molecules. In light of these findings, we hypothesize that the kinetics involve four sequential steps in which one creatinine molecule is progressively oxidized by one ferricyanide molecule at each step. The analysis presented here not only identifies the optimal reaction conditions, but also offers a rationale for the system limitation. More specifically, one must keep the ferricyanide concentration at least eight times (2×4) higher than the maximum creatinine concentration to be detected, taking into account the dilution factors when implemented in a sample matrix (such as urine). |
64c0f1de658ec5f7e5324bbc | 7 | Electrode design plays a critical role in determining the electrochemical performance, since signals are generally proportional to the electroactive area and the electrode arrangement can enhance signal quality by mitigating diffusion and migration pathways. This is particularly true for two-electrode detection schemes, as both the CE and WE design will affect overall detection efficiency, and our results suggest that there is a subtle balance between signal strength and sufficient counter reaction. |
64c0f1de658ec5f7e5324bbc | 8 | Using the static analyzer, we evaluated sensor performance on artificial urine samples spiked with creatinine, whilst varying the CE:WE area ratio (Figs. 5a, 5b and S9). Interestingly, although increasing the relative WE area amplifies the signal and lowers the detection limit (Fig. ), response linearity and analytical sensitivity are gradually lost due to the insufficient CE size (Fig. ). We found that a CE:WE of 2:1 showed optimal performance, with a detection limit of 0.56±0.07 mM and an analytical sensitivity of 4.83±0.25 μA mM -1 . Note that in addition to magnitude, the standard deviation of the sensitivity values indicates the response linearity of the system. When compared to the static analyzers, paper-based flow-injection analyzers offer a number of key advantages (Figs. 5e, 5f and S10). First, they allow the injection of multiple samples into a single device (at rates in excess of 1 sample per minute), substantially reducing reagent consumption (carrier flow rate of 7-12 μL min -1 , Fig. ) and sample volumes (down to 1 μL, Fig. ). Second, the device-to-device variation is eliminated since all measurements can be performed on a single device, leaving only variations resulting from the injection process, which is approximately 6 % in our FIA devices (Figs. and). Additionally, the signals (unit: Coulomb) are quantified by integrating currents over time as the sample flows through the electrode, thus averaging the noise from the measurement readout. Surprisingly, and despite the weaker signals generated by the smaller WEs, we observed detection limits below 0.32 mM regardless of the relative electrode area (Fig. ). This is due to ability of the FIA device to detect small deviations from the steady baseline established by the carrier flow. On the other hand, the sensitivity shares a similar trend with that of static analyzers, showing a reduced sensitivity with decreasing CE area (Fig. ). |
64c0f1de658ec5f7e5324bbc | 9 | The continuous flow sensor delivers a detection limit of 0.23±0.08 mM and sensitivity of 0.118±0.003 mC mM -1 up to 50 mM creatinine, with the concentration dependence showing only a 2.5% deviation from perfect linearity. We attribute the remarkable FIA performance to the high efficiency of in-flow detection with our integrated graphenic electrodes. More specifically, the porous and embedded nature of our laser-pyrolyzed electrodes, together with the fast kinetics of ferrocyanide at the interface (k0 = 0.011±0.003 cm s -1 ), result in overall detection efficiencies of up to 56 % (Fig. ). If we further account for the partial coverage of the WE in the channel, the observed efficiencies reached up to 90 % of the maximum attainable values, meaning that up to 90 % of the redox probes flowing through the WE were successfully detected. |
64c0f1de658ec5f7e5324bbc | 10 | It is also noteworthy that when two electrodes are placed in a parallel arrangement, the response time is significantly shorter than for sequential designs, which results in undesirable redox cycling (Figs. and). In summary, we have taken advantage of the rapid-prototyping capability offered by the laserpyrolyzed electrodes to investigate and optimize electrode design in our devices. Our findings reveal that the electrode areas and orientation directly affect the response in two-electrode detection. In FIA, the two-electrode detectors not only enable high-throughput sampling but also showcase improved sensing performance and high detection efficiencies. |
64c0f1de658ec5f7e5324bbc | 11 | A major advantage of FIA in testing clinical samples with unknown creatinine content is that the device does not require prior calibration. Instead, it incorporates internal calibration and control measurements that effectively remove the detrimental analytical effects caused by device-to-device or reagent variations. To examine the reliability and accuracy of our two-electrode FIA devices, we analyzed nineteen clinical urine samples collected in veterinary settings to determine their creatinine content and benchmarked our results against a commercial colorimetric assay (Table , Fig. ). A typical set of measurements starts with the injection of five standard creatinine solutions with known concentrations, followed by the urine samples on the same device (Fig. ). Using the internal standards, we directly determined the creatinine content in each clinical sample (Figs. and) which ranged from 7 to 47 mM, covering the entire clinical (high to low) range. Comparing our results to those benchmarked with the gold-standard commercial colorimetric kit using a laboratory plate reader (Fig. ), our paper-based two-electrode cartridges exhibit excellent agreement, with an average absolute deviation of only 11% (R 2 =0.95; Fig. ). In addition, we examined a selected panel of possible interferences at abnormally high concentrations (Table ) and apart from uric acid (+18 %), found insensitive response for all interfering compounds (<10 % signal deviation, Fig. ). |
64c0f1de658ec5f7e5324bbc | 12 | It is noteworthy that the real-time measurement of our FIA assay can quickly identify abnormal sampling. For example, in Fig. , sample #5 was unusually viscous and became pinned to the injection hole. The absence of signal allowed us to quickly identify the failure and resolve it by testing at a later stage, without interrupting the analysis workflow. In summary, our low-cost two-electrode cartridge demonstrated accurate high-throughput testing of urine samples in a clinical scenario. |
64c0f1de658ec5f7e5324bbc | 13 | The development and implementation of our two-electrode systems not only greatly simplify the fabrication process, but also leads to simplified electronic circuitry for driving and reading the electrochemical transducer. Indeed, due to the fixed potential of the CE, it is not necessary to variably adapt any potentials relative to a reference electrode, as is the case in three-electrode systems. This means fewer components are required to create and operate these two-electrode systems. Since open-source three-electrode potentiostats require at least $40 worth of parts, we see enormous potential for two-electrode systems in significantly lowering the costs associated with creating accessible signal readers. To this end, we designed a reader using only one operational amplifier to apply a constant voltage (361.76±0.05 mV) and monitor the current with 0.2 μA resolution and a range up to 211 μA (Figs. 7a, S18 and S19). The current range (with constant 0.1 % resolution) can be tuned by swapping a single resistance and tailored to a specific application. The system is controlled by an inexpensive microcontroller, resulting in a total part cost of less than $3 (Table ). Remarkably, we can harness the computing power of the reader to not only drive the transducer, but to also analyze results in real time and guide the user through the measurement process using a multicolor LED (Fig. , Supplementary Video 1). In a typical experiment, the reader remains idle until it detects the flow front from the carrier buffer (Figs. and) because of a change in resistivity. It then proceeds to detect the peaks and notifies the user when the current has sufficiently decayed, and the cartridge is ready for the next injection. We also incorporated a stop criterion, which currently simply relies on a time limit, but could detect critical system failures in the future. Remarkably, and despite its simplicity, the $3 reader delivers performance comparable with high-precision benchtop potentiostats (Fig. ). Furthermore, the smart reader streamlines the testing process for operators, and offers the flexibility of standalone operation or realtime communication of results to a display, laptop or smartphone via a USB connection (Fig. ). |
64c0f1de658ec5f7e5324bbc | 14 | In conclusion, we have demonstrated remarkable performance of two-electrode electrochemical detectors integrated in low-cost clinical analyzers. The two-electrode detectors are ideally suited for paper-based devices, where the fabrication of reference electrodes is cumbersome. The two-electrode sensors also excel at flow injection analysis, to their high detection efficiency and reliable sensing performance. We have confirmed the critical role of the counter electrode design, in order to ensure a sufficient number of redox probes for the reverse reaction. Importantly, the two-electrode system is capable of handling the entire development workflow of a novel assay. As a model system, we successfully developed an FIA assay for urinary creatinine, providing unprecedented insights into the ferricyanide-mediated reaction, and offering clinical performances comparable with commercial laboratory assays, at almost zero material cost. The cost competence is even further strengthened with our $3 smart reader owing to the simplified circuitry for driving two-electrode systems. We believe that our findings will be readily translatable to other ferricyanide-mediated assays such as glucose or lactate, and anticipate that the ultralow cost paper-based cartridges presented here will facilitate the development of quantitative assays at the point of need. |
63f39b8ffcfb27a31f1dc775 | 0 | Mitigation of anthropogenic CO2 accumulation is essential to tackle the current climate change and loss of biodiversity. Large scale global efforts are ongoing to develop CO2 conversion technologies for green fuel production. Solar-driven CO2 conversion is a promising approach to produce clean fuels and chemicals as it directly utilizes sunlight as the sole energy input. However, current CO2 utilization processes depend on pure and pressurized CO2 as reactant, whose production from post-combustion emission streams and air is energy intensive (~2 GJ tonCO2 -1 or 100 kJ molCO2 -1 ). Majority of this energy demand (~80%) is from desorption and compression steps following CO2 capture, involving heating large volumes of alkanolamine solutions and subsequent pressurization of the released gas (Figure ). Direct solar-driven utilization of captured CO2 species is therefore a more attractive way to reach net-zero carbon cycle, but barely explored due to their thermodynamic stability. Thermo-catalytic hydrogenation of captured CO2 has recently been reported at elevated temperatures (100-150 ºC). Electrochemical reduction of alkanolamine captured CO2 can also be achieved over metallic electrodes (Ag, Cu, etc.) at ambient temperatures. Recent studies have shown smallscale localized in situ CO2 gas adsorption followed by photocatalytic conversion in homogeneous and colloidal solutions. Nevertheless, sunlight-driven direct capture and utilization processes with industrially relevant amine/hydroxide capturing agents is lacking, presumably due to the overwhelming energy barrier to activate the trapped CO2-adducts (Figure ). Here we report an integrated CO2 capture and solar-driven photoelectrochemical (PEC) utilization process to produce syngas (mixture of CO and H2), a precursor for industrial liquid fuels and chemicals syntheses, from concentrated CO2 stream, simulated post-combustion flue gas, and atmospheric air. The process operates by combining CO2-to-fuel reduction with selective oxidation of waste plastic-derived ethylene glycol (EG) to glycolic acid (GA), which has applications in pharmaceutical, food and textile industries (Figure ). The system captures CO2 in an aqueous amine or glycolic hydroxide solution and the subsequent PEC conversion occurs in a two-compartment, two-electrode reactor equipped with a triple cation perovskitebased photocathode. Captured CO2 reduction is enabled by an immobilized molecular Cophthalocyanine catalyst at the photocathode. A bimetallic Cu26Pd74 alloy anode completes the circuit by catalyzing EG oxidation. Replacing anodic water oxidation (G 0 (H2O/O2) = +237 kJ mol -1 ) by EG oxidation (G 0 (EG/GA) ~ +20 kJ mol -1 ) makes the demanding captured CO2 reduction feasible with only sunlight, enabling the system to function even with a single visiblelight absorber without any external applied voltage, with simultaneous waste valorization. , where CO2 is first captured from post-combustion gases (step 1), followed by heat treatment to desorb CO2 (step 2). The released CO2 is then compressed and subsequently used for conversion (steps 3 and 4). b, Integrated one-step photoelectrochemical (PEC) approach reported in this study, where the post-capture solution is directly used for conversion using solar energy, to generate syngas as energy vector while at the same time upcycling plastic waste derived ethylene glycol to the commodity chemical glycolic acid. The abbreviations for the individual photocathode layers are in the Method section. |
63f39b8ffcfb27a31f1dc775 | 1 | A concentrated CO2 stream (99.995%) was first used to develop and optimize the PEC system. Different amines, including industrially-relevant monoethanolamine (MEA), diethanolamine (DEA), triethanolamine (TEA) and diazabicyclo[2,2,2]octane (DABCO) were employed for CO2 capture in aqueous medium at ambient temperature. The capture was done by purging concentrated CO2 through 1 M amine solution for 2 h (flowrate: 30 mL min -1 ). The solution was further purged with N2 (15 min) to remove any physically dissolved CO2 (Figure ). C nuclear magnetic resonance ( 13 C-NMR) spectra of the post-capture solutions revealed that under these conditions, MEA and DEA captured 0.75±0.07 and 0.77±0.10 mol CO2 per mol amine, respectively, as bicarbonate and carbamate species. In contrast, the tertiary amines TEA and DABCO captured CO2 only as bicarbonate salts (0.60±0.05 and 0.85±0.08 mol CO2 per mol amine, respectively, after 2 h; Scheme 1, eq 1-2; Figure ). Other than amines, aqueous and organic solutions of NaOH were also used for efficient CO2 capture (Scheme 1, eq 3). When concentrated CO2 was purged through an aqueous 1 M NaOH solution (2 h), NaHCO3 was formed quantitatively. Similarly, 1 M NaOH solution in organic EG captured 0.96±0.02 mol CO2 per mol NaOH as sodium glycol carbonate upon CO2 purging (Scheme 1, eq 4; Figure ). The CO2 captured species thus obtained were subsequently for conversion. |
63f39b8ffcfb27a31f1dc775 | 2 | ). Electrodes were prepared by first immobilizing CoPcNH2 on multi-walled carbon nanotubes (MWCNT) through π-π stacking, followed by drop-casting the composite on a graphite foil substrate (CoPcNH2@MWCNT; Figures ). Electrochemical reduction of aqueous captured CO2 solutions were then explored in a two-compartment, three-electrode configuration with the fabricated CoPcNH2@MWCNT electrode, a Ag/AgCl (sat. NaCl) electrode, and a Pt mesh as the working, reference, and counter (water oxidation) electrode, respectively. The catholyte was CO2 captured solutions with 0.1 M K2SO4 (supporting electrolyte; pH 7.8-8.3), the anolyte was 0.1 M K2SO4 (pH 7.6), and the compartments were separated by a bipolar membrane. Cyclic voltammetry (CV) scans in a TEA (1 M) captured CO2 medium in this setup showed an onset potential of -0.4 V vs. the reversible hydrogen electrode (RHE; Figures ). Subsequent controlled potential electrolysis (CPE) at different potentials (1 h) produced only CO and H2 as products (Figure ). The formed bicarbonate C-O bond during CO2 capture is thus cleaved during reduction, regenerating the amine. An optimum CO faradaic efficiency (FECO) of 46.2±2.0 % was obtained at -0.7 V vs. RHE (Figure ). CPE with the primary and secondary amine (MEA and DEA) captured CO2 solutions at this potential resulted in lower FECO of 10.2±1.7 and 16.5±1.5%, respectively (Figure ). Control experiments without an amine did not capture CO2 and consequently, produced no CO during CPE, confirming the role of amine in the CCU process (Figure ). |
63f39b8ffcfb27a31f1dc775 | 3 | Electrochemical reduction of aqueous NaOH captured CO2 species (NaHCO3, Na2CO3) at -0.7 V vs. RHE showed poor FECO (<3%, Figure ). Reduction of the captured CO2 in glycolic NaOH solution (Scheme 1, eq 4) was hence explored, after adding tetrabutylammonium tetrafluoroborate (TBABF4, 0.15 M) as supporting electrolyte and 20% v/v MeCN as co-solvent to ensure a homogeneous solution. The CoPcNH2@MWCNT catalyst, a CuxPdy alloy, and a Ag/AgNO3 (0.1 M n-Bu4NPF6 in MeCN) electrode was used as working, counter (EG oxidation), and reference electrode, respectively, and the applied potentials were converted to the Fc/Fc + scale. CV scans in this medium showed an onset potential of -1.7 V vs. Fc/Fc + . Subsequent CPE studies (for 10 h) revealed an optimum CO production at -1.85 V vs. Fc/Fc + with 19.0±1.4% FECO (Figures ). Isotopic labelling experiments with captured 13 CO2 in both the aqueous (TEA/H2O) and non-aqueous (NaOH/EG) medium showed only C labelled CO as reduction product during FTIR analysis of the headspace gas, confirming that CO was derived from captured CO2 (Figure ). ) with CoPcNH2@MWCNT cathode in a two-compartment three-electrode configuration. b, Product amount (normalized to geometric surface area of the electrode) and CO faradic efficiency (FECO) for TEA captured CO2 reduction at different potentials with CoPcNH2@MWCNT electrode. c, Captured CO2 amount (CO2 flowrate 30 mL min -1 ; capture duration, 2 h), and observed FECO for different amines after controlled potential electrolysis (CPE) at -0.7 V vs. the reversible hydrogen electrode (RHE) for 1 h (conditions for a-c: Catholyte, capture solution with 0.1 M added K2SO4; anolyte, 0.1 M K2SO4; cathode, CoPcNH2@MWCNT; anode, Pt mesh; bipolar membrane). d, FECO for reduction of NaOH captured CO2 products in aqueous medium at -0.7 V vs. RHE for 1 h. e, FECO obtained after CPE of CO2 captured by glycolic NaOH solution at different potentials (conditions: catholyte, capture solution with added 20% v/v MeCN and 0.15 M tetrabutylammonium tetrafluoroborate (TBABF4); anolyte, 0.6 M NaOH, 0.15 M TBABF4 in 20% v/v MeCN in ethylene glycol (EG); cathode, CoPcNH2@MWCNT; anode, Cu26Pd74 alloy; bipolar membrane). f, Fourier transform infra-red (FTIR) spectra of the headspace (CO region) after isotope labelling experiments with both TEA/H2O and NaOH/EG system using captured CO2 as reactant where the CO signal was obtained from an experiment with captured 12 CO2. All experiments were carried out at room temperature. |
63f39b8ffcfb27a31f1dc775 | 4 | The solar-driven conversion of captured CO2 (in TEA/H2O and NaOH/EG solutions, using concentrated stream) was then explored in a two-compartment cell. A photocathode was prepared for this purpose by interfacing the CoPcNH2@MWCNT catalyst with a triple cation lead halide perovskite (PVK) photoabsorber using a conducting graphite epoxy paste (PVK|CoPcNH2@MWCNT, Methods). The PVK has optimal bandgap (1.6-1.7 eV) to absorb broad-range of the solar spectrum (360-750 nm), and provides high open-circuit photovoltage (~1.1 V) to efficiently drive both half-reactions. A bimetallic Cu26Pd74 alloy electrodeposited on Ni foam substrate (Ni foam|Cu26Pd74, see Methods) was used as anode (Figure ), which facilitates alcohol oxidation under alkaline conditions. Operating conditions of the two-electrode PEC setup without external bias was determined from the overlap of individual CV curves of PVK|CoPcNH2@MWCNT photocathode (taken under 1 sun, AM 1.5G irradiation) and Ni foam|Cu26Pd74 anode (taken under 'dark' conditions) in threeelectrode configuration (Figures ). The overlap potentials (Voverlap) were 0.52 V vs. RHE and -0.85 V vs. Fc/Fc + in the TEA/H2O and NaOH/EG medium, with overlap current densities of 5.8 mA cm -2 and 0.27 mA cm -2 , respectively (Figures and). Accounting for the open circuit voltage (VOC) of the PVK devices (~1.05±0.03 V, Figure ), the potential experienced by CoPcNH2@MWCNT catalyst in the two-electrode setup without any external voltage is around -0.53 V vs. RHE and -1.9 V vs. Fc/Fc + in the TEA/H2O and NaOH/EG medium, respectively (calculated as Voverlap-VOC). Electrochemical analyses at these potentials confirmed activity of the CoPcNH2@MWCNT catalyst towards captured CO2 conversion (Figures ). |
63f39b8ffcfb27a31f1dc775 | 5 | For two-electrode PEC captured CO2 conversion coupled to EG oxidation, the catholytes were captured CO2 solution with additives (similar to Figure ) and the anolytes were 0.5 M EG in 0.5 M aqueous KOH (TEA/H2O) or 0.6 M NaOH, 0.15 M TBABF4 in 80/20 EG/MeCN solution (NaOH/EG). The two compartments were separated by a bipolar membrane to maintain the pH difference which introduced a chemical bias (~0.35 V) to the system. CV scans under solar-irradiation showed an onset voltage at -0.4 V for the TEA/H2O system with j ~4.9 mA cm -2 at zero applied voltage (Figure ). A stable photocurrent density of 1.1±0.3 mA cm -2 was obtained during PEC experiment without any external voltage under chopped light irradiation (1 sun, 50 min on, 10 min off, Figure ). After 10 h, syngas was detected in the cathode with 54.6±9.2 mol cm -2 CO and 106.6±8.4 mol cm -2 H2 (FECO 34.1±2.2% and FEH2 70.3±1.8%, Figure ). The turnover number of the molecular catalyst for CO formation (TONCO) was estimated 3657±591 (Figure ). No other reduction products were detected in NMR spectroscopic analysis of the catholyte. The high performance liquid chromatography (HPLC) analysis of the anolyte showed GA as only oxidation product with 85.8±16.2 mol cm -2 yield (FEGA 92.5±5.3%). The possibility of employing real-world PET waste as EG precursor was confirmed by using alkaline pre-treated commercial PET plastic bottle solution as anolyte (~0.2 M EG, see methods), which showed similar yields and FE for H2, CO, and GA formation after 10 h of PEC experiment without external voltage (Figure ). |
63f39b8ffcfb27a31f1dc775 | 6 | The solar-driven conversion of atmospheric CO2 was further investigated. Direct air capture (DAC) of CO2 and its conversion into value-added products is a promising technology to afford an overall negative carbon footprint when conversion is done with a renewable energy source such as sunlight. Moreover, DAC technologies can be easily decentralized to desired locations, opposed to capture from concentrated point sources requiring proximity to emissions. The direct capture and conversion of aerobic CO2 is especially challenging due to the ultra-low concentration, requiring aggressive capturing agents for suitable kinetics and consequently higher thermodynamic input for chemical reduction. We performed the atmospheric CO2 capture by pumping indoor air through capture solutions (1 M aqueous TEA or glycolic NaOH) at 1.8 L min -1 flowrate for 2 days using an aquarium pump (Figure ). The TEA/H2O solution captured only 0.02±0.01 mol CO2 per mol TEA after this time due to slow capture kinetics. In contrast, capture efficiency of the NaOH/EG solution was notable reaching 0.73±0.07 mol of captured CO2 per mol NaOH (Figures ). |
63f39b8ffcfb27a31f1dc775 | 7 | Electrochemical reduction of TEA captured atmospheric CO2 at -0.7 V vs. RHE with CoPcNH2@MWCNT cathode produced negligible CO due to very low captured CO2 availability. Contrarily, CPE of the atmospheric CO2 captured in NaOH/EG solution at -1.85 V vs Fc/Fc + produced CO with 2.9±0.3% FE, showing electrochemical conversion of direct air captured CO2 (Figure ). We note that despite a modest captured CO2 concentration (0.73 M), the FECO is lower compared to CO2 captured from a pure stream (FECO ~20%). This is likely due to the incomplete CO2 capture (owing to equilibrium shift in ultra-low CO2 concentration as per Le Chatelier's principle), which leads to higher alkalinity of the post-capture medium and a more challenging conversion process (supplemental discussion, Figure ). Solar-driven PEC reduction of captured atmospheric CO2 in TEA solution with PVK|CoPcNH2@MWCNT photocathode did not produce any CO. In contrast, the organic NaOH/EG medium was more suitable for the solar-driven conversion of atmospheric CO2. PEC experiment with CO2 captured in glycolic NaOH solution after DAC produced 2.1±0.5 mol cm -2 CO after 110 h under 1 sun irradiation with no applied voltage (Figures ). The CO production rate remained consistent over the reaction period with final FECO and TONCO reaching 2.3±0.4% and 141±38, respectively (Figure ). A control experiment without DAC produced tiny amount of CO due to epoxy degradation in EG medium, that was subtracted for all atmospheric CO2 conversion calculations (Figure ). HPLC analysis of the anolyte showed 61.6±8.3 mol cm -2 GA formation (FEGA 85.7±2.2%) that indicates the completion of overall PEC process. While further optimization is necessary to improve the overall CO formation efficiency, this proof-of-concept study demonstrates the viability of direct long-term solar-driven reduction of atmospheric CO2 to CO following DAC, using perovskite-PEC systems. Future work in this regard would benefit from the use of solar cells with higher VOC to match the stringent thermodynamic demands of air captured CO2 conversion. |
63f39b8ffcfb27a31f1dc775 | 8 | This work demonstrates capture and direct solar-driven utilization of CO2 to syngas, in combination with plastic-derived waste oxidation, in a two-compartment PEC setup. CO2 is captured from a concentrated stream or simulated flue gas or air in amine/hydroxide solution, and directly converted over a PVK|CoPcNH2@MWCNT photocathode under 1 sun irradiation. A Ni foam|Cu26Pd74 alloy anode completes the circuit by catalyzing EG oxidation to GA selectively. Replacing thermodynamically challenging water oxidation with EG oxidation allows the system to operate with a single light absorber without any externally applied voltage. Realworld plastic waste can be directly used as EG source after alkaline pre-treatment, facilitating waste valorization. Utilizing air as the CO2 source, pre-treated waste plastics as electron donors, and sunlight as the energy source, this proof-of-concept CO2 capture and solar-driven utilization system could be promising for future decentralized off-the-grid scalable solar fuels and chemical synthesis. |
63f39b8ffcfb27a31f1dc775 | 9 | To prepare the electrodes, MWCNT (5 mg) was dispersed in DMF (2.34 mL) using a probe sonicator for 20 min. Then, 600 L of freshly made CoPcNH2 solution in DMF (1 mM) and of Nafion TM (60 L, 5 % v/v in lower aliphatic alcohols and water) were added to the dispersion. The mixture was sonicated for an additional 15 min. 100 L of this catalyst ink was then drop-casted over an activated graphite foil substrate (active area ~0.84 cm 2 ). The ink was allowed to dry for a minimum two days at room temperature. The fabricated catalyst is denoted as CoPcNH2@MWCNT and characterised by SEM, SEM-EDX, ICP-OES analysis. |
63f39b8ffcfb27a31f1dc775 | 10 | The inverse structure triple cation mixed halide perovskite devices were prepared following a previously reported procedure. In brief, a hole transporting layer (HTL) of NiOx was first deposited on a FTO collated glass substrate by spin coating a solution of Ni(NO3)•6H2O (1 M), and ethylenediamine in EG (1 M), and was then annealed at 373 K for 30 min. A second hole transporting layer was deposited over the NiOx by spin coating a F4TCNQ doped PTAA solution inside a N2 filled glove box. A perovskite layer was then deposited on top of the HTLs. For this purpose, a cesium formamidinium methylammonium (CsFAMA) perovskite precursor solution was first prepared by adding a solution of FAMA0.22Pb1.32I3.2Br0.66 (1000 L) to DMF (510 L), DMSO (340 L) and 1-methyl-2-pyrrolidone (150 L), and then adding CsI in DMSO (1.5 M, 48 L). The perovskite layer was then deposited over the PTAA layer using a two-step spin coating process, first for 10 s at 1000 rpm, followed by 35 s at 6000 rpm. Chloroform was used as the antisolvent for the final 7 sec before the end. Afterwards, the PVK layer was annealed at 373 K for 30 min. Subsequently, a [6,6]-phenyl C61 butyric acid methyl ester (PCBM; 35 mg mL -1 in chlorobenzene) solution was spin coated on top of the perovskite layer at 3000 rpm for 45 sec as an electron transport layer (ETL). A PEIE film layer was then deposited over the PCBM coated perovskite device using a precursor solution (3.9 L mL -1 solution in isopropanol), which helps in stopping the interfacial degradation. Finally, a 100 nm conductive Ag layer was deposited over the PEIE layer using metal evaporation technique through a patterned mask, ensuring an active perovskite surface area of ~0.5 x 0.5 cm 2 . |
63f39b8ffcfb27a31f1dc775 | 11 | The perovskite devices were interfaced with the CoPcNH2@MWCNT catalyst to prepare the photocathode, which were subsequently encapsulated with epoxy to stop their degradation inside aqueous medium. For interfacing the devices with the catalyst, a conductive graphite epoxy (GE) paste was prepared by homogeneously mixing graphite powder with epoxy in 3:4 mass ratio. Araldite standard two-part epoxy was used for this purpose. The paste was then applied evenly over the Ag contact layer of the perovskite device, on top of which the graphite foil containing the CoPcNH2@MWCNT catalyst was attached. The device was then left to dry for 24 h to harden the GE layer, after which a Cu wire was attached for connection. Finally, the assembled device was encapsulated, and the edges were sealed using Araldite 5-min Rapid 2 component epoxy. The final PVK-based photocathodes are referred to as PVK|CoPcNH2@MWCNT. |
63f39b8ffcfb27a31f1dc775 | 12 | The Cu26Pd74 oxidation catalyst was synthesized by a dynamic H2 bubble template assisted galvanostatic electrodeposition method using an activated Ni foam as scaffold. The electrodeposition was carried out in a single compartment three-electrode configuration, where a leakless double junction Ag/AgCl electrode (sat. KCl, Metrohm, Switzerland) was used as a reference electrode, an activated Ni foam scaffold was used as the working electrode, and a Pt foil (~6 cm 2 area) was used as the counter electrode. The electrolyte solution contained a total 0.02 M solution of CuSO4•5H2O and Na2PdCl4 salts with a Cu 2+ :Pd 2+ molar ratio of 30:70. The galvanostatic electrodeposition was carried out by applying a current density of -2 A cm -2 for 40 s. The formed bimetallic catalyst was then washed with Milli Q ® water multiple times to remove any residual salt and acid and was subsequently dried under N2 flow. |
63f39b8ffcfb27a31f1dc775 | 13 | The FESEM images were acquired using TESCAN MIRA3 FEG-SEM instrument equipped with an Oxford Instruments Aztec Energy X-maxN 80 EDX system. The UV-vis spectra were recorded using a Varian Cary 50 UV-vis spectrophotometer. The ICP-OES measurements were performed on a Thermo Scientific iCAP 7400 ICP-OES DUO spectrometer at the Microanalysis Service, Yusuf Hamied Department of Chemistry, University of Cambridge. |
63f39b8ffcfb27a31f1dc775 | 14 | The CO2 capture form concentrated CO2, flue gas and ambient air was carried out by bubbling the respective CO2 containing gas through a capture solution. Briefly, the capture solution was prepared by dissolving the amine/hydroxide in water/ethylene glycol solvent to make a concentration of 1 M. Concentrated CO2 or simulated flue gas (15% CO2, 4% O2, balance N2) was then bubbled through the solution at 30 mL min -1 for 2 h or 6 h, respectively (Figure ). For CO2 capture from air, indoor air (Reisner Lab, Yusuf Hamied Department of Chemistry, University of Cambridge) was pumped through the capture solution at a flow rate of 1.8 L min -1 after passing it through a Drierite® column for drying, using a Pawfly Ultra Quiet Air Pump (MA-60) (Figure ). Afterwards, the capture solutions were purged with N2 for 15 mins to remove any physically dissolved CO2 or O2. 0.5 mL of the solution was analysed by 1 H and C NMR spectroscopy, after addition of 1,4-dioxane (aqueous systems) or imidazole (organic systems) as internal standard and D2O (aqueous)/DMSO-d6 (organic) as the deuterated solvent to characterize and quantify the captured products. The remaining solution was used as CO2 source for the electrochemical and PEC studies after addition of supporting electrolytes/solvents as previously mentioned. |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.