<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Amrita Ghag on Medium]]></title>
        <description><![CDATA[Stories by Amrita Ghag on Medium]]></description>
        <link>https://medium.com/@amritaghag08?source=rss-28aecc9c0df8------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*WFQcQRDnW7nlk7iL7tDmFA.png</url>
            <title>Stories by Amrita Ghag on Medium</title>
            <link>https://medium.com/@amritaghag08?source=rss-28aecc9c0df8------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Wed, 18 Nov 2026 20:26:01 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@amritaghag08/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Integrating Surrogate Modelling Techniques for Enhanced Fusion Plasma Control and Predictability]]></title>
            <link>https://medium.com/@amritaghag08/integrating-surrogate-modelling-techniques-for-enhanced-fusion-plasma-control-and-predictability-f48fc0741214?source=rss-28aecc9c0df8------2</link>
            <guid isPermaLink="false">https://medium.com/p/f48fc0741214</guid>
            <dc:creator><![CDATA[Amrita Ghag]]></dc:creator>
            <pubDate>Sun, 14 Jan 2024 22:00:14 GMT</pubDate>
            <atom:updated>2024-01-14T22:00:14.401Z</atom:updated>
            <content:encoded><![CDATA[<p><strong>A Research Paper</strong></p><h3>Abstract</h3><p>This research paper investigates the usage of surrogate modelling techniques in the field of plasma control in nuclear fusion tokamak plasmas. Nuclear fusion stands as a possible answer to a sustainable and clean energy source. However, it comes with additional complications, especially in plasma control. This paper explores the difficulties associated with plasma behaviour and the usefulness of surrogate models in overcoming these difficulties. It examines many approaches, such as Artificial Neural Networks, Bayesian Networks, and Gaussian Process Regression, and how effective they are in simplifying and complex plasma behaviours.</p><p>We take a dive into the evolution of surrogate modelling, comparing it to traditional modelling techniques while highlighting its advantages in computational efficiency. Additionally, understanding the complicated dynamics of plasma in fusion reactors is a major focus of the research. The interaction between surrogate modelling and plasma control is carefully looked at, demonstrating how these models can provide innovative opportunities for plasma control and simplification.</p><p>Finally, the paper not only looks at the significance of surrogate modelling, but it also suggests future research directions. These include the creation of more advanced AI-driven surrogate models, the investigation of new fusion reactor materials, and a better understanding of plasma instabilities. The entire combination of research shows how surrogate modelling can be used to achieve fusion energy.</p><h3><strong>Table of Contents</strong></h3><p><strong>1. Introduction</strong></p><p>1.1. Nuclear Fusion: Principles and Energy Potential</p><p>1.2. Plasma in Fusion: Properties, Importance, and Challenges</p><p>1.3. Tokamaks and Plasma Confinement: Technological Evolution</p><p><strong>2. The Emergence and Evolution of Surrogate Modelling</strong></p><p>2.1. Introduction to Surrogate Modelling: Concepts and Historical Development</p><p>2.2. Detailed Analysis of Surrogate Model Types and Their Mechanics</p><p>2.3. Comprehensive Review of Surrogate Models’ Benefits in Complex Systems</p><p>2.4. Comparative Analysis: Surrogate Models vs. Traditional Modelling Approaches</p><p><strong>3. Application of Surrogate Models in Enhancing Fusion Plasma Control</strong></p><p>3.1. Current Technology Being Used for Simulations</p><p>3.1.1. Rationale for Surrogate Models in Plasma Control</p><p>3.2. Methodologies for Implementing Surrogate Models in Plasma Physics</p><p>3.3. Success Story of Surrogate Models for Plasma Stability</p><p>3.4. Evaluating Performance: How Surrogate Models Optimize Reactor Operations</p><p><strong>4. Advancing Predictive Capabilities in Fusion Plasmas with Surrogate Models</strong></p><p>4.1. The Important Role of Surrogate Models in Predictive Analysis</p><p>4.2. Techniques for Enhancing Accuracy &amp; Reliability in Predictions</p><p>4.2.1. Predictive Success: Key Examples and Results</p><p>4.3. Surrogate Models Risks and Management</p><p><strong>5. The Future Landscape: Challenges and Innovations in Surrogate Modelling</strong></p><p>5.1. Addressing Current Limitations in Surrogate Modelling</p><p>5.2. Successful Research of Surrogate Models Done Outside of Fusion</p><p>5.3. Solutions to Current Limitations in Surrogate Modelling</p><p><strong>6. Conclusion</strong></p><p>6.1 Synthesis of Key Insights and Findings</p><p>6.2 Reflections &amp; Recommendations for the Future of Fusion Energy and Surrogate Modelling</p><p><strong>7. References</strong></p><h4>Introduction</h4><p><strong>1.1. Nuclear Fusion: Principles and Energy Potential</strong></p><p>Nuclear Fusion reactions power all stars. Nuclear Fusion is the process of two or more atomic nuclei combining to form a heavier nucleus. This process will usually include light elements, such as Deuterium and Tritium; hydrogen variants¹. This process releases energy. Because the overall mass of the resulting single nucleus is smaller than the mass of the two original nuclei, the remaining mass is converted into energy². Einsteins most notable equation <em>E=mc² (energy equals mass times the speed of light squared) </em>explains why this process naturally occurs. The equation states that mass and energy can be converted into one another, and in the case of fusion, will result in a huge number that reveals how a tiny amount of mass converts to enormous energy³.</p><p>Fusion energy has been seen as an attractive energy source. For context, the next upcoming decades are crucial in reducing our consumption of greenhouse gas emissions. This is due to the fact that the demand for energy has almost tripled, as a result of population growth and increased urbanization⁵. We need a scalable, sustainable, and carbon-free form of energy. Fusion energy meets all of these requirements.</p><p>The controlled fusing of atoms releases roughly 4 million times more energy than chemical reactions (coal, oil, gas). As previously mentioned, most fusion projects require Deuterium + Tritium. Deuterium is an abundant element. Tritium on the other hand is a rare isotope of Hydrogen, as well as a byproduct in fusion reactions. A topic of discussion for next time is <a href="https://medium.com/@amritaghag08/the-application-of-metal-hydrides-and-hydrogen-in-nuclear-fusion-reactions-tritium-management-and-1f5965c606dd">how to breed and recover tritium reliably in a fusion device</a>.</p><p>Fusion energy does not emit any harmful greenhouse gas emissions into the atmosphere. In fact, Helium is the main by-product in a reaction, a non-toxic gas. There is no long-lived nuclear waste, or any sort of risk to exploitation of nuclear weapons⁵.</p><p><strong>1.2. Plasma in Fusion: Properties, Importance, and Challenges</strong></p><p>Fusion occurs in the fourth state of matter: plasma, a keyword to todays topic. Plasma is a hot charged gas made up of positive ions and free-moving electrons. In order to fuse, nuclei must collide at extremely high temperatures, approximately ten million degrees Celsius. The extreme heat ultimately allows the nuclei to overcome their mutual electrical repulsion; once the nuclei are close together, the attractive nuclear force between them will outweigh the electrical repulsion and allow them to combine⁴.</p><p>To achieve fusion in a lab, three requirements must be fulfilled: extremely hot temperatures for high-energy collisions, sufficient plasma particle density to increase the likelihood of collisions, and sufficient confinement time to hold the plasma within a defined volume. A fusion plasma is heated and confined using strong magnetic fields, in devices such as Tokamaks.</p><p>However, it’s been clear that the perfect containment of a fusion plasma is impossible. Achieving the three requirements for fusion is a challenging task, and maintaining them is even more complex. To put it into perspective, fusion plasmas are extremely hot, exceeding temperatures of 100 million degrees Celsius. Plasma qualities and stability are still required to achieve viable fusion energy. The complexities of plasma behaviours, as well as the difficulties connected with plasma physics and confinement, all contribute to the difficulty of creating viable fusion energy. Despite these obstacles, the scientific argument for believing that fusion is possible remains strong, and scientists and engineers from around the world are continuing to take on research initiatives.</p><p>We know that a fusion grade plasma must meet three key requirements to achieve the conditions necessary for nuclear fusion. However, mimicking a natural process that occurs in space is a very complex situation and requires control mechanisms due to the unpredictable nature of plasma.</p><p>The real-time Plasma Control System <em>(PCS) </em>in fusion reactors must be able to control all aspects of a Tokamak, and stabilize any plasma instabilities that do occur. The PCS must remain as stable as possible, and be able to actively control and correct any lingering instabilities. The PCS must also be capable of detecting and responding to hardware defects and “out-of-the-ordinary” circumstances. Because fusion plasmas are fundamentally unstable and operate towards the edge of their performance envelope, the PCS requires very effective control performance and robustness.</p><p>The main theme here is that fusion grade plasmas are highly uncontrollable and unpredictable. A major challenge of these plasmas are plasma instabilities, as they disrupt the stability and performance of a fusion reaction. Key types include:</p><ol><li><strong>Kink Instability:</strong> This instability involves a distortion in the shape of the plasma, leading to a helical or kinked structure. It can cause the plasma to become disrupted, impacting the stability of the fusion process⁶. It happens as a result of transverse displacements of a plasma column’s cross-section from its centre of mass, with no change in plasma properties. When a “kink” develops in a column, the magnetic forces on the inside of the kink become imbalanced, causing the disruption to grow.</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/904/1*cehdhUAqCKJfwFR5N_wGag.png" /></figure><p>2. <strong>Ballooning: </strong>A<strong> </strong>type of plasma instability caused by internal pressure that often occurs in tokamak fusion power reactors. The term “ballooning” refers to the shape and action of the instability, which is similar to the elongations created when a long balloon is squeezed. Under typical experimental conditions, the plasma travels through weak and strong places inside the container repeatedly, which can result in the ballooning instability. The difference in pressure between the core and the edge region can cause the ballooning instability to take place. The ballooning instability’s narrow fingers of plasma are capable of accelerating and driving an explosive release of energy, which is why the instability is also known as the “explosive instability.” In a plasma, ballooning instability occurs when one component of the plasma expands, simulating the shape of a balloon. This instability is primarily determined by the magnitude of the pressure gradient in the plasma as well as the state of the magnetic field that confines the plasma⁷.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*DxAA7KKvsmFQ707m5jMmJw.png" /></figure><p>3. <strong>Tearing Mode Instability: </strong>Tearing modes are instabilities that tear and reattach magnetic field lines at various resonant surfaces in the plasma to form radially-localized magnetic chains. Tearing modes are driven by radial current and pressure gradients inside the plasma and can be unstable even when the plasma is optimally stable. Tearing modes weaken plasma confinement because heat and particles can flow extremely quickly from one side of a magnetic island chain to another by streaming along magnetic field-lines rather than distributing themselves over magnetic flux-surfaces⁸.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/600/1*_SCmNJAtgvs27H6I-bmOZw.gif" /></figure><p>These instabilities can have a negative impact on plasma confinement, operational stability, and overall efficiency, making their understanding and management critical for the development of stable and efficient fusion energy systems.</p><p><strong>1.3. Tokamaks and Plasma Confinement: Technological Evolution</strong></p><p>As previously mentioned, to confine fusion plasmas, a device called a “Tokamak” is employed. A Tokamak (Fig. 1) is a machine that confines a plasma using magnetic fields in a donut shape that scientists call a torus. Magnetic field coils confine plasma particles in a tokamak, allowing the plasma to reach the temperatures required for fusion. One set of magnetic coils produces an intense “toroidal” field that is directed the whole circumference of the torus. A central solenoid generates a second magnetic field oriented along the “poloidal” direction, which is the shortest path around the torus. The combination of the two field components produces a twisted magnetic field that confines the particles in the plasma. An outer poloidal field is generated by a third set of field coils, forming and positioning the plasma⁹.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/620/1*49eDBmIM0RBwqK5CCxUw_A.jpeg" /><figcaption><a href="https://www.energy.gov/science/doe-explainstokamaks"><strong>Fig 1:</strong></a><strong> </strong>A depiction of the different parts of a Tokamak</figcaption></figure><p>However, the magnetic confinement technique Tokamaks use present several challenges. And the theme here is that they are very prone to turbulence and other instabilities. Not only this but the environment within the fusion device is one of the harshest ever imagined, with plasma-facing materials subjected to extreme temperatures, high heat flux, neutron irradiation, and bombardment by deuterium, tritium, and helium¹⁰.</p><p>Magnetic field perturbations are another factor that can significantly affect plasma confinement in Tokamaks. Plasma instabilities, such as kink, ballooning, and tearing modes, can cause magnetic field perturbations, causing disruptions in the plasma and affecting its confinement. Changes in the magnetic field can cause a loss of plasma confinement, enabling heat and particles to escape. This can affect the efficiency of the fusion process and have an impact on the plasma’s overall stability¹¹.</p><p>Finally, a complex range of obstacles can be caused by the dynamic and frequently turbulent interaction between plasma and the materials that make up the inside of fusion reactors. These challenges have an enormous impact on the operational lifespan of these reactors as well as their overall safety. For many reasons, this complex plasma-material interaction (PMI) is an important factor to take into account in the field of fusion energy research and development.</p><p>The retention of tritium in the reactor materials is one of the more urgent issues brought on by PMI. When tritium gets trapped in materials, it increases the reactor’s operating costs and poses a risk to public safety because it can be radioactive at that point¹².</p><p>The selection of materials that can endure the harsh conditions seen in a fusion reactor is one of the difficulties. It is hard to find materials that can decrease tritium retention, endure radiation, and not be damaged by plasma interactions. The proper materials are necessary for the safe and efficient operation of fusion reactors, hence this selection process is crucial.</p><h4><strong>2. The Emergence and Evolution of Surrogate Modelling</strong></h4><p>We take a thorough look at surrogate modelling in this chapter;<em> “The Emergence and Evolution of Surrogate Modelling,” </em>which is a revolutionary approach in computer science that provides efficient replacements for conventional or traditional modelling techniques. In<em> “2.1. Introduction to Surrogate Modelling: Concepts and Historical Development,”</em> which opens this section, we explore the basic framework of surrogate models, categorizing them as efficient but simplified substitutes for more complicated systems.</p><p>The section continues in <em>“2.2. Detailed Analysis of Surrogate Model Types and Their Mechanics,”</em> where we look at several methods of surrogate modelling, such as Artificial Neural Networks (ANNs), Bayesian Networks, and Gaussian Process Regression (GPR). We’re later moved to the benefits of using surrogate models in<em> “2.3. Comprehensive Review of Surrogate Models’ Benefits in Complex Systems,”</em> especially in situations involving a large amount of computational power.</p><p>In the final section, <em>“2.4. Comparative Analysis: Surrogate Models vs. Traditional Modelling Approaches,”</em> surrogate models are reviewed + compared with standard modelling methods. In comparison to the more sophisticated, physics-based traditional models, this section highlights the less computational cost offered by surrogate models. In in addition to demonstrating the value and flexibility of surrogate modelling, this chapter also lays the groundwork for understanding its significant influence on a range of scientific and engineering fields, most notably of course; fusion energy research.</p><p><strong>2.1. Introduction to Surrogate Modelling: Concepts and Historical Development</strong></p><p>The term <em>“surrogate model” </em>is derived from the English language and is used to describe a model that serves as a substitute or proxy for another, often more complex, model or system. (<a href="https://www.etymonline.com/search?q=surrogate+model">8</a>) The term <em>“surrogate”</em> itself comes from the Latin word<em> “surrogatus”,</em> which means <em>“substitute”</em> or <em>“deputy”¹³.</em> In the context of modelling, a surrogate model takes the place of a more resource-intensive or time-consuming model to provide faster or more efficient approximations for various purposes- such as optimization, prediction, or analysis.</p><p>Surrogate models are simple analytical models that mimic the input/output behaviour of complex systems¹⁴. By building these models that mimic the behaviour of the original, more advanced ones, it is possible to predict outcomes based on fresh input data. In engineering, design optimization, uncertainty quantification, sensitivity analysis, and other areas where direct evaluation of the original model is unfeasible owing to high computational costs or other limitations, surrogate models are often used. By building these models that mimic the behaviour of the original, more advanced ones, it is possible to predict outcomes based on fresh input data. In engineering, design optimization, uncertainty quantification, sensitivity analysis, and other areas where direct evaluation of the original model is unfeasible owing to high computational costs or other limitations, surrogate models are often used. Surrogate models are generally made by using a small number of well selected input-output data points and are typically constructed using a data-driven, bottom-up process. These models are helpful in helping to understand the overall behaviour of complex systems and in helping to make reasonable choices without requiring thorough analyses of the original complicated models: They are simplified yet more efficient models.</p><p><strong>2.2. Detailed Analysis of Surrogate Model Types and Their Mechanics</strong></p><p>In applied mathematics, statistics, and machine learning, Gaussian process regression (GPR), often referred to as Kriging, is a frequently used technique for building surrogate models, interpolation, classification, supervised learning, and active learning. Assuming that a partially observed function (of time or space) is an example of a Gaussian process (GP), GPR builds a statistical model of that function. GPR comes in a variety of forms, such as basic, regular, and universal Kriging. GPR shares many similarities with machine learning kernel machines, but it offers additional information because it gives an estimate of the unknown.<br>The data are modelled as a GP with a constant mean and a predetermined shape of the stationary covariance function in the usual Kriging. The stationarity assumption lowers the complexity of the model and the number of hyper parameters¹⁵.</p><p>Another key type of surrogate model are <em>Bayesian networks. </em>Bayesian networks are a type of Probabilistic Graphical Model that can be used to build models from data and/or expert opinion¹⁶. There are a number of random or unknown variables in Bayesian networks. It’s vital to explore how the variables’ together distribution can be characterized using what we know about their natural interactions through conditional distributions. This technique describes their interrelationship using graph theory. If data on the random variables is available, we design a Bayesian network model that depicts their relationship clearly; Bayesian networks are a combination of probability theory and graphs. Prediction is one of the primary purposes of Bayesian networks¹⁷.</p><p>Artificial Neural Networks (ANN) are the final significant surrogate modelling technology to briefly discuss. ANNs are biologically inspired computational networks since they are based on brain and nervous system research (Fig. 2).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/659/0*VoPY6HRP0P8tojz3.gif" /><figcaption><a href="https://www.sciencedirect.com/topics/earth-and-planetary-sciences/artificial-neural-network#:~:text=Second%20Edition)%2C%202019-,Artificial%20Neural%20Networks%20(ANNs),phenomenon%20or%20a%20decision%20process."><strong>Fig 2:</strong></a><strong> </strong>Sample artificial neural network architecture</figcaption></figure><p>ANN models imitate the electrical activity of the brain and nervous system. The neurodes (processing elements) are often organized in a layer or vector, with the output of one layer serving as the input to the next layer and maybe further layers. A neurode may be linked to all or a subset of the neurodes in the adjacent layer, replicating synaptic connections in the brain. Weighted data signals entering a neurode imitate the electrical activation of a nerve cell and, as a result, information flow within the network or brain. The input values to a processing element, in, are multiplied by a connection weight which replicates neural pathway strengthening in the brain. Learning is simulated in ANNs by the modification of connection strengths or weights¹⁸.</p><p><strong>2.3. Comprehensive Review of Surrogate Models’ Benefits in Complex Systems</strong></p><p>Now that we know some of the key times of surrogate modelling techniques, we can explore the advantages of these systems: The advantages of surrogate modelling in complex systems are multifaceted and very impactful, especially in the fields requiring thorough computational resources and processes. Here is a detailed exploration of these advantages:</p><ol><li><strong>Reduced Computational Cost: </strong>As computationally cheaper substitutes for physically expensive simulations and experiments, surrogate models are what we need for increased efficiency. This benefit greatly reduces the time and resources commonly required for complex system tests, making it especially important in situations where the exploration of large design spaces is necessary¹⁹.</li><li><strong>Efficient Design Exploration: </strong>Design space exploration is revolutionized by the use of surrogate models. Through the prediction of different combinations’ results, these models significantly lessen the need to build numerous physical prototypes. This helps save costs and maintain resources in addition to quickening the design process overall²⁰.</li><li><strong>Informed Decision-Making: </strong>The last of many advantages we will explore. Surrogate models offer quick and accurate approximations of system behaviours, which are very helpful for well-informed decision-making. These models provide a solid foundation for decision-making, bridging the gap between theory and real-world testing, in situations where full simulations or extended experiments are unfeasible²¹.</li></ol><p><strong>2.4. Comparative Analysis: Surrogate Models vs. Traditional Modelling Approaches</strong></p><p>Traditional modelling techniques, which are frequently physics-based, are used to accurately represent real-world processes. These models frequently require a lot of processing power, especially when used at higher and higher resolutions to find finer structural details²². They typically come in handy in situations when measuring an outcome of interest is difficult and an outcome model is used in its place²³.</p><p>As we know, surrogate models are just condensed versions of these complicated, higher-order models. Because they are built to resemble the original models’ behaviour, outcomes can be predicted using fresh input data²⁰. When evaluating whether the original model is computationally expensive or when it’s unknown how input and output are actually related, surrogate models are applied. They are built with a bottom-up, data-driven method²³.</p><p>Surrogate models have a lower computational cost than traditional models, which is their primary advantage. They give a practical replacement for expensive physical experiments and simulations, allowing effective complex system optimization and design space exploration²⁰. Additionally, surrogate models provide freedom in selecting any statistical framework that is appropriate for the modeller or the particular situation²⁴.</p><h4><strong>3. Application of Surrogate Models in Enhancing Fusion Plasma Control</strong></h4><p>In this chapter of “Application of Surrogate Models in Enhancing Fusion Plasma Control”, we look at how traditional computational methods are giving way to the unique use of surrogate models in nuclear fusion. We begin by discussing the limitations of present simulation technologies in accurately modelling plasma dynamics within fusion reactors, emphasizing the computing difficulties connected to traditional computer simulations.</p><p>The chapter explores the justification for using surrogate models, mainly those run on Reinforcement Learning (RL), and how these advanced techniques give more adaptable and efficient solutions for plasma control in tokamaks. We look into necessary implementation approaches for these models, such as the use of transfer learning for disruption prediction, and the integration of these models into existing plasma control systems. This is all done through a series of case studies and past research paper findings, highlighting the effectiveness of surrogate models and their potential for plasma control.</p><p><strong>3.1. Current Technology Being Used for Simulations</strong></p><p>Traditional computer simulations face a large amount of computational limitations. Nuclear fusion requires extremely sophisticated computations to estimate the temperature and density profiles of plasma. Even the most powerful supercomputers find it difficult to meet such enormous processing demands. For example, modern supercomputers cannot solve issues involving plasma turbulence by using brute force, therefore other methods such as machine learning are required for optimization²⁵.</p><p>Tokamaks are extremely detailed and incorporate a number of different physical principles, including radiation, electromagnetic, structural forces, heat transfer, and chemical transport. A system like this presents a multi-scale, multi-physics, system of systems difficulty for control, requiring equally as complex simulations. Because of the additional layers of complexity and computational load, current models are not enough for timely and realistic engineering design²⁶.</p><p>Traditional computer simulations need to use <em>high-performance computing </em>(HPC) facilities, which have computers with powerful processors and large amounts of memory. This is mainly due to the fact that these models require solving complex mathematical equations that must properly describe multiple physical processes, which is an extremely resource-dependant operation. Furthermore, the simulations performed by these models are time-consuming, typically taking weeks to complete. This time is linked to the level of depth and precision these simulations strive for, which requires the processing of massive volumes of data and the performance of multiple complex calculations, even on high-speed computer systems. Furthermore, these traditional models are distinguished by sophisticated software infrastructures. They include a wide range of components and dependencies, such as specialized libraries, data input/output systems, and parallel processing frameworks.</p><p>In addition to showing the status quo — the inefficiency of traditional computer simulations — they also highlight the fluid dynamics complexity of plasma. Since plasma dynamics varies from the <em>Maxwellian distributions</em> used in fluid dynamics, they are difficult to represent. The modelling of several dimensions in both real and velocity space is necessary for current models, which is a challenging and computationally demanding operation²⁶. The detailed and specifics of surrogate model approaches are significantly more necessary than ever.</p><p><strong>3.1.1 Rationale for Surrogate Models in Plasma Control</strong></p><p>It has been stated numerous times that traditional fusion computer simulations are generally too slow for real-time applications in plasma control. These simulations are less suitable for real-time control, which is an important factor in nuclear fusion experiments, due to their high processing requirements and the long periods of time necessary in changing the plasma state, as plasma is very unpredictable.</p><p>To further persuade the use of surrogate models for plasma control, there has been research linked to the benefits of using these systems; specifically using Reinforcement-Learning (RL) based surrogate models.</p><p>A notable research paper in the field, <em>“Magnetic control of tokamak plasmas through deep reinforcement learning”,</em> done by <em>nature </em>has shown a flexible approach for designing tokamak magnetic confinement controllers (Fig. 3). When compared to traditional designs, the controllers created by their architecture have far simpler structural frameworks. RL-driven design generates a single network controller as opposed to a group of controllers.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/685/0*H63H32P65zP8_Bov.png" /><figcaption><a href="https://www.nature.com/articles/s41586-021-04301-9#:~:text=The%20controllers%20designed%20by%20our,creates%20a%20single%20network%20controller"><strong>Fig. 3: </strong>All creds to nature</a></figcaption></figure><p>More information from this paper shows that RL enhances flexibility and lowers design effort by making it simple to specify performance objectives and adapting to various plasma forms and configurations. The “sim-to-real” gap can be bridged and a transition from engineering-driven control to AI-driven optimization can be made possible by applying these models directly to real tokamak hardware after they have been trained on simulators as well. Finally, the models are appropriate for active reactor management and optimization because they can forecast outcomes and withstand variable operating conditions²⁷.</p><p><strong>3.2. Methodologies for Implementing Surrogate Models in Plasma Physics</strong></p><p>The three primary stages of the RL-designed magnetic controller’s operation are setting the experimental goals, working with a tokamak simulator to determine the best control strategy, and putting this strategy into practice on tokamak hardware instantly²⁷. This approach indeed closes the gap between simulation and practical implementation, as previously mentioned.</p><p>A <em>deep parameter-based transfer learning</em> method has been applied for disruption prediction in tokamaks. This method consists of training a model on one tokamak and transferring it to another with significantly different characteristics. For example, a model trained on the J-TEXT tokamak was successfully transferred to the EAST tokamak, achieving similar performance, ultimately with less training data²⁷.</p><p>A key aspect of plasma control in tokamaks is disruption prediction, which has the main goal of shielding the reactor from damage. Rule-based methods, which concentrate on understanding disruption processes, and data-driven methods, which use past data, are the two primary approaches. But it’s still difficult to gather enough information to train reliable predictors in high-performance regimes. But as previously mentioned, the transfer learning approach gets around this problem by making use of pre-existing data from other tokamaks. It has demonstrated effectiveness in models such as the Hybrid Deep-Learning (HDL) architecture and Fusion Recurrent Neural Network (FRNN), which have been trained on mixed discharges from different tokamaks and effectively used to predict disruptions in different machines²⁷.</p><p>The performance of the transfer learning model in the source domain is a major factor in determining its success in the target domain. Using data from the source tokamak, the model is first generated with high performance and then applied to the target tokamak by a<em> freeze&amp;fine-tune</em> transfer learning technique. In practice, the model is tested on the same discharges from the target domain using various methodologies to predict disruptions in the target tokamak. This procedure include fine-tuning the model for the intended purpose and selecting hyper-parameters in an organized way. The most successful method is the freeze and fine-tune approach, which provides a means of integrating information from the source and target tokamaks²⁷.</p><p><strong>3.3. Success Story of Surrogate Models for Plasma Stability</strong></p><p>The <em>DeepMind</em> project on accelerating fusion science through learned plasma control is a very accomplished advancement in surrogate modelling for nuclear fusion research. This project, in collaboration with the <em>Swiss Plasma Center</em>, marks the first time a deep reinforcement learning (RL) system has been developed to autonomously control the magnetic coils of a tokamak.</p><p>This method is different from the current plasma-control systems, which need individual controllers for every magnetic coil. DeepMind’s approach learns the best voltages for each coil directly from sensors, then using a single neural network to control them all. This device showed how to control different characteristics of the plasma, forming shapes that are useful for cooling and producing energy, such as a “snowflake” with many “legs.” A “droplet” shape was even stabilized by the system, which had never been done in a tokamak before²⁸ (<strong>Fig. 4)</strong>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/553/0*p6i13WEB9oYaMfAX.png" /><figcaption><a href="https://www.epfl.ch/research/domains/swiss-plasma-center/research/tcv/"><strong>Fig. 4:</strong></a><strong> </strong>Image depicting the magnetic control of tokamak plasma through deep reinforcement learning</figcaption></figure><p>In plasma control, the RL system showed outstanding accuracy and performance, with root-mean-square error (RMSE) values falling substantially below target levels. This involved managing both complicated, time-varying goals and basic plasma properties²⁷.</p><p>Without a doubt, Google DeepMind has made incredible contributions to the field and demonstrated the potential of any AI technique for plasma control, including surrogate modelling and RL techniques. The following are additional insights from the project:</p><ul><li>DeepMind’s RL method was extremely data-efficient, which was essential considering how computationally demanding it is to simulate plasma states.</li><li>By using a simulated environment for extended training and optimization prior to applying the results to real-world circumstances, this strategy reduced risks and eliminated the need for expensive physical tests²⁷, ultimately saving more time and money.</li><li>Because the AI system can learn and adapt on its own, it requires less human input to build new plasma configurations.</li></ul><p>DeepMind’s exploration into nuclear fusion research signifies not just an achievement in managing plasma inside a tokamak but also a significant advancement in using AI to deal with difficult scientific problems. The project’s success demonstrates how RL can potentially be used to achieve great precision and flexibility in plasma control, which could speed up fusion research and also speed up other fields.</p><p><strong>3.4. Evaluating Performance: How Surrogate Models Optimize Reactor Operations</strong></p><p>To sum up the learnings; the use of surrogate models greatly improves the efficiency of Tokamak reactor operations by allowing quick and precise predictions of plasma dynamics. These models provide an effective substitute for the numerical algorithms that are usually used in plasma physics, which are tedious and time-consuming. But let’s go more in-depth into a particular class of surrogate models: Fourier Neural Operators (FNOs), which are based on deep learning and have shown to be very successful.</p><p>FNOs are a deep learning-based technique that predicts plasma evolution accurately in both simulations and experiments. They outperform standard solvers while maintaining good accuracy. FNOs can handle many variables included in partial differential equations, which is required for capturing non-linear effects in plasma dynamics³⁰.</p><p>The use of FNOs in Tokamak reactor operations provides significant advantages such as high-speed calculations, prediction accuracy, and flexibility in dealing with complex plasma dynamics. The FNOs are designed to collect input from a variety of sources, including high-speed video footage, and are capable of forecasting future plasma states, which improves the capacity to construct efficient control feedback loops for fusion reactors³⁰ <strong>(Fig. 5)</strong>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*DskNZoaM9VGhY-XamGyHyA.png" /><figcaption><a href="https://arxiv.org/pdf/2311.05967.pdf"><strong>Fig. 5 (+ description): </strong></a>Fast camera image of a MAST plasma during an Edge-Localized-Mode, where plasma filaments are ejected outwards due to MHD instabilities at the plasma edge</figcaption></figure><p>Understanding and managing plasma behaviour, particularly instabilities such as Edge-Localized Modes (ELMs), is so important in Tokamak reactors. ELMs are plasma filaments that burst from the plasma edge and reach the first wall of the Tokamak, causing considerable operational issues. FNOs effectively simulate these complex plasma dynamics by simulating similar non-linear equations, which are important in mitigating the impacts of ELMs³⁰.</p><p>FNOs in Tokamak reactor operations require specific setups tailored to the unique requirements of plasma simulations. The FNO-2d setup is used to simulate plasma dynamics. In this scenario, the FNO analyzes field values throughout a 2D grid, using the initial fields as input and predicting field values at later times³⁰. This method allows for continuous calculation of field evolution over time.</p><p>Finally, the multi-variable FNO is specifically developed to address the complex and connected structure of variables in plasma dynamics; such as density, temperature, and electricity. This feature ensures that all important variables are taken into account within a single model, providing an in-depth and physically consistent knowledge of plasma behavior. In terms of performance, as previously mentioned, the FNO achieves high accuracy, as measured by Mean Squared Error (MSE)³⁰.</p><h4><strong>4. Advancing Predictive Capabilities in Fusion Plasmas with Surrogate Models</strong></h4><p>In this chapter, we emphasize the importance of surrogate models in predictive analysis, particularly in understanding and regulating plasma blobs, which are critical to plasma stability and containment in magnetic confinement fusion. The Gas-Puff Imaging (GPI) technique is highlighted as an innovative tool for evaluating plasma blobs, along with machine learning methods such as blob tracking, RAFT for optical flow computation, e.t.c. These methods improve the accuracy of plasma behaviour predictions.</p><p>The chapter then goes into methods for improving prediction accuracy and reliability, with a focus on deep reinforcement learning in plasma management. The design of DRL reward functions and their impact on trained agent behaviour is looked at, with emphasis on how different reward components correspond to specific criteria of an “ideal” agent. The chapter also includes major examples and findings from tokamak magnetic control research, showing how changes in reward functions affect the agent’s performance in plasma control.</p><p>Finally, the chapter discusses risks and management techniques related with the use of surrogate models in the management of fusion plasmas. It emphasizes the difficulties in precisely predicting plasma behaviour due to its complexity, as well as the dangers of disruptions that can harm the fusion device. To avoid these concerns, the chapter recommends using advanced machine learning algorithms such as Fourier Neural Operators, as well as doing thorough research and testing of surrogate models. These strategies have a chance of improving model accuracy, lowering computing costs, and improving control strategies.</p><p><strong>4.1. The Important Role of Surrogate Models in Predictive Analysis</strong></p><p>Predictive analysis in the context of Tokamak reactors, with regard to plasma <em>blobs</em>, is an important area of fusion research. Plasma blobs are important elements of a Tokamak’s plasma edge, and understanding their behaviour becomes essential for managing plasma instabilities to improve tokamak performance.</p><p>A plasma blob or bubble is a region of space containing plasma, an ionized gas containing free electrons and ions. Plasma can be formed through the ionospheric heating process or by the interaction of energetic particles with the Earth’s magnetic field. Plasma bubbles can affect radio wave propagation, which can have an influence on communication and navigation systems³¹.</p><p>A plasma bubble, also known as a blob, can have a significant impact on plasma confinement and stability in <em>magnetic confinement fusion</em> (MCF). As previously mentioned, magnetic confinement is a mechanism used in fusion reactors to keep plasma from coming into contact with the reactor walls.</p><p>As a result, we must seek new ways to decrease their development. This is owing to the fact that blobs have ruined the fusion plasma in the past. And in tokamaks, these blobs built up at the edge of fusion plasmas and drain the heat from the edge³¹. There has been research to a link between blobs and changes in the magnetic field, keeping the fusion plasma solely in the core.</p><p>The technique of <em>Gas-Puff Imaging </em>(GPI) is used to study these blobs. It involves infusing a small amount of neutral gas into the plasma and capturing the visible-light emissions produced by the plasma’s interaction with this gas cloud, helping with visualizing the processes that happens at the plasma edge³².</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*AKr-NFQqE0XHqQt8YBW8eQ.png" /><figcaption><a href="https://www.energy.gov/science/fes/articles/track-turbulence-tokamaks-researchers-turn-machine-learning"><strong>Fig. 6:</strong></a><strong> </strong>Fusion plasma with GPI found at plasma edge</figcaption></figure><p>Machine learning offers an innovative solution to this predictive analysis, allowing for <em>blob-by-blob</em> tracking for each frame. Machine learning technology is effective at identifying and tracking objects in photos, and many models are available for this purpose. Blob-tracking gives thorough, time- and space-resolved information on blobs and is a useful technique for estimating blob statistics³². As a result, it represents a possible tool for building and operating future fusion power devices.</p><p>The models are trained using synthetic data, allowing for complete understanding and control over features and noise. Synthetic data can represent a greater range of scenarios than real data, including edge cases such as blob splitting and merging. Also, the experimental GPI videos are made up of a sequence of grayscale images that have been standardized and remapped to a higher resolution for better model interpretation³².​</p><p>For predictive analysis, many advanced tracking techniques are used. Using the <em>Recurrent All-Pairs Field Transforms </em>(RAFT) approach, optical flow is computed by building multi scale 4D correlation volumes for each pair of pixels. The optical flow field is then repeatedly updated using these volumes, allowing for detailed tracking of pixel movements across time. For segmenting and understanding blobs in data, the <em>Mask R-CNN with Bayesian Optimization </em>technique is applied. Bayesian optimization is used to fine-tune the hyper parameters of the <em>Mask Region-based Convolutional Neural Network </em>(Mask R-CNN). With a very detailed two-level optimization procedure, it ensures that the model accurately and precisely splits the plasma blobs³².</p><p><em>Global Motion Aggregation</em> (GMA) and <em>Flow Walk </em>are the following two steps in this four-step predictive analysis procedure. GMA first estimates hidden motions within plasma pictures. It accomplishes this by identifying long-distance correlations between pixels and collecting motion characteristics. This method uses a transformer model, which excels at dealing with large-scale and complex data structures. The Flow Walk method relies on learning pixel trajectories. It does this by calculating transition matrices between frames using multi scale contrastive random walks, useful for identifying small, pixel-level changes in images³².</p><p>Finally, to evaluate performance, two key metrics are used:</p><p><em>Endpoint Error</em> (EPE): This is a commonly used metric for determining optical flow accuracy. It quantifies the difference between the estimated and ground truth flow fields.</p><p><em>Volumetric IoU </em>(VIoU): VIoU is used to examine the quality of object detection. It computes the intersection and union areas of predicted and ground truth blob masks, taking brightness level into consideration as an additional dimension.</p><p><strong>4.2. Techniques for Enhancing Accuracy &amp; Reliability in Predictions</strong></p><p>The most effective approach of using <em>deep reinforcement learning </em>(RDL) involves many key steps. Reinforcement learning is a subset of machine learning that models control problems as interactions between an agent and its surroundings. The agent can be thought of as the control mechanism, and the environment as the system to be controlled. The agent receives a collection of signals referred to as observations and sends a set of control signals referred to as actions³³.</p><p>RL algorithms strive to maximize a generically defined reward signal, whereas classical control algorithms seek to minimize the error of an actively measured quantity. During training, the agent’s behaviour is guided by the reward maximization goal. Although, the reward value is not computed during installation. The performance of the controller in classical control algorithms can be modified by explicitly setting control gains and altering trade-off weights for<em> multi-input multi-output</em> (MIMO) systems. In RL, on the other hand, the reward function is crucial to the learnt controller’s behaviour³³. To modify controller behaviour, the reward function must be carefully designed. Let’s look at how changing the reward’s design can trigger desired behaviours in the resulting trained agent.</p><p>Different requirements of an ideal agent correspond to reward components. Each reward component is calculated by reducing the required value from the value reported by the simulated environment. This difference is subjected to non-linear scaling and processing, resulting in an effective reward for that component. A non-linear combination of the various component rewards is used to compute the overall reward³³.</p><p><strong>4.2.1. Predictive Success: Key Examples and Results</strong></p><p>All information from this section will build off from the last.</p><p>The research done in “<em>Towards practical reinforcement learning for tokamak magnetic control” </em>has<em> </em>combined the reward component values using a weighted SmoothMax function. In other circumstances, an individual reward component is made from many related error variables, such as shape error at various control points. The SmoothMax function is also used to integrate these faults into a single scalar component reward³³.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*sIQjnfdO3pxFyTZUeB1uxQ.png" /></figure><p>The <em>wi</em> variable directly reveals the significance of each component. The value <em>a</em> influences the choices between components that are “easy” and “hard” to meet. If the <em>a</em> value is substantially less than zero, it indicates that the agent’s reward is almost equivalent to the component on which it does the least well, whereas a value close to zero indicates that all components are in equilibrium. A value of <em>a </em>greater than zero is improper because it would emphasize components that are highly regulated at the expense of components that are poorly controlled³³. Reward components are not required to be measured from sensor measurements, resulting in greater flexibility in their design. Reward components can also be multi-modal to drive behaviour away from unpleasant or poorly modelled parts of state-space by the simulator. In this paper, the use of a SoftPlus was also employed:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*YxwaaFuMSV-hPLTaagpTDg.png" /></figure><p>The combination of good and bad parameters work together to scale the reward signal into an ideal region. If the true value is worse than the bad, the reward rapidly decays towards zero, whereas if it is at or better than good, the reward saturates at one. If there is still confusion, think that the parameter algorithm influences the “sharpness” of reward scaling between good and bad reference points. For getting a final scalar reward, the set of scalar reward components is blended with a SmoothMax operator³³. <strong>Fig. 7</strong> depicts these functions as well as the impact of hyper-parameter selection.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*XxWQb2XRtwq6t-ohb_E2jQ.png" /><figcaption><a href="https://arxiv.org/pdf/2307.11546.pdf"><strong>Fig. 7 (+description): </strong></a>Visualization of the influence of reward hyperparameters</figcaption></figure><p>A variety of parameter selections (should) produce comparable results because they represent incremental changes to the reward. However the reliance on gradient descent and the lack of a perfect global optimizer complicate matters. Setting limits for what defines good and bad performance can make identifying a zone with important rewards difficult. Setting a broad parameter for bad performance may make it easier to detect a reward signal, but it can also impact the discovery of precise control due to the decreased variation in reward with progress. Setting stricter reward criteria might therefore be more suited for cases where the beginning conditions are already close to the target state³³, reducing the need for the algorithm to assist both goal discovery and fine-tuned precision.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*bX9nSZPup8xVcUKpZpyGNA.png" /><figcaption><a href="https://arxiv.org/pdf/2307.11546.pdf"><strong>Table 1: </strong></a>Results from reward adjustment in the simple stabilization task</figcaption></figure><p>Table 1 shows the results from this experiment. It demonstrates that the reward can have an important effect on the final trained agent’s performance. Focusing on the shape mistake, we see that the Narrow Reward, with its highly demanding static reward function, had the most impact³³. The more precise reward mechanism provides a substantial incentive for the controller to increase accuracy in this simple activity.</p><p>As discussed, traditional tokamak control used precomputed feedforward coil currents and feedback loops for plasma stabilization, which was effective but needed major engineering and design efforts for new plasma designs²⁷. RL, on the other hand, reduces the control system architecture, replaces layered controls with a single network controller, and takes in state reconstruction, ultimately shortening the controller development period.​</p><p><strong>4.3. Surrogate Models Risks and Management</strong></p><p>Because of the complexity and unpredictability of plasma behaviour, the use of surrogate models in tokamaks for plasma management presents certain risks. Surrogate models have been used to reduce the computational requirements of simulating plasma behaviour, although their accuracy and dependability vary.</p><p>One associated risk of surrogate models includes accuracy; they may not correctly predict plasma behaviour in experimental settings. Despite progress in simplified, physics-based model building, the ability of these models to predict plasma behaviour in burning-plasma environments requires extensive research³⁴. This is because the models implemented in these surrogate models are frequently based on scientific methods and integrated techniques that may not fully capture the complex nature of plasma behaviour³⁴.</p><p>Another danger is the inability to control plasma shape and position. While surrogate models can be used to create experience for reinforcement learning algorithms to train on, the dynamics of the plasma’s shape are better understood than other features of the plasma, and can consequently be modelled and controlled reasonably precisely³⁵.</p><p>Finally, as for risks, disruptions are a key barrier for tokamak development since they limit the tokamak’s operational zone and can damage the entire fusion device, preventing the completion of a functioning plasma reactor. The lack of first-principles modelling of plasma disruption makes disruption a more serious issue³⁴.</p><p>The following will be methods we could potentially reduce the risks listed above, whether they were previously stated or not.</p><p>To improve surrogate model accuracy, advanced machine learning approaches such as neural operators, specifically Fourier Neural Operators (FNOs) as mentioned before, can be used. These models can predict plasma evolution quickly and accurately by learning directly from the solution operator of partial differential equations in a <em>discretization-convergent</em> way³⁶. They can forecast at any resolution, are more data-efficient, and converge faster than standard models, which is essential to capturing complicated plasma behaviours.</p><p>Extensive study and benchmarking of surrogate models, such as the FNO, against more traditional models helps in evaluating their performance and potential areas for improvement. This includes assessing the impact of various hyper parameters across the model and training to ensure that the models are suitable for the complicated circumstances of plasma control³⁶. Even developing reduced models that can capture important dynamics can be helpful for reducing the risk of disruptions and controlling plasma shape and location. These models are less computationally costly and can provide useful insights for control strategies³⁶.</p><h4><strong>5. Challenges and Innovations in Surrogate Modelling</strong></h4><p>In this chapter, we look at the overcoming present limitations and exploring future potential in surrogate modelling. Due to expenses and high-dimensionality of systems, one key problem observed is the production of diiverse datasets required for training surrogate models to accommodate for these systems. In high-dimensional environments, traditional surrogate modelling methods such as Kriging and Gaussian Processes have scaling difficulties. Recent advances in Machine Learning (ML) and Deep Learning (DL) show promise in this field. We take a look at how to counteract these current limitations for the most effective method of surrogate modelling.</p><p>The chapter also discusses the problem of oversampling in current datasets derived from experimental parameter spaces, which results in a significant amount of irrelevant data for the surrogate model. To tackle this, new modelling strategies such as response feature technology, which allows surrogate models to learn from small training datasets, are proposed in the research paper. This method is very useful in the fusion field as electromagnetic simulations are a necessity. Bayesian neural networks are also identified as an approach to uncertainties in surrogate models, where accuracy and reliability in fusion are extremely crucial. Also, the importance of hyper parameter tuning in surrogate modelling is emphasized, as it has a significant impact on model efficiency and performance.</p><p>Overall, this chapter presents a thorough review of the challenges and new solutions in the field of surrogate modelling, highlighting its potential and expanding significance in a variety of industries such as plasma management for fusion reactors, healthcare, and aerospace.</p><p><strong>5.1. Addressing Current Limitations in Surrogate Modelling</strong></p><p>A major obstacle is the generation of large and diverse training datasets required for training surrogate models. High-fidelity codes are frequently computationally expensive for these datasets. For example, a recommended model, Active Deep Ensembles for Plasma Turbulence (ADEPT), shows the difficulty of covering a wide volume in parameter space for training due to the prohibitive cost associated with higher resolution codes³⁸. Traditional surrogate modelling methods, such as Kriging and the Gaussian Process, have scaling constraints, particularly in high-dimensional domains. Recent breakthroughs in Machine Learning (ML), but mainly Deep Learning (DL), show promise in this field. DL models thrive at learning in high-dimensional domains and extracting important features automatically. Even with DL, the challenge of selectively sampling and adaptive step sizing must be addressed for effective surrogate modelling³⁹.</p><p>Current datasets generated from experimental parameter spaces are frequently oversampled, which means that a large amount of the data may be useless to the model. This difficulty is shown by a study in the field of electromagnetic (EM) simulation tools. In this scenario, a single full-wave simulation can take a long time, making the cost of multiple EM studies excessive. Surrogate models, which are faster and cheaper but still accurate <em>meta models</em>, are used to combat this, as we know. However, the use of these surrogate models can be affected by the problem of dimensionality. To address this, a new modelling technique incorporating response feature technology may be used, allowing for surrogate models to learn from extremely little training datasets⁴⁰.</p><p>Performance-driven modelling frameworks are another technique to dealing with dimensionality difficulties in surrogate modelling. The most recent advances in this field, such as reference-design-free constrained modelling, have significantly lowered the computational expenses made in obtaining reference designs, which was a key portion of the process in previous techniques. This surrogate modelling strategy enhances the efficacy of this method even further by using response features technology, which handles the modelling task at the level of system response features rather than frequency characteristics in their original form⁴⁰.</p><p>Handling stable and unstable plasma states in tokamak turbulence surrogate models is a difficult task. Growing modes and turbulent transport are triggered only over a particular threshold in driving gradients, which is dependent on plasma conditions, in this turbulence. Previous techniques used a physics-based loss function to maintain consistency with the critical threshold behaviour of tokamak turbulence, which supported controlled extrapolation to negative values when output fluxes were zero³⁸.</p><p>To overcome this inefficiency, ADEPT, as previously mentioned, can be used. ADEPT decreases the need for simulations in stable parts of the parameter space. When compared to non-physics driven techniques, this leads in increased data efficiency. The ADEPT workflow consists of a two-stage active learning (AL) framework where a classifier acts as a surrogate of the critical gradient threshold, determining if a given input would result in growing modes. The output turbulent transport fluxes are then predicted by a regressor. This method enhances accuracy by focusing on inputs with the largest output uncertainty of the neural network³⁸.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/953/0*QX98a1dJD73CASRl.png" /><figcaption><a href="https://ar5iv.labs.arxiv.org/html/2310.09024"><strong>Fig. 8 (+description) :</strong></a><strong> </strong>Diagram of the two-stage physics-informed AL workflow. This strategy is repeated until the computational budget has been exhausted or the performance of the surrogates is deemed actionable.</figcaption></figure><p>Other machine learning approaches, in addition to ADEPT, are being used to better comprehend tokamak plasma turbulence. Gas-Puff Imaging (GPI) is used to visualize processes at and near the plasma barrier. GPI data processing, which can include up to one million frames each experiment, is too large for human analysis. Machine learning offers a new method by tracking plasma blobs “blob-by-blob” in each frame, providing significant, time- and space-resolved information on the blobs³².</p><p>Calibration and training of neural network uncertainties in surrogate models is important in scientific and engineering disciplines when predicted accuracy and reliability are involved. This calibration is especially difficult because uncertainties regularly grow when forecasts move away from the training distribution. However calibrating these uncertainties is so crucial for the model’s prediction dependability³⁸.</p><p>Using Bayesian neural networks (BNNs) is one key technique to tackling this issue. This technique suggests that while a “correct” model exists, it cannot be found due to a lack of data. Instead, a variety of models are explored, each one weighted by its future likelihood. It works by randomly dropping nodes during training and using the full network for predictions, but it produces a predictive probability distribution on the predictions by keeping dropping activated at prediction time⁴¹. This method, however, has drawbacks. One being the level of prediction uncertainty being decided by user-specified hyper parameters that lack clear meaning. We can combat this by implementing a consistent requirement on prediction uncertainties over a validation dataset³⁸. This involves using a calibration method to ensure that the uncertainty in a forecast accurately represents the probability that the prediction is correct.</p><p>In the context of surrogate modelling, hyper parameter tuning is an essential component that has a considerable impact on the models’ efficiency and performance. For the best possible model, hyper parameter adjustment should ideally be carried out throughout every stage of machine training³⁸. However, due to the large computing expenses involved, this method is frequently avoided… why? The process of modifying model parameters that are not immediately learned from data is known as hyper parameter tuning. These variables have a significant impact on the learning process and the model’s capacity to extrapolate from training data. Re-tuning the hyper parameters at every stage helps ensure that the model is always optimized for the current dataset in active learning scenarios⁴².</p><p>Failure to tune hyper parameters in surrogate models can result in poorer performance. As previously mentioned, without regular tuning, the model may be unable to react to new data patterns or changes in data distribution. This is especially problematic for plasma control, where surrogate models are used for sophisticated simulations and decision-making processes where model correctness and durability are critical⁴³.</p><p>A very large section of this research paper touched on the advantages of FNOs, however they also present limitations for surrogate modelling.</p><p>With the development of sustainable fusion reactors, Fourier Neural Operators (FNOs) have become an important tool in surrogate modelling for plasma dynamics within Tokamak reactors, as already known. When predicting plasma dynamics from complex plasma physics models, FNOs provide an enormous speedup — six orders of magnitude quicker than traditional simulators — while maintaining good accuracy⁴⁴. However one limitation of FNOs is their low capacity to make long-term timing predictions. While effective for short-term roll-outs, their performance varies when long-term predictions are added to the mix. The training process of the FNOs is mostly responsible for this variation. Prediction errors accumulate over longer time roll-outs in such a complex system, resulting in greater variances from actual plasma behaviour as the prediction window increases⁴⁵. Other model configurations, such as U-NET structures have been found to outperform regular FNOs in accuracy and stability when predicting turbulence over longer time periods. These models excel at handling complex flows and are critical for obtaining stable and accurate forecasts over long periods of time⁴⁶.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*pwUHI5GbLqcyjEhWqwuTFA.png" /><figcaption><a href="https://www.researchgate.net/publication/372625706_Towards_Long-Term_predictions_of_Turbulence_using_Neural_Operators/figures"><strong>Fig. 9 (+description) :</strong></a><strong> </strong>Diagram of the UNO architecture. Each block of the decoder is an FNO whose output is in a domain smaller than the previous. The middle processor layer is a standard FNO layer on the reduced domain.</figcaption></figure><p><strong>5.2. Successful Research of Surrogate Models Done Outside of Fusion</strong></p><p>The industry of <em>Soft Computing </em>has greatly benefit from the use of surrogate models. Copious amounts of research into recent advances and applications of surrogate modes for finite element method computations have been reviewed and further researched into. Surrogates have been effectively used by researchers in a variety of technical and scientific domains, not just nuclear fusion, due to their capacity to deal with challenges and other computational requirements. As we all know, efficient use of surrogates can result in significant savings in computational resources and time. A recent paper analyses important publications that used surrogate modelling for finite element method-based computations. In their words, they describe the function of surrogate modelling, sampling, and model validation techniques, as well as the many surrogate types. Then, in detail, cover the main areas where surrogate models are used: prediction, sensitivity analysis, uncertainty quantification, and surrogate-assisted optimization, as well as current breakthroughs and applications⁴⁷. Let’s take a dive into the problem that occurs and briefly discuss how surrogate models aided.</p><p>Numerical analysis methods such as the <em>finite-element method </em>(FEM), <em>computational fluid dynamics </em>(CFD), and s<em>tructural finite-element analysis</em> (FEA) are consistently used to analyze complex systems &amp; structures where obtaining an analytical solution may be difficult, or at least near impossible. According to the paper, such situations are becoming increasingly common in evaluating and optimizing the design, reliability, and maintenance of complex systems + structures in a wide range of industrial applications; including aerospace, automotive, architecture, biomedical engineering, chemical engineering, and many others⁴⁷.</p><p>However it’s without a doubt that these computer simulations are exceedingly computationally intensive, as we already know from the discussion about nuclear fusion above. As a result, we may find this comparable trend across a number of industries. These engineering challenges based on computer models also require the computing of thousands of simulations in order to develop an appropriate solution, requiring an enormous computational budget, as we are well aware of. Furthermore, due of their accuracy, different challenges in creating computer simulations can exist regardless of how much computer power is available. Even recent advances in parallel and pooled computing approaches, which simply put, perform multiple calculations or process executions at the same time, are not appear very useful⁴⁷.</p><p>The <em>aerospace </em>industry has recognized that the use of surrogate models is a way of speeding up design search and optimization in many fields of engineering that require the use of expensive computer simulations. This research paper acknowledges that surrogate models can be used to identify optimal solutions quicker and provide insights into the links between design elements and performance measures. They are also less susceptible to noise and uncertainties, allowing them to provide accurate forecasts even with insufficient data. Artificial intelligence, particularly machine learning, had proved that it may be used as a substitute approach in aerodynamic design. For example, neural networks for predicting the aerodynamic coefficients of airfoils and wings, or airfoil inverse design have been developed and demonstrated to speed up the design process, with the added benefit of efficiently and accurately predicting solutions to nonlinear problems⁴⁸. There are two techniques to modelling aerodynamic data with neural networks studied in the paper: aerodynamic response prediction and flow field reconstruction⁴⁸. The employment of appropriate techniques to develop aerodynamic data models and produce the expected amount of the response parameters in the case of given design parameters is referred to as aerodynamic response predictions. This network type can be applied in both direct and inverse design processes⁴⁸. Both multilayer perceptron (MLP) neural networks and convolutional neural networks (CNN) can be used for predicting a constant aerodynamic response. However, the use of MLP or CNN is determined by the task at hand and the available data⁴⁸. Several MLPs have been created throughout history to predict aerodynamic coefficients, as previously mentioned, ranging from predicting a single coefficient for an airfoil to predicting multiple coefficients for an airfoil, to even predicting aerodynamic coefficients of wing configurations⁴⁸.</p><p>Finally, the last industry that has been taking great advantage and benefiting from surrogate models is the <em>healthcare </em>industry. For context, a very prevalent issue at the forefront of bio-mechanical research is the development of quick algorithms for evaluating material properties from clinical data. disease, such as myocardial infarction, has a significant impact on human quality of life. Early detection and treatment can significantly reduce the occurrence and mortality from heart disease, and with the advancements of medical image technology, such as magnetic resonance imaging (MRI), in recent years, it has offered crucial information for non-invasively describing the structure and function of the human heart⁴⁹. However, imaging data alone does not convey every aspect of heart function.</p><p>We have already reviewed the FEM method, which has actually been successful in recent years, for the application of cardiac modelling. FEM is used for material parameter estimation of a heart model, solving a constrained optimization problem by minimizing the mismatch between limited measured data and FEM model <strong>(Fig. 10) </strong>predictions by identifying potential material parameters⁴⁹. However, often the problem is computationally costly and difficult to solve, often leading to computational demand and unknown parameters. Surrogate models have been created for fast parameter inference in bio-mechanical cardiac models to reduce computational expenses in a constrained optimization problem⁴⁹. According to the research, myocardial property could be a possible biomarker for predicting ventricular pump function recovery following a heart attack. Estimation of cardiac material parameters from image-based models has piqued researchers’ interest, either through the formulation of a gradient-based inverse problem or through the use of machine learning (ML)-based surrogate techniques for quick parameter inferences. Furthermore, the behaviour of the left ventricular (LV) in response to changes in material properties, stresses, and boundary conditions can be predicted in real time using ML models, which can then be applied to the design of medical equipment for monitoring heart condition⁴⁹. There was a study done to propose deep learning as a surrogate of FEM for stress analysis, and the trained model was capable of predicting aortic stress distributions by replacing the complex structural finite-element analysis with an average error of less than 1% for ML-predicted stress distribution. This is all done owing to surrogate modelling.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/850/1*iy4n93o2AlqelQqtCZc_vA.png" /><figcaption><a href="https://www.researchgate.net/publication/344264656_Impact_of_Aortic_Stenosis_on_Myofiber_Stress_Translational_Application_of_Left_Ventricle-Aortic_Coupling_Simulation/figures?lo=1&amp;utm_source=google&amp;utm_medium=organic"><strong>Fig. 10:</strong></a><strong> </strong>Solid Finite-Element Model of the Human Heart</figcaption></figure><p>This section was designed to provide an in-depth overview of real-world surrogate model applications. The entire research paper has revolved around surrogate modelling strategies for tokamak fusion plasma controls, understanding the different various types of instabilities, techniques, why they are so effective, and how they work. However, this part focuses primarily on how applicable they are in everyday life, whether in the healthcare or aerospace industries. It was also beneficial as we gained an understanding of how surrogate models work in various domains while also learning about their larger-scope goal: optimization of tasks working with difficult computational tasks.</p><p><strong>5.3. Solutions to Current Limitations in Surrogate Modelling</strong></p><p>One of the main limitations in surrogate modelling is the ‘curse of dimensionality’, referring to the exponential increase in computational demand as the number of system parameters grows. To solve this, dimensionality reduction (DR) techniques are used to develop more cost-effective surrogates by converting high-dimensional data into a lower-dimensional space, without significantly impacting the information quality. This simplifies high-dimensional problems and makes surrogate models more manageable⁵⁰. DR can be accomplished through feature removal, selection, or separation. Feature removal is the process of decreasing the input dimension space by removing irrelevant dataset features. Although eliminating those traits simplifies future computations, no information is gained. Meanwhile, feature selection is the process of identifying and sorting features based on their information contribution to the total dataset using statistics. Finally, feature extraction methods generate new independent features by combining the original dataset features⁵⁰.</p><p>According to a research paper published in Data Science and Engineering⁵⁰, dimensionality reduction techniques can be classified as linear or non-linear. For example, linear DR approaches use <em>linear kernelization</em> and<em> original variable </em>combinations to transform data to a lower-dimension feature space. On the other hand, non-linear DR approaches typically produce a lower-dimensional representation of the data while preserving the distance between data points⁵⁰. In the next few paragraphs, we will look at the specific techniques towards tackling limitations in surrogate modelling, as per this paper.</p><p>The first technique we’ll look at is <em>unsupervised linear dimensionality</em> <em>reduction</em>, which is used during the pre-processing step of surrogate modelling. This approach is trained on unlabeled data and take input features into account while developing lower-dimensional representations. <em>Factor analysis</em> (FA) is another unsupervised dimensionality reduction technique for discovering correlations among a set of random variables seen or quantified for each datapoint in a group. The method looks for the number of factors and the underlying structure of a set of variables in order to reduce dimensionality⁵¹. The following information will now be about how we can address the current limitations of surrogate modelling for fusion plasmas.</p><p>Using unsupervised linear DR approaches during the surrogate modelling pre-processing step in plasma control would consist of simplifying the input features for the surrogate models by training them on the unlabeled data frequently gained in tokamak experiments. Tokamaks are supplied with several sensors that generate massive volumes of data. Linear DR can be used to reduce the dimensionality of this sensor data, making it easier for surrogate models to interpret and analyze. We can equip the factor analysis technique to find correlations between different plasma parameters. This may help in finding underlying structures and prominent events in plasma behaviour, which are crucial for effective tokamak control and prediction. Factor analysis can also greatly reduce the complexity of surrogate models by defining the number of factors and the structure of plasma variables. This results in more efficient models that can work in real-time, which is ultimately very critical for plasma control systems.</p><p>In fusion research, we can use feature selection to determine the most informative and relevant plasma parameters. This guarantees that surrogate models concentrate on the most important features of plasma behaviour, enhancing accuracy as well as computational efficiency. Another advantage of feature removal is the removal of excessive or irrelevant data points from plasma control datasets. This simplification can lower the computational load on the surrogate models significantly. Finally, because tokamak plasma behaviour is highly complex and nonlinear, nonlinear DR approaches can be very effective. These methods can result in more precise plasma control and prediction. Non-linear DR techniques can be used to display high-dimensional plasma data in a lower-dimensional environment, helping in understanding and making sense of complicated plasma behaviours.</p><p>Overall, in section <strong>5.1</strong>, we discussed the existing limits of surrogate modelling. We have now looked at techniques for avoiding these limitations, which are specific to the control of tokamak fusion plasmas. Applying these DR techniques to the field of plasma control involves using them to simplify &amp; interpret the complex data generated in these systems. By reducing the dimensionality, while keeping critical information, these techniques ultimately enhance the performance of surrogate modelling, resulting in more efficient predictive tools.</p><h4>6. Conclusion</h4><p>In this final chapter, we present a complete synthesis of the research’s important insights and conclusions. It summarizes the main points made from each chapter and the application to plasma management in tokamaks using surrogate models. We also look at the key recommendations for enhanced surrogate modelling that we can use for the chance at a clean future.</p><p><strong>6.1 Synthesis of Key Insights and Findings</strong></p><p><em>Chapter 1: Introduction to Nuclear Fusion and Plasma</em></p><ul><li>Nuclear Fusion is the process of two or more atomic nuclei (Deuterium + Tritium) combining to form a heavier nucleus. This energy source is attractive due to its scalability, sustainability, and carbon-free emissions.</li><li>Fusion occurs in an ionized fluid called plasma, however achieving the extreme conditions of this fourth state of matter is extremely challenging. It comes with a variety of plasma instabilities, such as kink, ballooning, and tearing mode instabilities.</li><li>Solutions for Tokamaks to develop real-time plasma control prediction systems is essential for maintaining plasma instabilities and harnessing fusion energy.</li></ul><p><em>Chapter 2: Evolution of Surrogate Modelling</em></p><ul><li>Surrogate models are efficient + simplified alternatives to complex systems as seen in fusion energy. By building these models that mimic the behaviour of the more advanced systems, it is possible to predict outcomes based on fresh input data with the use of surrogate modelling.</li><li>A key type of surrogate model is Bayesian networks: a type of Probabilistic Graphical Model that can be used to build models from data and/or expert opinion</li><li>Surrogate models provide a plethora of advantages. Some notable ones include reduced costs, increased efficiency, and informed decision-making. All of which could aid in the research of fusion energy.</li></ul><p><em>Chapter 3: Surrogate Models in Fusion Plasma Control</em></p><ul><li>Nuclear fusion research requires extremely sophisticated computations to estimate the temperature and density profiles of plasma. However, even the most powerful supercomputers find it difficult to meet such enormous processing demands.</li><li>Surrogate models have been proven to benefit plasma prediction. Reinforcement Learning based surrogate models have enhanced flexibility and simplified the Tokamak behaviour.</li><li>The <em>DeepMind</em> project on accelerating fusion science through learned plasma control is a very accomplished advancement in surrogate modelling for nuclear fusion research, as it marks the first time a deep reinforcement learning system has been developed to autonomously control the magnetic coils of a tokamak.</li></ul><p><em>Chapter 4: Advancing Predictive Capabilities with Surrogate Models</em></p><ul><li>Plasma blobs, a region of space containing plasma, are important elements of a Tokamak’s plasma edge, and understanding their behaviour becomes essential for managing plasma instabilities to improve tokamak performance as they impact plasma confinement.</li><li>Machine learning offers an innovative solution to this predictive analysis, allowing for <em>blob-by-blob</em> tracking for each frame.</li><li>The most effective approach of using <em>deep reinforcement learning </em>(RDL) involves many key steps. RL algorithms strive to maximize a generically defined reward signal, whereas classical control algorithms seek to minimize the error of an actively measured quantity.</li><li>One risk of surrogate models is the inability to control plasma shape and position. Rather the dynamics of the plasma’s shape are better understood than other features of the plasma, and can consequently be modelled and controlled reasonably precisely.</li></ul><p><em>Chapter 5: Challenges and Innovations in Surrogate Modelling</em></p><ul><li>A major obstacle is the generation of large and diverse training datasets required for training surrogate models, as they can be expensive to repeatedly train on. Performance-driven modelling frameworks are a technique for dealing with dimensionality difficulties in surrogate modelling, as they have significantly lowered the computational expenses made in obtaining reference designs.</li><li>Many industries outside of fusion energy have greatly benefited from surrogate modelling. One notable example is the aerospace industry. Neural Networks have been used for predicting the aerodynamic coefficients of airfoils and wings, with the added benefit of efficiently and accurately predicting solutions to nonlinear problems.</li><li>One of the main limitations in surrogate modelling is the ‘curse of dimensionality’, referring to the exponential increase in computational demand as the number of system parameters grows. However, this can be solved using dimensionality reduction techniques. These techniques develop more cost-effective surrogates by converting high-dimensional data into a lower-dimensional space, without significantly impacting the information quality.</li></ul><p><strong>6.2 Reflections &amp; Recommendations for the Future of Fusion Energy and Surrogate Modelling</strong></p><p>Despite technological challenges, the promise of fusion energy as a sustainable, abundant source is more important than ever in our attempts to mitigate climate change. Its ability to provide a continuous, carbon-free energy supply makes it even more attractive. Surrogate modelling, on the other hand, may be our only option to reach fusion energy. Surrogate models are useful as traditional computer simulations struggle with complex systems in fusion energy, such as reactor design or plasma physics. Now that the knowledge of fusion energy and surrogate modelling has been completely laid, the following section will outline key recommendations to reach our goal of a clean future.</p><p><em>Integration of AI &amp; ML for Surrogate Modelling</em></p><p>Future research should concentrate on incorporating innovative AI and ML tech into surrogate modelling. Deep learning, reinforcement learning, and neural networks tailored specifically for plasma control in fusion reactors are all great choices. Also, attention should be focused on improving surrogate models’ predictive capabilities, for accuracy, speed, and the ability to handle complicated datasets.</p><p><em>Plasma Instabilities and Confinement Methods</em></p><p>Conducting more research to better understand plasma instabilities such as kinking, ballooning, and tearing should be prioritized. The goal of the research should be to create surrogate modelling strategies to mitigate these instabilities. Also, investigating more efficient magnetic confinement techniques would increase plasma stability and containment efficiency.</p><p><em>Collaboration Amongst Industries</em></p><p>We saw how various industries have made tremendous development for surrogate models. Surrogate modelling research could thrive if these scientists collaborated and applied surrogate modelling techniques to other areas of research.</p><p>In conclusion, this research paper has provided extensive research into the use of surrogate modelling tools to improve plasma control and predictability, in nuclear fusion and tokamak reactors. They have shown great potential and by managing the complexity of plasma behaviour and other systems obstacles. This research paper highlighted the innovative influence that surrogate models can have on the development of fusion energy. Of course despite technological, scientific, and computational challenges, fusion energy with the help of surrogate models holds the possibility of contributing to a more sustainable world.</p><h4><strong>References</strong></h4><p>PHYSICS-INFORMED KRIGING: A PHYSICS-INFORMED GAUSSIAN PROCESS REGRESSION METHOD FOR DATA-MODEL CONVERGENCE. Retrieved from <a href="https://engineering.lehigh.edu/sites/engineering.lehigh.edu/files/_DEPARTMENTS/ise/pdf/tech-papers/19/19T-026.pdf">https://engineering.lehigh.edu/sites/engineering.lehigh.edu/files/_DEPARTMENTS/ise/pdf/tech-papers/19/19T-026.pdf</a></p><p>Wikipedia, the free encyclopedia. Retrieved from <a href="https://surface.syr.edu/etd/227/">https://surface.syr.edu/etd/227/</a></p><p>Wikipedia, the free encyclopedia. Retrieved from <a href="https://www.linkedin.com/pulse/surrogate-modeling-powerful-tool-engineers-elijah-toppo/">https://www.linkedin.com/pulse/surrogate-modeling-powerful-tool-engineers-elijah-toppo/</a></p><p>Wikipedia, the free encyclopedia. Retrieved from <a href="https://arxiv.org/pdf/2311.05967.pdf">https://arxiv.org/pdf/2311.05967.pdf</a></p><p>Wikipedia, the free encyclopedia. Retrieved from <a href="https://ar5iv.labs.arxiv.org/html/2310.09024">https://ar5iv.labs.arxiv.org/html/2310.09024</a></p><p>Wikipedia, the free encyclopedia. Retrieved from <a href="https://ar5iv.labs.arxiv.org/html/2211.10360">https://ar5iv.labs.arxiv.org/html/2211.10360</a></p><p>(2017, November 9). YouTube: Home. Retrieved from <a href="https://www.etymonline.com/search?q=surrogate+model">https://www.etymonline.com/search?q=surrogate+model</a></p><p>(2017, November 9). YouTube: Home. Retrieved from <a href="https://www.linkedin.com/pulse/artificial-intelligence-24-surrogate-models-behind-digital-jaokar/">https://www.linkedin.com/pulse/artificial-intelligence-24-surrogate-models-behind-digital-jaokar/</a></p><p>(2017, November 9). YouTube: Home. Retrieved from <a href="https://arxiv.org/pdf/2307.11546.pdf">https://arxiv.org/pdf/2307.11546.pdf</a></p><p>(2017, November 9). YouTube: Home. Retrieved from <a href="https://ar5iv.labs.arxiv.org/html/2311.05967">https://ar5iv.labs.arxiv.org/html/2311.05967</a></p><p>(2017, November 9). YouTube: Home. Retrieved from <a href="https://ar5iv.labs.arxiv.org/html/2010.13749">https://ar5iv.labs.arxiv.org/html/2010.13749</a></p><p>(2017, November 9). YouTube: Home. Retrieved from <a href="https://pubs.aip.org/aip/pof/article-abstract/35/7/075145/2903750/Long-term-predictions-of-turbulence-by-implicit-U?redirectedFrom=fulltext">https://pubs.aip.org/aip/pof/article-abstract/35/7/075145/2903750/Long-term-predictions-of-turbulence-by-implicit-U?redirectedFrom=fulltext</a></p><p>(2019, February 24). Materials Challenges in Superconducting Tokamaks. Retrieved from <a href="http://large.stanford.edu/courses/2019/ph241/dayrit2/">http://large.stanford.edu/courses/2019/ph241/dayrit2/</a></p><p>(2023, November 13). , — YouTube. Retrieved from <a href="https://www.sciencedirect.com/topics/engineering/surrogate-model">https://www.sciencedirect.com/topics/engineering/surrogate-model</a></p><p>(2023, November 13). , — YouTube. Retrieved from <a href="https://www.sciencedirect.com/topics/computer-science/bayesian-networks">https://www.sciencedirect.com/topics/computer-science/bayesian-networks</a></p><p>(2023, November 13). , — YouTube. Retrieved from <a href="https://www.sciencedirect.com/topics/earth-and-planetary-sciences/artificial-neural-network">https://www.sciencedirect.com/topics/earth-and-planetary-sciences/artificial-neural-network</a></p><p>(2023, November 13). , — YouTube. Retrieved from <a href="https://www.nature.com/articles/s41586-021-04301-9">https://www.nature.com/articles/s41586-021-04301-9</a></p><p>(2023, November 13). , — YouTube. Retrieved from <a href="https://www.sciencedirect.com/science/article/pii/S2590123023003407">https://www.sciencedirect.com/science/article/pii/S2590123023003407</a></p><p>(2023, November 13). , — YouTube. Retrieved from <a href="https://iopscience.iop.org/article/10.1088/1741-4326/ac64b2">https://iopscience.iop.org/article/10.1088/1741-4326/ac64b2</a></p><p>(2023, November 13). , — YouTube. Retrieved from <a href="https://www.nature.com/articles/s41598-022-08710-2">https://www.nature.com/articles/s41598-022-08710-2</a></p><p>(2023, November 13). , — YouTube. Retrieved from <a href="https://www.nature.com/articles/s41598-023-32027-3">https://www.nature.com/articles/s41598-023-32027-3</a></p><p>(2023, November 13). , — YouTube. Retrieved from <a href="https://royalsocietypublishing.org/doi/10.1098/rsos.201121">https://royalsocietypublishing.org/doi/10.1098/rsos.201121</a></p><p>(2023, November 13). , — YouTube. Retrieved from <a href="https://link.springer.com/article/10.1007/s41019-022-00193-5">https://link.springer.com/article/10.1007/s41019-022-00193-5</a></p><p>(2023, November 13). , — YouTube. Retrieved from <a href="https://link.springer.com/chapter/10.1007/978-3-319-40415-8_17">https://link.springer.com/chapter/10.1007/978-3-319-40415-8_17</a></p><p><em>Advantages of fusion</em>. (n.d.). ITER.org. Retrieved from <a href="https://www.iter.org/sci/Fusion">https://www.iter.org/sci/Fusion</a></p><p>Arroyo, M. (2021, September 22). <em>,</em>. , — YouTube. Retrieved from <a href="https://ar5iv.labs.arxiv.org/html/2302.06542">https://ar5iv.labs.arxiv.org/html/2302.06542</a></p><p><em>Ballooning instability</em>. (n.d.). Wikipedia. Retrieved from <a href="https://en.wikipedia.org/wiki/Ballooning_instability">https://en.wikipedia.org/wiki/Ballooning_instability</a></p><p>Barbarino, M. (2023, August 3). <em>What is nuclear fusion | IAEA</em>. International Atomic Energy Agency. Retrieved from <a href="https://www.iaea.org/newscenter/news/what-is-nuclear-fusion">https://www.iaea.org/newscenter/news/what-is-nuclear-fusion</a></p><p>Bonnet, A. (2023, August 22). <em>Fine-tuning Models: Hyperparameter Optimization</em>. Encord. Retrieved from <a href="https://encord.com/blog/fine-tuning-models-hyperparameter-optimization/">https://encord.com/blog/fine-tuning-models-hyperparameter-optimization/</a></p><p><em>DOE Explains…Nuclear Fusion Reactions</em>. (n.d.). Department of Energy. Retrieved from <a href="https://www.energy.gov/science/doe-explainsnuclear-fusion-reactions">https://www.energy.gov/science/doe-explainsnuclear-fusion-reactions</a></p><p><em>Focus on Physics: How E = mc2 Helps Us Understand Nuclear Fission and Fusion</em>. (2016, December 21). National Science Teachers Association. Retrieved from <a href="https://www.nsta.org/blog/focus-physics-how-e-mc2-helps-us-understand-nuclear-fission-and-fusion">https://www.nsta.org/blog/focus-physics-how-e-mc2-helps-us-understand-nuclear-fission-and-fusion</a></p><p>Greenwald, M. (2022, April 27). <em>Machine learning, harnessed to extreme computing, aids fusion energy development</em>. MIT News. Retrieved from <a href="https://news.mit.edu/2022/machine-learning-harnessed-extreme-computing-aids-fusion-energy-development-0427">https://news.mit.edu/2022/machine-learning-harnessed-extreme-computing-aids-fusion-energy-development-0427</a></p><p>Hampton, J. (2023, May 25). <em>The Grand Challenge of Simulating Nuclear Fusion: An Overview with UKAEA’s Rob Akers</em>. HPCwire. Retrieved from <a href="https://www.hpcwire.com/2023/05/25/the-grand-challenge-of-simulating-nuclear-fusion-an-overview-with-ukaeas-rob-akers/">https://www.hpcwire.com/2023/05/25/the-grand-challenge-of-simulating-nuclear-fusion-an-overview-with-ukaeas-rob-akers/</a></p><p>Hassabis, D. (2022, February 16). <em>Accelerating fusion science through learned plasma control</em>. Google DeepMind. Retrieved from <a href="https://deepmind.google/discover/blog/accelerating-fusion-science-through-learned-plasma-control/">https://deepmind.google/discover/blog/accelerating-fusion-science-through-learned-plasma-control/</a></p><p>Hill, C. (2020, July 29). <em>Studying the Interaction of Plasma with Irradiated Tungsten and Tungsten Alloys in Fusion Devices (F43021)</em>. International Atomic Energy Agency. Retrieved from <a href="https://www.iaea.org/newscenter/news/studying-the-interaction-of-plasma-with-irradiated-tungsten-and-tungsten-alloys-in-fusion-devices-f43021">https://www.iaea.org/newscenter/news/studying-the-interaction-of-plasma-with-irradiated-tungsten-and-tungsten-alloys-in-fusion-devices-f43021</a></p><p><em>Introduction to Bayesian networks</em>. (n.d.). Bayes Server. Retrieved from <a href="https://www.bayesserver.com/docs/introduction/bayesian-networks/">https://www.bayesserver.com/docs/introduction/bayesian-networks/</a></p><p>Jones, H., &amp; Chen, S. (2021, August 3). <em>Surrogate Models: A Comfortable Middle Ground?</em> SOA. Retrieved from <a href="https://www.soa.org/digital-publishing-platform/emerging-topics/surrogate-models/">https://www.soa.org/digital-publishing-platform/emerging-topics/surrogate-models/</a></p><p><em>Kink instability</em>. (n.d.). Wikipedia. Retrieved from <a href="https://en.wikipedia.org/wiki/Kink_instability">https://en.wikipedia.org/wiki/Kink_instability</a></p><p><em>magnetic field perturbation: Topics by Science.gov</em>. (n.d.). Science.gov. Retrieved from <a href="https://www.science.gov/topicpages/m/magnetic+field+perturbation">https://www.science.gov/topicpages/m/magnetic+field+perturbation</a></p><p><em>Nuclear fusion</em>. (n.d.). Wikipedia. Retrieved from <a href="https://en.wikipedia.org/wiki/Nuclear_fusion">https://en.wikipedia.org/wiki/Nuclear_fusion</a></p><p><em>Numerical Simulations and Algorithms for High-Performance Computing</em>. (n.d.). Frontiers. Retrieved from <a href="https://www.frontiersin.org/research-topics/58479/numerical-simulations-and-algorithms-for-high-performance-computing">https://www.frontiersin.org/research-topics/58479/numerical-simulations-and-algorithms-for-high-performance-computing</a></p><p><em>Office of Science — DOE Explains…Tokamaks</em>. (n.d.). Department of Energy. Retrieved from <a href="https://www.energy.gov/science/doe-explainstokamaks">https://www.energy.gov/science/doe-explainstokamaks</a></p><p><em>Offline Model-Based Reinforcement Learning for Tokamak Control</em>. (n.d.). Proceedings of Machine Learning Research. Retrieved from <a href="https://proceedings.mlr.press/v211/char23a/char23a.pdf">https://proceedings.mlr.press/v211/char23a/char23a.pdf</a></p><p><em>(PDF) Towards Long-Term predictions of Turbulence using Neural Operators</em>. (2023, July 29). ResearchGate. Retrieved from <a href="https://www.researchgate.net/publication/372625706_Towards_Long-Term_predictions_of_Turbulence_using_Neural_Operators">https://www.researchgate.net/publication/372625706_Towards_Long-Term_predictions_of_Turbulence_using_Neural_Operators</a></p><p><em>Recent advances and applications of surrogate models for finite element method computations: a review | Request PDF</em>. (2022, July 17). ResearchGate. Retrieved from <a href="https://www.researchgate.net/publication/362061545_Recent_advances_and_applications_of_surrogate_models_for_finite_element_method_computations_a_review">https://www.researchgate.net/publication/362061545_Recent_advances_and_applications_of_surrogate_models_for_finite_element_method_computations_a_review</a></p><p><em>Surrogate Aerodynamic Wing Modeling Based on a Multilayer Perceptron</em>. (2023, February 5). MDPI. Retrieved from <a href="https://www.mdpi.com/2226-4310/10/2/149">https://www.mdpi.com/2226-4310/10/2/149</a></p><p><em>Surrogate Models</em>. (n.d.). Berkeley Lab Computing Sciences. Retrieved from <a href="https://cs.lbl.gov/what-we-do/machine-learning/surrogate-models/">https://cs.lbl.gov/what-we-do/machine-learning/surrogate-models/</a></p><p><em>TEARING MODE DYNAMICS IN TOKAMAK PLASMAS</em>. (n.d.). Richard Fitzpatrick. Retrieved from <a href="https://farside.ph.utexas.edu/EPEC-documentation/epec.html">https://farside.ph.utexas.edu/EPEC-documentation/epec.html</a></p><p><em>To Track Turbulence in Tokamaks, Researchers Turn to Machine Learning</em>. (2023, May 1). Department of Energy. Retrieved from <a href="https://www.energy.gov/science/fes/articles/track-turbulence-tokamaks-researchers-turn-machine-learning">https://www.energy.gov/science/fes/articles/track-turbulence-tokamaks-researchers-turn-machine-learning</a></p><p><em>What is Surrogate Model</em>. (n.d.). Deepchecks. Retrieved from <a href="https://deepchecks.com/glossary/surrogate-model/">https://deepchecks.com/glossary/surrogate-model/</a></p><p>Zohuri, B. (n.d.). <em>Java Based Distributed Learning Platform</em>. David Publishing. Retrieved from <a href="https://www.davidpublisher.com/Public/uploads/Contribute/6437a748a5e6b.pdf">https://www.davidpublisher.com/Public/uploads/Contribute/6437a748a5e6b.pdf</a></p><p><em>If you have any questions, please feel free to reach out to me at amritaghag08@gmail.com or find me on </em><a href="https://www.linkedin.com/in/amrita-ghag/"><em>Linkedin</em></a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f48fc0741214" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Enhancing Fusion Reactors Through the Lawson Criterion]]></title>
            <link>https://medium.com/@amritaghag08/enhancing-fusion-reactors-through-the-lawson-criterion-b868e973d357?source=rss-28aecc9c0df8------2</link>
            <guid isPermaLink="false">https://medium.com/p/b868e973d357</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[the-knowledge-society]]></category>
            <category><![CDATA[nuclear-energy]]></category>
            <dc:creator><![CDATA[Amrita Ghag]]></dc:creator>
            <pubDate>Fri, 29 Dec 2023 13:31:30 GMT</pubDate>
            <atom:updated>2023-12-29T13:47:20.008Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*hgk5ZqYW_17ksv2X.jpg" /></figure><blockquote>“The sun, — the bright sun, that brings back, not light alone, but new life, and hope, and freshness to man — burst upon the crowded city in clear and radiant glory.”</blockquote><p>The sun’s massive energy continuously upheld its position as the <strong>unquestionable</strong> star of our solar system.</p><p>In Greek mythology, the handsome <em>Helios</em> was born to the Titans Hyperion and Theia. He rode his chariot drawn by fire-breathing horses across the sky, delivering brightness all over the earth, accompanied by a magnificent burst of sunbeams. Helios, the brother of Selene (the Moon Goddess) and Eos (the Dawn Goddess), made this celestial journey every day. Helios became connected with Apollo, the god of light, over time.</p><p>The Romans continued to worship various sun gods during their long reign, but they substituted the Greek name for sun, Helios, with the Sol in Latin, a root word that still refers to the sun today- in the term “solar system.” In ancient Rome, the most powerful sun god was Sol Invictus, which meant “Unconquered Sun.” (1)</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*3YuOY8fcScJs56_-.jpg" /><figcaption><strong>“Helios”. </strong>Retrieved from; <a href="https://www.pinterest.ca/pin/256845984983868381/">https://www.pinterest.ca/pin/256845984983868381/</a></figcaption></figure><p>Since the sun has been present for <strong>4.6 billion years,</strong> there have been so many stories told about it in a variety of civilizations and religions; this is just <strong>one</strong> example.</p><p><em>So was Helios </em><strong><em>actually </em></strong><em>the deliverer of light, or was there something else that happened in the sun that made it shine- </em><strong><em>what powers the sun?</em></strong></p><h3>TL:DR</h3><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FKrLDP1xsMEo%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DKrLDP1xsMEo&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FKrLDP1xsMEo%2Fhqdefault.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/c66bbf269a61914f259f81abe2d5bab3/href">https://medium.com/media/c66bbf269a61914f259f81abe2d5bab3/href</a></iframe><h3>Nuclear Fusion ☀️</h3><p>Nuclear fusion is the process of colliding light atomic nuclei to form heavier nuclei, while releasing an enormous amount of energy <strong>without</strong> the carbon emissions + long-lasting nuclear waste associated with other energy sources.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*EO8XF1W1_wmdwxzz.jpeg" /></figure><p>Fusion stands out as an abundant and appealing energy source because it has the ability to provide almost <strong>four million times</strong> more energy than conventional chemical reactions, and four times more energy than nuclear fission reactions at equal mass. Its long-term viability and low environmental effect are like a dream come true for reducing greenhouse gas emissions and pollution 🤩 (2)</p><blockquote>Nuclear fusion is the energy that powers the sun… so how do we replicate it on earth?</blockquote><p>Fusion energy scientists believe that tokamaks are the <strong>leading</strong> plasma confinement machine for future fusion power plants. In a tokamak, magnetic field coils confine plasma particles to allow the plasma to achieve the conditions necessary for fusion.</p><p>We still haven’t been 100% successful when it comes to full blown fusion reactions, there’s still a <strong>long</strong> way to go. But <em>why</em>?</p><h3>Fusion plasma has to be heated to at least 100 M°C on earth 🔥</h3><p>Fusion plasma has to be heated to at least 100 M°C on earth</p><p>An obstacle called the<em> Coulomb barrier,</em> caused by the strongly repulsive <em>electrostatic forces</em> between the positively charged nuclei, prevents them from fusing under normal circumstances. (3)</p><p>The nuclei in a fusion reaction have <strong>positive charges, </strong>meaning they have to collide at extremely high temperatures to overcome this barrier. Obviously, the core of the <strong>sun</strong> has extreme heat and densities, at about 15 million °C <em>(Fun fact: It’s </em><strong><em>150x</em></strong><em> more dense than water!💧) </em>These<em> </em><strong>extreme</strong> conditions cause the electrons in the atoms to completely detach from the atomic nucleus, and create the ionized fluid we know as <strong>plasma. </strong>By definition, plasma is a <em>soup </em>of hot gases, with positively charged atomic nuclei + negatively charged electrons moving at high speeds.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/460/0*HCg7btcHA_SWHlaQ.gif" /><figcaption><strong>The Coulomb Barrier. </strong>Retrieved from: <a href="https://www.pas.rochester.edu/~blackman/ast104/reactions.html">https://www.pas.rochester.edu/~blackman/ast104/reactions.html</a></figcaption></figure><p>But <strong>pressure</strong> also plays an extremely important role; without the high pressures from a fusion reaction, the hot plasma in the suns core would ultimately just explode into space. The pressure from the reaction is <strong>250 billion </strong>atmospheres <em>(a lot) </em>at the suns core, and squeezes the nuclei so that they’re 10-¹⁵ m close to each other 😮. (3)</p><p>At this intense proximity, the original nuclear force that binds protons + neutrons together in the nucleus becomes <strong>dominant </strong>and pulls incoming particles together…. <strong>nuclear fusion.</strong></p><p>But space is different than earth. In space, we have massive <em>gravitational forces</em> that cause nuclei to be crowded together very densely. Meaning, collisions happen very frequently, which is something we have to add to the checklist of fusion requirements ✅. Scientists found that we need about 10³⁸ collisions <strong>per second </strong>to keep the sun going, while the core is about 10⁶⁴ collisions per second… that is <strong>a lot.</strong></p><p>Our challenge on earth is actually harnessing an efficient fusion reaction with the similar conditions to the sun; meaning we would need <strong>extremely high temperatures, maintain high density, and be able to do this for a long period of time.</strong></p><h4>e²=1.44 MeV-fm</h4><p>Estimations have been made that the <strong>minimum </strong>temperature actually required to make a fusion reaction is e²=1.44 MeV-fm… let’s talk about it</p><p>In this equation, e would be the charge of a proton, and fm is the separation between two protons.</p><blockquote>The equation is derived from the <em>mass-energy equivalence principle</em>, expressed by Einstein’s famous equation of <em>E=mc², </em>where <em>E</em> is <strong>energy</strong>, <em>m</em> is <strong>mass</strong>, and <em>c</em> is the <strong>speed of light.</strong> In nuclear reactions, a <strong>small</strong> amount of mass is converted into energy, and the <em>E</em><strong><em>²</em></strong> term in the equation represents the <strong>squared</strong> energy released.</blockquote><p>The <em>“MeV-fm”</em> part of the equation tells us that the constant is expressed in units of <em>mega-electron volts times femtometer</em>s. This unit combination is from the conversion factor between energy + length scales in the microscopic world of nuclear reactions.</p><p>In practical terms, this equation is often used in nuclear physics calculations to tell us the energy being released in nuclear reactions, as it also estimates the binding energy of atomic nuclei. And now, we can see the connection between the <em>Coulomb barrier </em>and the <em>nuclear mass formula, </em>when accounting for the energy required to <strong>overcome</strong> electrostatic repulsion in fusion reactions. (4)</p><h3>The Lawson Criterion… continued</h3><blockquote>I wrote an <a href="https://medium.com/@amritaghag08/triple-product-decoding-the-most-important-metric-in-fusion-reactors-b89a49c84565">article</a> on this topic, more about the triple-product though! Check it out if you’re interested in seeing what my latest project in the field was :)</blockquote><p>The <strong><em>Lawson Criterion </em></strong>is<strong><em> </em></strong>rooted in the principles of plasma and nuclear physics, and provides a mathematical framework which is very very important for achieving sustained nuclear fusion reactions.</p><p>In a nutshell, the criterion contains the equilibrium of three key parameters:</p><ul><li>Plasma density (<em>n</em>)</li><li>Temperature (<em>T</em>)</li><li>Energy confinement time (<em>τ</em>)</li></ul><p>Mathematically, the Lawson Criterion is expressed as <em>nTτ </em>≥ *. The product <em>nTτ</em> <strong>must</strong> <strong>surpass</strong> a threshold for sustained fusion reactions <em>(the actual numerical value of the threshold depends on the details of the fusion system being considered).</em></p><h4>Critical Ignition Temperature for Fusion</h4><p>The overall fusion temperature given by setting the average <em>thermal energy = coulomb barrier, </em>is too high of a temperature. Fusion is initiated through particles that are very high energy <em>(due to the </em><a href="http://hyperphysics.phy-astr.gsu.edu/hbase/kinetic/kintem.html#c3"><em>Maxwellian distribution</em></a><em>). </em>Instead, the ignition temperature is lowered further by the fact that some particles have energies <strong>below </strong>the coulomb barrier, and can tunnel through the barrier.</p><p>If what I just said was a bunch of mumbo jumbo, don’t worry 😉</p><p>The estimated height of the coulomb barrier is based on the distance where the nuclear force can <strong>overcome</strong> the coulomb repulsion. When trying to model the <strong>probability</strong> of nuclear fusion, the usual way to go is to model it as a <em>“cross-section” </em>for the reaction to occur. (5)</p><p>This means that the cross-section can be taken to mean the probability for nuclear fusion to occur. Modelling this cross-section involves taking into account the probability for <strong>tunnelling</strong> through the coulomb barrier. This probability is <strong>higher</strong> for higher energy particles. Also, because the energy of impact between the particles for fusion is determined by their corresponding velocities, the model calculation for nuclear fusion output involves averaging over <strong>all</strong> relative velocities. The fusion cross-section as a function of average particle energy is plotted as a result of this modelling.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/251/0*LrxICgspITLMv8oV.png" /><figcaption><strong>Figure 1: Critical ignition temperature for fusion expressed using the “cross-section”. </strong>Retrieved from; <a href="http://hyperphysics.phy-astr.gsu.edu/hbase/NucEne/coubar.html#c3">http://hyperphysics.phy-astr.gsu.edu/hbase/NucEne/coubar.html#c3</a></figcaption></figure><h4>Confinement Time for Fusion</h4><p>Confinement time in tokamaks is defined as the <strong>time</strong> the plasma is maintained at a temperature <strong>above</strong> the critical ignition temperature (6). To obtain more energy from the fusion than what’s been put in to heat the plasma, the plasma must be held up to this temperature for some sort of minimum length time. <em>Calculations of that minimum time are:</em></p><p>τ = (2 x 10¹⁴)n/s -&gt; Deuterium-Tritium Fusion</p><p>τ = (5 x 10¹⁵)n/s -&gt; Deuterium-Deuterium Fusion</p><p>n is the <strong>ion density</strong> in the plasma. The product of the <em>ion density x confinement time</em> required for fusion is called the Lawson criterion- hopefully you’ve made that connection now 😉</p><h4>Ion Density for Fusion</h4><p>You need a <strong>high</strong> enough temperature to overcome the coulomb barrier for nuclear fusion. To do this, a significant density of ions <strong>must be maintained </strong>to make the probability of collisions high enough to obtain a positive energy net result. (4)</p><p>The density required for a net energy yield is associated with the confinement time ⬆️ for the hot plasma, so the <strong>minimum</strong> condition for an efficient fusion reaction is typically stated in terms of the product of the ion density and confinement time, called the Lawson criterion.</p><p>It feels like I’ve been explaining the same thing <strong>over and over again</strong>; the Lawson Criterion. But, there’s actually an identical theme when it comes to everything I’ve been talking about -&gt; they’re all <strong>products</strong> of the Lawson Criterion; <strong>density, temperature, and energy confinement time.</strong></p><p>Together, these factors make up the <em>triple-product</em>. As I’ve explained, the triple product provides a <strong>quantitative</strong> measure of the <strong>combined impact</strong> of plasma density (<em>n</em>), temperature (<em>T)</em>, and energy confinement time (<em>τ</em>), telling us whether the conditions are suitable for a <strong>sustainable</strong> fusion reaction.</p><p><em>We have the baseline of what makes up a sustainable fusion reaction. But what are the problems associated with the reactions?</em></p><h3>An Overview of the Challenges in a Fusion Reaction 🫠</h3><p>Out of the two <em>magnetic confinement fusion (MFE) systems; </em>tokamaks &amp; stellarators, I’ll be touching on <strong>tokamaks.</strong></p><p>A challenge for these toroidal MFE systems is that they need to <strong>minimize damage</strong> to material surfaces inside the vacuum vessel. Today, most modern tokamaks use a “<em>divertor</em>”; a set of open field lines outside the last closed flux surface, which channel <em>exhaust plasma </em>(unconfined plasma) past a specific point onto a set of plates (steeply angled to the magnetic field to <strong>maximize</strong> the surface area where the flowing plasma is in contact with.) (7)</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/390/0*tR73c1CNfT3dS8Rs.jpg" /><figcaption><strong>The Divertor. </strong>Retrieved from: <a href="https://www.iter.org/mach/Divertor">https://www.iter.org/mach/Divertor</a></figcaption></figure><p>However, even with the divertor in place, material surfaces facing the plasma can still experience erosion + damage due to the intense heat, radiation, and particle intensity.</p><p>Additional challenges also lie in the major possibility of <em>plasma instabilities</em>, which must in general be avoided. For example,<em> Edge Localized Modes</em> (“ELM’s”) are a class of instability that can lead to enhanced particle and energy flows onto the divertor plates. As mentioned before, the divertor is such an important component in tokamaks responsible for directing exhaust plasma <strong>away</strong> from the core plasma. Due to these instabilities, it ultimately becomes more sensitive and cannot do it’s job efficiently; resulting in increased heat loads and erosion. The impact of ELMs extends <strong>beyond</strong> the divertor, influencing the <strong>overall</strong> performance of the fusion device. (6)</p><p>In general, plasma instabilities play a bigger risk to the confinement of the plasma itself. If these instabilities are not effectively controlled, they can lead to disruptions in plasma confinement, and destruct the <strong>Lawson Criterion </strong>by losing <strong>thermal + magnetic control</strong> over the plasma, potentially causing an end to the overall fusion reaction.</p><p>I want to jump in and talk about what’s <strong>currently </strong>being done to monitor these inefficiencies, that will help us understand tokamak behaviour in the long term, and how analyzing these processes helps with building tokamaks.</p><h3>What’s Currently Being Done 🧐</h3><p>Computer simulations definitely play an important role in understanding + predicting the specific behaviour of tokamak plasmas. The main focus of these simulations is to look into the complex dynamics of high-temperature plasmas confined within the toroidal magnetic fields of tokamaks. Advanced numerical techniques and high-performance computing allow researchers to model the behaviour of plasma and explore how magnetic fields <strong>shape</strong> and <strong>control</strong> it.</p><p><em>What do they do? </em>A specific area of interest in tokamak simulations is the study of ELMs (what I mentioned earlier). Computer simulations dedicated to ELMs help researchers identify strategies to decrease their effects, and design divertor systems capable of handling enhanced particle and energy flows.</p><p>Normal computer simulations in nuclear fusion take so long ⏰ and cost 💰 so much due to the <strong>complexity</strong> of the physical processes involved in nuclear fusion. The simulations involve plasma temperature, rotation, and density. (8) With such complex and difficult plasma physics problems in tokamaks, it’s becoming more and more impractical to use computer simulations directly for simulating these systems.</p><p>But a new era has come, especially for the intersection of AI x Fusion. ML algorithms have been introduced to <strong>enhance </strong>the control of tokamak behaviour. Simulations incorporating machine learning techniques explore advanced control strategies, optimizing the tokamak’s response to different operating conditions.</p><p>But, what I’ve been working on are coded functions associated to the Lawson Criterion. By using code to understand this theory, we would be improving the stability and performance of tokamak plasmas, and there could maybe be a chance for fusion very soon 👀</p><h3>Coding for Lawson’s Criterion 🤓</h3><p>Coding for Lawson’s Criterion is so important for <strong>enhancing</strong> the efficiency + efficacy of fusion reactors. By coding for these simulations based on Lawson’s Criterion, researchers can adjust parameters such as plasma density, temperature, and energy confinement time (<em>triple-product) </em>to identify the <strong>best </strong>conditions for efficient fusion reactions. 🔥</p><p><strong>Traditional</strong> simulations rely on numerical models that may struggle to capture the exact physics of plasma in fusion reactors with high accuracy. These simulations are valuable for understanding <strong>general</strong> trends.. but they can be computationally demanding and actually require simplifications that can compromise overall accuracy.</p><p>By paying more attention to the parameters of triple-product, coded simulations can provide a more focused + resource-efficient exploration of the parameter space. It may also reduce the need for computer simulations. These simulators are used to measure accuracy of the physical parameters (temperature, density). But coding for these processes + plotting to see the curve, whether that be a downward curve or upward curve, does the same task… much more efficiently.</p><p>I also found that coding for Lawson’s Criterion can be optimized with a new layer of <em>machine learning</em> applications, for more intelligent simulations. The ML algorithms can analyze patterns + interactions within simulation data, and help to predict conditions for sustained fusion reactions. But, this is something I’ll be working on later… for now, I’ve been creating a 1-D linear model based off of functions in the Lawson’s Criterion.</p><p>Essentially, it’s coded so that you can <strong>manipulate </strong>a certain parameter, and see how your energy output changes from that. By plugging in certain numbers (of course they would have to be reasonable for a fusion reaction), it will plot what your energy output looks like.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Hl1ZJREkxXf30XkxWIR3OA.png" /></figure><p>What I’ve done here is define my variables in a fusion reaction; <strong>density (n), temperature (T), energy (E), and energy confinement time (tau). </strong>Then, I made sure it would plot the <strong>energy output </strong><em>(line 7).</em></p><p>The parameter/variable I decided to <strong>adjust </strong>and test for was <strong>density. </strong>In <em>line 12, </em>you can see how I’ve changed the parameter to instead be a range, of 10¹⁸-10¹⁹ (average particle density in a reaction). Finally, I asked for an energy output plot from these density inputs. Below, it shows how fusion power changes with this density to see how we can build the <strong>best </strong>reactor.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*CBHXqfhhph7385M61jnh_g.png" /></figure><p>We can repeat what we’ve done for density to temperature + time to see how fusion power varies.</p><p><em>But do you see the flaw in this?</em></p><p>So far, it can only plot for one variable that’s been adjusted. Something I do want to work on in the future is plotting for more than one variable, and introducing the ML algorithm I have plans on building out.</p><h3>A Benchmark for Improvement 🎉</h3><p>Meeting the Lawson criterion in an experiment is such a big milestone because it shows scientists that the fusion reactions are producing <strong>more</strong> energy than what’s being <strong>lost</strong>. This is so so exciting because it proves the fact that fusion <strong>can </strong>become a possible energy source.</p><p>When the Lawson criterion is met, researchers can use it as a <strong>benchmark</strong> to analyze the performance of their fusion experiments and see what they’re doing right, and what to improve for next time.</p><p>The sun has always been a central role in human history and mythology for so many civilizations. Its bright energy has been supporting life on Earth for billions of years. The sun has brought inspiration to many historical figures + stories, such as <strong>Helios in Greek mythology.</strong></p><p>Forgetting these mythical interpretations of the sun’s role, the main question has now been answered: <em>what powers the sun?</em> We now know that the sun’s energy is from the nuclear fusion reactions happening deep within its core.</p><p>But who knows, maybe a long long <strong>long </strong>time ago, Helios really did ride his chariot controlled by fire-breathing horses to deliver light all across the earth.</p><p><em>Maybe we’ll never know if that really used to be the case….</em></p><p>Thank you so much for reading! If you liked this, check out my other stories on Medium that I’ve published :)</p><p>Check out my <a href="https://www.linkedin.com/in/amrita-ghag/"><strong>Linkedin</strong></a></p><p>If you want to see more of what I’m doing, <a href="https://medium.com/@amritaghag08"><strong>Subscribe</strong></a> here!</p><p>Follow my <a href="https://substack.com/@amritaghag"><strong>monthly newsletter</strong></a> for more updates and progress ⭐️</p><h3>References</h3><ol><li><a href="https://www.sciencefriday.com/articles/science-diction-sun/#:~:text=The%20ancient%20Greeks%20personified%20the,the%20goddess%20of%20the%20dawn">https://www.sciencefriday.com/articles/science-diction-sun/#:~:text=The%20ancient%20Greeks%20personified%20the,the%20goddess%20of%20the%20dawn</a>.</li><li><a href="https://www.iter.org/sci/Fusion">https://www.iter.org/sci/Fusion</a></li><li><a href="https://www.intechopen.com/chapters/64653">https://www.intechopen.com/chapters/64653</a></li><li><a href="https://www.sfu.ca/~boal/390lecs/390lec17.pdf">https://www.sfu.ca/~boal/390lecs/390lec17.pdf</a></li><li><a href="http://hyperphysics.phy-astr.gsu.edu/hbase/NucEne/coubar.html#c3">http://hyperphysics.phy-astr.gsu.edu/hbase/NucEne/coubar.html#c3</a></li><li><a href="https://www.osti.gov/servlets/purl/1984761">https://www.osti.gov/servlets/purl/1984761</a></li><li><a href="https://www.osti.gov/servlets/purl/1984761">https://www.osti.gov/servlets/purl/1984761</a></li><li><a href="https://www.computerweekly.com/feature/Computer-simulation-at-work-for-the-future-of-nuclear-fusion">https://www.computerweekly.com/feature/Computer-simulation-at-work-for-the-future-of-nuclear-fusion</a></li><li><a href="https://www.pas.rochester.edu/~blackman/ast104/reactions.html">https://www.pas.rochester.edu/~blackman/ast104/reactions.html</a></li><li><a href="https://www.iter.org/mach/Divertor">https://www.iter.org/mach/Divertor</a></li></ol><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b868e973d357" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Triple-Product: Decoding the Most Important Metric in Fusion Reactors]]></title>
            <link>https://medium.com/@amritaghag08/triple-product-decoding-the-most-important-metric-in-fusion-reactors-b89a49c84565?source=rss-28aecc9c0df8------2</link>
            <guid isPermaLink="false">https://medium.com/p/b89a49c84565</guid>
            <category><![CDATA[articifical-intelligence]]></category>
            <category><![CDATA[nuclear-fusion]]></category>
            <category><![CDATA[the-knowledge-society]]></category>
            <dc:creator><![CDATA[Amrita Ghag]]></dc:creator>
            <pubDate>Sat, 25 Nov 2023 21:29:56 GMT</pubDate>
            <atom:updated>2023-12-23T13:17:40.200Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*LHlyyr2Mu6cS87xOQqmWgA.jpeg" /><figcaption><a href="https://usfusionenergy.org/science-fusion#:~:text=Plasmas%20must%20meet%20three%20conditions,criterion%2C%20or%20the%20triple%20product">https://usfusionenergy.org/science-fusion#:~:text=Plasmas%20must%20meet%20three%20conditions,criterion%2C%20or%20the%20triple%20product</a>.</figcaption></figure><p>Nowadays, going on a plane is a breeze (minus going through security 🫣). Picture it: you’re flying to Hawaii — you sit and relax over a <strong>smooth flight</strong>, are looked after by your flight attendants, and fed pretty good food for a designated period of time before you reach your destination — what a great system, right?</p><p>The <strong>aerodynamic efficiency</strong> + smooth airflow over the aircraft’s surfaces are primarily influenced by many parameters; such as the design of the airframe, the shape of the wings, and the surface finish. Small details like a smooth and glossy finish helps in the transition of air across the surface of the wing, reducing skin friction drag and ultimately contributing to the overall aerodynamic performance of the aircraft (1).</p><p><em>Okay so what am I getting at?</em></p><p>All of these different factors in the process of making a plane have a <strong>direct</strong> influence on the efficiency of the flight, and to make sure that you get to Hawaii in <strong>one piece.</strong></p><p>I want to shift gears a little bit; I’m really interested in <em>nuclear fusion, </em>and creating more efficient fusion reactors. But to do that, we need to understand <strong>how</strong> different factors affect the efficiency of your machine. Hopefully you get it now that you’ve flown to Hawaii 🧘‍♀️</p><p>But you can’t just increase the efficiency of a fusion reactor with no understanding of the input parameters needed to do so. Let’s take a look at how different parameters in a nuclear fusion reactor <strong>directly</strong> impact the reactors ignition — a concept called <em>Triple-product.</em></p><h3>TL:DR</h3><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FB9pth89M67c%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DB9pth89M67c&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FB9pth89M67c%2Fhqdefault.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/b36bfa7b6075696ff4a6dacfcc60c96c/href">https://medium.com/media/b36bfa7b6075696ff4a6dacfcc60c96c/href</a></iframe><h3>The Building Blocks of a Fusion Reactor: The Lawson Criterion</h3><p>The concept of nuclear fusion is simple- all we really want is <strong>clean </strong>energy on earth. But to do so, nuclear reactors <em>(tokamaks) </em>need to produce significantly <strong>more</strong> energy than they consume, so of course it’s important to understand which inputs we need to achieve that goal- <em>but what are they?</em></p><p>Well, John D. Lawson already thought of that, back in 1955! He wondered what conditions needed to be met in order to get enough energy from a fusion reaction to use in larger scaled projects.</p><p>The Lawson criterion is expressed in terms of the <strong><em>triple product</em></strong> (<em>nTτ</em>), where <em>n</em> is the plasma density, <em>T</em> is the plasma temperature, and <em>τ (pronounced “tau”) </em>is the energy confinement time. (<strong>3 </strong>terms = <strong>triple</strong> product)</p><blockquote>The Lawson criterion is given by the following inequality:</blockquote><blockquote><strong><em>nTτ </em>≥ Lawson criterion</strong></blockquote><p>To simplify, the Lawson criterion is essentially a criteria/threshold that fusion plasma must surpass for the fusion reactor itself to become self-sustaining. If the triple product falls <strong>below</strong> the Lawson criterion: the rate of energy being lost from the plasma <strong>exceeds</strong> the rate of energy production, and the <strong>reaction cannot be sustained.</strong></p><h4>Big Q (<em>Q</em>)</h4><p>The ratio <em>Q</em> is a measure of how well a fusion reactor is performing in terms of energy output compared to the energy input. For sustained fusion reactions, <em>Q</em> needs to be greater than 1. Achieving <em>Q</em>&gt;1 means that the fusion reactions are producing <strong>more</strong> energy than is being externally supplied to maintain the plasma conditions.</p><p>The Lawson criterion, expressed in terms of the <strong><em>triple product</em></strong>, sets the foundation for achieving self-sustained fusion. Meeting, or even surpassing the Lawson criterion ensures that the conditions for sustained fusion are met. Ultimately, achieving a reaction of <em>Q</em>&gt;1 shows a net <strong>positive</strong> energy output, signifying a successful and potentially practical fusion reactor (2)!</p><p>The relationship between the Lawson Criterion, the triple product, and <em>Q</em> is intertwined, which were put in place to help us understand when fusion reactions can happen effectively!</p><h3>The Triple-Product x Lawson’s Criteria</h3><p>So, the triple product is a broader concept, like saying “we need these three things to make fusion work,” and the Lawson criterion is a specific measurement or rule within that concept, specifying <strong>exactly</strong> how much of those three things we really need.</p><p>I mainly want to focus on <em>triple-product</em>, because it consists of the factors <em>(T, n,τ)</em> that make a fusion reaction possible.</p><p>Essentially, the triple product, in the context of nuclear fusion, <strong>quantifies</strong> the performance of a fusion grade plasma, measuring how good the overall fusion reactor is (3).</p><p>It reflects the combination of <strong>three</strong> very important measurements in a fusion reactor I mentioned before: <strong>temperature</strong> <em>(T)</em>, <strong>density</strong> <em>(n)</em>, and <strong>time</strong> <em>(</em>τ<em>)</em>. By maximizing these three conditions in the plasma made of the right fuels, the fusion reactor has a great efficiency, and ultimately leads to the release of more energy.</p><blockquote>Therefore, the <strong>higher</strong> the triple product, the <strong>greater</strong> the energy released!</blockquote><p>Triple-product also allows for the comparison of different fusion concepts + tracks the progress of fusion research towards energy gain and commercial uses.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*OOggQfFAwnSHWKNM51zdsQ.png" /><figcaption>Retrieved from; <a href="https://www.youtube.com/watch?v=MCs86Ous1f4">https://www.youtube.com/watch?v=MCs86Ous1f4</a></figcaption></figure><p>Something really cool about triple product is that it’s independent of the specific machine used to create the fusion plasma, so it can be used to compare performance across different kinds of approaches to fusion (4). For example, it can be applied to quantities in <strong><em>magnetic confinement</em></strong> machines <em>(tokamaks, stellarators)</em>, or<em> </em><strong><em>inertial confinement</em></strong> machines <em>(laser fusion).</em></p><p>Between the <strong>1950s</strong> and the early <strong>2000s</strong>, the maximum triple product achieved by fusion energy experiments <strong>grew</strong> dramatically. During post WWII times in the 1950s, the US, the Soviet Union, and the UK received well funded efforts to design and build fusion energy systems <em>(See Figure 1)</em>. A few early experiments initially reported promising results of high temperature plasmas and fusion neutrons, results of the D-T fusion reaction.</p><p>…..But it was soon realized that this was a <strong>misinterpretation</strong> of the data. The source of the fusion reactions were areas of <em>plasma instabilities</em>, <strong>not</strong> from the hot plasma alone. The actual achieved triple products were<strong> very low</strong>, lower than what was required for a full nuclear power plant. By the end of the 1950s, the research was declassified and the field entered into a brief period of defeat 😢.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/448/1*RiFslt41ckKPXWqAx6neCQ.jpeg" /><figcaption><strong><em>Figure 1: Photograph of the ZETA pinch experiment in the UK from the mid 1950s.</em></strong><em> Description + image r</em>etrieved from; <a href="https://www.fusionenergybase.com/article/measuring-progress-in-fusion-energy-the-triple-products">https://www.fusionenergybase.com/article/measuring-progress-in-fusion-energy-the-triple-products</a></figcaption></figure><p>Fast forward to today, a <strong>rush</strong> of tokamak construction followed, and the achieved triple products <strong>increased</strong> dramatically throughout the rest of the 20th century. The ability to drive higher triple products in tokamaks was the result of so many advancements in the understanding of plasma physics, improved technologies, and an increase in tokamak size <em>(See Figure 2)</em>. The increase in size might not seem important, but larger tokamaks actually hold on to heat for a longer time, increasing their energy confinement times and therefore also their triple products (5).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*8evCDrIJhQoMIX6_LBcHwQ.jpeg" /><figcaption><strong>Figure 2: <em>Photograph of the JET tokamak in 1991. </em></strong><em>Description + image r</em>etrieved from; <a href="https://www.fusionenergybase.com/article/measuring-progress-in-fusion-energy-the-triple-products">https://www.fusionenergybase.com/article/measuring-progress-in-fusion-energy-the-triple-products</a></figcaption></figure><p>I mentioned the <strong>specific</strong> parameters that triple product measures- let’s take a look at how exactly each one is measured, why, and the specifics of it all!</p><h4>Temperature 🌡️</h4><p>Fusion reactor plasmas need to be <strong>hot 🔥 </strong>. While fusion in the sun’s core occurs at a temperature of <strong>15 million degrees Celsius,</strong> fusion on Earth has to <strong>exceed</strong> this temperature, by <strong>several</strong> <em>orders of magnitud</em>e to account for our low atmospheric pressure (6).</p><p>Fusion reactors (specifically <em>tokamaks</em>) consistently heat plasmas to temperatures of <strong>100 million degrees Celsius</strong> or more. And these temperatures are necessary to provide the nuclei with enough energy to overcome repulsion, so that they <strong>fuse</strong> (hence nuclear <strong>fusion!)</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*YZNtR5S-EDulJIPBlQASEg.jpeg" /><figcaption>Retrieved from; <a href="https://usfusionenergy.org/science-fusion#:~:text=Plasmas%20must%20meet%20three%20conditions,criterion%2C%20or%20the%20triple%20product">https://usfusionenergy.org/science-fusion#:~:text=Plasmas%20must%20meet%20three%20conditions,criterion%2C%20or%20the%20triple%20product</a>.</figcaption></figure><p>Achieving + maintaining high plasma temperatures is <strong>essential</strong> for sustaining fusion reactions. In fusion, particles need to overcome the <em>Coulomb barrier- </em>essentially just a repulsive force due to the positive charges of atomic nuclei- to collide and fuse. The triple-product here is that <strong>higher temperatures = increased particle velocities</strong>, allowing for more successful collisions between nuclei, and ultimately an increased chance of fusion.</p><blockquote>Temperatures in fusion research are written in units of keV (thousand electron volts), instead of degrees Celsius. <strong>1keV ≈ 11,600,000°C.</strong></blockquote><p>Due to these large temperature requirements, it’s evident that effective plasma heating is such a big and important milestone in any fusion model. This whole idea explains why the 1keV temperature result of the 1968 tokamak was so exciting- although 1keV is too low for any sort of fusion mode, it’s a factor of ten of what’s required (7), which is still pretty impressive. Even in the present a temperature of 1keV is very exciting, when combined with reasonable parameters for plasma confinement <em>(See Figure 3)</em>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1021/1*dQfjq1eVsmBEWSIxEcVOFA.png" /><figcaption><strong>Figure 3: <em>Achieved triple products and temperatures for selected experiments.</em></strong><em> Full description + image retrieved from; </em><a href="https://www.fusionenergybase.com/article/measuring-progress-in-fusion-energy-the-triple-products">https://www.fusionenergybase.com/article/measuring-progress-in-fusion-energy-the-triple-products</a></figcaption></figure><p>It’s also important to know that many important fusion experiments don’t actually rely solely on high temperatures as a goal. Rather, their goals are set in place to be small achievements to eventually reach those levels; for example studying plasma instabilities, creating new confinement approaches, or exploring the limits of other emerging technologies. But each of these goals are <strong>connected</strong> to enabling high triple products down the line (8).</p><h4>Density 🏋️</h4><p>In a fusion reaction, there has to be enough atoms actually fusing… meaning that there are<strong> a lot </strong>of atoms needed for a fully sustained reaction.</p><p>The density <em>(n) </em>revolves around the number of particles per unit volume in the plasma, influencing the probability of particle collisions required for making fusion possible.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*fpKpymk2TbLLB4G3Yp9L-g.jpeg" /><figcaption>Retrieved from; <a href="https://usfusionenergy.org/science-fusion#:~:text=Plasmas%20must%20meet%20three%20conditions,criterion%2C%20or%20the%20triple%20product">https://usfusionenergy.org/science-fusion#:~:text=Plasmas%20must%20meet%20three%20conditions,criterion%2C%20or%20the%20triple%20product</a>.</figcaption></figure><p>The more dense a plasma is, there is an <strong>increased</strong> chance for particle collisions, the most important requirement for fusion reactions. The distance of the particles within the plasma also determines the frequency of collisions. These distances heavily rely on the <strong>density </strong>of the atoms, influencing the overall efficiency of fusion.</p><p>To put it all into perspective, a plasma density of about 10¹⁴ particles per cubic centimetre is <strong>250,000 times</strong> thinner than the earth’s air mantle 😮 (9). This extremely <strong>low density</strong> means that plasma in fusion has an overall power density only a bit bigger than an ordinary light bulb, despite its high temperature!</p><p>Keeping the plasma density high enough makes sure that the atoms do fuse, and to do so, the plasma is surrounded by very big <em>electromagnets. </em>These create magnetic fields are <strong>10,000 x</strong> stronger than the earth’s magnetic field 🌎.</p><p>However a <strong>large </strong>increase in density for the plasma can create different types of <strong>collisions<em> </em></strong>between nuclei and electrons, which can create large amounts of <em>radiation</em> (10). The radiation (called <em>bremsstrahlung</em>) takes energy from the plasma and prevents fusion from happening 😣.</p><h4>Time ⏰</h4><p>Of course the more time a reaction runs, the closer we get to harnessing fusion power. Most recently, a nuclear fusion reactor research facility successfully harnessed plasma at <strong>70 million degrees</strong> Celsius for as long as 17 minutes, 36 seconds! (11)</p><p>When adding time into density + temperature, there <strong>has </strong>to be enough time to let the fusion reaction to actually happen, and to cool down. The more and more the plasma is heated <em>(T),</em> the number of reactions increase overall, and so does the time (<em>τ)</em>!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*0Mw3JOXwYnLdE4Pqo5wm0w.jpeg" /><figcaption>Retrieved from; <a href="https://usfusionenergy.org/science-fusion#:~:text=Plasmas%20must%20meet%20three%20conditions,criterion%2C%20or%20the%20triple%20product">https://usfusionenergy.org/science-fusion#:~:text=Plasmas%20must%20meet%20three%20conditions,criterion%2C%20or%20the%20triple%20product</a>.</figcaption></figure><p>Time is represented by the <em>energy confinement time. </em>The energy confinement time measures the average time particles spend within the plasma <strong>before</strong> losing their energy. It also measures how effectively the plasma can confine and maintain the high-temperature conditions required for a full fusion reaction.</p><p>In a fusion reaction, 80% of the energy is carried by the neutrons, while the other 20% energy is carried by the helium nuclei in the plasma. The newly formed helium (20%) bounces around the vessel, colliding with unburnt nuclei used for fuel, and heats them up, reducing the need for the external heating systems. But it takes a pretty long <strong>time </strong>for this to happen depending on the density + temperature of the plasma (12).</p><p>Overall, after a certain time, a specific value of the fusion product called <em>ignition </em>becomes <strong>self-sustaining</strong>. This means that the heat generated by the reaction is enough to keep the plasma hot, and the external heating systems can be turned off 🥵. This self-sustaining state is a <strong>key</strong> milestone in the development of nuclear fusion, and it’s all done using temperature, density, and time.</p><p>We can see that all three products live in harmony, and have to work together if we want to see fusion energy becoming a reality. Now that the <strong>groundwork </strong>for the <em>Lawsons Criterion </em>has been laid and understanding the role of all three products required for a fusion reaction, I really want to show my <strong>linear regression </strong>model which <strong>predicts</strong> the relationship between different parameters, and how they affect the overall <em>plasma power </em>of a reactor!</p><h3>Linear Regression Analysis of Lawson’s Criterion + Impact on Ignition</h3><p>The data set I used to predict values was based off of a <a href="https://www.kaggle.com/datasets/rajashekhar3009/atmospheric-plasma-spray"><em>Plasma_ML_Dataset.csv</em></a>, which included parameters like<strong> plasma power, particle velocity, particle temperature, e.t.c. </strong>And, since the whole basis of this project is on triple product, I chose this particular data set because it included <strong>temperature <em>(t).</em></strong></p><p>Since this was a dataset based off of some pretty hard analysis’ in fusion, creating the predictions was not very easy.</p><h4><em>So how exactly did I build the ML prediction model?</em></h4><p>Using Python through Jupyter Notebook, I was able to import the dataset, which included <strong>all the original </strong>factors in a fusion plasma.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*nBrcEADBw_3m-D4WlgS6zw.png" /></figure><p>I defined “Plasma Power” as what I wanted the ML model to predict, and dropped all unnecessary parameters. I made three different scatterplots using all the parameters I wanted to analyze: <strong>Particle Temperature,</strong> <strong>Particle Velocity, and Primary Gas Flow Rate, </strong>to see what effect each parameters had on <strong>Plasma Power. </strong><em>I got the following results:</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/563/1*3XPxH0HNRwWf2M_-R__dyA.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/563/1*Lkcw8riGjt-muCE1TDOu5Q.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/563/1*8xVdW1T-4L7ycW166-qaQA.png" /></figure><p>What I found the most interesting was how much each parameter was so different, very inconsistent, and hard to predict <strong>for humans</strong>. For example, you would think that the hotter the particle, the more plasma power — right? <strong>Wrong: </strong>in the first scatterplot, we can see that the highest temperature is 2900K, but it has the second lowest plasma power. <em>So maybe the ML can make better + more accurate predictions than us to solve these uncertainties!</em></p><p>With the machine learning p.o.v, you have some inputs, and you can make a linear model that predicts continuous values as well. So in this case, our input variables were <strong>continuous values</strong> to predict the output of how much plasma power is <strong>actually</strong> being produced 🔥, an essential component in a fusion reactor.</p><p>After defining all important variables + cleaning up the data, the model spit out the “test predictions” as the following:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*uCm-un7STmt-lWJucsGvqw.png" /></figure><p>What this means is that the model predicted values based off of <strong>other</strong> values, and came up with those outputs. Obviously, the Sklearn Plasma Predictions and Test Predictions are not very close,<em> so what did I do?</em></p><p>I asked the model for the <strong>“best predictions” </strong>to ensure that the model was as accurate as can be, where the best predictions + Sklearn Plasma Predictions were identical, meaning these predictions are the most accurate.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*t1FuFlxyApb-NojJBXKMqA.png" /></figure><h4>Visualizing the Data 🔍</h4><p>If you want to see a more washed down + simplified version of Particle Temperature <em>(remember we chose this specific variable because of the </em><strong><em>triple-product), </em></strong>VS. Plasma Power to visualize the data and to see how accurate the model was:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*vyLs56GjWBk3jBfqUtjIrA.png" /></figure><p>We can see that the highest point of plasma power actually does <strong>not</strong> have the hottest particle temperature, instead, a plasma with less power has the highest temperature. Our model predicted something very close to that, how the highest particle temperature <strong>did not </strong>produce the largest amount of plasma power.</p><p>Hopefully as you read this, you started to make the connection between the triple-product and how it’s essentially a <strong>performance indicator</strong> for a fusion reactor (or in this case — plasma power)! We have such unpredictable values for humans to predict, based off of triple product, so our ML algorithm helps us to make the best predictions, <strong>enhancing</strong> the triple product!</p><p>You landed in Hawaii just in time for checking into your hotel! You think about your overall flight; there were no errors, no delays, a little turbulence 😅, and no mechanical malfunctions within the plane itself.</p><p>One day, hopefully fusion reactors can be the same — perform <strong>perfectly, </strong>with little to none turbulence (<em>haha get it?). </em>We want to measure this performance using all the factors of the <strong>Triple Product, </strong>which make up the <strong>Lawson’s Criterion. </strong>And one day, we <strong>will</strong> get there!</p><p>…<em>but for now, I could </em><strong><em>really</em></strong><em> use a trip to Hawaii</em></p><p>Thank you so much for reading! If you liked this, check out my other stories on Medium that I’ve published :)</p><p>Check out my <a href="https://www.linkedin.com/in/amrita-ghag/"><strong>Linkedin</strong></a></p><p>If you want to see more of what I’m doing, <a href="https://medium.com/@amritaghag08"><strong>Subscribe</strong></a> here!</p><p>Follow my <a href="https://substack.com/@amritaghag"><strong>monthly newsletter</strong></a> for more updates and progress ⭐️</p><h3>References</h3><ol><li><a href="https://www.faa.gov/sites/faa.gov/files/07_phak_ch5_0.pdf">https://www.faa.gov/sites/faa.gov/files/07_phak_ch5_0.pdf</a></li><li><a href="https://www.energyencyclopedia.com/en/nuclear-fusion/thermonuclear-fusion/lawson-criterion">https://www.energyencyclopedia.com/en/nuclear-fusion/thermonuclear-fusion/lawson-criterion</a></li><li><a href="https://www.tandfonline.com/doi/full/10.1080/15361055.2023.2198049#:~:text=The%20triple%20product%20of%20density,has%20no%20inherent%20physical%20significance">https://www.tandfonline.com/doi/full/10.1080/15361055.2023.2198049#:~:text=The%20triple%20product%20of%20density,has%20no%20inherent%20physical%20significance</a>.</li><li><a href="https://www.fusionenergybase.com/article/measuring-progress-in-fusion-energy-the-triple-products">https://www.fusionenergybase.com/article/measuring-progress-in-fusion-energy-the-triple-products</a></li><li><a href="https://usfusionenergy.org/science-fusion#:~:text=Plasmas%20must%20meet%20three%20conditions,criterion%2C%20or%20the%20triple%20product">https://usfusionenergy.org/science-fusion#:~:text=Plasmas%20must%20meet%20three%20conditions,criterion%2C%20or%20the%20triple%20product</a>.</li><li><a href="https://www.ipp.mpg.de/15144/zuendbedingungen#:~:text=a%20plasma%20density%20of%20about,than%20an%20ordinary%20light%20bulb">https://www.ipp.mpg.de/15144/zuendbedingungen#:~:text=a%20plasma%20density%20of%20about,than%20an%20ordinary%20light%20bulb</a>.</li><li><a href="https://en.wikipedia.org/wiki/Timeline_of_nuclear_fusion#:~:text=%5B%20Record%5D%20China&#39;s%20Experimental%20Advanced%20Superconducting,fusion%20energy%20however%20requires%20i.a">https://en.wikipedia.org/wiki/Timeline_of_nuclear_fusion#:~:text=%5B%20Record%5D%20China&#39;s%20Experimental%20Advanced%20Superconducting,fusion%20energy%20however%20requires%20i.a</a>.</li><li><a href="https://euro-fusion.org/fusion/fusion-conditions/">https://euro-fusion.org/fusion/fusion-conditions/</a></li></ol><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b89a49c84565" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Plasma Instabilities in Fusion Reactors: A Surrogate Model Approach to Optimization]]></title>
            <link>https://medium.com/@amritaghag08/plasma-instabilities-in-fusion-reactors-a-surrogate-model-approach-to-fusion-reactor-optimization-d0f88afa246e?source=rss-28aecc9c0df8------2</link>
            <guid isPermaLink="false">https://medium.com/p/d0f88afa246e</guid>
            <category><![CDATA[nuclear-fusion]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[the-knowledge-society]]></category>
            <dc:creator><![CDATA[Amrita Ghag]]></dc:creator>
            <pubDate>Mon, 23 Oct 2023 03:20:55 GMT</pubDate>
            <atom:updated>2023-10-27T15:45:07.567Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*s6C0YU8LFv1HPsV7-Lg3RA.png" /></figure><p><strong><em>TL:DR</em></strong></p><ul><li>Nuclear fusion energy is produced using a <strong>Tokamak, </strong>essentially a machine that confines plasma with magnetic coils</li><li>While it does confine plasma, the plasma is so unstable<strong>, </strong>and creates plasma <strong>instabilities</strong> + <strong>turbulences</strong></li><li>To currently simulate this unstable plasma, scientists use <strong>computer simulations</strong>. However, the problem with these simulations is that they take <strong>way</strong> too much time, money, and resources</li><li>The use of AI, specifically an application called <strong>Surrogate Modelling, </strong>will help to predict + simplify these simulations, and make the fusion process so much less complex 💡</li></ul><p><strong>The Sun</strong>. This big ball in the sky has captivated human imagination for centuries. Long before the scientific understanding of fusion, many cultures and civilizations admired its warmth, light, and overall presence.</p><p>In ancient times, many cultures thought of the Sun as a divine and central force in their overall understanding of the world, an endless flame that always came back. However, it wasn’t until the peak of modern science that we began to uncover the true nature of the sun 🔭 .</p><p>The journey to understanding fusion and its connection to the Sun’s huge energy output was <strong>very</strong> gradual. In the early 20th century, as our understanding of atomic physics got better, scientists began to unlock the secrets of the atomic nucleus… they realized the sun isn’t just an endless flame. This discovery laid down the foundation for understanding basic fusion — the process where atomic nuclei combine to release huge amounts of energy.</p><p>The realization that fusion was the energy source driving the Sun ignited so much curiosity among scientists. Over the years, extensive research and experimentation led to the development of controlled nuclear fusion reactors, with the goal of recreating the Sun’s energy-generation process under controlled conditions.</p><p>Scientists realized that nuclear fusion holds tremendous promise for our planet- it offers abundant, sustainable, and clean energy with <strong>zero</strong> greenhouse gas emissions. Its safety advantages, little to none radioactive waste, and reduced carbon footprint honestly makes it seem unreal.</p><p>The problem with nuclear fusion has been the part of handling and confining superheated plasma. Plasma’s unpredictable behaviour frequently results in turbulence and instability, which has a direct interference with the fusion process, part of the reason why we haven’t harnessed this energy.</p><p>Humanity went from essentially not knowing what the sun was, to almost mimicking the energy it runs on. We just need a little <strong>push</strong> to help us move forward from not knowing how to control plasma.</p><p>Now, humanity has the development of incredible technologies that have expanded our ability to explore and discover. From the microscope to the telescope, and from electron microscopes to particle accelerators, technology has greatly enhanced our capacity to revolutionize anything we want.</p><p><em>So what technology is it going to be that pushes us to the next energy revolution?</em> We can answer that as soon as we understand basic concepts and roadblocks in fusion.</p><h3>A big donut leading us to the next Energy Revolution?</h3><p>“Тороидальная Камера с Магнитными Катушками”</p><p>….. or “Toroidal Chamber with Magnetic Coils”, is where the word <em>Tokamak</em> was derived from — the Russian acronym “Токамак”. (<a href="https://en.wiktionary.org/wiki/tokamak">1</a>)</p><p>A Tokamak is essentially a big donut 🍩. It’s a machine that confines <em>plasma, </em>using magnetic fields in a torus, a donut like shape. Tokamaks are the key to harnessing Nuclear Fusion 🔑 .</p><p><em>Magnetic fields, </em>specifically, are used to get the plasma to the necessary conditions of Nuclear Fusion- which is very hot. It does this by confining the plasma particles using magnetic field coils.</p><blockquote>Tokamaks can sustain plasma currents at the mega-ampere level, which is equivalent to the electric current in the most powerful bolts of lightning.</blockquote><p>Tokamaks are <strong>very</strong> powerful, and it’s all thanks to the magnetic fields! Who knew a big donut could hold such hot plasma?</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/942/1*70in5pzuq3I1n2LcZGHnbw.png" /></figure><p>The most important part of Tokamaks are their magnetic fields, and every field has a set job. Now here comes the good part…</p><p>One set of magnetic coils generates an <strong>intense</strong> <em>toroidal</em> <em>field</em>, near the long way around the torus. This field is the primary mechanism of confinement of the plasma particles.</p><p>The<em> central solenoid</em> is a magnet that carries electric current. It creates a second magnetic field directed along the “poloidal” direction (the short way around the torus). The central solenoid is the backbone of the tokamak. It causes most of the necessary magnetic field changes needed to start the plasma, create the plasma current, and keep that current going throughout the burning process ⚡️ .</p><p>These two fields make a twisted magnetic field that confines the particles in the plasma. Then, a third set of field coils generates an <em>outer poloidal</em> field that shapes and positions the plasma. (<a href="https://www.energy.gov/science/doe-explainstokamaks#:~:text=A%20tokamak%20is%20a%20machine,for%20future%20fusion%20power%20plants.">2</a>)</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/620/0*aHcGMB0MaDGB7Jsm" /><figcaption><strong>Figure 1: A diagram of the parts that make up a tokamak. </strong>Retrieved from; <a href="https://www.energy.gov/science/doe-explainstokamaks#:~:text=A%20tokamak%20is%20a%20machine,for%20future%20fusion%20power%20plants">https://www.energy.gov/science/doe-explainstokamaks#:~:text=A%20tokamak%20is%20a%20machine,for%20future%20fusion%20power%20plants</a></figcaption></figure><p>The toroidal magnetic field is required to keep the plasma confined and away from the walls of the tokamak. These machines carry so much potential- but there’s just one <strong>problem</strong>; plasma has to be dense for fusion to actually happen, but the temperature of the plasma gets so high that anything it touches would melt instantly. These are known as <em>plasma instabilities……</em></p><h3>Plasma Physics: A complex Factor in Fusion Optimization</h3><p>Plasma is a gas where the atoms are <em>ionized</em>, so that the electrons and ions are separately free (<em>See Figure 2)</em>. This ionization occurs when the temperature is hot enough 🔥. (<a href="http://silas.psfc.mit.edu/introplasma/chap1.html#tth_sEc1.1">3</a>)</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/290/1*urqnniMsQ1d2egyiSv7FCg.png" /><figcaption><strong>Figure 2: An example of a plasma (right) after being heated significantly enough to become ionized and escape from its gaseous phase (left). </strong>Full description + image retrieved from; <a href="https://www.researchgate.net/publication/236143825_Computational_Methods_for_a_One-Directional_Plasma_Model_with_Transport_Field/figures?lo=1">https://www.researchgate.net/publication/236143825_Computational_Methods_for_a_One-Directional_Plasma_Model_with_Transport_Field/figures?lo=1</a></figcaption></figure><p>But plasma is not a very easy matter to work, as there are many <em>turbulence</em> risk factors involved.</p><h4>Plasma Turbulence: Why we Haven’t Fully Achieved Nuclear Fusion</h4><p><em>Plasma turbulence</em> prevents the plasma from reaching the temperatures needed to split the atomic nuclei — which prevents those nuclei from fusing together, and makes the plasma <strong>very</strong> unstable. (<a href="https://news.mit.edu/2013/understanding-turbulence-in-plasmas-0429#:~:text=One%20simple%20phenomenon%20explains%20why,lose%20much%20of%20its%20heat.">4</a>)</p><p>And these plasma instabilities are one major reason why fusion is not possible yet 🤯. Keeping the core of a plasma hot enough that fusion can occur while preventing the tokamak walls from melting from that heat is <strong>really</strong> hard. (<a href="https://www.energy.gov/science/fes/articles/taming-plasma-edge-reducing-instabilities-tokamaks#:~:text=The%20Science,instead%20of%20a%20steady%20flow.">5</a>)</p><p>To generate power efficiently, tokamaks would need to produce <strong>large</strong> amounts of power. But to do so, they would need to operate for long periods, with the plasma running for at least a couple of minutes.</p><p>Many design calculations show that extra bursts of heat from impulsive plasma instabilities would <strong>severely</strong> damage the tokamaks’ surfaces. To prevent this damage, fusion tokamaks must operate <strong>without</strong> these instabilities…. <em>but how?</em></p><h4>What’s Currently Being Done to Approximate Plasma Behaviours?</h4><p>Computer simulations have become a very important part in modern science, their results are as accurate as real life experiments. Therefore, computer simulations are used as a <em>complementary</em> tool for basically anything to generate required data for research and development, rather than conducting expensive experiments.</p><p>But a computer simulation can be expensive in terms of necessary resources and being time-consuming, depending on what exactly it has to simulate. In the case of Nuclear Fusion… that’s <strong>a lot</strong> of money and time <strong>⏰.</strong></p><p>Normal computer simulations in nuclear fusion take so long and cost so much due to the complexity of the physical processes involved in nuclear fusion. The simulations involve plasma temperature, rotation, and density. (<a href="https://www.computerweekly.com/feature/Computer-simulation-at-work-for-the-future-of-nuclear-fusion">6</a>) With such complex and difficult plasma physics problems in tokamaks, it’s becoming more and more impractical to use computer simulations directly for simulating these systems.</p><p>But to accurately approximate these simulations, we need a more feasible option than computer simulations in fusion energy.</p><p><em>What if we used one humanity’s most revolutionary technologies to solve such complicated problems- why not use it to transform the next energy revolution; Nuclear Fusion?</em></p><p>Artificial Intelligence (AI) is the groundbreaking frontier of technology, a field where machines emulate human intelligence to perform tasks, learn from data, and adapt to new situations<em>- the technology that humanity can use to our advantage for fusion.</em></p><p>It’s important to know the technology that’s been wired into our phones, computers, and even music streaming apps 🎶. Artificial intelligence has been on the rise in the past decade; companies like Google, Facebook, and others invested heavily in AI research, leading to breakthroughs in image recognition, natural language processing, and more. Now, it’s being used in autonomous vehicles, healthcare diagnostics, finance, and so many other industries.</p><blockquote>💡 Artificial Intelligence involves the development of algorithms and computer programs that can learn from data and improve their performance over time.</blockquote><p>AI is probably the most complex and shocking creations of humanity yet (<a href="https://www.forbes.com/sites/cognitiveworld/2019/06/19/7-types-of-artificial-intelligence/?sh=71454c99233e">7</a>). And it’s such a broad field, I mean there’s so many branches to it. We have Machine Learning, Deep Learning, Natural Language Processing, e.t.c. But don’t worry, I’m only here to touch on one <strong>specific</strong> application of AI for tokamaks.</p><p><em>Surrogate models.</em></p><h3>Surrogate models ≠ Surrogate mothers 🤰</h3><p>Haha don’t be confused with<em> surrogate mothers</em>... because I know I was 😳</p><p>The term <strong>“surrogate model”</strong> is derived from the English language and is used to describe a model that serves as a substitute or proxy for another, often more complex, model or system. (<a href="https://www.etymonline.com/search?q=surrogate+model">8</a>) The term “surrogate” itself comes from the Latin word “surrogatus,” which means “<strong>substitute</strong>” or “deputy.” In the context of modelling, a surrogate model takes the place of a more resource-intensive or time-consuming model to provide faster or more efficient approximations for various purposes- such as optimization, prediction, or analysis.</p><blockquote>The surrogate model is also known as an “emulator” ✨</blockquote><p>A surrogate model is a trained statistical/mathematical model that replaces computer simulations in predicting the behaviour of a specific system. (<a href="https://www.researchgate.net/publication/338312476_Surrogate_Model-Based_Engineering_Design_and_Optimization">9</a>) In the case of Nuclear Fusion, it’s used for the <strong>optimization </strong>of fusion reactors. The concept of AI x Fusion is still very new, but companies are slowly integrating the use of it, but the overall concept is still <strong>very unknown, </strong>which is why it hasn’t been done in the past.</p><p>Surrogate modelling is the process of constructing a statistical model that is used to bypass expensive/long simulations and accurately predict the output of the simulations. But the objective of a surrogate model is not to eliminate computer simulations entirely, but instead to reduce its need as much as possible. Surrogates must be properly trained in order to replicate the output from the original simulation. Training these models requires a machine learning (ML) based data-driven process. (<a href="https://browse.arxiv.org/pdf/2210.00073.pdf">10</a>)</p><p>In this process, approximate functions are used from pre-existing data 💻. These approximate functions are then used to generate output data that are relative to the original simulation results. The approximation functions of a surrogate model are a simplified version of computer simulations, and are <strong>much easier</strong> to analyze than the actual differential equations that describe the system. In this way, training surrogate models is <strong>much</strong> lower than that of the original simulation model, and surrogate models are much cheaper than computer simulations in terms of time-consumption. (<a href="https://browse.arxiv.org/pdf/2210.00073.pdf">10</a>)</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Hm5kyti7zS3m2IimaILCag.png" /><figcaption><strong>Figure 3: Basic concept of surrogate modelling and its correlation to computer simulation. </strong>Retrieved from; <a href="https://browse.arxiv.org/pdf/2210.00073.pdf">https://browse.arxiv.org/pdf/2210.00073.pdf</a></figcaption></figure><p>In <em>Figure 3,</em> we can see that a surrogate model replaces the original output functions<em> fi(x) </em>with <strong>approximation functions</strong> <em>Fi(x). </em>The precision/accuracy of the approximate functions indicates how effectively the model is trained. This correlates directly with the surrogate model’s performance.</p><p><em>But how should they be trained?</em></p><p>A pre-existing sample/data set is used at the very beginning to train the surrogate model. After training the model with the pre-existing data, the <br><strong>approximation functions</strong> are used to make special learning functions that can identify a new sample for further training. With this new sample, new computer simulations are run to generate new training data (<em>See Figure 4)</em>. Using the improved dataset, the surrogate model is re-trained. The process is repeated until the surrogate model is as <strong>accurate</strong> as possible. (<a href="https://books.google.ca/books/about/Engineering_Design_via_Surrogate_Modelli.html?id=ulMHmeMnRCcC&amp;redir_esc=y">11</a>)</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*1E_9Drmhi6b0ijAhfHtYIA.png" /><figcaption><strong>Figure 4: The process of training a surrogate model. </strong>Retrieved from; <a href="https://browse.arxiv.org/pdf/2210.00073.pdf">https://browse.arxiv.org/pdf/2210.00073.pdf</a></figcaption></figure><p>We can see that surrogate modelling is essentially just a <strong>special</strong> type of <em>supervised machine learning</em> that has many applications. Therefore, the training process of surrogate modelling is entirely within the domain of artificial intelligence, and it would essentially <strong>cut out </strong>the<strong> time + cost</strong> normal computer simulations use.</p><blockquote>💡“Supervised machine learning is a type of machine learning used in the context of surrogate models. In supervised machine learning, the model is trained on a labeled dataset, where the input data and the corresponding output data are known. The goal of supervised machine learning is to learn a mapping between the input data and the output data, so that the model can make accurate predictions on new, unseen data.”</blockquote><p>There are so many <strong>different</strong> machine learning approaches that can be used in surrogate modelling… we’ll be looking at <em>Artificial Neural Networks </em>(ANN)</p><h3>Understanding Artificial Neural Networks in the Context of Surrogate Modelling</h3><p><em>Artificial Neural Networks</em> (ANN), aka <em>Neural Networks</em> (NN) is a modelling <br>technique inspired by the neural network of a brain 🧠. (<a href="https://www.sciencedirect.com/science/article/abs/pii/S0098135414001719">12</a>)</p><p>I mentioned before how there are so many ML algorithms that can be applied to surrogate modelling. ANN’s are the <strong>most efficient + simple</strong> to control nonlinear systems (which in this case are the plasma instabilities in tokamaks). These networks are based off of <em>nodes</em> — an interconnected collection of artificial neurons. Each connection acts like the <br>synapses of a brain and can transmit signals. These signals represent real numbers. (<a href="https://www.sciencedirect.com/science/article/pii/S2666912922000046">13</a>)</p><p>As seen in <em>Figure 5, </em>the network has three layers; and <strong>input layer, hidden later, and output layer. </strong>The signal made by one layer is treated as the input for the layer <strong>next</strong> to it.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*FBK62stnZ_UyTDRdLM4PzA.png" /><figcaption><strong>Figure 5: A 3-Layer ANN Structure. </strong>Retrieved from; <a href="https://browse.arxiv.org/pdf/2210.00073.pdf">https://browse.arxiv.org/pdf/2210.00073.pdf</a></figcaption></figure><p><em>So what picture </em><strong><em>actually </em></strong><em>mean? Don’t worry — I got you!</em></p><p>Neural networks have an incredible ability to take meaning from complicated data. They can be used to extract patterns + detect trends that are too complicated to be noticed by humans or computer simulations 👁️ .</p><p>Think of it like this: A trained neural network is as an “expert” in the category of information you give it to analyze. But there’s so many other advantages;</p><ol><li><strong>Adaptive learning:</strong> ANN’s have the ability to learn how to do tasks based on the data given for training or from initial experience</li><li><strong>Self-Organization: </strong>An ANN can create its own organization or representation of the information it receives while you train it</li><li><strong>Fault Tolerance: </strong>A partial destruction of the network may lead to fault performance. However, some network capabilities can work, even with some major network damage! (<a href="https://ijritcc.org/index.php/ijritcc/article/view/2920/2920">14</a>)</li></ol><p>Let’s take a few steps back 🤚. Neural networks solve problems with a different approach compared to conventional computer simulations. They use an<em> algorithmic approach; </em>meaning the computer follows a <strong>specific</strong> set of instructions in order to solve a problem.</p><p>As I mentioned before, neural networks process information similar to how the human brain does. The brain learns from experience — and so do ANN’s. They consist of hundreds of simple processing units which are wired together in a complex communication network.</p><p>Each node in a layer serves as simplified models of neurons. These nodes transmit a signal when they receive a strong enough input signal from the connected nodes.</p><h4>An Artificial Neural Network is more than a Bunch of Neurons …</h4><p>An ANN is a functional neural network that consists of arranging these neurons into layers and making connections between these layers.</p><p>The <strong>input layer</strong> of neurons <strong>receives data</strong> from input files (pre-existing data). <strong>The output layer transmits information </strong>to external sources. Between these two layers, multiple hidden layers can exist. These internal layers consist of so many interconnected neurons arranged in different structures. The inputs and outputs of each hidden neuron are given to other neurons. (<a href="https://ijritcc.org/index.php/ijritcc/article/view/2920/2920">14</a>)</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/938/1*JAicfRGELcaNswoOfcuTeg.png" /><figcaption><strong>Figure 6: A simplified diagram of a neural network. </strong>Retrieved from; <a href="https://ijritcc.org/index.php/ijritcc/article/view/2920/2920">https://ijritcc.org/index.php/ijritcc/article/view/2920/2920</a></figcaption></figure><h4>Training an Artificial Neural Network: The Supervised Learning approach</h4><p>There are two approaches to training -<em>supervised and unsupervised. </em>In the context of surrogate models for fusion, we’ll be looking at <em>supervised learning.</em></p><p>The inputs + the outputs are given in supervised learning. The ANN processes the inputs and compares those resulting outputs to the desired outputs. Then these errors are put back through the network, causing the whole system to adjust. This process is repeated many times.</p><p>But training is not just a technique; it involves “<strong>intuition</strong>” and conscious analysis to prevent the network from being overly trained. Initially, an ANN adapts itself to the broad statistical patterns in the data 📊. Later on, it continues to gather knowledge about other data aspects that could seem unimportant. In some systems, this finalized network is then turned into hardware so that it can be faster, and they don’t lock themselves in but rather continue to learn while in production use. (<a href="https://ijritcc.org/index.php/ijritcc/article/view/2920/2920">14</a>)</p><p>AI has been rapidly improving in the past decade, so the topic of AI x Nuclear Fusion shouldn’t be too unfamiliar, <em>right</em>? I mean, are there <strong>other </strong>applications<strong> </strong>of AI for fusion optimization?</p><h3>Deep Reinforcement Learning Algorithms Vs Surrogate Models</h3><p><em>Deep reinforcement learning </em>(DRL)<em> algorithm</em>s are a type of machine learning that involves an agent interacting with an environment to learn something, and it gets rewarded in this process. (<a href="https://iopscience.iop.org/article/10.1088/1361-6587/acc60f">15</a>) These algorithms are applicable to fusion reactors, and are used to control the plasma in a tokamak, optimize the stability, confinement, and energy exhaust of the plasma. (<em>See Figure 7)</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/685/1*3fQNUs446aUtgbGMKFn1gw.png" /><figcaption><strong>Figure 7: Depiction of the learning loop used for DRL algorithms. </strong>Retrieved from; <a href="https://www.nature.com/articles/s41586-021-04301-9">https://www.nature.com/articles/s41586-021-04301-9</a></figcaption></figure><p>Currently, <strong>Google Deepmind </strong>is using DRL algorithms to <strong>control </strong>and <strong>regulate </strong>the plasma, however the topic of today is <strong>simulating </strong>plasma instabilities, so make sure to keep in mind that these are two different processes. 💡</p><p>We know that surrogate models can be used to <strong>approximate</strong> the behaviour of a system using a smaller number of simulations or experiments. And surrogate models only require a smaller amount of data to train the model, which can be computationally efficient in the context of nuclear fusion research.</p><p>DRL algorithms and surrogate models have different applications and requirements, but let’s be honest- <em>who’s better?</em></p><h4>Comparing the Efficacy: Surrogate Models vs. Deep Reinforcement Learning for Simulations in Fusion Research — Who’s Better?</h4><p>Let’s look at a variety of factors I’ve come up with to see which is more efficient in tokamaks. Remember, we want the most efficient approach as nuclear fusion is such a complex field!</p><ol><li><strong>Data requirements: </strong>Deep reinforcement learning algorithms require a large amount of data to be trained on, which can be computationally expensive in the context of nuclear fusion research. Surrogate models only require a smaller amount of data to train the model, which is <strong>very</strong> <em>computationally </em>efficient<em>.</em></li><li><strong>Accuracy:</strong> Surrogate models are very accurate when predicting the output for new input data. This is important in fusion research, where accurate predictions are necessary for optimizing plasma stability in tokamaks.</li><li><strong>When the objective is to optimize a specific outcome: </strong>Surrogate models can be designed to optimize a specific outcome, whereas DRL algorithms are more suited for learning complex decision-making processes.</li><li><strong>Costs + Time:</strong> Surrogate models can provide a faster + more efficient way to simulate the behaviour of complex systems, while reducing the computational cost significantly. I mentioned how deep reinforcement learning algorithms require more computational resources; these resources can take up months to process (the input data require many hours of computer time to simulate one second of real time) and cost too much per each simulation. (<a href="https://www.deepmind.com/blog/accelerating-fusion-science-through-learned-plasma-control">16</a>)</li></ol><p>In the context of fusion research, the feasibility of surrogate models is <strong>so</strong> important due to the <strong>complexity</strong> and cost of experimentation and simulations. With that being said, surrogate models are like a beacon of hope for fusion; they require significantly less data, achieve high accuracy in predicting complex system behaviour, and can help researchers and operators make informed choices in real-time 👩‍🔬.</p><h3>How Would Applying Surrogate Models to Fusion Work?</h3><p>I’ve just laid the foundation out for <strong>all</strong> the basics you have to know, covering essential topics like the nature of fusion reactors and their associated challenges, the inefficiencies of current computer simulations, the potential for improvement through the integration of AI, the fundamentals of ANNs, and the significant efficiency of surrogate models compared to existing AI technologies.</p><p>With these basics in place, our next step is to <strong>really </strong>understand the <em>practical</em> <em>applications</em> of surrogate models in fusion research, and the future they bring.</p><p>Here’s how surrogate models use data from fusion reactors to predict plasma turbulence and instabilities, specifically using Artificial Neural Networks (ANNs):</p><ol><li><strong>Data collection: </strong>First, a dataset is created by running a set of simulations. This dataset contains input parameters such as; plasma density, temperature, magnetic field strength, and corresponding output values such as; growth rate, mode structure for the plasma turbulence/instability. This data collection process does not require a large data set.</li><li><strong>Actually training the surrogate model: </strong>The dataset is then used to train an ANN (capable of learning complex relationships between inputs and outputs). The ANN is trained to approximate the behaviour of the simulations, creating a “surrogate” for the original model.</li><li><strong>Validation and testing:</strong> The trained surrogate model is then validated using a <strong>separate</strong> dataset to ensure its accuracy + reliability. This dataset can include data from different operating conditions, or rather experimental setups to test the surrogate model’s output.</li><li><strong>Prediction and analysis</strong>: Once the surrogate model is validated, it can be used to predict plasma turbulence and instabilities for new input parameters! Researchers can analyze the model’s predictions to gain insights into the underlying physics and make informed decisions about reactor operation and design.</li></ol><p>If you’re more of a visual learner, then check this out….</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*SqMKODVZgeTFAOZITEDtEQ.png" /><figcaption><strong>Figure 8: In statistical learning, it is standard practice to divide the data into training, validation and testing sets, where the training data is used to fit the model. </strong>Image + description retrieved from; <a href="https://www.osti.gov/pages/servlets/purl/1835548">https://www.osti.gov/pages/servlets/purl/1835548</a></figcaption></figure><p>The dark blue solid arrows on the left are essentially experimentally measured signals such as plasma core density, one-dimensional profile signals (plasma density profile), and two-dimensional signals (the magnetic fields), are all also used as inputs. The surrogate model then outputs the plasma instability information, such as the mode growth rate, frequency, and global mode structure shown in the lower left panel. (<a href="https://sun.ps.uci.edu/people/zlin/bib/dong21.pdf">17</a>).</p><p>Again, the instability properties are used as inputs to control algorithms, as shown by the red arrows, and the ANN does the rest of the work by using its analytical and predictive properties.</p><h4>Surrogate Models in Real-Time</h4><p>Surrogate’s are based on <em>mechanistic models</em> and use data from a physical asset to predict how the digital twin will respond and then take appropriate control actions.</p><p>Machine learning processes will be used in the dataset in an <br>effort to analyze the dynamics model and make adjustments to the digital twin, which will be repeated until 100% accuracy.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/924/1*PJeFNSelyEXMQNPKR3145g.png" /><figcaption><strong>Figure 9: Framework of interaction between surrogate model and the input data. </strong>Retrieved from; chrome-<a href="https://browse.arxiv.org/pdf/2210.00073.pdf">https://browse.arxiv.org/pdf/2210.00073.pdf</a></figcaption></figure><p><strong>The sun</strong>, as we know it, was simply once just known as a ball in the sky that comes and goes, and provided light, heat, and warmth. This endless flame, has revealed itself as a fusion reactor, shedding light on the immense potential of this energy source.</p><p>The realization of nuclear fusion’s promise for Earth, with its abundance, sustainability, and low environmental impact, has sparked hope for a cleaner, more prosperous future.</p><p>Although the challenge of controlling plasma in particular is difficult, our progress in technological advancements, particularly in AI, clearly demonstrates human creativity and resourcefulness. As we find ourselves on the cusp of a potential <strong>energy revolution</strong>, the promise of technological innovation, shown by surrogate models, provides so much hope.</p><p>Just as our ability to understand the Sun’s true nature evolved over centuries, our journey to unlock the full potential of nuclear fusion is an ongoing process. With surrogate models and continued scientific exploration, the dream of practical fusion energy production on Earth is now definitely within reach.</p><p>But the pursuit of knowledge and innovation shows our willingness for a brighter, cleaner, and more sustainable future ☀️ .</p><p>Thank you so much for reading! If you liked this, check out my other stories on Medium that I’ve published :)</p><p>Check out my <a href="https://www.linkedin.com/in/amrita-ghag/"><strong>Linkedin</strong></a></p><p>If you want to see more of what I’m doing, <a href="https://medium.com/@amritaghag08"><strong>Subscribe</strong></a> here!</p><p>Follow my <a href="https://substack.com/@amritaghag"><strong>monthly newsletter</strong></a> for more updates and progress ⭐️</p><h3>References</h3><ol><li>n.d. chap1.html. <a href="http://silas.psfc.mit.edu/introplasma/chap1.html#tth_sEc1.1.">http://silas.psfc.mit.edu/introplasma/chap1.html#tth_sEc1.1.</a></li><li>n.d. Adaptive sequential sampling for surrogate model generation with artificial neural networks <a href="https://www.sciencedirect.com/science/article/abs/pii/S0098135414001719.">https://www.sciencedirect.com/science/article/abs/pii/S0098135414001719.</a></li><li>n.d. Convolutional Neural Network applications in additive manufacturing: A review <a href="https://www.sciencedirect.com/science/article/pii/S2666912922000046.">https://www.sciencedirect.com/science/article/pii/S2666912922000046.</a></li><li>n.d. View of Research Paper on Basic of Artificial Neural Network. <a href="https://ijritcc.org/index.php/ijritcc/article/view/2920/2920.">https://ijritcc.org/index.php/ijritcc/article/view/2920/2920.</a></li><li>n.d.<a href="https://iopscience.iop.org/article/10.1088/1361-6587/acc60f."> https://iopscience.iop.org/article/10.1088/1361-6587/acc60f.</a></li><li>“‘;;.’” 2019. 7 Types Of Artificial Intelligence <a href="https://www.forbes.com/sites/cognitiveworld/2019/06/19/7-types-of-artificial-intelligence/?sh=71454c99233e.">https://www.forbes.com/sites/cognitiveworld/2019/06/19/7-types-of-artificial-intelligence/?sh=71454c99233e.</a></li><li>“‘;;.’” 2019. ‘“Surrogate”. <a href="https://www.etymonline.com/search?q=surrogate+model.">https://www.etymonline.com/search?q=surrogate+model.</a></li><li>“‘;;.’” 2019. ‘;;’ — YouTube. <a href="https://browse.arxiv.org/pdf/2210.00073.pdf.">https://browse.arxiv.org/pdf/2210.00073.pdf.</a></li><li>“‘;;.’” 2019. Engineering Design via Surrogate Modelling: A Practical Guide <a href="https://books.google.ca/books/about/Engineering_Design_via_Surrogate_Modelli.html?id=ulMHmeMnRCcC&amp;redir_esc=y.">https://books.google.ca/books/about/Engineering_Design_via_Surrogate_Modelli.html?id=ulMHmeMnRCcC&amp;redir_esc=y.</a></li><li>“Accelerating fusion science through learned plasma control.” 2022. Google DeepMind. <a href="https://www.deepmind.com/blog/accelerating-fusion-science-through-learned-plasma-control.">https://www.deepmind.com/blog/accelerating-fusion-science-through-learned-plasma-control.</a></li><li>Chandler, David L. 2013. “Understanding the turbulence in plasmas.” MIT News. <a href="https://news.mit.edu/2013/understanding-turbulence-in-plasmas-0429.">https://news.mit.edu/2013/understanding-turbulence-in-plasmas-0429.</a></li><li>“Deep learning based surrogate models for first-principles global simulations of fusion plasmas.” 2021. GTC.<a href="https://sun.ps.uci.edu/people/zlin/bib/dong21.pdf."> https://sun.ps.uci.edu/people/zlin/bib/dong21.pdf.</a></li><li>“DOE Explains…Tokamaks.” n.d. Department of Energy. <a href="https://www.energy.gov/science/doe-explainstokamaks.">https://www.energy.gov/science/doe-explainstokamaks.</a></li><li>Mathieson, SA. 2022. “Computer simulation at work for the future of nuclear fusion.” Computer Weekly. <a href="https://www.computerweekly.com/feature/Computer-simulation-at-work-for-the-future-of-nuclear-fusion.">https://www.computerweekly.com/feature/Computer-simulation-at-work-for-the-future-of-nuclear-fusion.</a></li><li>“Surrogate Model-Based Engineering Design and Optimization | Request PDF.” n.d. ResearchGate. <a href="https://www.researchgate.net/publication/338312476_Surrogate_Model-Based_Engineering_Design_and_Optimization.">https://www.researchgate.net/publication/338312476_Surrogate_Model-Based_Engineering_Design_and_Optimization.</a></li><li>“Taming the Plasma Edge: Reducing Instabilities in Tokamaks.” 2022. Department of Energy. <a href="https://www.energy.gov/science/fes/articles/taming-plasma-edge-reducing-instabilities-tokamaks.">https://www.energy.gov/science/fes/articles/taming-plasma-edge-reducing-instabilities-tokamaks.</a></li><li>“tokamak.” n.d. Wiktionary, the free dictionary. <a href="https://en.wiktionary.org/wiki/tokamak.">https://en.wiktionary.org/wiki/tokamak.</a></li></ol><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d0f88afa246e" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Stacking Success: Elevate Tech Conference — Layers of Insightful Talks, People-Packed Encounters…]]></title>
            <link>https://medium.com/@amritaghag08/stacking-success-elevate-tech-conference-layers-of-insightful-talks-people-packed-encounters-a84cb2fefe80?source=rss-28aecc9c0df8------2</link>
            <guid isPermaLink="false">https://medium.com/p/a84cb2fefe80</guid>
            <category><![CDATA[tech]]></category>
            <category><![CDATA[elevate]]></category>
            <category><![CDATA[the-knowledge-society]]></category>
            <dc:creator><![CDATA[Amrita Ghag]]></dc:creator>
            <pubDate>Sat, 30 Sep 2023 10:54:37 GMT</pubDate>
            <atom:updated>2023-10-17T20:49:39.380Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*7OwINwZcr0xlztpAluBjYQ.jpeg" /><figcaption>The Sandwich of Elevate.</figcaption></figure><h3>Stacking Success: Elevate Tech Conference — Layers of Insightful Talks, People-Packed Encounters, Mindset Transformations, Savoury Advice, and Tasty Takeaways</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1006/1*QnlUvmCWNSfQzeZtvpZh7A.png" /><figcaption>Tks Fam ❤️</figcaption></figure><p>The title is pretty self explanatory, right 😅?</p><p>In the world of technology, going to a tech conference honestly feels like taking a bite out of a richly layered sandwich 😮‍💨. Each layer represents a distinct aspect of the experience, combining to create a flavour profile that leaves a lasting impression!</p><p>The Elevate Tech Conference was no exception, offering a diverse assortment of talks, memorable encounters, transformative mindset shifts, expert advice, and valuable takeaways. Let’s unwrap the layers of insights, connections, and inspirations that made this event a delicious sandwich 🥪</p><blockquote>Elevate unites the world’s innovators to solve society’s greatest challenges. It’s a place to hear from world-class speakers, enjoy evening socials, build your influence, and connect with the community.</blockquote><p>The Main Stage talks at Elevate Tech Conference was like the foundation of the sandwich- it provided the sturdy base where the layers of knowledge and inspiration were built, as experts and influencers delved deeper into revolutionary ideas that will build the way to our future. But my favourite one had to be Chris Hadfield’s 👨‍🚀</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*GreutdecHJ0Q5mCh5I5lTw.png" /></figure><p>I don’t want to spoil too much, but Chris’ talk was on <em>The Speed of Light</em>, in the main stage. I loved his talk the most because;</p><ol><li><strong>Chris Hadfield is Chris Hadfield</strong>, but that’s not the reason I loved his talk. I loved his humour and passion for the general subject of space, and I truly learned so much more as a result of this authenticity 🚀</li><li><strong>The future of space exploration </strong>is something that I think about pretty often- where are we headed next, where is exploration taking us, what are we finding? Well, Chris Hadfield definitely had me on the edge of my seat when talking about all we understand in space, and what the future holds for us 🌌</li></ol><p>From inspirational fireside chats, like in <em>Women in Tech, </em>that ignited my curiosity to smaller talks in other buildings, like<em> Becoming an Exceptional Engineer</em>, that sculpted my skills, and thought-provoking panel discussions, like <em>The Future of Food, </em>that brought diverse perspectives to the table 😅, all of the talks were the <strong>bread and butter</strong> of my conference experience. They provided the essential framework, just like the slices of bread in a sandwich, holding everything together and offering a solid structure for the layers of insights to come 🍞</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*S0SAdKICk9L-dBqwG_SXeQ.png" /></figure><p>But as any good sandwich lover knows, it’s the fillings that truly define the taste. So, let’s look at the people I met in the community, the mindset shifts that occurred, the expert advice I would give to anyone, and the mouthwatering takeaways that made this Elevate Tech Conference an unforgettable experience.</p><blockquote><em>“Elevate will serve as a meeting place wherein discussion will swirl around the solutions to society’s greatest challenges.” </em><strong><em>— Daily Hive</em></strong></blockquote><p>People-packed encounters at the Elevate Tech Conference were like the vibrant and diverse ingredients in between the layers of a delicious sandwich 😋. These interactions were the heart and soul of the event, and was honestly my favourite part; interacting with the TKS community.</p><p>Entrepreneurs, tech enthusiasts, innovators, and experts could all meet to make meaningful connections and ideas, each adding their unique flavour to the overall experience. The conference was the ideal place to make valuable connections for any reason. While I met some extraordinary people, I mainly want to emphasize how important it was for me to interact with the TKS community.</p><p>I have to say, I have never been so thankful and blessed to be a part of such an amazing family, everyone at Elevate treated me with so much kind…. so thank you ❤️</p><p>I met a few alumni + activates I would have never met before, and even got to know them on a personal level during lunch/dinners. Or, even meeting the innovates was so refreshing, seeing new faces and giving advice, as I was just in their spot last year.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*24ZFtuGTjVnCHFMwcEK6ZQ.png" /></figure><p>These people really stuck with me throughout the whole experience, and you know who you are, and I honestly can’t thank everyone enough, but it was the community that made this experience a memorable one!</p><p>And so, just like the layers in a sandwich complement each other to create a yummy meal, the people-packed encounters at the conference complemented the other aspects of the event, leaving me with a sense of fulfillment and the promise of future hangouts 🧑‍🤝‍🧑.</p><blockquote>Elevate was able to shine a light on the tech icons who dream big, the innovative minds who take great risks, and the fearless creatives who envision a future that includes us all.</blockquote><p>Mindset transformations at Elevate were just like the subtle changes in flavour that happen when you eat a sandwich 🤤. Throughout the event, we had the opportunity to embrace new ideas, and ultimately reshape the way we approached the tech industry and our own personal development.</p><p>Something I thought was amazing was how speakers were able to show their growth mindset 🧠, as they shared their journeys of innovation and resilience, highlighting the importance of adaptability and a willingness to embrace change. For me, these lessons served as a <strong>catalyst</strong> to reevaluate my own attitudes toward challenges and setbacks, because we should all see obstacles as stepping stones toward success.</p><p>And going back to the Chris Hadfield talk ( I just loved it that much 🫣), I remember the way he mentioned to <strong>embrace curiosity</strong>, because ultimately that’s what scientists do everyday to get one step closer to space exploration 🌟.</p><p>It really got me thinking, without human curiosity, where would we all be today? For me, curiosity is not only the starting point for innovation but also the driving force that keeps the tech industry and our own personal development evolving. Just like a sandwich can be a delightful blend of unexpected flavours, curiosity can lead us down unexpected paths, bettering our lives and expanding our horizons!</p><p>As I went from talk to talk, and even just talking with others, I realized that this knowledge and innovation left me with a different mindset — more resilient, innovative, and adaptable; I understood how curiosity was the base for all this. These mindset transformations added a depth of flavour to the overall conference experience, making it an event that would resonate with me long after it finished!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*xUFyCdxDBNwbHJO5-ABYoA.png" /></figure><blockquote>The Elevate festival unites icons and innovators at the intersection of arts and technology.</blockquote><p>Okay… I think. I can say I had an awesome experience. But I want to share my advice to anyone who may attend Elevate next year, or who just wants to learn more about conferences in general.</p><ul><li><strong>Don’t be afraid to go out and meet people </strong>🗣️</li></ul><p>Let me just say this, people don’t bite! I know that everyone is scared of talking to these adults who are super successful, but they were your age once too, and they’re most likely excited to talk to someone as <strong>curious</strong> as they were. It’s so important to connect with people, because who knows, they could be your future boss, friend, or mentor.. you’ll never know if you don’t try!</p><ul><li><strong>Don’t miss the talks </strong>🎤</li></ul><p>For me, the ideal conference experience was <strong>going to talks</strong> + connecting with others, if there was a talk I was super interested in, I made time for it, because listening to innovators and their experience about how they got here can be a lesson to others… so if there’s a talk you really want to go to, then go! This is also where you customize your sandwich of knowledge!</p><ul><li><strong>Find <em>your</em> group of people </strong>🧑‍🤝‍🧑👭👬</li></ul><p>The reason that Elevate was such a memorable experience for me was because of my friends + the TKS community. You shouldn’t be in a big group when networking, but during lunch or sometimes during talks, it’s nice to have a couple friends who were there to laugh and adventure with me! Being involved in this community is very important, so this is where you start! Just like the perfect sandwich, the right advice and people can transform an ordinary journey into an extraordinary one.</p><h3>Overall 🔑 Takeaways — tl;dr</h3><ul><li>My experience at Elevate was like a sandwich, each part of meeting people and going to talks were like the layers of the sandwich, all adding up on one another to make a tasty meal!</li><li>Don’t miss any talks, because those are what made me understand curiosity + perseverance of young innovators- including you and me</li><li>Talk to people, eat some good food, share a laugh, have some deep conversations. Whatever it is, just make sure you’re making this experience as savoury as it can be.</li></ul><p>Just as every ingredient in a sandwich contributes to its overall taste, each element of this conference played a huge role in shaping my understanding of the tech industry and our place within it. The Main Stage talks was a sturdy foundation, the people-packed encounters enriched our network, mindset transformations ignited our potential, savoury advice seasoned our aspirations, and the tasty takeaways left us with a satisfying sense of fulfillment, I couldn’t be any happier.</p><p>Elevate was a great place of knowledge, inspiration, and connection that left an indefinite mark on all who attended. It left us hungry for more knowledge, more connections, and more opportunities to shape the future. This conference left us with appreciation for everything, and the fact that the journey of tech and innovation has endless possibilities, waiting to be savoured at every turn 🤩.</p><p>I want to thank Elevate for giving me this opportunity ❤️</p><h3>More Pictures!</h3><p>I just have so many to share 🤗</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*K78ELhO4Va8j_RUvIKg60Q.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/846/1*nNxNaEmH8x_0-VIEwEYPPw.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/986/1*rVhJhaaqfc-vvj16Okf8fA.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/974/1*TLNanUrPmDuE_zdrz9hS4g.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/982/1*Ox8jEfyd4AR5IJ76L5uRrQ.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*HwTZm6vm2-DFrk5eTJmV5Q.png" /></figure><p>Thank you so much for reading! If you liked this, check out my other stories on Medium that I’ve published :)</p><p>Check out my <a href="https://www.linkedin.com/in/amrita-ghag/"><strong>Linkedin</strong></a></p><p>If you want to see more of what I’m doing, <a href="https://medium.com/@amritaghag08"><strong>Subscribe</strong></a> here!</p><p>Follow my <a href="https://substack.com/@amritaghag"><strong>monthly newsletter</strong></a> for more updates and progress ⭐️</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a84cb2fefe80" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Enhancing Stellarator Efficiency Through AI-Driven Optimization]]></title>
            <link>https://medium.com/@amritaghag08/enhancing-stellarator-efficiency-through-ai-driven-optimization-b047c6ea849d?source=rss-28aecc9c0df8------2</link>
            <guid isPermaLink="false">https://medium.com/p/b047c6ea849d</guid>
            <category><![CDATA[nuclear-energy]]></category>
            <category><![CDATA[the-knowledge-society]]></category>
            <category><![CDATA[fusion-energy]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[Amrita Ghag]]></dc:creator>
            <pubDate>Sat, 02 Sep 2023 18:15:46 GMT</pubDate>
            <atom:updated>2023-09-02T19:02:03.447Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*e5YtdI8vp6bVqJZ9YSseJg.png" /></figure><p><strong>Stellarator</strong>….that sounds like a fancy word right? Imagine trying to explain what it means to your little siblings <strong>😅</strong></p><p>Okay, so why is this word important? Well, it has something to do with sustainable energy, specifically a concept that has fascinated scientists for decades: <strong>nuclear fusion.</strong></p><p>I know, it sounds pretty badass right? <strong>N-U-C-L-E-A-R — F-U-S-I-O-N</strong>. This remarkable process holds the promise of unlocking immense energy by fusing atomic nuclei. With that being said, check out my <a href="https://medium.com/@amritaghag08/the-sun-powers-all-life-but-what-powers-the-sun-900c9333dc7c">article</a> that explains the concept way more into depth!</p><p>However, while the potential is amazing, the path to harnessing fusion’s power has been riddled with scientific, engineering, and technological challenges.</p><p>At the forefront of this journey stands the stellarator — a sophisticated fusion device that aims to confine and stabilize superheated plasma for fusion reactions. The quest to make stellarators a practical energy source has sparked innovation and curiosity. But today, we’re on the cusp of a new chapter in this journey…….</p><p>Imagine a future where these sophisticated devices are given access to artificial intelligence (AI), something you don’t see everyday 🤩<strong>.</strong></p><p>And no, I’m not talking about “ChatGPT” type AI (the AI community is going to kill me with that comment 😭 ), but AI that can understand intricate patterns, forecast behaviour, and improve processes. I mean, what if we could combine the analytical strength of algorithms with the brilliance of plasma physics?</p><p>In this article, we’ll look into the captivating world where fusion science intersects with AI. I’ll explain the mechanics of stellarators, their potential, and the challenges they can deal with.</p><p>Buckle up as we embark on a voyage of discovery. As we navigate the fusion of science and technology, we’ll look into a future where the fusion of fusion and AI will illuminate the path to sustainable and limitless energy. Welcome to <strong>“Enhancing Stellarator Efficiency Through AI-Driven Optimization,</strong>” where the stars themselves can hold the key to our energy evolution!</p><h3>Stellarators and the Quest for Fusion</h3><p>In the realm of nuclear fusion research, <strong>stellarators, </strong>there’s that fancy word again, are <em>intricately</em> designed devices aimed at harnessing controlled fusion reactions. A stellarator is a device that confines plasma in the shape of a donut, known as a torus, using magnetic fields (1). These magnetic fields let scientists manipulate plasma particles and generate ideal conditions for fusion reactions. Stellarators generate twisting magnetic fields that wrap all the way around the donut shape using incredibly powerful electromagnets 🍩 .</p><p>A <em>key</em> challenge in fusion lies in replicating the intense conditions related to stars, where hydrogen nuclei overcome their repulsion to fuse and generate energy through high-speed collisions. Stellarators (now that we know what this word means, I’m going to be using it a lot more 🔥 <strong>), </strong>present a distinctive solution by confining plasma within elaborate magnetic field arrangements, offering a new pathway to crack this fusion puzzle.</p><p>At the core of a stellarator’s design is the <em>intricate</em> arrangement of <strong>magnetic coils (See Figure 1)</strong>. These coils produce a magnetic field that twists and winds in intricate patterns, generating a containment vessel for the plasma. Unlike other fusion devices like <strong>tokamaks — </strong>Stellarators’ cousin, which rely on a timed magnetic field made by electrical currents, stellarators use steady-state magnetic field configuration. This removes the need for large electrical currents and the associated instabilities, letting in longer plasma confinement.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*zLo2aZ0RO1BXEDkShKHelA.png" /><figcaption><strong>Figure 1:</strong> Schematic image of permanent magnet stellarator with plasma in yellow. Red and blue indicate permanent magnets with simplified coils surrounding the vessel. <a href="https://scitechdaily.com/science-made-simple-what-is-a-stellarator/#google_vignette">Credit: Coaxing Zhu</a></figcaption></figure><p>Stellarators solve a <em>major</em> difficulty in fusion research: the need for long confinement times to generate the conditions needed for fusion events. By maintaining plasma stability over long periods of time, stellarators produce a condition where atomic nuclei have a greater chance of colliding and fusing, resulting in the desired energy output.</p><p>But, the complexity of stellarator designs presents its own set of challenges. Achieving the perfect magnetic field requirements for stable plasma confinement needs lot’s of engineering and precision. The interplay between magnetic fields, plasma pressure, and particle motion must be carefully balanced to prevent instabilities that could disrupt the fusion process.</p><p>Stellarators have progressed from theoretical concepts to practical experimental devices over the history of fusion research. They have distinct advantages over other fusion techniques, including natural steady-state functioning, less reliance on electrical currents, and potential compatibility with a wide range of fusion fuels. But still, improving stellarator efficiency is a driving factor in fusion research, and an idea that I want to pitch!</p><p>In the next few sections, we’ll look deeper into the intersection of science and artificial intelligence, seeing how these complicated devices can be further enhanced by AI-driven optimization…..</p><h3>The Fusion of Science and AI</h3><p>Imagine plasma physics as the tango <strong>💃 🕺</strong> of charged particles, a field where high-energy matter unfolds in a lot of complexity. It’s here that nuclear fusion’s secrets are unlocked, where atoms collide and unleash energy like cosmic <strong>fireworks</strong>.</p><p>But, within this <em>enchanting </em>choreography lies a puzzle — a puzzle AI can solve.</p><p>It’s not just about processing data; it’s about recognizing the little details that escape the human eye, testing out connections that would otherwise stay hidden. This ability makes AI into the perfect partner in crime for enhancing stellarator efficiency in fusion energy detective 🕵️‍♀️ .</p><p>Picture stellarators — very detailed machines designed to control the wild energy of plasma. Their purpose: to confine <em>superheated</em> matter for the conditions necessary for fusion.</p><p>However, this is no easy task. Plasma’s unpredictable behaviours needs constant adjustment to magnetic fields, just like steering a ship through a stormy sea 🌊 . This is where AI shines, as its data-crunching abilities breathe life into this complex problem. The strength of AI is in deciphering intricate relationships from massive datasets. AI can reveal these hidden patterns, correlations, and anomalies that escape from us!</p><p>AI doesn’t just stop at deciphering data — it predicts. It foresees how changes in magnetic fields impact plasma’s dance, foreseeing the consequences before they happen, which I’ll explain more in the next section! This predictive ability holds the key to real-time <em>optimization</em>. As AI learns from past experiments, it anticipates plasma’s every move. It adjusts magnetic fields on the spot, making sure plasma remains in harmony, untouched by any disruptions.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*5FhIeS2vETgxazX8z2Fx3A.png" /><figcaption><strong>Figure 2</strong></figcaption></figure><h3>The Role of Data Collection and Sensors</h3><p>In the <strong>heart</strong> of the fusion quest is the challenge of maintaining the equilibrium of plasma in stellarators. This is where AI sensors go beyond simple observation, becoming an interface for revealing details of plasma. Picture these AI sensors as expert interpreters, translating the complex language of plasma into concrete data points.</p><p>These AI sensors are not inactive observers; they can predict plasma behaviour based on real-time data collecting (2). These sensors understand subtle variations in temperature, density, and behaviour in the plasma dynamics. Variations in plasma temperature are similar to changes in tone in a piece of music, representing the dynamic motion of particle interactions. These variations are then captured by AI sensors, giving great insight into the <em>thermodynamic</em> landscape of plasma within the stellarator.</p><p>Let me explain more about how this works; density changing inside of plasma can affect the particle interactions. These AI sensors capture these density fluctuations, transforming them into identifiable data patterns. AI sensors give really good information on the interplay between particles and magnetic fields!</p><p>Stellarators use complex magnetic fields to confine the plasma. AI sensors are placed around the plasma containment area to monitor the strength, stability, and configuration of the magnetic fields. This data would help optimize the field for efficient plasma confinement <strong>(See Figure 3)</strong>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*0vgSUWOc3bMhwhIvXrwt0w.png" /><figcaption><strong>Figure 3: </strong>A diagram I created to help you understand the process these AI Sensors would go through</figcaption></figure><p>And, the importance of these AI sensors goes beyond interpretation. Their predictive ability shines through, closing the gap between theoretical understanding and practical observation. The data they collect is given to AI algorithms, letting them predict future plasma instabilities and make real-time modifications to maintain stability.</p><h3>Machine Learning Algorithms and Plasma Behaviour</h3><blockquote>A machine learning algorithm is a computer-like method that lets computers learn patterns, anticipate outcomes, or make judgements based on data without being explicitly programmed (3).</blockquote><p>It’s a set of mathematical rules and approaches that a computer system uses to improve its performance on a specific task through data analysis &amp; adaptation. When given fresh, unknown data, machine learning algorithms <em>evaluate</em> and analyze it to identify underlying patterns, correlations, and trends, which they then use to make predictions or judgements. These algorithms are frequently used in a variety of areas, including as image and speech recognition, natural language processing, recommendation systems, and scientific research; just like the investigation of complicated phenomena such as plasma behaviour in fusion research..which is what I will be explaining!</p><p>Their <strong>analytical</strong> skills let them filter through large datasets with a critical eye 👁️ , surpassing the limits of the human brain when it comes to understanding detailed trends…but let me explain more scientifically..</p><p>These algorithms, which come with a range of mathematical procedures and statistical techniques, can process a large number of data points in an <em>organized</em> way. This processing goes beyond basic calculating, and instead can recognize underlying patterns that are embedded within messy and complex datasets. This unique ability let’s machine learning algorithms gain detailed information from data, exposing hidden connections that were previously unknown.</p><p>These algorithms serve as channels, connecting different plasma characteristics. Their ability to unravel the threads that connect temperature variations, density fluctuations, and behavioural <em>complexities</em> adds to more understanding of plasma behaviour. Machine learning algorithms can serve as translators in this process, transforming raw data into a good story that researchers can interpret and use to improve the efficiency of stellarators <strong>🤯.</strong></p><h3>Predictive Maintenance and Proactive Strategies</h3><p>Within the framework of enhancing stellarator efficiency, something important emerges — the <strong>integration</strong> of AI-driven predictions and proactive strategies. AI’s machine learning capability takes on the position of a “watchdog” 🐕 , monitoring and examining lots of measurements associated with stellarator operations <strong>(See Figure 4)</strong>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*e5YtdI8vp6bVqJZ9YSseJg.png" /><figcaption><strong>Figure 4</strong></figcaption></figure><p>But, AI comes with many <strong>unpredictable</strong> things! AI-guided procedures stop unplanned delay by catching possible difficulties in their early stages — a characteristic that might severely compromise experimental and operational efficacy. This knowledge lets operational teams plan repair activities during pre-scheduled downtimes, minimizing disruptions to stellarator operations. AI-driven foresight reduces inefficient resource allocation by avoiding issues before they get really bad 😰. This corresponds to improved operating <em>dynamics</em>, reduced expense, and, eventually, increased the overall efficacy of stellarator operations.</p><p>As AI’s predictive powers grow, the field of fusion research has the potential to lead to <strong>critical</strong> discoveries. The combination of AI-driven predictive maintenance and proactive measures <em>accentuates</em> the partnership between AI’s analytical precision and stellarator operations’ complicated <em>labyrinth</em>. This incorporation of AI-driven predictive maintenance and proactive measures is a cornerstone in the landscape of increasing stellarator efficiency. This combination not only highlights AI technology’s potential, but also its dramatic impact on fusion research ☀️ . This amazing collaboration of two emerging technologies promises to reduce disruptions, increase operational efficiency, and extend equipment <em>longevity</em> — a mixture that holds the key to unleashing the full potential of stellarator technology 😳.</p><h4>Collaboration and Multidisciplinary Innovation</h4><p>This little sub-section 😮 is to explain how important it is for everyone to work together!</p><p>….. Think of stellarators like complicated machines that need different knowledge.</p><p>There are scientists who <strong>know</strong> how plasma behaves, engineers who <strong>make</strong> the machines, and AI experts who make smart <strong>predictions</strong>. When these experts work together, they can make stellarators <strong>perform</strong> even better.</p><p>This teamwork is not just for fusion. It also helps AI technology. AI can understand plasma behaviour better, which means better control over the machines. Also, using stellarators to test AI helps improve how AI works, and ChatGPT will finally update! 😅/jk</p><p>This kind of teamwork makes both fusion and AI research <em>better</em>. When different experts share their ideas, they can find new and better ways of doing things. It’s like when people with different skills work together to make something amazing. This teamwork can bring us closer to making fusion energy a real and useful thing!</p><h3>The Road Ahead and the Promise of Fusion</h3><p>The <em>convergence</em> of AI-driven efficiency and its seamless integration into stellarator technology is an incredible leap in fusion energy development.</p><p>Amplifying the efficiency of stellarators is more than just a technical achievement; it represents an important step towards the achievement of controlled and sustained fusion reactions — a major milestone on the path to harnessing the hidden potential of fusion energy!</p><p>The greater efficiency translates into <strong>increased</strong> energy output compared to input, overcoming a long-standing stumbling barrier for fusion.</p><p>The effects go far beyond lab successes and academic honours though….</p><p>The increased efficiency of stellarators not only accelerates the advancement of fusion energy, but also moves it closer to the limit of <strong>practical implementation</strong>. This closeness to reality underlines the potential of a scalable and dependable energy source with large power generation. The practicality of fusion energy has huge impacts on global energy security and environmental sustainability. It has the potential to free civilizations from the limitations of finite fossil fuels while also significantly contributing to climate change mitigation — a brave step towards charting a sustainable future for our planet 🌎 .</p><p>In addition, the introduction of efficient fusion energy has the potential to <em>catalyze</em> harmful disruptions across businesses and economies. The emergence of a nearly limitless and environmentally friendly energy source could cause major disruptions in industries ranging from transportation and manufacturing to <strong>agriculture 👩‍🌾 </strong>and <strong>healthcare 👩‍⚕️ </strong>. It has the ability to transform economic landscapes, moving them towards low-carbon, environmentally sensitive paths that not only solve energy demand but also shift society towards sustainable growth and development.</p><p>Still, the path to the promise of fusion energy is filled with obstacles and needs continuous research, technical innovation, and <em>multidisciplinary</em> collaboration. The combination of AI and stellarator technologies represents the spirit of this endeavour — a mix of innovation and collaboration that pulls the fusion industry forward.</p><h3>Conclusion: Bridging the Gap Between Vision and Reality</h3><p>Stellarator.. now that you’ve read this article, I hope that word doesn’t sound fancy anymore 😅.</p><p>With my <strong>proposal</strong> + ongoing research, we have discovered new dimensions of efficiency and performance increase through the <em>symbiotic</em> combination of AI and stellarator skills, characterized by real-time adaptation, predictive foresight, and data-driven precision. This fusion has the potential to push stellarator efficiency beyond the limits of existing concepts, providing a future where the promise of fusion energy becomes a reality.</p><p>As we come to the end of my article, a powerful call to action arises 👊 .</p><p>Fusion research is still evolving, thanks to advances in AI, stellarator design, and multidisciplinary collaboration. So, I want everyone to stay up to date on the <strong>advances</strong> made in fusion research and the quick improvements that continue to transform the fusion landscape! Because who knows, one day something big could happen with these emerging technologies 🥳.</p><p>……Man, Stellarators, gotta love them!</p><p><strong>tl;dr: </strong>Check out my <a href="https://youtu.be/5j4Mwohc0zY">video</a> I made which summarizes this whole idea!</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.loom.com%2Fembed%2F3cf4d40776a94a6f9c81ab969a48d283&amp;display_name=Loom&amp;url=https%3A%2F%2Fwww.loom.com%2Fshare%2F3cf4d40776a94a6f9c81ab969a48d283%3Fsid%3D79f4cd5c-0936-4c07-9fec-12113e46268b&amp;image=https%3A%2F%2Fcdn.loom.com%2Fsessions%2Fthumbnails%2F3cf4d40776a94a6f9c81ab969a48d283-00001.gif&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=loom" width="1920" height="1440" frameborder="0" scrolling="no"><a href="https://medium.com/media/dd459f4ec5f250619fc4e39fb5d94ec8/href">https://medium.com/media/dd459f4ec5f250619fc4e39fb5d94ec8/href</a></iframe><p>Thank you so much for reading! If you liked this, check out my other stories on Medium that I’ve published :)</p><p>Check out my <a href="https://www.linkedin.com/in/amrita-ghag/"><strong>Linkedin</strong></a></p><p>If you want to see more of what I’m doing, <a href="https://medium.com/@amritaghag08"><strong>Subscribe</strong></a> here!</p><h3><strong>Sources</strong></h3><ol><li><a href="https://www.energy.gov/science/doe-explainsstellarators">https://www.energy.gov/science/doe-explainsstellarators</a></li><li><a href="https://www.pppl.gov/news/2021/artificial-intelligence-speeds-forecasts-control-fusion-experiments">https://www.pppl.gov/news/2021/artificial-intelligence-speeds-forecasts-control-fusion-experiments</a></li><li><a href="https://www.techtarget.com/whatis/definition/machine-learning-algorithm">https://www.techtarget.com/whatis/definition/machine-learning-algorithm</a></li><li><a href="https://iopscience.iop.org/article/10.1088/1741-4326/ac29d0/ampdf">https://iopscience.iop.org/article/10.1088/1741-4326/ac29d0/ampdf</a></li><li><a href="https://www.popularmechanics.com/science/energy/a39124779/deepmind-ai-controls-plasma-nuclear-fusion-reactor/">https://www.popularmechanics.com/science/energy/a39124779/deepmind-ai-controls-plasma-nuclear-fusion-reactor/</a></li><li><a href="https://www.nature.com/articles/s41586-021-04301-9">https://www.nature.com/articles/s41586-021-04301-9</a></li><li><a href="https://www.newscientist.com/article/2308243-deepmind-uses-ai-to-control-plasma-inside-tokamak-fusion-reactor/">https://www.newscientist.com/article/2308243-deepmind-uses-ai-to-control-plasma-inside-tokamak-fusion-reactor/</a></li><li><a href="https://onlinelibrary.wiley.com/doi/full/10.1002/ctpp.202300060">https://onlinelibrary.wiley.com/doi/full/10.1002/ctpp.202300060</a></li><li><a href="https://www.weforum.org/agenda/2023/06/the-renaissance-of-the-stellarator-what-this-means-for-fusion-energy/">https://www.weforum.org/agenda/2023/06/the-renaissance-of-the-stellarator-what-this-means-for-fusion-energy/</a></li><li><a href="https://scitechdaily.com/science-made-simple-what-is-a-stellarator/#google_vignette">https://scitechdaily.com/science-made-simple-what-is-a-stellarator/#google_vignette</a></li></ol><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b047c6ea849d" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Application of Metal Hydrides and Hydrogen in Nuclear Fusion Reactions: Tritium Management and…]]></title>
            <link>https://medium.com/@amritaghag08/the-application-of-metal-hydrides-and-hydrogen-in-nuclear-fusion-reactions-tritium-management-and-1f5965c606dd?source=rss-28aecc9c0df8------2</link>
            <guid isPermaLink="false">https://medium.com/p/1f5965c606dd</guid>
            <category><![CDATA[the-knowledge-society]]></category>
            <category><![CDATA[science]]></category>
            <category><![CDATA[nuclear-fusion]]></category>
            <dc:creator><![CDATA[Amrita Ghag]]></dc:creator>
            <pubDate>Tue, 22 Aug 2023 22:04:15 GMT</pubDate>
            <atom:updated>2023-08-25T15:57:48.399Z</atom:updated>
            <content:encoded><![CDATA[<h3>The Application of Metal Hydrides and Hydrogen in Nuclear Fusion Reactions: Tritium Management and Beyond</h3><h3>Abstract</h3><p>Nuclear fusion has emerged as a promising clean and sustainable energy source, offering a virtually limitless supply of power with minimal environmental impact. At the core of fusion reactions lies the efficient management of tritium, a radioactive isotope of hydrogen and a key fuel for fusion. This <strong>research paper</strong> explores the application of metal hydrides and hydrogen in nuclear fusion reactions, with a particular focus on tritium management. It delves into the use of metal hydrides for tritium storage, breeding, and processing, as well as the crucial role of hydrogen isotopes in plasma heating and safety mechanisms. Additionally, the paper discusses potential future applications and advancements in harnessing metal hydrides and hydrogen to advance nuclear fusion as a practical energy solution.</p><h3>Table of Contents</h3><h4><strong><em>Introduction</em></strong></h4><p>1.1 Background</p><p>1.2 Significance of Tritium Management in Fusion</p><p>1.3 Motivation for Metal Hydrides and Hydrogen in Fusion</p><h4><strong><em>2. Tritium Management in Fusion Reactions</em></strong></h4><p>2.1 Tritium as a Fusion Fuel</p><p>2.2 The Role of Metal Hydrides in Tritium Storage</p><p>2.3 Tritium Breeding with Metal Hydrides</p><p>2.4 Tritium Processing with Metal Hydrides</p><h4><strong><em>3. Hydrogen in Plasma Heating</em></strong></h4><p>3.1 Role of Hydrogen in Fusion Reactions Vs Fission Reactions</p><p>3.2 Plasma Heating Techniques</p><p>3.3 Importance of Hydrogen in Plasma Confinement</p><h4><strong><em>4. Safety Measures with Metal Hydrides and Hydrogen</em></strong></h4><p>4.1 Safety Merits in Fusion Reactors</p><p>4.2 Metal Hydrides as Passive Safety Mechanisms</p><h4><strong><em>5. Current Research and Experimental Findings</em></strong></h4><p>5.1 Overview of Current Fusion Research</p><p>5.2 Advancements in Metal Hydride Technology: 1980’s-1990’s</p><p>5.3 Experimental Studies on Tritium Management: Methods that don’t work</p><h4><strong><em>6. Synthesis Methods of Metal Hydrides: How are they Formed?</em></strong></h4><p>6.1 Reactive Milling</p><p>6.2 Mechanochemical Methods</p><p>6.3 Electrochemical Synthesis</p><h4><strong><em>7. The Manifestation of Cold Fusion</em></strong></h4><p>7.1 Introduction to Cold Fusion</p><p>7.2 The Connection Between Cold Fusion and Metal Hydrides</p><p>7.3 Efficiency and Optimization of Metal Hydrides</p><h4><strong><em>8. Conclusion</em></strong></h4><p>8.1 Importance of Continued Research and Development</p><p>8.2 What Will the Future Hold for Metal Hydrides?</p><h3>Introduction</h3><p><strong>1.1 Background</strong></p><p>Nearly all human-made forms of power production depend <em>significantly</em> on hydrogen. The energy from the sun and the hydrogen in the biosphere, which once gave life to the universe, are what give rise to the hydrocarbons in fossil fuels. Humans have relied on the combustion of hydrocarbons (coal, wood, and later fossil fuels) as their <em>primary</em> energy source for thousands of years. Modern and cutting-edge technology still produces electricity using hydrogen. Proton-proton fusion processes in the sun provide energy that heats the earth and fuels the production of electricity from wind and solar energy (14). To produce an electrical current, fuel cells most frequently rely on the recombination of hydrogen and oxygen. Hydrogen and its isotopes, <strong>Deuterium</strong> and <strong>Tritium</strong>, are <em>essential</em> to the process of Nuclear fusion as well. Deuterium-tritium (DT) fusion is usually regarded as the most practical kind of fusion because it has the highest reaction cross section at the lowest centre of mass energy <em>(see Figure 1)</em>. Nearly all advanced fusion ideas will continue to incorporate some combination of hydrogen, which all offer many advantages.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/706/1*sJvxXqNiinw5dpPl0wukXg.png" /><figcaption>Figure 1</figcaption></figure><p><strong>1.2 Significance of Tritium Management in Fusion</strong></p><p>The most prevalent element in the universe, hydrogen, has isotopes called deuterium and tritium. All hydrogen isotopes have one proton, but tritium and deuterium each have two neutrons, making their ion weights higher than protium, the only hydrogen isotope without any neutrons <em>(see Figure 2)</em>. The Deuterium-Tritium reaction is the best fuel for Fusion reactors (15). Luckily, deuterium is widely available. Deuterium makes up about 1 out of every 5,000 hydrogen atoms in seawater. This shows there are tonnes of deuterium in our oceans. Tritium, on the other hand, is a radioactive isotope that is uncommon in nature and decays very quickly.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*NDh7Zkjxx3URmIk4tDPWOA.png" /><figcaption>Figure 2</figcaption></figure><p>Let’s get a little more in-depth as to why tritium is a scarce source and why it is so crucial to manage it during the fusion process. Natural processes in the high atmosphere result in tiny amounts of tritium when cosmic rays strike nitrogen molecules. Tritium is absorbed into water, which then rains down on Earth. Tritium only makes up around a <em>billionth</em> of a billionth of one percent of the three hydrogen isotopes, which shows how scarce it is on Earth. However, there is still <em>some</em> hope because tritium can be produced.Tritium is short-lived. It is radioactive and begins to undergo radioactive decay and transform into non-radioactive helium as soon as new tritium is created. After 12.3 years, half of the initial amount of tritium is gone. This makes it impossible to synthesize large quantities of tritium in advance and store it. The bar <em>(see Figure 3</em>) shows how much tritium is usually expected of a Fusion reactor to consume every day. The bars on the right show how much tritium could be produced. Tritium’s absence as a fuel source in nature has long been known to fusion scientists. Additionally, tritium’s importance as a fuel component for fusion reactors has been known for decades. Given that Tritium is a scarce supply, effective methods of handling it in the Fusion process are crucial.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*QjqGqPUiT1rYSQoo6XbrJA.png" /><figcaption>Figure 3</figcaption></figure><p><strong>1.3 Motivation for Metal Hydrides and Hydrogen in Fusion</strong></p><p>Tritium processing with metal hydrides is like a smart way to handle a special kind of hydrogen, called tritium, in fusion science. It’s important because tritium is radioactive and can be tricky to manage. Metal hydrides act like helpful containers for tritium. They can absorb tritium and keep it safe. When we need to use the tritium for fusion, the metal hydrides can release it in a <em>controlled</em> manner. This helps us avoid any accidents and also lets us reuse the tritium for more fusion reactions. These metal hydrides also make sure the tritium is really clean and pure, which is super important for making fusion happen. So, tritium processing with metal hydrides is like a smart and safe way to use tritium in fusion research.</p><p>Let’s get more scientific… Using metal hydrides and hydrogen in fusion research is driven by several key motivations. As previously mentioned, Tritium is a <em>crucial</em> fuel for fusion reactions and needs efficient management. Metal hydrides offer promising solutions for tritium storage, breeding, and processing, ensuring a self-sustaining fuel cycle and a reliable supply of fuel. Safety is a concern in fusion, and metal hydrides can act as passive safety mechanisms, securely containing tritium and minimizing the risk of environmental contamination. According to <strong>ThoughtCo,</strong> “Metal hydrides are metals which have been bonded to hydrogen to form a new compound.” Any hydrogen compound that is bonded to another metal element can effectively be called a metal hydride. The most common examples of metal hydrides include aluminium, boron, lithium borohydride and various salts. Metal hydrides have long sparked the curiosity of scientists who are interested in moderating reactions in high-temperature fusion reactors. While pressurized water is typically used as a moderator in standard reactors, metal hydrides have an equivalent or higher concentration of hydrogen and can operate at high temperatures without the high pressure that water needs.</p><blockquote>According to <strong>energy.gov</strong>, “The high hydrogen density and moderating efficiency of metal hydrides enables smaller reactor cores that can operate more efficiently and reduce waste products.”</blockquote><p>Metal hydrides serve as <em>effective</em> storage materials by forming chemical compounds with hydrogen atoms. When hydrogen, in the form of tritium or deuterium, is introduced to the metal hydride material, it undergoes a process of <strong>absorption</strong> and <strong>dissolution</strong>. The hydrogen atoms diffuse into the crystal structure of the metal hydride, where they react with the metal to create metal hydride compounds. This storage mechanism allows metal hydrides to store hydrogen at relatively low pressures and moderate temperatures, making them useful in fusion reactors. Metal hydrides’ ability to efficiently store and release hydrogen on demand is critical for maintaining a steady and reliable supply of fuel, as well as ensuring safety in handling and containing the radioactive tritium used in fusion reactions.</p><h3>Tritium Management in Fusion Reactions</h3><p>In this chapter on Tritium Management in Fusion Reactions, we delve into the pivotal role of tritium, a swiftly decaying hydrogen isotope, as a fusion fuel. While various isotopes can facilitate fusion, the deuterium-tritium (DT) reaction stands out as the most efficient, currently employed by the <strong>ITER</strong> reactor.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/762/1*5xCy1m4g5OwE0N5hSc5M6Q.png" /></figure><p>Tritium, present in limited quantities in nature, can be generated through fusion interactions involving <em>lithium</em>. Proper containment and handling protocols are essential, given its tendency to permeate materials. The subsequent sections explore innovative approaches to tritium management, particularly focusing on the role of metal hydrides in tritium storage. Metal hydrides, adept at chemically absorbing hydrogen isotopes, offer secure and efficient tritium containment. These materials can be heated to release tritium, providing a safe alternative for storage and transportation. Furthermore, metal hydrides facilitate hydrogen compression, isotope separation, and gas purification through controlled heat-induced processes, enhancing worker safety and reducing environmental emissions. By exploring these aspects, this chapter aims to contribute to the effective management and usage of tritium in fusion reactions.</p><p><strong>2.1 Tritium as a Fusion Fuel</strong></p><p>Although other light element isotopes can be used to achieve fusion, the<strong> deuterium-tritium </strong>(DT) reaction has been recognized as the most efficient for fusion devices. And, ITER is using this reaction to power its reactor. As you may recall, Tritium is a rapidly decaying element of hydrogen that exists in small amounts in nature (1). However, tritium can be produced during the fusion reaction by interaction with lithium: tritium is produced, or “bred,” when neutrons leaving the plasma interact with lithium contained in the tokamak’s blanket wall. However, Tritium can pose safety risks due to its radioactivity and potential health and environmental hazards. Emitting low-energy beta particles during decay, tritium can penetrate the human body through ingestion, inhalation, or skin absorption, potentially causing cellular damage and health issues. Prolonged exposure to tritium increases the risk of radiation-related health problems, including cancer and genetic mutations. Still, with responsible management and adherence to safety protocols, tritium can be used safely in controlled environments, which brings us to our next section..</p><p><strong>2.2 The Role of Metal Hydrides in Tritium Storage</strong></p><p>Hydrides are materials, specifically metals, that can chemically absorb hydrogen (and its isotopes deuterium and tritium) into their metal lattice structure, acting as a very effective pump at room temperature <em>(see Figure 4) </em>(2). Metal hydrides often hold hydrogen at higher densities than liquid hydrogen. Because these metal hydrides must be heated to recover the hydrogen (or deuterium tritium fuel), they are a <em>very</em> safe way to store and transport tritium and can be used for storage instead of tanks, to store hydrogen as a gas. They can also be used to pump/compress hydrogen, separate hydrogen isotopes, and purify hydrogen from other gases. All of this can be done through the use of heat, with no or few mechanical moving parts. Using metal hydride to process tritium can reduce the risk of tritium emission into the environment while also increasing worker safety.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*juZ_W5Y6kal6xbdXPJautQ.png" /><figcaption>Figure 4</figcaption></figure><p><strong>2.3 Tritium Breeding with Metal Hydrides</strong></p><p>Metal hydrides reversibly react with hydrogen isotopes, and the process is controlled by temperature and hydrogen pressure <em>(see Figure 5)</em>. Pressure or temperature can be used as the driving force in the equation to affect the reaction direction.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/600/1*I-adFK0EnQKjdFCXMprJpA.png" /></figure><p>Most metal hydrides have a very high hydrogen density, which can be as high as liquid hydrogen. Metal hydride selection is influenced by characteristics such as hydrogen absorption capability, stability, and compatibility. When three different forms of hydrogen (protium — aka regular hydrogen, tritium, and deuterium) are mixed together, their pressures may not necessarily match. The proportions of these kinds of substances in the gas and solid phases can change when they are mixed with a metal hydride. Depending on the metal hydride, this difference can be significant or little. This unique difference in proportions might help in distinguishing between different forms of hydrogen. A metal like palladium, for example, can be used to differentiate various hydrogen types since it has a considerable effect on their proportions.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/884/1*zK-RlUAI84AVJyFT8ackMA.png" /><figcaption>Figure 5</figcaption></figure><p><strong>2.4 Tritium Processing with Metal Hydrides</strong></p><p>Tritium processing starts with tritium capture, in which metal hydrides act as chemical sponges, absorbing tritium from sources such as water that has been tritiated (contaminated) or fusion reactor exhaust gases. The absorbed tritium produces stable compounds inside the metal hydride’s crystal structure, allowing secure containment and preventing unintentional tritium emission.</p><p>The next step is tritium extraction, which is done through controlled desorption. Tritium molecules are released from their captive condition by sending the metal hydrides to controlled increases in temperature, allowing for recovery and reuse as fusion fuel. This temperature-driven desorption technique provides for exact control of tritium emission rates while reducing operational risks.</p><p>Metal hydrides are also important in tritium purification. This involves taking out tritium selectively from mixes including other isotopes, guaranteeing the purity and quality of tritium fuel. The capacity of metal hydrides to selectively collect tritium contributes to the production of high-purity tritium fuel, which is required to maintain efficient fusion reactions.</p><h3>Hydrogen in Plasma Heating</h3><p>In this chapter dedicated to <strong>Hydrogen</strong> in Plasma Heating, we explore the pivotal role of hydrogen in <em>both</em> fusion and fission reactions. Moving forward, we delve into Plasma Heating Techniques, a crucial aspect of fusion. To achieve the extreme temperatures required for fusion, diverse methods, including magnetic field-induced rotation and external heating sources, are applied to heat plasma particles. External heating methods, such as neutral beam injection and high-frequency electromagnetic waves, complement ohmic heating, transferring <em>significant</em> amounts of energy to the plasma and bringing it closer to fusion conditions. Moreover, the chapter sheds light on the<em> Importance of Hydrogen in Plasma Confinement</em>, highlighting how hydrogen’s light mass, heating efficiency, and positive charge contribute to maintaining plasma equilibrium, stability, and ionization, ultimately facilitating the intricate fusion processes.</p><p><strong>3.1 Role of Hydrogen in Fusion Reactions Vs Fission Reactions</strong></p><p>As we know, the primary fuel source in nuclear fusion is <strong>hydrogen</strong> isotopes, deuterium and tritium. Under high temperatures and pressures, deuterium and tritium are mixed to produce energy through fusion processes. These processes, which mimic the sun’s natural energy production, hold the possibility of producing clean, nearly <em>endless</em> power.</p><p>Hydrogen is important in nuclear fission because it acts as a moderator and coolant within nuclear reactors. Heavy atomic nuclei, such as uranium or plutonium isotopes, divide into smaller fragments during fission, releasing energy and more neutrons. The actual moderation happens when hydrogenous materials, such as water or graphite, successfully slow down these quick neutrons. The controlled slowing of neutrons works as a coolant, helping to <em>improve</em> the efficiency and safety of nuclear energy production.</p><p><strong>3.2 Plasma Heating Techniques</strong></p><p>One of the major requirements for fusion is to heat the plasma particles to <em>extremely</em> high temperatures. For example, several heating methods will be applied together in ITER to heat the plasma in the machine’s core to <em>150 million degrees Celsius</em>, in the near future.</p><p>The rotation of the magnetic fields required to regulate the plasma within the tokamak causes a heating effect. Through induction, magnetic fields generate a<em> high-intensity </em>electrical current, and as this current flows through the plasma, electrons and ions become electrified and collide.</p><blockquote>According to <strong>iter.org</strong>, “Collisions create “resistance” that results in heat, but paradoxically as the temperature of the plasma rises, this resistance — and therefore the heating effect — decreases”.</blockquote><p>Ohmic heating, or the transfer of heat through high-intensity current, has a limited capacity. Additional heating methods must be supplied from sources outside the tokamak device in order to reach even greater temperatures and create the conditions required for fusion reactions <em>(see Figure 6)</em>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/486/1*yycBfea4koc400iy3gTb-Q.png" /><figcaption>Figure 6</figcaption></figure><p>According to<strong> iter.org</strong>, “Two families of external heating methods — neutral beam injection and high-frequency electromagnetic waves — will complement ohmic heating to bring the ITER plasma to temperature.” The process of introducing a neutral beam involves high-energy particles into the plasma. Charged deuterium particles are accelerated to the necessary energy level outside of the tokamak. And, an ion beam neutralizer is used to remove the electrical charge from these accelerated ions, for safety reasons. The high-velocity neutral particles can then be introduced into the plasma’s core, where they quickly collide and give the plasma particles a chance to absorb their energy. This method allows for the transfer of <em>millions</em> of watts of heating energy to the plasma, bringing its temperature closer to that at which fusion can take place!</p><p><strong>Imagine</strong> it like how a microwave heats up food in a microwave oven. In a similar way, high-frequency waves are used in fusion reactors to transfer energy to the plasma. These waves speed up the movement of charged particles, making them move around more chaotically and become hotter. In ITER, the same idea is applied using three types of waves, each designed to match the frequencies of plasma ions and electrons inside the machine. This strategic approach maximizes the transfer of heat, helping to achieve the high temperatures needed for fusion reactions to take place effectively.</p><p><strong>3.3 Importance of Hydrogen in Plasma Confinement</strong></p><p>The confinement of hydrogen-based plasma, in which particles are heated to high temperatures and ionized, is critical for maintaining the high-energy conditions required for fusion processes.</p><p>The light mass of hydrogen allows for effective heating and ionization, helping to reach the high temperatures needed for fusion. The kinetic energy released by heating hydrogen ions drives the fusion process, resulting in <em>massive</em> amounts of energy produced. Also, the positive charge of hydrogen contributes to overall plasma stability by balancing the negative charge of electrons.</p><p>The importance of hydrogen goes beyond its use as a fusion fuel. It actively contributes to the preservation of plasma equilibrium and the prevention of instabilities. The complex interaction between hydrogen ions and electromagnetic fields within the confinement device, which is usually a tokamak or a stellarator, helps control the shape, temperature, and density of the plasma. Hydrogen contributes to the stability of the plasma by properly controlling these properties.</p><h3>Safety Measures with Metal Hydrides and Hydrogen</h3><p>In this chapter addressing Safety Measures with Metal Hydrides and Hydrogen, we delve into the safety attributes to fusion reactors, <strong>emphasizing</strong> their inherently secure nature. Fusion power, unlike fission, promises a clean and safe energy source, without greenhouse gas emissions, damaging pollutants, or long-lived radioactive waste. Transitioning to Metal Hydrides as Passive Safety Mechanisms, we explore how these versatile materials play a pivotal role in enhancing safety in fusion reactors. Metal hydrides exhibit remarkable tritium retention capabilities, serving as efficient tritium “sinks” that curb uncontrolled emissions and associated radiological hazards. Additionally, their ability to function as passive heat sinks aids in managing overheating scenarios, preventing pressure escalation and maintaining reactor stability. This chapter thus underscores the promising potential of metal hydrides to regulate tritium and mitigate thermal risks, contributing to the secure operation of fusion reactors.</p><p><strong>4.1 Safety Merits in Fusion Reactors</strong></p><p>Fusion power will not produce greenhouse gases, other damaging pollutants, or long-term radioactive waste. A reactor that generates power by nuclear fusion is very safe: Under any circumstances, a nuclear chain reaction <em>cannot</em> occur. Furthermore, no long-lived extremely radioactive compounds are produced. The neutrons produced by the <strong>DT</strong> fusion reaction will interact with materials near the reactor, but careful selection of these materials will ensure that fusion power produces no long-term legacy of radioactive waste. According to<strong><em> </em>Safety and Environment aspects of Tokamak- type Fusion Power Reactor- An Overview</strong><em>,</em> “The safety of Tokamak type fusion reactor operations includes intrinsic termination of the fusion reaction (hence no runaway reactions) in case of power transient, excessive fuel supply, loss of fuel supply, excessive heating, uncontrolled perturbation of the plasma, low quantity of fuel in the plasma and hence the energy of the fusion reaction, low energy density and large surfaces allowing heat transfer and use of significant heat sink masses, vacuum-tight chambers”, and the list goes on….</p><p>The fundamental differences in fusion reactor physics and technology make a fission-type nuclear meltdown or runaway reaction <em>unlikely</em>. Fusion is a naturally safe process. A fusion reactor will have only a limited quantity of fuel (less than four grams) at any given time. The reaction is dependent on a continual supply of fuel; any disruption in this process causes the reaction to stop immediately. According to <strong>iter.org</strong>, “even in the event of the total loss of the cooling function, ITER’s confinement barriers would not be affected. The temperatures of the vacuum vessel that provides the first confinement barrier would under no circumstances reach the melting temperatures of the materials.” <em>(see Figure 7)</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/670/1*u_oyZx6MmEskcrBAQsbBsQ.png" /><figcaption>Figure 7</figcaption></figure><p><strong>4.2 Metal Hydrides as Passive Safety Mechanisms</strong></p><p>Metal hydrides, which have strong tritium retention capacities, act as efficient tritium “sinks” by rapidly creating stable tritium compounds (3). This regulated binding reduces the likelihood of uncontrolled tritium emission, potentially minimizing radiological concerns. These materials absorb excess hydrogen in overheating circumstances, reducing pressure escalation and preserving reactor stability. The self-regulating nature of metal hydride hydrogen absorption and release allows for an autonomous safety reaction that does <em>not</em> rely on external systems. Furthermore, metal hydrides’ fundamental stability, longevity, and dependability support their function as long-term safety components. Overall, the ability of metal hydrides to regulate tritium while also mitigating heat-related dangers makes them promising passive safety measures in fusion reactors.</p><p>While nearly all metals can be generated to react with hydrogen under certain conditions, only a few do so <strong>“reversibly”</strong> at room temperature and at atmospheric pressures. These substances are referred to as “reversible” metal hydrides that can absorb and release hydrogen (4). They can do it again and again. Metal hydrides that are reversible can be <em>pure</em> <em>metals;</em> such as palladium, titanium, or zirconium. When storing hydrogen, reversible metal hydrides have several benefits over other methods that aren’t as safe + effective. For hydrogen, hydrides have an exceptionally high volumetric density. This means that a <em>large</em> amount of hydrogen can be packed into a small, compact space. In fact, most metal hydrides can store hydrogen many times more compactly and safer than high-pressure gas, and at times even more compactly than liquid hydrogen. This is because the hydrogen atoms in a hydride are more closely bonded to metal atoms than they can be to themselves as a gas or a liquid. Hydrides can hold hydrogen at <em>extremely</em> low pressures, providing a higher level of safety. Hydrides frequently react solely with hydrogen, making them ideal for use in a variety of separation techniques.</p><h3>Current Research and Experimental Findings</h3><p>In this chapter dedicated to Current Research and Experimental Findings, we embark on a comprehensive exploration of <strong>recent advancements</strong> in fusion research and experimental outcomes. We start off by highlighting the groundbreaking achievement in nuclear fusion, as the <strong>US Department of Energy</strong> generated net energy from a fusion reaction at the<strong> Lawrence Livermore National Laboratory</strong> in California, a historic milestone after over <em>six decades</em> of global research and substantial investments. We delve into the monumental ITER project, a global endeavour striving to harness fusion energy, supported by numerous nations, which promises to revolutionize energy production. With a focus on Metal Hydrides as Passive Safety Mechanisms, we investigate how these materials contribute to tritium regulation and heat management, enhancing safety in fusion reactors. Furthermore, we delve into pioneering studies on Tritium Management, exploring viable and ineffective methods. The intricate challenges and potential pitfalls of storing tritium in liquid form underscore the significance of metal hydrides’ superior<em> safety</em> and <em>efficacy</em>. Through this chapter, we embark on a journey through the latest research endeavours and experimental findings, shedding light on the evolving landscape of fusion energy research.</p><p><strong>5.1 Overview of Current Fusion Research</strong></p><p>On December 13, 2022, the head of the <strong>US Department of Energy</strong> (DOE) and other federal scientific officials revealed that a nuclear fusion reaction experiment they operated at the Lawrence Livermore National Laboratory in California had generated net energy (5). This meant that the reaction produced more energy than was used to start the process. It is the <em>first</em> time in human history that this milestone has been reached, following more than 60 years of global study, development, engineering, and experimenting with <em>billions</em> of dollars invested over many decades. Since the 1950s, the United States government has invested heavily on nuclear fusion research, with an average annual investment of $700 million in the last decade. The European Union contributed 45% of the total cost of the most ambitious ITER project, and it is the world’s largest nuclear fusion experiment, with 35 countries participating (6).</p><p>You’ll keep hearing the word “ITER”, so let me explain what it means. ITER means “the way” or “the path” in Latin, and it is a global nuclear fusion research and engineering <strong>megaproject</strong> aimed at producing energy via fusion. The completion of the main reactor and the first plasma is scheduled for late 2025. ITER will be the <em>largest</em> magnetic confinement plasma physics experiment in the world, as well as the world’s largest experimental Tokamak nuclear fusion reactor. ITER aims to be the largest of over <em>100</em> fusion reactors, with ten times the plasma volume of any other tokamak currently in operation.</p><p>ITER is being built in collaboration with 35 nations. It is supported and administered by the European Union, China, India, Japan, the Russian Federation, South Korea, and the United States. <em>(see Figure 8)</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/275/1*U4tR2Avm9_jlx_rYGVsOAQ.jpeg" /><figcaption>Figure 8</figcaption></figure><p>Building a nuclear fusion reactor that can resist the <em>enormous temperatures and pressures </em>produced by this process is one of the century’s most difficult engineering challenges. The fuel, which is composed of hydrogen isotopes (Deuterium and Tritium), must be heated to around 100 million degrees Celsius. A fully ionized gas-plasma is generated at that hotter-than-the-sun temperature.The plasma will then be ignited, resulting in nuclear fusion. Scientists are currently researching two approaches to nuclear fusion (7):</p><ol><li><em>Inertial confinement fusion systems: </em>A pea-sized deuterium-tritium fuel pellet is compressed to extraordinarily high densities using ion or laser beams. When a crucial point is achieved, shock wave heating ignites the pellet. This approach would allow fusion power reactors to ignite fuel pellets multiple times per second. The heat produced is then used to generate steam, which powers turbines that generate electricity.</li><li><em>Magnetic confinement systems:</em> Electromagnets are used to keep the plasma fuel contained. The tokamak device, one of the most promising possibilities, houses the plasma in a doughnut-shaped cylinder. A large electric current is created in the plasma, causing it to heat up. Auxiliary systems such as microwaves, radio waves, or accelerated particles are also used to heat the plasma. Temperatures of several hundred million degrees Celsius are reached during the process!</li></ol><p><strong>5.2 Advancements in Metal Hydride Technology: 1980’s-1990’s</strong></p><p>Hydrogen has the <em>smallest</em> relative molecular mass of any gas monomer, but it has an extremely high calorific value of <strong>142 mJ/kg</strong>, which is over three times that of gasoline of the same mass. In comparison to other energy sources, hydrogen releases only water after combustion and <em>no additional greenhouse gases</em> or poisonous chemicals, making it the cleanest energy source known to man. As a result of the foregoing benefits of hydrogen and the global energy growth trend, the use of hydrogen as a form of energy combustion or electricity production has become an energy industry “hotspot.” (8)</p><p>In 1981, a facility conducted the first small-scale metal hydride test. A few years later, in a new tritium loading procedure, the first metal hydride applications in tritium facilities were introduced. The facility requires both short-term storage and tritium compression. <strong>Lanthanum-nickel-aluminium</strong> (LANA) hydride was chosen for this application. It was discovered that by replacing a little quantity of aluminium for some of the nickel in the lanthanum-nickel hydride, the operational pressure of the hydride system could be controlled. Later, in the mid-1980s, another metal hydride material was introduced into the facilities loading system in order to build a higher pressure, or second stage, compressor. With the exception of valves, the entire system was fairly simple and reliable.</p><p>A pump was another early application of metal hydrides in tritium facilities. A metal hydride pump, or separator, was installed in 1987 to supply pumping to the freezing distillation tanks as well as to purify inert gases (helium, argon, nitrogen) from the hydrogen isotope stream. The feed system to the freezing tanks, which are used to separate the different isotopes of hydrogen, was always above normal pressure prior to the use of the pump/separator. This was done to ensure that no gases like nitrogen or oxygen leaked into the system. The pump/separator used a very low-pressure hydride (palladium) capable of successfully separating hydrogen isotopes from other gases and delivering pure hydrogen isotopes directly to the tanks.</p><p>Following the <em>successful</em> adaptation of metal hydrides in the current facilities, a new tritium plant was planned that would make more extensive use of metal hydrides. Construction on the new facility, initially known as the <strong>Replacement Tritium Facility</strong> (RTF), began in 1986 (9). The RTF was created to take advantage of cutting-edge technology in order to improve operating safety, heighten safeguards and security, and reduce tritium leaks into the environment. The tritium facilities needed a new way for safely and compactly storing tritium for lengthy periods of time, 7–10 years, in the mid-1990s. To overcome this issue, a new Hydride Storage Vessel (HSV) was created. The hydride material in the HSV was a kind of titanium. The use of this material allowed extra tritium to be stored at <em>extremely</em> low pressure, eliminating the need for an external cooling system to remove any of the tritium’s heat of decay. Metal hydrides were used in this application to give a safe and compact solution to the long-term tritium storage problem, avoiding the usage of more dangerous and costly high pressure gaseous storage <em>(see Figure 9).</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/355/1*XnyN4SnRVjtUaUr8B4K97A.jpeg" /><figcaption>Figure 9</figcaption></figure><p><strong>5.3 Experimental Studies on Tritium Management: Methods that don’t work</strong></p><p>We already know that storing tritium in metal hydrides is the most effective way for tritium management. But what about a method that <em>doesn’t</em> work?</p><p>Storing tritium in liquid form, specifically as tritiated water (T2O), provides a complicated combination of obstacles and potential drawbacks in terms of both safety and the environment. As it decays, tritium, a radioactive isotope of hydrogen, emits low-energy beta radiation. The radiation emitted when tritiated water is stored poses concerns to human health and safety. Although <em>easily</em> insulated by materials such as plastic or glass, tritium’s weak radiation demands rigorous handling techniques, specialized storage facilities, and stringent safety regulations (10). Furthermore, liquid tritium storage systems are subject to leaks or unintentional spills, which could result in the discharge of radioactive tritium water into the environment. Such releases can have far-reaching implications, such as soil and groundwater <em>contamination</em>, which can harm both terrestrial and aquatic ecosystems. Also, the disposal of tritiated water poses a considerable challenge, requiring reliable long-term containment methods to avoid unintended environmental contamination. With all of these safety and environmental concerns, storing liquid tritium is a <em>bad</em> idea. With the possibility of leaking, the tritium will also run out before any experiments can be performed.</p><h3>Synthesis Methods of Metal Hydrides: How are they formed?</h3><p>In this chapter focusing on Synthesis Methods of Metal Hydrides, we delve into <em>three</em> distinct techniques employed for the synthesis of metal hydrides. We begin with an exploration of <strong>Reactive Milling</strong>, a mechanical process that involves grinding metal powders in a hydrogen-containing atmosphere, driving hydrogen diffusion into the metal lattice and facilitating hydride formation. <strong>Mechanochemical Methods</strong> are then examined, harnessing mechanical energy for chemical reactions within solid materials, resulting in diverse metal hydride compositions and structures through increased surface area and altered kinetics. <strong>Electrochemical Synthesis</strong> is introduced as a controlled and efficient process, using an electrochemical cell to drive metal hydride formation through electrolysis of a metal electrode in the presence of hydrogen ions. Through an in-depth analysis of these techniques, we uncover the intricate processes underlying the formation of metal hydrides and their potential implications in <em>various</em> applications.</p><p><strong>6.1 Reactive Milling</strong></p><p>Reactive milling involves the mechanical mixing and grinding of metal powders in the presence of a hydrogen-containing atmosphere. The mechanical energy generated during milling facilitates the diffusion of hydrogen atoms into the metal lattice, resulting in the <em>formation</em> of metal hydrides. This technique overcomes the limitations of traditional hydrogenation methods, such as slow kinetics and incomplete hydrogen absorption. Through plastic deformation, fracturing, and welding of metal particles, reactive milling accelerates the hydrogenation process and enables the formation of metal hydrides at lower temperatures and shorter reaction times.</p><p>When compared to traditional approaches, reactive milling greatly accelerates the <em>hydrogenation process</em>, resulting in higher reaction rates and enhanced kinetics. This is because high-energy collisions between metal powders and milling small increase hydrogen diffusion and absorption. Metal hydrides synthesized via reactive milling frequently have increased hydrogen storage capacities, making them excellent candidates for hydrogen storage systems in a variety of applications. Milling parameters such as milling time, speed, and environment can be changed to adjust the properties of the metal hydride, such as hydrogenation/dehydrogenation kinetics and thermodynamic stability. Reactive milling can reduce the activation energy required for the hydrogenation process, allowing for hydride production at lower temperatures.</p><p><strong>6.2 Mechanochemical Methods</strong></p><p>Mechanochemical synthesis methods harness mechanical energy, generated through processes like milling or grinding, to initiate chemical reactions within solid materials. Mechanochemical synthesis can be carried out under many conditions, including cryogenic (very cold) temperatures, inert or reactive environments, and higher gas pressures or in a solvent (11). As a result, mechanochemistry is a very adaptable approach that can be used to create a wide spectrum of unique materials, like metal hydrides. Mechanochemistry has been often used to prepare new forms of hydrides, greatly expanding the diversity of compositions and structures available. By increasing surface area, promoting efficient mixing, and reducing activation energy, mechanochemical synthesis alters the kinetics and thermodynamics of reactions, leading to the formation of new materials and phases.</p><p><strong>6.3 Electrochemical Synthesis</strong></p><p>Electrochemical synthesis of metal hydrides involves using an electrochemical cell to drive the formation of metal hydrides through the electrolysis of a metal electrode in the presence of hydrogen ions (protons). This process allows for controlled and efficient production of metal hydrides, which have applications in hydrogen storage, energy conversion, and catalysis. In this method, a metal electrode (the anode) is immersed in an electrolyte <em>(see Figure 10) </em>solution containing hydrogen ions. When a voltage is applied to the electrode, hydrogen ions are reduced at the electrode surface, resulting in the absorption of hydrogen atoms into the metal lattice. Over time, the accumulation of hydrogen leads to the formation of metal hydrides (12). The electrochemical synthesis of metal hydrides offers <em>several</em> advantages, including <em>precise control</em> over reaction conditions, the ability to tailor the properties of the resulting hydrides, and the potential for scalability. Additionally, this method can be used to synthesize metal hydrides that might not be easily accessible through other synthesis routes.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/975/1*1JWKLoQgR93_qia6arJLnA.jpeg" /><figcaption>Figure 10</figcaption></figure><h3>The Manifestation of Cold Fusion</h3><p>In this chapter delving into The Manifestation of Cold Fusion, we explore the intriguing concept of nuclear reactions occurring at or near room temperature, known as <strong>cold fusion</strong>. Early experiments associated with cold fusion used palladium electrodes and deuterium gas, forming metal hydrides. This introduces the link between cold fusion and metal hydrides, where certain experiments hinted at nuclear reactions near metal electrodes, often composed of transition metals like palladium. Speculation proposes the fusion of deuterium nuclei within a metal lattice, potentially giving rise to metal hydrides that could play a role in facilitating or catalyzing the fusion process. While the validity of cold fusion remains debated, it serves as a thought-provoking connection that underscores the potential role of metal hydrides in fusion power applications.</p><p><strong>7.1 Introduction to Cold Fusion</strong></p><p>Cold fusion is a proposed nuclear reaction that would occur at, or near, <em>room temperature</em>. It gained widespread attention and controversy in the late 20th century due to claims of achieving nuclear fusion at much lower temperatures and pressures than those required by traditional fusion reactions (13). Some early experiments that were claimed to achieve cold fusion involved the use of palladium electrodes and deuterium (a heavy isotope of hydrogen) gas, which can potentially form <strong>metal deuterides</strong>, a type of metal hydride.</p><p>But… What does this have anything to do with metal hydrides in the fusion process? Well, this would bring us to the next section….</p><p><strong>7.2 The Connection between Cold Fusion &amp; Metal Hydrides</strong></p><p>The connection between cold fusion and metal hydrides lies in the potential formation of metal hydrides as a result of the fusion process. In <em>certain</em> cold fusion experiments, researchers reported the formation of anomalous isotopes and excess heat, which they attributed to nuclear reactions occurring in or near metal electrodes, often made of <strong>palladium</strong> or other transition metals.</p><p>In some speculative scenarios proposed by proponents of cold fusion, the process of cold fusion could involve the fusion of deuterium nuclei (heavy hydrogen) within a metal lattice, potentially leading to the formation of metal hydrides. These metal hydrides could be involved in facilitating or catalyzing the fusion process.</p><p>While cold fusion has been extensively debated in the scientific community, it shows us how metal hydrides can be used in any situation involving fusion power.</p><h3>Conclusion</h3><p>In this final chapter, we conclude by highlighting the Importance of Continued Research and Development in the realm of metal hydrides for tritium management. Given the critical role of tritium in various industries, including nuclear power and medical applications, the safe confinement and handling of this radioactive isotope are paramount. Metal hydrides have emerged as a promising solution, effectively capturing and securely storing tritium while minimizing the risk of leakage or dispersion. Ongoing research in this area holds the potential to enhance tritium adsorption and desorption capabilities through careful adjustments to chemical composition, internal structure, and surface properties of metal hydrides, thus enabling efficient and controlled tritium release when needed.</p><p><strong>8.1 Importance of Continued Research and Development</strong></p><p>As I have mentioned many times, managing tritium, a radioactive isotope of hydrogen, is a <em>critical</em> concern in a variety of industries, including nuclear power generation, fusion research, and the manufacture of isotopes for medical applications. Tritium’s radioactivity requires severe safety precautions to prevent environmental pollution and associated health risks. Metal hydrides have emerged as a promising alternative for tritium management in this context, providing a diverse and effective method of capturing, storing, and securely handling tritium.</p><p>Continued research and development in the field of metal hydrides for tritium management is <em>very</em> important. Metal hydrides offer a reliable and useful method for confining tritium, a rare material, and they do it well. The risk of tritium leakage or dispersion is substantially decreased by protecting it inside a stable metal hydride matrix. This function is essential for protecting containment systems against damage and limiting the release of tritium into the environment.The continued study of metal hydrides could result in the creation of new substances with improved tritium adsorption and desorption capabilities. Metal hydrides’ ability to efficiently absorb and release tritium can be improved by carefully adjusting their chemical make-up, internal structure, and surface characteristics. The ability to build efficient tritium storage systems and allow controlled tritium release when necessary, such as for fusion reactors or isotope production facilities, is made possible by such improvements.</p><p><strong>8.2 What Will the Future Hold for Metal Hydrides?</strong></p><p>The future of metal hydrides holds <em>great</em> promise as these versatile materials continue to find applications across various fields. Metal hydrides have already demonstrated their potential in hydrogen storage, clean energy, and advanced materials.</p><p>Metal hydrides are also finding applications in other areas of clean technology. They have the potential to revolutionize <strong>battery</strong> technology by serving as advanced materials for energy storage in rechargeable batteries. Metal hydride-based batteries could offer higher energy densities, longer lifespans, and improved safety compared to conventional lithium-ion batteries. This could lead to <em>breakthroughs</em> in electric vehicle range, grid storage, and portable electronics.</p><p>Metal hydrides are being researched for their distinct characteristics and capabilities in the field of cutting-edge materials. Novel sensors, catalysts, and even materials with changeable mechanical properties might all use them. In the electronics and aerospace industries, for instance, many metal hydrides have outstanding heat-absorbing properties that have paved the way for sophisticated thermal management systems.</p><p>As research and innovation in materials science continue to progress, we can anticipate the discovery of new metal hydride compositions, improved synthesis methods, and <em>enhanced</em> understanding of their properties. These developments will likely unlock new applications and expand the capabilities of metal hydrides across industries, especially in the fusion world!</p><p>In this research paper, we have explored the diverse applications of metal hydrides and hydrogen in nuclear fusion reactions, with a primary focus on tritium management. The efficient handling of tritium is essential for the safe and sustainable operation of fusion devices. Metal hydrides offer unique capabilities in tritium storage, breeding, and processing, while hydrogen isotopes contribute to plasma heating and safety measures. The integration of metal hydrides and hydrogen in fusion research shows great promise in advancing fusion technology toward practical energy production. As we continue to study and develop these applications, nuclear fusion holds the potential to transform the global energy landscape, offering a cleaner and more abundant energy solution for the future.</p><p><strong>tl;dr:</strong> Check out my video I made which summarizes this whole topic!</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FG3VciENUSaA%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DG3VciENUSaA&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FG3VciENUSaA%2Fhqdefault.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/427d52aa26b26a6da7d1daa36673f86e/href">https://medium.com/media/427d52aa26b26a6da7d1daa36673f86e/href</a></iframe><p>Thank you so much for reading! If you liked this, check out my other stories on Medium that I’ve published :)</p><p>Check out my <a href="https://www.linkedin.com/in/amrita-ghag/"><strong>Linkedin</strong></a></p><p>If you want to see more of what I’m doing, <a href="https://medium.com/@amritaghag08"><strong>Subscribe</strong></a> here!</p><h3>Sources</h3><ol><li><a href="https://news.newenergytimes.net/2021/10/10/the-fusion-fuel-discrepancy-the-scientific-facts/">https://news.newenergytimes.net/2021/10/10/the-fusion-fuel-discrepancy-the-scientific-facts/</a></li><li><a href="https://www.energy.gov/science/doe-explainsdeuterium-tritium-fusion-reactor-fuel">https://www.energy.gov/science/doe-explainsdeuterium-tritium-fusion-reactor-fuel</a></li><li><a href="https://digital.library.unt.edu/ark:/67531/metadc721353/m2/1/high_res_d/787459.pdf">https://digital.library.unt.edu/ark:/67531/metadc721353/m2/1/high_res_d/787459.pdf</a></li><li><a href="https://link.springer.com/article/10.1557/mrs.2013.132">https://link.springer.com/article/10.1557/mrs.2013.132</a></li><li><a href="https://www.energy.gov/articles/doe-national-laboratory-makes-history-achieving-fusion-ignition">https://www.energy.gov/articles/doe-national-laboratory-makes-history-achieving-fusion-ignition</a></li><li><a href="https://www.researchgate.net/journal/Journal-of-Physics-Conference-Series-1742-6596">https://www.researchgate.net/journal/Journal-of-Physics-Conference-Series-1742-6596</a></li><li><a href="https://paw.princeton.edu/inbox/inertial-fusion-vs-magnetic-inertia">https://paw.princeton.edu/inbox/inertial-fusion-vs-magnetic-inertia</a></li><li><a href="https://drpress.org/ojs/index.php/HSET/article/view/10114">https://drpress.org/ojs/index.php/HSET/article/view/10114</a></li><li><a href="https://inis.iaea.org/collection/NCLCollectionStore/_Public/26/047/26047517.pdf">https://inis.iaea.org/collection/NCLCollectionStore/_Public/26/047/26047517.pdf</a></li><li><a href="https://patents.google.com/patent/WO2006050077A2/en">https://patents.google.com/patent/WO2006050077A2/en</a></li><li><a href="https://www.sciencedirect.com/topics/materials-science/mechanochemical-synthesis#:~:text=Mechanochemical%20synthesis%20is%20a%20processing,coupled%20on%20a%20molecular%20scale">https://www.sciencedirect.com/topics/materials-science/mechanochemical-synthesis#:~:text=Mechanochemical%20synthesis%20is%20a%20processing,coupled%20on%20a%20molecular%20scale</a>.</li><li><a href="https://www.researchgate.net/publication/235437188_Electrochemical_Behavior_of_Metal_Hydrides">https://www.researchgate.net/publication/235437188_Electrochemical_Behavior_of_Metal_Hydrides</a></li><li><a href="https://www.sciencedirect.com/topics/engineering/cold-fusion-phenomenon">https://www.sciencedirect.com/topics/engineering/cold-fusion-phenomenon</a></li><li><a href="https://www.sciencedirect.com/science/article/abs/pii/S0022311521006164?fr=RR-2&amp;ref=pdf_download&amp;rr=7f4293170cada20e">https://www.sciencedirect.com/science/article/abs/pii/S0022311521006164?fr=RR-2&amp;ref=pdf_download&amp;rr=7f4293170cada20e</a></li><li><a href="https://www.iter.org/sci/FusionFuels">https://www.iter.org/sci/FusionFuels</a></li><li><a href="https://www.thoughtco.com/metal-hydrides-2340044">https://www.thoughtco.com/metal-hydrides-2340044</a></li><li><a href="https://www.iter.org/sci/PlasmaHeating">https://www.iter.org/sci/PlasmaHeating</a></li><li><a href="https://www.iter.org/mach/safety#:~:text=The%20fundamental%20differences%20in%20the,grams)%20at%20any%20given%20moment">https://www.iter.org/mach/safety#:~:text=The%20fundamental%20differences%20in%20the,grams)%20at%20any%20given%20moment</a>.</li><li><a href="https://h2tools.org/bestpractices/material-compatibility">https://h2tools.org/bestpractices/material-compatibility</a></li><li><a href="https://inis.iaea.org/collection/NCLCollectionStore/_Public/05/115/5115988.pdf">https://inis.iaea.org/collection/NCLCollectionStore/_Public/05/115/5115988.pdf</a></li><li><a href="https://www.iter.org/mach/TritiumBreeding">https://www.iter.org/mach/TritiumBreeding</a></li><li><a href="https://www.sciencedirect.com/science/article/abs/pii/S036031991831406X">https://www.sciencedirect.com/science/article/abs/pii/S036031991831406X</a></li><li><a href="https://www.iaea.org/newscenter/news/what-is-deuterium">https://www.iaea.org/newscenter/news/what-is-deuterium</a></li><li><a href="https://iopscience.iop.org/article/10.1088/1742-6596/2454/1/012001/pdf">https://iopscience.iop.org/article/10.1088/1742-6596/2454/1/012001/pdf</a></li></ol><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1f5965c606dd" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Kicking to Glory: The 2023 Taekwondo World Championship in Italy]]></title>
            <link>https://medium.com/@amritaghag08/kicking-to-glory-the-2023-taekwondo-world-championship-in-italy-c8ff04b8d461?source=rss-28aecc9c0df8------2</link>
            <guid isPermaLink="false">https://medium.com/p/c8ff04b8d461</guid>
            <category><![CDATA[taekwondo]]></category>
            <category><![CDATA[hard-work-pays-off]]></category>
            <dc:creator><![CDATA[Amrita Ghag]]></dc:creator>
            <pubDate>Thu, 03 Aug 2023 16:54:34 GMT</pubDate>
            <atom:updated>2023-08-03T17:06:07.150Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*-0H5MrycZTf2ZTzW7hu8ww.jpeg" /></figure><p>Recently, I’ve been pretty inactive on Linkedin… and I want to explain why.</p><p>I was flown to Italy this July to compete in the 2023 Taekwondo World Championship!</p><p>Not many of you know, so I’ll start from the top. I have been doing Taekwondo competitively since I was 4 years old, and received my black belt in 2019 at age 11. I have been training my whole life, and this December 2022, after many months of hard work and training, I was tested for my Second Degree Black Belt! For about a year and a half now, Taekwondo has also been my part-time job, as I teach classes to my students. Taekwondo has taught me so much, one thing for sure is time management. Doing school, Tks, having a job, and participating in Taekwondo (along with some other extracurriculars), has been the hardest part about this year. But I persevered, which is how I ended up competing in the world championship!</p><p>Before November of 2022, I already thought that Taekwondo was my life and I couldn’t train more, until the news of the world championship arrived, and I was one of the few chosen to participate. By January of 2022, I was training rigorously in class 3–4x a week, for about 2–3 hours each time. I was of course really tired, but I knew I couldn’t just give up, I had to push myself even more if I wanted to win…</p><p>Before I knew it, for the past six months, I have been waking up at 5 AM, lifting weights and going on runs to build muscle+endurance for the championship. After that, I would of course go to school, and do it all again the next day. It was a very hard process, and I don’t think a lot of people could do it, but I stuck with it, and it was all worth it in the end. In the evening, I would teach Taekwondo, and then train. In the evenings <strong>alone</strong>, I would spend almost 10 hours a week training/teaching (not including the morning). I worked my a** 🫣 off to get where I was, as I even participated in other local tournaments, training to become the best version of myself.</p><p>I want to be transparent, and can say that my mental health was declining. At one point, I was in a pretty bad place and wasn’t exactly sure what I wanted, I was so stressed from the pressure of work and my peers, and from Taekwondo. But I’ve always had a passion for Taekwondo, it was the place I came to when I had no idea what to do. I was so motivated on the World Championship, and I couldn’t let anything get in my way, so I continued training.</p><p>However, I want to share how important Taekwondo is to me, as a similar event to this happened in Spring of 2022. I was in a bad mental state, but Taekwondo was the one place that I could actually go to, the people there are amazing, and the tenets of Taekwondo are what really stuck out to me as well; Courtesy, Integrity, Perseverance, Self Control, and Indomitable Spirit. “A Tenet is an opinion, principle or doctrine that a person holds or maintains as true, and the 5 tenets of Taekwondo should serve as a guide for all serious students of the art”. Honestly, I never gave much thought into these tenets before. But after giving more thought into them, I realized that I actually do resonate with them, in all aspects of my life. I think these are great values that everyone should follow, and I could go on and talk about them forever, but here’s a website that explains it a little more — <a href="https://gajoktkd.com/home/5-tenets-of-taekwondo-oath/">https://gajoktkd.com/home/5-tenets-of-taekwondo-oath/</a></p><p>My instructors have definitely helped me through this process. They saw a change in me, and guided me in training, so I definitely have to thank them. School work was getting very stressful though, but I managed, and yes, I did train during exam week too 🫣. Some people definitely thought I was doing too much, but I didn’t think so. They had no idea what I was going through. And there were times that my family/friends even thought I was crazy, but this wasn’t about them- this was my dream, not theirs, and I had to keep going. This journey has definitely shaped me into who I am today, a resilient and hard working young girl, who had a vision and did everything in her power to reach it.</p><p>Flash forward to July 2023, and I was in Italy representing my country, Canada… and let me just say, we did it 🤯. I came back home with a Gold Medal 🥇in Team patterns! All my hard work definitely paid off, and all the nosebleeds I got were worth it too 😓. It’s safe to say that I put my blood, sweat, and tears into this tournament!</p><p>There were many other competing countries, including Poland, Italy, UK, The Netherlands, e.t.c. I was so proud to be representing Canada along with my other teammates, this was an experience I will truly never forget, and am so thankful for the opportunity.</p><p>Italy was an amazing destination as well! It’s so rich in culture, and the food was 🤩🤩🤩. For the 3 weeks I stayed there, I learned so much about its history and people, and can definitely say that this was one of my favourite destinations, even though it was so hot 🥵 !</p><p>I wanted to share this because this has been pretty much my life this whole year, and now I can finally relax a little bit for the summer. Many people do not know about this at all, this is my first time sharing this. I wanted everyone to know that the journey of a thousand miles does not start with one step, it starts with your mind (-creds to my TKWD Instructor). During <strong><em>my</em></strong> journey of a thousand miles, I mentioned how my mental health would sometimes suffer, but then I realized that it all starts somewhere else, in your mind. You need to have a good mindset going into the journey, something I realized later. What you put into the world, you get out of it, and I hope that my journey of a thousand miles has inspired you or taught you something about hard work and perseverance ⭐️</p><p>Check out all the pics below from the trip + tournament :)</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*5xBGLPAQuP1FvDS6oOk2bw.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*xhc8c_N0ic9r5ROtNLfSsA.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*yuqjKJOhykPfGHPzwP-Exg.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*XXYkjEqhZPO0FirMNTDa6Q.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*pvT2_qx098llQZ1dfwYIBg.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*kgvgaIEvLqZ7Rc6wevQp-w.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*c46MX49Dz8UtA6Iq7RAARw.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*CorhzTClU50969cuNzgcug.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*WJuvJqRB6Ci64r_nq-UWDw.png" /></figure><p>Thank you so much for reading! If you liked this, check out my other stories on Medium that I’ve published :)</p><p>Check out my <a href="https://www.linkedin.com/in/amrita-ghag/"><strong>Linkedin</strong></a></p><p>If you want to see more of what I’m doing, <a href="https://medium.com/@amritaghag08"><strong>Subscribe</strong></a> here!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c8ff04b8d461" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Nuclear Fusion: Hot or Cold?]]></title>
            <link>https://medium.com/@amritaghag08/nuclear-fusion-hot-or-cold-c18826c497fe?source=rss-28aecc9c0df8------2</link>
            <guid isPermaLink="false">https://medium.com/p/c18826c497fe</guid>
            <category><![CDATA[coldfusion]]></category>
            <category><![CDATA[nuclear-fusion]]></category>
            <category><![CDATA[the-knowledge-society]]></category>
            <dc:creator><![CDATA[Amrita Ghag]]></dc:creator>
            <pubDate>Sat, 25 Mar 2023 15:24:46 GMT</pubDate>
            <atom:updated>2023-08-09T20:02:54.062Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*y_ll3oS-mDYhd9F8lJW2-g.jpeg" /></figure><p>Most people, including scientists and politicians, now recognize that a major energy crisis is on the horizon. Human populations consume enormous amounts of energy, and as the population grows and living standards rise, we will require even more. Unfortunately, all of the current energy sources have significant long-term drawbacks. Oil is efficient, but it contributes to climate change and will eventually run out. Coal is abundant but polluting. Solar energy is appealing, but it is only as reliable as a sunny day — and it is currently very expensive! A clean, dependable energy source that will not run out anytime soon would solve our energy problems and transform the world.</p><p>Nuclear fusion is an alternative energy source that will transform our planet 🌍. But what if I told you that there is another version of nuclear fusion known as cold fusion, which was coined in the 1980s by Stanley Pons and Martin Fleischmann.</p><p>Not only do I want to touch on what cold fusion is and how it is different from nuclear fusion, but I also want to touch on the history of cold fusion and how it came to be, as it is a fascinating part of fusion energy 🔍.</p><h3>TL;DR</h3><ul><li>Cold fusion is the polar opposite of nuclear fusion (it works without heat🔥)</li><li>For cold fusion to work, a palladium-Deuterium interaction is required (the particles of the substances collide with each other and create energy 🔋)</li><li>Cold fusion was coined in 1989 by two chemists named Stanley Pons and Martin Fleischmann. 🧑‍🔬</li><li>The two chemists concluded that by putting a rod of palladium and a coil of platinum in heavy water, an electrical current will separate the heavy water into oxygen and deuterium. (This releases energy) 🧪</li><li>However, once the two chemists published their theory, other scientists began fact-checking this with months of research. 📝</li><li>No scientist had the same result as Pons and Fleischmann, and the two chemists were named “crazy”, along with their theory, in the science community. 🤯</li></ul><h3>A Deep Dive into Cold Fusion 🔬</h3><p>Cold Fusion is not a very complicated topic! In fact, it works in the same way as nuclear Fusion, just the opposite temperature. Instead of working at extremely high temperatures, cold fusion only requires room temperature, hence the name “cold” fusion.</p><h4>Deuterium Based Fuel</h4><p>Deuterium is the fuel used to facilitate a cold fusion reaction. Deuterium is a Hydrogen isotope, meaning that the two atoms ⚛️ (Hydrogen and Deuterium) have the same number of protons (1) but different numbers of neutrons: Hydrogen has none whereas Deuterium, has one (see Fig 1). Because isotopes are so similar, their chemical properties are nearly identical — which is why Deuterium, like Hydrogen, is able to be absorbed into palladium. Furthermore, palladium is used in the cold fusion process to separate Hydrogen from deuterium, releasing energy. In studies conducted by Stanley Pons and Martin Fleischmann, they used Deuterium as it produced 30x more energy than that of Hydrogen fusion.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/608/1*otK7ZjpfvooOhPNvDScyFg.png" /><figcaption><strong>Fig 1: </strong>A Hydrogen atom contains only one proton in its nucleus, whereas deuterium, a rarer isotope of Hydrogen, contains both a proton and a neutron.</figcaption></figure><h4>Electrolysis</h4><p>Electrolysis is a process used to create Deuterium gas by splitting heavy water molecules (Deuterium gas is water with Deuterium particles, but for Electrolysis to occur, the Deuterium particles are replaced by Hydrogen particles). When an electric current is passed through heavy water in a fusion cell, chemical reactions occur at the metal rods. These metal rods are used to crush the Hydrogen gas, releasing a substantial amount of pressure. To increase the water’s ability to conduct electricity and speed up the process, Lithium Deuteroxide (LiOD) was added. Heavy water (D2O) reacts with the electrons in the current at the palladium rod to produce Deuterium gas and Deuterium Oxide ions.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/712/1*1Mr43mrXfzztY6sin1IhaQ.png" /><figcaption><strong>Fig 2: </strong>A visual of Stanley Pons’ and Martin Fleischmanns idea for a cold fusion chamber</figcaption></figure><h4>Palladium-deuterium interaction</h4><p>Palladium is a crystal and has a crystalline structure (see Fig 3). Because of this, its atoms are organized in a repeating pattern. When Deuterium is absorbed into palladium, it is drawn into the cell centres and can then diffuse through the crystal, hopping from cell centre to cell centre. These Deuterium particles will then collide with molecules in their environment, transferring some of their energy as heat to the palladium and water. This is how cold fusion is created, without the use of extreme heat in contrast to nuclear fusion.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/844/1*zj4_TyssEAWQqlsbYJR4rw.png" /><figcaption><strong>Fig 3: </strong>A visual representation of a Palladium-deuterium interaction (the particles)</figcaption></figure><p>Now that we understand the science behind cold fusion, how does it differ from nuclear fusion️? ☀</p><h3>Nuclear Fusion vs Cold Fusion 🔥❄️</h3><p>It is important to understand nuclear fusion since it was the fundamental idea from which cold fusion originated. Fusion is a nuclear reaction where two smaller nuclei join together to form a larger nucleus. When that large nucleus becomes unstable, it quickly disintegrates and releases energy. The main problem is that, because the initial nuclei are all positively charged, they repel each other strongly as they approach (Positive charges will only attract negative charges). As a result, only nuclei with high kinetic energy ⚡ approach close enough to fuse. On Earth, we can use particle accelerators or extreme temperatures (50 million+ degrees Celsius) in order to produce high-speed nuclei. However, it is important to note that we have not successfully achieved these conditions on Earth or harnessed nuclear fusion.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/700/1*aisgsEbdpSPHiXHtRukLIw.jpeg" /><figcaption><strong>Fig 4: </strong>How nuclear fusion is created, and how it is converted into energy</figcaption></figure><p>Cold fusion, as we know, is designed to occur in the absence of these extreme heat conditions. When Deuterium is dissolved in a solid using palladium metal, cold fusion would release measurable energy from fusion reactions at or near room temperature. The idea, which dates back to the 1920s, is that Hydrogen and its isotopes can dissolve in such high concentrations in certain solids that the Hydrogen nuclei approach each other closer than even in solid Hydrogen. Furthermore, the negative electrical charges from electrons slightly cancel the repulsion between the nuclei (see Fig 5).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/696/1*k-PoNfpc9y8YaGcUujAxdg.png" /><figcaption><strong>Fig 5: </strong>The negative charges from electrons</figcaption></figure><p>These distinctions between the two fusions are obvious. Nuclear fusion occurs in the presence of plasma, which heats the atoms to the point of fusion. Cold fusion, on the other hand, uses the electrical charges of atoms to fuse them, without the use of heat. Because electricity is conserved, the energy that was previously used to push electrons along is turned to heat energy.</p><p>Now that we’ve gone over the science behind cold fusion and the distinction between nuclear fusion and cold fusion, let’s discuss the history if cold fusion! Though cold fusion seems revolutionary, this wasn’t the case. Let’s dig a little deeper into the history of cold fusion, and why it didn’t work.</p><h3>The History of Cold Fusion 📜</h3><p>Cold fusion became popular in the late 1980s thanks to two chemists named Stanley Pons and Martin Fleischmann 👨‍🔬. The two men announced to the world that they had created low-temperature nuclear fusion after five years of research and hundreds of thousands of dollars. Not only were they able to demonstrate the existence of cold fusion, but their findings suggested that they would be able to power the world once their experiment was scaled up.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/468/1*T7HJfvZpbJEMc58sZKgjGw.png" /><figcaption><strong>Fig 6: </strong>Stanley Pons (left) and Martin Fleischmann (right)</figcaption></figure><p>The experiment itself was straightforward: put a rod of palladium and a coil of platinum in heavy water (water where Deuterium replaces the hydrogen) <strong>Fig 2</strong>. A current of electricity will separate the heavy water into oxygen and deuterium. This Deuterium will absorb into palladium, where the atoms will become so tightly packed together that they will eventually fuse. The two witnessed firsthand how palladium can absorb unusually large amounts of Hydrogen — about 900 times its own volume — during those experiments. Think about a vacuum; that’s like sucking up hundreds of pounds of dirt with the vacuum!</p><p>This incredible absorption power is caused by a chemical reaction on the palladium’s surface that draws Hydrogen inside the metal. Because Hydrogen and Deuterium are so similar, the same reaction occurs with Deuterium — it can also be sucked up in surprisingly large amounts by palladium. Because Deuterium absorbed by palladium shrinks dramatically in volume, Pons and Fleischmann thought that the deuterium atoms must be squished together inside the palladium. He began to wonder if a similar process could be used to bring Deuterium atoms together in order for them to fuse and release energy.</p><h4>The Ingenious Idea</h4><p>They tucked their fusion ideas away until the fall of 1983, when the two began discussing the possibility of using chemical processes to trigger a nuclear process. They decided to conduct a full-fledged experiment to put the theory to the test. Working in Pons’ lab, the two created what they called a fusion cell. This cell was made up of two metal pieces, one palladium and one platinum, submerged in a container of heavy water. They knew that zapping the cell with electricity would set off a chemical reaction known as electrolysis, in which the heavy water molecules would split, releasing Deuterium gas and oxygen. A chemical reaction could then absorb the Deuterium into the palladium.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/412/1*63Bp9aZ5NBYJIKkLe9qkPQ.png" /><figcaption><strong>Fig 7: </strong>Final design for a cold fusion chamber</figcaption></figure><p>Pons and Fleischmann hypothesized that once inside the palladium, the Deuterium atoms would be forced so close together that they would fuse, releasing massive amounts of energy as heat. Pons and Fleischmann continuously monitored the temperature of the cell during its operation. After some data analysis, they discovered that the cell was producing about 100 times more heat than chemistry alone could account for! They interpreted the excess heat as proof of fusion. Pons and Fleischmann were eager to test their idea further, excited by the possibility that they had discovered a low-cost way to harness fusion for energy production!</p><p>However, many people are aware that cold fusion contradicts well-known physics. There was no way the palladium could bring atoms close enough to fuse and release energy. Fearing competition, the two scientists rushed to publish a scientific paper and went so far as to hold a press conference to announce their findings. Their paper went through a limited peer-review process in which many experimental errors were overlooked. Pons and Fleischmann also refused to share specifics about their setup with the scientific community. This meant that their experiments could not be replicated and tested accurately in other laboratories around the world.</p><h4>Idea Into Action</h4><p>Months after the announcement, other institutions began conducting their own versions of the Pons-Fleischmann experiment, with the majority of them failing ❌ to produce any fusion energy. However, some scientists supported cold fusion and even claimed to have made it work in their own laboratories. Tritium, a known byproduct of fusion reactions, was produced in their experiments. But, these researchers were accused of dishonesty.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/638/1*1MdoevrUkGocTbrbiP43zQ.png" /><figcaption><strong>Fig 8: </strong>An unknown scientist working in a lab to re-create the Pons-Fleischmann experiment</figcaption></figure><p>In addition to attempting to replicate Pons and Fleischmann experiment, which had been derailed due to a lack of information, scientists attempted to validate the work in other ways, scrutinizing the cold fusion paper for potential sources of error. Many of the issues they discovered would have been caught in a thorough peer review, and some errors were surprisingly simple. Scientists noticed, for example, that Pons and Fleischmann had not stirred the heavy water inside their fusion cells; not stirring the water in a fusion cell causes uneven heat distribution and inaccurate temperature measurements.</p><p>Others continued to try to replicate the findings by experimenting with various experimental combinations in the hopes of finding the one used by Pons and Fleischmann.</p><p>While most research groups saw no evidence of fusion, a few claimed to have seen excess heat and/or neutrons coming from their fusion cells. However, these groups disagreed on the conditions required for fusion. Some discovered that the nuclear reactions took months to begin, while others discovered results in just a few hours. And, in many cases, these groups were unable to replicate their own findings.</p><p>How could very similar experiments produce such different results? Some of the outcomes were unintentional. Several of Pons and Fleischmann confirmations had to be retracted due to errors, such as forgetting to connect a key wire in the experimental setup. Other inconsistencies were caused by differences in data analysis as well. 🔑</p><p>Over the next few months, scientists conducted the most sophisticated and sensitive experiments on cold fusion, but they were unable to find any evidence to support it. The cold fusion case was not looking good. However, there was still a chance that the discovery couldn’t be replicated — not because cold fusion wasn’t happening, but because other scientists weren’t exactly replicating the original experiment’s conditions.</p><p>Maybe Pons and Fleischmann were doing something special in their experiment that they weren’t revealing or were unaware of, and this special something was what led to cold fusion.</p><p>Independent experts searching for fusion products from Pons and Fleischmann’s fusion cells would be the best way to test this. Many scientists offered to collaborate, but their proposals were turned down. Pons and Fleischmann actively obstructed tests that could have revealed whether or not their hypothesis was correct.</p><h4>A Bad Reputation</h4><p>The science community was now confused. They were unable to replicate Pons and Fleischmann’s cold fusion process, as the two scientists refused requests to collaborate with others.</p><p>In a last-ditch effort to validate the cold fusion results, University of Utah professor Michael Salamon was granted access to Pons’ lab to conduct experiments looking for neutrons released by Pons and Fleischmann’s own fusion cells. If any experiment could be guaranteed to replicate the original’s conditions, this would be it…. During his five-week long test, Salamon was unable to detect any neutrons. 😲</p><blockquote>Pons and Fleischmann lost their scientific reputation and respect after failing to replicate their original results and provide any additional proof of cold fusion. They eventually left the USA for a fresh start in France.</blockquote><p>Some scientists are still looking at the feasibility of cold fusion. Even if an idea fails, science doesn’t give up on it if it has potential. After all, all scientific knowledge is unreliable. Hence, despite the overwhelming evidence that what Pons and Fleischmann saw was not cold fusion, some researchers are still looking into the possibility of cold fusion ❄️ ……</p><p>Check out this video I made conducting an experiment on Cold Fusion!</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FXZPYec-wn6M%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DXZPYec-wn6M&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FXZPYec-wn6M%2Fhqdefault.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/3f4e073a2ddffecf55cee88999552421/href">https://medium.com/media/3f4e073a2ddffecf55cee88999552421/href</a></iframe><p>Thank you so much for reading! If you liked this, check out my other stories on Medium that I’ve published :)</p><p>Check out my <a href="https://www.linkedin.com/in/amrita-ghag/"><strong>Linkedin</strong></a></p><p>If you want to see more of what I’m doing, <a href="https://medium.com/@amritaghag08"><strong>Subscribe</strong></a> here!</p><h3>Sources</h3><p><em>Cold fusion: A case study for scientific behaviour</em>. (n.d.). Understanding Science. Retrieved March 21, 2023, from<a href="https://undsci.berkeley.edu/wp-content/uploads/2022/08/cold_fusion.pdf"> https://undsci.berkeley.edu/wp-content/uploads/2022/08/cold_fusion.pdf</a></p><p>Ritter, S. K. (2016, November 7). <em>Cold-fusion-died-25-years</em>. C&amp;EN. Retrieved March 21, 2023, from <a href="https://cen.acs.org/articles/94/i44/Cold-fusion-died-25-years.html">https://cen.acs.org/articles/94/i44/Cold-fusion-died-25-years.html</a></p><p><em>The science behind cold fusion</em>. (n.d.). Understanding Science. Retrieved March 21, 2023, from <a href="https://undsci.berkeley.edu/the-science-behind-cold-fusion/">https://undsci.berkeley.edu/the-science-behind-cold-fusion/</a></p><p><em>What is the current scientific thinking on cold fusion? Is there any possible validity to this phenomenon?</em> (1999, October 21). Scientific American. Retrieved March 21, 2023, from <a href="https://www.scientificamerican.com/article/what-is-the-current-scien/">https://www.scientificamerican.com/article/what-is-the-current-scien/</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c18826c497fe" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Nuclear Fusion at the Forefront of Deep Space Exploration]]></title>
            <link>https://medium.com/insights-of-nature/nuclear-fusion-at-the-forefront-of-deep-space-exploration-cfeb93325834?source=rss-28aecc9c0df8------2</link>
            <guid isPermaLink="false">https://medium.com/p/cfeb93325834</guid>
            <category><![CDATA[fusion-rockets]]></category>
            <category><![CDATA[space-exploration]]></category>
            <category><![CDATA[the-knowledge-society]]></category>
            <category><![CDATA[nuclear-fusion]]></category>
            <category><![CDATA[astronomy]]></category>
            <dc:creator><![CDATA[Amrita Ghag]]></dc:creator>
            <pubDate>Mon, 02 Jan 2023 19:42:37 GMT</pubDate>
            <atom:updated>2024-03-27T23:30:13.526Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*XCBmueCQ743EeDwoVyQNKQ.jpeg" /></figure><h4>By Amrita Ghag</h4><p>Humanity has made huge advances in space exploration; Yuri Gagarin was the first man in space in 1961, inside the Vostok 1 Capsule. Or, we have Apollo 11 (1969), when Neil Armstrong was the first man on the moon 🌕. It’s incredible to consider that these developments by humanity were made not so long ago! We also can’t forget how many things went wrong in the process. Nothing we were doing was effective. Perhaps we weren’t using the right technology, maybe we needed something more efficient.</p><p>Well, I would like to propose the idea of Nuclear Fusion based rockets 🚀. Rockets that would help in deep space exploration, like going to Mars or beyond!</p><p>A new space race is emerging, with higher stakes and more ambitious goals than simply returning to Earth’s orbit or the Moon. Fusion rockets will help us reach these goals, here’s why.</p><h3><strong>Why Nuclear Fusion? </strong>☀️</h3><p>Nuclear Fusion has always had such immense potential, it’s an abundant energy source from the sun with no consequences at all! However, we humans have never fully learned the ways of Nuclear Fusion, as we haven’t efficiently harnessed it yet.</p><p>The reason for this is because Fusion occurs when the atoms are heated to an extremely high temperature 🔥, and then they form Plasma. — Plasma is an extremely hot substance that makes up the Sun and is also the main ingredient for Nuclear Fusion to occur. Under all this heat, the nuclei create energy and fuse to create Nuclear Fusion. But, nuclei also have to be under small spaces to compact, this is one of the many reasons we haven’t harnessed Nuclear Fusion on Earth yet.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*3h1XpnKUB9ylIdYo4uH1ug.jpeg" /><figcaption>Enjoy this diagram I created!</figcaption></figure><p>Before we start sending rockets into space, we need to efficiently harness Nuclear Fusion on Earth. But, I’m not going to get into that side of Fusion, but instead going to be talking about the future of Fusion… why we need to start implementing Fusion based rockets 🚀!</p><h3><strong>A Deep Dive into our Current Rocket Ships </strong>🛰️</h3><p>Space rocket ships are not simple, as they’re our only way of getting out of Earth’s atmosphere. Rocket ships don’t work the same as other flying vehicles, for example, airplanes ✈️, as rockets don’t use wings and propellers, knowing that they would not work in space. For rockets to fly, they generate thrust, using the principle of action and reaction. Exhaust flames are then released by explosive chemicals out of the back of the rocket at a very high speed.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/709/1*pCvwlNu96yXy1dXBJQYW-Q.gif" /></figure><p>Rockets must also be very delicate, balanced, and controlled to be able to withstand the powerful forces of going from earth’s atmosphere 🌎 to space. To do so, two things happen; first, as the rocket generates thrust using the explosive chemicals, then, fuel and oxidant (chemical that lets fuel to combust) undergo violent chemical reactions. The remaining expanded gases from the explosion are pushed out of the back of the rocket through a nozzle — a specially designed exhaust that directs the hot, high-pressure gas produced during combustion into a stream that travels at hypersonic speeds.</p><p>Do you remember Isaac Newton’s third law of motion? Well, it states “that every action has an equal and opposite reaction”. In this case, when the “action” (the force that drives the exhaust out of the rocket to the nozzle), must be balanced by the “equal and opposite reaction” (the thrust pushing the rocket forward — kind of like the rocket itself).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/709/1*cxWOt0mHZTfjy-tfHNsgdQ.gif" /></figure><h3><strong>How a Rocket’s Motors Work </strong>⚙️</h3><p>Modern-day rockets have advanced so much compared to the rockets from 100 years ago today, the way they function is much more complex and efficient 🔥. However, there are two main types of rockets, <strong>simple solid rockets</strong>, and <strong>liquid-fueled rockets.</strong></p><p><strong>Simple solid rockets </strong>are relatively less complex, as their main purpose is as a booster to provide extra thrust at launch. These rockets follow the same principles of using fuel and oxidant; the pace at which fuel is burned, and ultimately the amount of thrust, can be regulated. Once started, a solid rocket will burn continuously until its fuel is expanded.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/259/1*4esM1BqbyYTbGACIz5ToDQ.png" /></figure><p>Now, the more widespread <strong>liquid-fueled rocket</strong>s are far more complicated. Usually, they use a pair of propellant tanks; one for the fuel and one for the oxidant. These tanks are connected to a combustion chamber using a complex maze of pipes. To deliver the liquid propellant into the chamber, high-speed “turbopumps” use their motor systems to deliver the propellant into the injection system.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/259/1*BhcoY4pQAYCoBylVjL_0iQ.jpeg" /></figure><p>Inside the combustion chamber, an ignition mechanism is used to begin combustion — usually, a jet of high-temperature gas or an electric spark 💥 does the trick. However, something that is common is too much fuel/oxidant mixture in the rocket. If there are too many of these chemicals in the combustion chamber, a delayed ignition can generate enough pressure to blow the whole rocket apart!</p><h3><strong>Interplanetary Travel </strong>🧭</h3><p>Now, the last thing to mention for our modern-day rocket ships’ deep dive is interplanetary travel; how do rockets go from orbit to deep space 🌌?</p><p>Any space mission begins with a launch from the surface of the Earth into a low orbit that is around 124 miles above the majority of the atmosphere. Since friction from Earth’s upper atmosphere is very low and gravity is almost as strong here as it is on the surface, a rocket’s upper stage can maintain a stable and circular path where the forces of gravity and the rocket desire to fly off in a straight line balance each other out.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/706/1*V56ynoGx2B86BhVnnG6gag.gif" /></figure><p>Many satellites and spacecraft are limited to this low Earth orbit, but those that are meant to leave Earth and explore the rest of the solar system require an additional boost in speed to reach light speed 💨, the speed at which they will never again be attracted to our planet by gravity. Injecting a spacecraft into a <strong>Hohmann transfer orbit</strong>, an elliptical orbit around the sun, is necessary to get it from one planet to another with the bare minimum of requirements. When a rocket ship reaches its target object, it may enter its final orbit using only gravity, or it may need to employ a rocket push in the opposite direction to produce a stable orbit. Usually, this is done by simply rotating the spacecraft around in space and firing the motor.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/249/1*szByOiAEYVrMDdWxADgmFA.png" /><figcaption>The Hohmann Transfer Orbit</figcaption></figure><p>So.. what was the point of this deep dive? Well, I’m not sure if it’s obvious to you, but these current rockets have significant drawbacks, such as the potential for explosion and the inability to effectively conduct risk-free deep space exploration. I went in-depth on our current rocket ships, but what about <strong>fusion-based rockets?</strong></p><p>The US magnetic fusion effort has struggled with a lack of urgency and sporadic financing for this challenging objective for many years, harnessing Nuclear Fusion ☀️. Fusion is hardly ever mentioned when it comes to future energy production in American circles. But, what if there was another, more critical, special, and even more significant application for fusion, what if we applied our knowledge of Nuclear Fusion into space?</p><h3><strong>A Deep Dive into Fusion Rockets </strong>☀️🚀</h3><p>Although there have not yet been any successful models of a Fusion Rocket, many companies are still researching into how it will work, how long it may take, and what technologies they would use to pursue this rocket. One famous company that is working to make this possible is NASA.</p><p>NASA has been working on two types of Nuclear Fusion systems for space travel;<strong> Nuclear Thermal Propulsion</strong> &amp; <strong>Nuclear Electric Propulsion</strong>.</p><h3><strong>Nuclear Thermal Propulsion Systems </strong>♨️</h3><p>This system has a high power and average efficiency. It heats a gas, such as hydrogen, in a miniature nuclear fusion reactor, and then accelerates that heated gas via a rocket nozzle to produce thrust. According to NASA engineers, a journey to Mars propelled by nuclear thermal propulsion would be 20%–25% quicker than one using chemical propulsion.</p><p>Nuclear thermal propulsion systems can produce 100,000 Newtons of thrust and are more than twice as efficient than chemical propulsion systems, generating twice as much thrust with the same mass of propellant — that’s enough force to get a car 🏎️ from 0–60 mph in about a quarter of a second!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/315/1*b-CaULEAH6m8f9QIvS_CQA.png" /></figure><h3><strong>Nuclear Electric Propulsion Systems </strong>💡</h3><p>Nuclear electric propulsion is the name of the second nuclear-powered rocket technology. The idea is to employ a high-power fusion reactor to generate electricity, which would then power an electrical propulsion system like a Hall thruster. However, nuclear electric systems have not yet been developed. This would be about three times more effective than a nuclear thermal propulsion system in terms of performance. Given the nuclear reactor’s high power output, a large number of separate electric thrusters may be run at once to provide a significant amount of thrust.</p><p>For exceptionally long-range missions, nuclear electric systems would be the greatest option because they don’t require direct heat from the sun ☀️(as the plasma in the reactor is heating the atoms), are incredibly efficient, and can produce a reasonable amount of thrust. Nuclear electric rockets are very promising, but a lot of technical issues need to be resolved before they can be used.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/285/1*5sSVeJaRZ-ahQTS5k4dA6Q.jpeg" /></figure><h3><strong>The FDR </strong>🚀</h3><p>Using the two technologies listed above, NASA has been trying to create a Fusion rocket! NASA’s <strong>Fusion Driven Rocket</strong> (FDR) provides a ground-breaking method of Fusion propulsion where the power source releases its energy directly into the propellant, not requiring conversion to electricity. It uses a solid lithium propellant that doesn’t need a lot of tankage. With little direct contact with the spacecraft, the propellant is heated quickly and propelled to a high exhaust velocity, preventing damage to the rocket and lowering the thermal heat load and radiator bulk.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/217/1*BWlPRFYywDJKWpY9aG0qiQ.jpeg" /></figure><p>The FDR is also thought to be feasible with little extrapolation from current technology, at high specific power, at a manageable mass scale, and consequently at a low cost. If achieved, it would not only make it possible for humans to travel between planets in space but also make it a common occurrence.</p><p>The research on the magnetically driven implosion of metal foils into a magnetized plasma target to create fusion conditions is the key to realizing all of this.</p><p>This work’s logical progression results in a technique that uses metal 🧲shells both as the propellant and to create the Fusion conditions. To compress the target plasma to Fusion conditions, a number of low-mass, magnetically driven metal liners are inductively forced to converge radially and axially and form a thick blanket around it. The metal blanket that surrounds the spaceship absorbs almost all of the radiant, neutron, and particle energy from the plasma, shielding it from the Fusion process and avoiding the requirement for a substantial radiator mass. High thrust is produced at the ideal access point by the expansion of this hot, ionized metal propellant through a magnetically insulated nozzle. Thus, the energy produced by the Fusion process ☀️ is used very effectively!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/267/1*P2DDQq51KMBg7MT0JKbe-Q.jpeg" /></figure><p>So… I’ve just given so much information about how Fusion rockets are better than normal rockets; they’re more safe, reliable, cost-friendly, and efficient. This is something I believe in, and I know it will be implemented in the future! Now, although we still haven’t efficiently harnessed Nuclear Fusion on Earth, I still believe that Fusion power will be implemented in space, let’s keep an eye on it because we’ll have to wait and see! 🤩</p><p>Check out my <a href="https://www.linkedin.com/in/amrita-ghag/"><strong>Linkedin</strong></a></p><p>If you want to see more of what I’m doing, <a href="https://medium.com/@amritaghag08"><strong>Subscribe</strong></a> for more!</p><h3><strong>TL;DR</strong></h3><ul><li>Nuclear Fusion is a complex, abundant alternative energy with no consequences</li><li>It is energy from the sun, made with nuclei and plasma!</li><li>We haven’t harnessed this energy efficiently on Earth, but it has tremendous potential in space tech</li><li>Our modern rockets are not efficient, there is always a risk of failure</li><li>With fusion-based rockets, we would be able to go to Mars and back for fun!</li><li>This is because they are ultimately more safe, reliable, cost-friendly, and efficient</li><li>If we accomplished fusion based rockets, we would be able to deep research space!</li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=cfeb93325834" width="1" height="1" alt=""><hr><p><a href="https://medium.com/insights-of-nature/nuclear-fusion-at-the-forefront-of-deep-space-exploration-cfeb93325834">Nuclear Fusion at the Forefront of Deep Space Exploration</a> was originally published in <a href="https://medium.com/insights-of-nature">Insights of Nature</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>