Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Fusion energy : burning questions Jakobs, M.A. Published: 14/11/2016 Document Version Publisher’s PDF, also known as Version of Record (includes final page, issue and volume numbers) Please check the document version of this publication: • A submitted manuscript is the author’s version of the article upon submission and before peer-review. There can be important differences between the submitted version and the official published version of record. People interested in the research are advised to contact the author for the final version of the publication, or visit the DOI to the publisher’s website. • The final author version and the galley proof are versions of the publication after peer review. • The final published version features the final layout of the paper including the volume, issue and page numbers. Link to publication Citation for published version (APA): Jakobs, M. A. (2016). Fusion energy : burning questions Eindhoven: Technische Universiteit Eindhoven General rights Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights. • Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain • You may freely distribute the URL identifying the publication in the public portal ? Take down policy If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim. Download date: 19. Jun. 2017 Fusion Energy - Burning Questions PROEFSCHRIFT ter verkrijging van de graad van doctor aan de Technische Universiteit Eindhoven, op gezag van de rector magnificus, prof.dr.ir. F.P.T. Baaijens, voor een commissie aangewezen door het College voor Promoties in het openbaar te verdedigen op maandag 14 november 2016 om 16.00 uur door Merlinus Ambrosius Jakobs geboren te Eindhoven Dit proefschrift is goedgekeurd door de promotoren en de samenstelling van de promotiecommissie is als volgt: voorzitter: 1e promotor: copromotoren: leden: prof.dr. K.A.H. van Leeuwen prof.dr. N.J. Lopes Cardozo dr. R.J.E. Jaspers dr.ir. L.P.J. Kamp prof.dr.ir. D.M.J. Smeulders Prof.Dr. D. Reiter (Heinrich Heine Universität Düsseldorf) dr. D.J. Ward (Culham Centre for Fusion Energy) prof.dr.ir. B. Koren Het onderzoek dat in dit proefschrift wordt beschreven is uitgevoerd in overeenstemming met de TU/e Gedragscode Wetenschapsbeoefening. The witches’ ride to the devil’s castle, where we meet only ourselves, ourselves, ourselves. . . Dag Hammarskjöld Waymarks A catalogue record is available from the Eindhoven University of Technology Library. Jakobs, Merlijn Fusion Energy - Burning Questions Eindhoven: Technische Universiteit Eindhoven, 2016. ISBN: 978-90-386-41-63-8 NUR 926 Cover: Original image ‘The Wizard’ CC BY 2.0 by Sean McGrath Solar image by ESA/NASA/SOHO Photo montage by SuperNova Studios Typeset by the author using LATEX 2ε . c 2016 Merlijn Jakobs i Summary Nuclear fusion is the process in which two atomic nuclei are joined together to form a heavier one, thereby releasing a large amount of energy. It is the energy source of all stars in our universe. Its application as an energy source on earth would have several appealing properties like a virtually inexhaustible fuel, inherent safety and the absence of long-lived radioactive waste. It is therefore an attractive candidate to contribute to the world energy supply. Currently the first power producing fusion reactor ITER is under construction in southern France, and, if successful, a first generation of electricity producing demonstration reactors is foreseen to follow in the 2040-2050 time frame. Present day fusion reactors require external heating power to achieve the high temperatures needed for fusion, but energy-producing reactors will have to rely (to a large extent) on self-heating by the alpha particles that are produced in the deuterium-tritium fusion reaction. A fusion reactor will therefore ’burn’, much like an ordinary wood-burning stove. You fill it with fuel, kindle it (i.e. inject heat until the ignition temperature is reached) and once ignited the system will find an equilibrium ’burn’ temperature. The only thing the operator has to do is to regularly add new fuel and remove the ash (i.e. the helium that is produced in the fusion reaction). This thesis deals with the properties of these burn equilibria, what determines their fusion power and position in the operational space of the reactor, and how the system reacts to a perturbation of its equilibrium state. There are several parameters that govern the burn equilibria in a burning plasma. One of the most important is the energy confinement time τE , a measure for how fast energy is lost from the plasma. Because it is difficult to calculate the energy transport in a fusion plasma from first principles, often scaling laws are used which express the energy confinement time in engineering or physics parameters. We have found an expression relating the electron density ne at the operating points to the temperature, by eliminating τE from the equations using such a scaling law. We showed that the so-called burn contours, i.e. the contours in the operational space of the reactor spanned by the plasma density and temperature, are exactly ii Summary the same for all reactors, apart from a normalisation factor of the density which contains the design values of the reactor, such as its dimensions and magnetic field strength. This finding implies that the results of the analyses of the burn equilibria are generic, i.e. are of application to any reactor design that follows the same τE scaling. One of the salient results of the analysis is that, for a given reactor, the power output will generally not increase if the energy confinement is improved. Good confinement - one of the central goals of fusion research - is still a highly desirable property as it allows smaller reactors to ignite and burn, but in the existing conceptual power plant designs an improvement of confinement does not bring any benefit. This also means that the fusion output power of such a reactor will respond only weakly to (small) changes in τE , disqualifying it as a useful control parameter. However, reducing τE too much, say by 30% or so, will quench the reactor. This result is directly connected to a second parameter that has a big influence on the operating points of a reactor, the ratio between energy and particle confinement time ρ = τp /τE . Generally, energy and particle transport are linked, which would result in ρ ≈ 1. However, particles that hit the wall can return to the plasma, but they lose their energy in the process. This is called (edge) recycling and is the main reason that ρ is expected to be between 5 and 10 in a reactor. The value of ρ determines the accumulation of helium ash in the plasma, and the fusion power output reacts strongly to variations in ρ. This makes it a candidate to control the fusion power of a burning reactor, if a means can be found to effectively change the value of ρ, for instance by changing the rate at which particles are pumped from the reactor exhaust. It should be kept in mind, however, that the efficiency of the reactor is highest at low values of ρ (say < 5), while the burn can become unstable when ρ nears 10 (as we shall see) and no burn is possible for ρ > 15. This would suggest aiming for a high value of ρ, but it is not that simple unfortunately. A fusion reactor needs to breed the tritium it consumes from lithium, as tritium does not occur naturally on earth. The tritium breeding ratio, the amount of tritium bred divided by the consumed amount, just exceeds one, requiring tritium losses to be minimised. One of the ways of doing this is reducing the number of cycles tritium needs to make through the reactor before it fuses. The tritium burn-up fraction, the amount of tritium that fuses before being exhausted from the plasma, therefore needs to be as high as possible, which requires a long particle confinement time, or high value of ρ. The first demonstration reactors will most likely still require some amount of external power (to drive the plasma current, with heating only a side effect), and this changes the shape and position of the burn contours in the reactor operating space. Most importantly, it increases the fusion power output, but in most cases not enough to compensate for the conversion losses associated with the generation Summary iii of the heating power. The plasma in a reactor will always contain some impurities and the inclusion of those in the analysis shows that especially impurities with a low atomic number Z have a big impact on the fusion power, because they are very effective at diluting the fuel. The amount of impurities can increase through a change in the source, or by being better confined because of an increase in ρ. The latter would have a double effect: both the helium and the impurity concentration will increase, which has an even stronger impact on the fusion power. The upside to this effect is that the fusion power becomes more sensitive to the external heating power for higher values of ρ and impurity content. We have analysed the stability of the operating points and, although the system possesses many interesting properties (including saddle points, several different bifurcation points, limit cycles, and damped or growing oscillations), the upshot is that (virtually) all reactor relevant operating points are stable except for ρ > 10. However, the addition of external heating also stabilises these equilibria, so stability considerations will most likely only have implications for the case of a reactor design with little or no external heating. Finally, we show that the current form of the τE scaling law can result in bizarre predictions when applied to burning plasmas. First of all, ignition should be possible at arbitrarily low densities, arbitrarily low power and arbitrarily small reactor size. Secondly, a small change in the density or power dependence of the scaling law, which has a negligible effect on the predicted value of τE , results in wildly different operating points and fusion power. These unphysical results are the consequence of the coupling between the density and the heating power in a burning plasma, which leads to a singularity in the burn condition for a particular combination of the n- and P -dependence in the τE scaling. This might be a point of academic interest only, were it not for the strange coincidence that the family of 5 scaling laws that are used in the ITER physics basis, all happen to exhibit precisely this pathology. Put very succinctly, these scalings laws approximately have τE ∝ n0.4 and P −0.7 , and this means that if for P the fusion power Pfus ∝ n2 is substituted, the well-known triple product nτE T becomes independent of density and confinement time, i.e. it reduces to T . We have no explanation for the fact that the ITER scaling laws all happen to have this peculiar behaviour, the data base on which they are based does not contain burning plasmas at all. Summarising, this thesis shows that the particle confinement is an attractive candidate for burn control, whereas the energy confinement is not. The operating points for future reactors are stable and their stability is increased by the addition of external heating power. The stability properties of the burn point are, however, complex and might need to be considered in the design of a fusion reactor. The applicability of current τE scaling laws to burning plasmas is questionable at best, and an effort should be undertaken to obtain data points for burning plasmas. iv v Contents Summary i 1 Introduction 1.1 Ignition and burn . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2 Burn stability and sensitivity . . . . . . . . . . . . . . . . . . . . . 1.3 Research questions . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 2 4 5 2 Theory 2.1 The fusion reaction . . . . . . . . . . . . . 2.2 The tokamak . . . . . . . . . . . . . . . . 2.2.1 Operational limits . . . . . . . . . 2.3 Transport and confinement . . . . . . . . 2.3.1 Classical transport . . . . . . . . . 2.3.2 Neo-classical transport . . . . . . . 2.3.3 Anomalous or turbulent transport 2.3.4 L and H mode . . . . . . . . . . . 2.3.5 Sawtooth crashes . . . . . . . . . . 2.3.6 Energy confinement time . . . . . 2.3.7 Scaling laws . . . . . . . . . . . . . 2.3.8 Particle transport and confinement 2.4 Helium transport . . . . . . . . . . . . . . 2.4.1 Helium profile . . . . . . . . . . . . 2.5 Tritium breeding and burn-up fraction . . 2.6 Power balance . . . . . . . . . . . . . . . . 2.7 Burn equilibria . . . . . . . . . . . . . . . 2.8 Reactor studies . . . . . . . . . . . . . . . 2.9 Stellarators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 7 9 11 12 12 13 14 15 17 17 18 19 23 23 25 27 31 33 35 vi 3 Burn equilibria 3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2 Burning plasmas . . . . . . . . . . . . . . . . . . . . . . . 3.2.1 Introduction . . . . . . . . . . . . . . . . . . . . . 3.2.2 Theory . . . . . . . . . . . . . . . . . . . . . . . . 3.2.3 Results . . . . . . . . . . . . . . . . . . . . . . . . 3.2.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . 3.3 Burn equilibria with impurities and Pext . . . . . . . . . . 3.3.1 Introduction . . . . . . . . . . . . . . . . . . . . . 3.3.2 Temperature domain of a burning plasma . . . . . 3.3.3 Helium fraction with external heating . . . . . . . 3.3.4 Burn equilibria with external heating . . . . . . . . 3.3.5 Impurities . . . . . . . . . . . . . . . . . . . . . . . 3.3.6 Power output with external heating and impurities 3.3.7 The effect of Pext on net electric output . . . . . . 3.3.8 Uncertainties in scaling laws . . . . . . . . . . . . . 3.4 Discussion and conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37 37 39 39 41 41 48 49 49 49 51 52 54 60 61 62 64 4 Burn stability 4.1 Introduction . . . . . . . . . . . . . . . . . . . 4.2 Theory . . . . . . . . . . . . . . . . . . . . . . 4.2.1 Burn equations . . . . . . . . . . . . . 4.2.2 Stability of a two-dimensional system 4.2.3 Bifurcation theory . . . . . . . . . . . 4.3 Reduced system . . . . . . . . . . . . . . . . . 4.3.1 Derivation . . . . . . . . . . . . . . . . 4.3.2 Jacobian matrix of the reduced system 4.3.3 Normalisation . . . . . . . . . . . . . . 4.3.4 Reduced system stability . . . . . . . 4.3.5 Physical interpretation . . . . . . . . . 4.3.6 Low temperature stability . . . . . . . 4.3.7 High temperature stability . . . . . . 4.3.8 Phase portrait . . . . . . . . . . . . . 4.3.9 Stability for different scaling laws . . . 4.3.10 Stability with external heating . . . . 4.3.11 Reactor comparison . . . . . . . . . . 4.4 Full system . . . . . . . . . . . . . . . . . . . 4.4.1 Jacobian matrix of the full system . . 4.4.2 Full system stability . . . . . . . . . . 4.4.3 Eigenvectors and eigenvalues . . . . . 4.4.4 Low temperature stability . . . . . . . 4.4.5 High temperature stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67 67 69 69 71 72 73 73 74 75 76 78 81 82 83 85 87 92 92 92 95 97 98 99 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii 4.5 4.4.6 Stability for different scaling laws . . . . . . . . . . . . . . . 102 4.4.7 Reactor stability comparison with external heating . . . . . 102 Discussion and conclusions . . . . . . . . . . . . . . . . . . . . . . . 106 5 Sensitivity of burn contours to form of scaling 5.1 Introduction . . . . . . . . . . . . . . . . . . . . 5.2 Theory . . . . . . . . . . . . . . . . . . . . . . . 5.3 Results . . . . . . . . . . . . . . . . . . . . . . . 5.3.1 Operating contours . . . . . . . . . . . . 5.3.2 Density and power coupling . . . . . . . 5.4 Discussion and conclusions . . . . . . . . . . . . laws . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109 109 110 113 113 115 119 6 Discussion and conclusions 121 7 Outlook and recommendations 125 Appendix A Partial derivatives for the Jacobian 127 A.1 Reduced system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127 A.2 Full system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129 Appendix B Derivation of ne as function of T 131 Appendix C Neoclassical confinement time 133 Appendix D Alternative scaling for the confinement time 135 Bibliography 137 Acknowledgements 147 Curriculum Vitae 149 1 Chapter 1 Introduction Fusion is a fascinating phenomenon. The ’simple’ and elegant process of joining two elements into a heavier one has lit up our universe since the birth of the first stars and, in the case of our sun, enabled life to evolve on earth. Ever since we understood this seemingly unlimited source of energy, the harnessing of its power has stood as one of the great challenges of physics. And we are fortunate to live in a time when our collective efforts are about to culminate in the first demonstration of controlled fusion as an energy source. Successful operation of the large test reactor ITER will hopefully lead to the construction of one or more demonstration reactors, which for the first time will provide fusion electricity to the grid. To fuse two nuclei the Coulomb repulsion, due to their respective charges, needs to be overcome. This can be achieved by heating up the fuel to, typically, 150 million degrees centigrade, or 15 keV1 . The success of future fusion reactors depends on the ability of the fusion process to maintain this temperature with little or no external power, so the basic question is: can we create a fusion reactor that works like a stove? You put in fuel, heat it until it reaches the ignition temperature and after that it will burn indefinitely, as long as you refuel on time and remove the ash. When thinking about the design of such a reactor, several questions arise. How high are the ignition and burn temperatures (which are generally not the same)? What is the power output? How much ash can be tolerated in the machine? Is the system stable? What happens in case of a disturbance? Do we need to control it? And if this is the case, can we? 1 The electronvolt (eV) is a unit that is often used in plasma physics and corresponds to approximately 11000 degrees Kelvin. 2 1.1 Chapter 1 Introduction Ignition and burn At the temperatures required for fusion, the fuel has become a plasma, the fourth state of matter. In a plasma, the nuclei are stripped of their accompanying electrons and form a soup of charged particles. This has the advantage that it can be contained in a magnetic field that reduces the heat loss from the plasma by several orders or magnitude. Also, it prevents the plasma from touching the walls of the reactor. Charged particles can travel freely along the field lines, like beads on a string, but perpendicularly to them they are restricted to a gyrating motion. If the field lines were to touch the reactor wall, there would be excessive heat and particle losses, so to avoid these the field lines in a fusion reactor are bent such that they close on themselves, resulting in a toroidally shaped magnetic field. The sun creates energy by fusing hydrogen atoms into helium [1], but the fusion reaction used in reactors on earth is between the hydrogen isotopes deuterium (D) and tritium (T) because this reaction has a higher chance of occurring for a given temperature. The products of the reaction are an alpha particle (helium nucleus), a neutron and an amount of energy: D+ + T+ −→ He2+ + n + 17.6 MeV. (1.1) The energy is released in the form of kinetic energy of the alpha particle (3.52 MeV) and the neutron (14.1 MeV), with the lions share going to the neutron because of conservation of momentum. The neutron escapes the magnetic field unhindered and is absorbed in the wall, where its energy is converted into heat. This heat is then extracted and used to power a generator. The alpha particle on the other hand, is confined by the magnetic field and will heat the plasma by transferring its kinetic energy through collisions with plasma particles. It is this process that will have to provide most of the heating power in a fusion reactor. The plasma loses energy through radiation and conduction, and at low temperatures these losses outweigh the alpha heating power from the fusion reactions. This means that external heating is required to make up the deficit, which is undesirable from an economic point of view because it lowers the efficiency of the plant. Fortunately, it turns out that for a large enough reactor, the fusion power increases faster with temperature than the radiation and conduction losses. So at a certain temperature the external heating can be switched off and the plasma heats itself. This is a precarious balance, because the fusion power has a stronger response to variations in temperature than the radiation and conduction losses. This means that a small temperature perturbation will grow, making this an unstable equilibrium. Intuitively, it can therefore be thought of as the ignition point. Bear in 1.1 Ignition and burn 3 mind that the temperature at this point is also determined by the reactor, not only the fuel. A positive temperature perturbation will be kept in check by the conduction losses, that will ultimately outweigh the fusion power. There is therefore a second, stable, equilibrium at higher temperature, which corresponds to the natural understanding of a ‘burn point’. This dynamic behaviour is represented in figure 1.1, which shows the time derivative of the temperature (Ṫ ) as a function of temperature (T ) for a hypothetical reactor. For low temperatures Ṫ is negative, indicating the need for external heating, until at the ignition temperature it crosses zero. Beyond that Ṫ is positive, which means that the temperature in the reactor will increase on its own accord until the second zero crossing at the burn temperature. Ṫ (keV/s) 0.5 ignition 0 burn −0.5 0 5 10 T (keV) 15 Figure 1.1: The time derivative of the temperature (Ṫ ) plotted against the temperature for a hypothetical fusion reactor. For low temperatures, Ṫ is negative and external heating is required. The curve then crosses the horizontal axis and Ṫ becomes positive, so the temperature of the plasma will increase by itself, until the stable temperature is reached at the second zero crossing. There is a limit to the amount of fuel (deuterium and tritium) and ash (helium) that a reactor can contain, as is the case in a normal stove. Because the fusion power scales with the fuel density squared, one wants to operate close to this limit. Every helium particle takes the place of two fuel particles, thus lowering the power output, and therefore needs to be removed from the plasma after it has had time to transfer its energy. Because the helium and fuel are mixed, selective removal of one particle species is complicated. Consequently, rapid removal of helium results in a low burn-up fraction of the fuel, because it is exhausted from the plasma before it has had time to fuse. The fuel can of course be separated from the helium and be recirculated, 4 Chapter 1 Introduction but this process is inevitably accompanied by some losses. As tritium does not occur naturally (it has to be made, or ’bred’, in the reactor) and the maximum tritium breeding ratio (defined as the average number of tritium atoms bred per fusion reaction) is only slightly larger than one, these losses can be ill afforded. Reducing the fuel recirculation on the other hand, by keeping the particles in the plasma longer, will result in a higher burn-up fraction. But this comes at the cost of a lower power output because the helium concentration will also increase. Furthermore, there is also a limit on the plasma pressure, often referred to as the Troyon or β-limit [2]. The exact value of this limit depends on the shape of the plasma, but exceeding it inevitably leads to the development of a magnetohydrodynamic (MHD) instability that changes the geometry of the magnetic field and causes the plasma to disrupt, potentially damaging the reactor. 1.2 Burn stability and sensitivity A fusion plasma is a very dynamical system. There is regular redistribution of particles, energy and current by the sawtooth instability, possible changes in transport due to the interaction of fast alpha particles with the magnetic field or turbulence, or changes in power output and confinement due to the gradual build-up of helium ash in the plasma core. Such phenomena will nudge the plasma out of its burn point and the question is: where will it go from there? Will the plasma drift away from its burn point? Will it return to the previous equilibrium? In either case, will it cross operational limits on these excursions, such as the β-limit? Can we control these excursions? What happens to the fusion power? In short: will a fusion reactor burn like a candle or will it make uncontrollable excursions in temperature and power? The latter is of course highly undesirable behaviour, since not only would the utility companies not like it, it would also put harder requirements on the plasma facing components and structural materials in the reactor. Furthermore, the energy transfer from alpha particles to the plasma is not instantaneous but happens gradually. The time scale of this transfer depends on plasma parameters such as density, temperature and composition and introduces a time delay between variations in density and temperature, and the heating power delivered to the plasma. This could introduce oscillatory behaviour or change the stability properties of the burn equilibria. To model the performance of a fusion reactor, descriptions of the energy and particle losses are needed. The common approach is to use scaling laws that predict the energy confinement time τE and particle confinement time τp (measures of how fast the plasma loses its thermal energy and its particles, respectively), taking machine and plasma parameters as input. This allows the study of burn equilibria as a function of density, temperature, energy and particle transport, and investigation of the sensitivity of, for instance, the fusion power to these 1.3 Research questions 5 parameters. However, these scaling laws have been developed using fusion reactors in which alpha heating of the plasma was (almost) completely absent, and caution needs to be exercised when applying them to burning plasmas. 1.3 Research questions The main question this thesis tries to answer is the following: What are the properties of burn equilibria in fusion reactors? Whereby with properties we mean: • the contours in the operational space of the reactor spanned by plasma temperature and density where stable burn is possible; • the dependencies of these contours on parameters that are under operator control, such as the density, and those that are much less so, such as the particle and energy confinement time and plasma purity; • the stability of the burn under perturbation of these parameters, and the level of perturbation that can be tolerated before the burn quenches. We’ll articulate these aspects in four sub-questions below. What parameters determine the temperature and composition of the plasma at the burn equilibria and how sensitive is the system with respect to these parameters? To keep the cost of electricity down, we want to maximise the power output of the reactor which requires operation close to the density limit, limiting its effectiveness as an actuator for control of the power. In a burning plasma, the only other parameters at the disposal of the operator are the energy and helium removal, leading to the question How does the power output of a burning plasma respond to changes in energy confinement or particle transport? Not only the position, but also the stability of the equilibrium is of importance, because it determines the level of control that is needed. And while a burn equilibrium might be stable, the evolution of the system in phase space in response to a perturbation might still lead to a violation of an operational limit, be it a fundamental physics limit for the plasma, or a material limit for the reactor. It is therefore of importance to know how the system responds to perturbations of the equilibrium and whether this leads to a reduction in the accessible operating space: What are the stability properties of the operating points? The last point concerns the use of scaling laws for the energy confinement time. While it is common practice to use them to predict the performance of future experiments, they are based on databases without burning plasma entries. 6 Chapter 1 Introduction Applying them to burning plasmas might uncover sensitivities that are not present in externally heated plasmas. How sensitive are the burn equilibria to errors in the scaling laws for the energy confinement time? This thesis is organised as follows. Chapter 2 provides the theoretical framework of burning plasmas based on existing literature, followed by an analysis of burn equilibria - and the influence of density, particle and energy confinement on these equilibria - and the effect of external heating and impurities in chapter 3. Subsequently, chapter 4 provides a linear stability analysis of burn equilibria, for both a two dimensional and a four dimensional system. The sensitivity of burn equilibria with respect to scaling laws is investigated in chapter 5. The final chapters, 6 and 7, provide the conclusions and outlook towards possible future research. 7 Chapter 2 Theory 2.1 The fusion reaction Fusion is merging of two atomic nuclei into a heavier particle. For reaction products up to iron, the mass of the resulting nucleus is slightly smaller than the sum of the masses of the fusing particles. This mass difference m is converted into energy (E), described by Einstein’s famous E = mc2 with c the velocity of light [3]. So in principle a lot of reactions could be used as an energy source, but there are some factors that limit the choice to only one realistic candidate. Firstly, there is a tradeoff between overcoming the Coulomb barrier and the time the particles are close enough to interact. Atomic nuclei carry a positive charge and repel each other. To overcome this repulsion, the particles need to have enough kinetic energy1 . Although a higher initial velocity will bring the particles closer together, thereby increasing the chance that they will fuse, it also reduces the time they spend in each others vicinity which reduces the fusion probability. It turns out that the fusion probability, or cross-section σ, has a maximum and the particle energy at which this optimum occurs is reaction specific. The repulsive force between two particles with charge Z scales with Z 2 , while the kinetic energy scales only with the mass of the nucleus, which ∝ Z. Particles with higher charge need a higher velocity to overcome the Coulomb barrier, thus making it harder to fuse them. And indeed, low Z particles generally have higher cross-sections. For a given element however, reactions with heavier isotopes are favoured because for equal energies they have a lower velocity. It is no surprise therefore that fusion reactions involving light elements like hydrogen and helium have the highest cross-sections, or reactivity. The reactivity is the integral of the product of velocity and cross-section of the reaction 1 Another way of overcoming the Coulomb barrier is to create a very high pressure, which is the case in the core of stars and for inertial confinement fusion. Because this thesis deals with magnetic confinement fusion, we will not discuss this further. Chapter 2 Theory 8 over a Maxwellian temperature distribution. This is relevant in case the reactions take place in a plasma where the energy of the individual particles follows a (Maxwellian) distribution function. Figure 2.1 displays the reactivity, denoted hσvi, for the deuterium-tritium (DT), the DD and the 3 HeD reactions based on the fitting formulas provided by Bosch and Hale [4]. 10−20 DT D D 10−29 He D 10−26 3 hσvi (m3 /s) 10−23 10−32 0 10 101 102 103 T (keV) Figure 2.1: The reactivity of three fusion reactions involving hydrogen isotopes. The DT reaction has the highest reactivity for temperatures up to several hundred keV. Please note that, although plotted here up to 1 keV, the parametrisation of the reactivities from [4] is only valid below 100 keV for the DT and DD reactions, and below 190 keV for the 3 HeD reaction. A second consideration when picking a fuel is availability. The 3 HeD reaction has the advantage that it is (mainly) aneutronic, which reduces the radioactive activation of the machines, increases the lifetime of components, enables more neutron susceptible technologies and diminishes the need for neutron shielding. Unfortunately, 3 He is exceedingly rare on earth and thus seems unlikely to be used for fusion on a commercial scale2 . Moreover, the 3 HeD reaction requires temperatures that are an order of magnitude higher than the DT reaction, which is problematic because of the β-limit (see section 2.2.1). 2 There are significant resources of 3 He on the moon though, which might become accessible in the future [5]. 2.2 The tokamak 9 The DD reaction has neither the advantage of being aneutronic, nor of having the highest reactivity. On top of that, the energy released per reaction is, at 3.70 MeV, much lower than for DT (17.6 MeV) or 3 HeD (18.3 MeV). This leaves the DT reaction as the only realistic candidate at this moment. While deuterium is a naturally occurring isotope of water and there is plenty available on earth, this is not the case for tritium. Tritium has a half life of 12.3 years and does therefore not occur naturally, so it has to be produced artificially. This can be done by irradiating lithium with the neutrons released in the DT reactions, and will be covered in more detail in section 2.5. 2.2 The tokamak The high temperatures needed for fusion require the fuel to be kept away from the walls of the reactor and while there are several ways in which this can be achieved, the most promising approach for reactor development relies on magnetic fields to confine and position the plasma. Charged particles can move freely along magnetic field lines, but are restricted in their perpendicular motion due to the Lorentz force. To avoid end losses, the magnetic field is usually bent into a toroidal shape. The most successful reactor concept to date is the tokamak, invented in Russia in the 1950s by Sacharov and Tamm. It derives its name from the Russion acronym for ’toroidal chamber with magnetic coils’: тороидальная камера с магнитными катушками (toroidal’naya kamera s magnitnymi katushkami). The results from experiments on the first tokamak, T1, were presented to the world at the second Geneva Conference on the Peaceful Uses of Atomic Energy in 1958 [6], although the device was at that time still unnamed. A schematic representation of a tokamak can be found in figure 2.2. A tokamak consists of a toroidally shaped vacuum vessel, which is surrounded by coils that generate a toroidal magnetic field (Bφ , see figure 2.2). The curved nature of the field causes the particles to drift, necessitating a helical transform of the field lines. This is achieved by running a current through the plasma, which induces a poloidal magnetic field (Bθ ). The resulting field lines have a helical shape and form a set of nested flux surfaces which are isothermals and isobars (a detailed derivation of the magnetic equilibrium in a tokamak can be found in [7], but the intuitive picture is that particles are free to travel along the field lines, smoothing out variations in pressure and temperature). The safety factor q is defined as the number of toroidal turns a field line has to make to complete one poloidal turn. In a cylindrical approximation this is given by q = rBφ /RBθ . The plasma current is driven by operating the plasma as the secondary winding of a transformer, the primary of which is the central solenoid placed in the central opening of the vacuum vessel. A final set of coils generates a vertical field that prevents the plasma from expanding, shapes it, and positions it in the vacuum 10 Chapter 2 Theory Figure 2.2: A schematic representation of a tokamak, with the vacuum vessel omitted for clarity. The toroidal field coils are shown in light blue, the poloidal field coils in silver, the central solenoid in green and the plasma in purple. Image courtesy of EFDA-JET. 2.2 The tokamak 11 vessel. To exhaust the helium produced in the fusion reaction and to create a well defined plasma-wall interaction region, most tokamaks are equipped with a socalled divertor. Usually located at the bottom of the vacuum vessel, this is a region where the field lines intersect the wall. The transition surface between closed flux surfaces and open field lines is called the separatrix and the plasma outside this surface is referred to as the scrape-off layer. Current tokamaks rely on external heating to create the necessary condition for fusion. The ratio between the fusion power Pfus and external heating power Pext Q= Pfus , Pext (2.1) is often used to gauge reactor performance. In case of a burning plasma in which the alpha particles provide the required heating power, Q is infinite. 2.2.1 Operational limits In equilibrium, the pressure gradient ∇p in the plasma has to be balanced by the Lorentz forces arising from the plasma currents and the magnetic field ∇p = J × B, (2.2) with J the current density and B the magnetic field. Because the magnetic field coils constitute a large fraction of the cost of a fusion reactor, ideally the ratio between plasma and magnetic pressure β= p B 2 /2µ0 (2.3) would be one, so there is no ’wasted’ magnetic pressure. Unfortunately this value is unattainable due to the existence of MHD instabilities, or modes as they are often referred to. The maximum value of β that can be achieved in a tokamak, before large scale MHD modes become unstable, can be expressed as βmax = g IM , aB (2.4) and is generally referred to as the Troyon limit. Here IM is the plasma current in megaAmperes, a the minor radius in meters and B the magnetic field on axis in Tesla. Extensive stability calculations for a wide range of pressure and current profiles by Troyon et. al [2] found the value of g to be 0.28 N/A2 , in fusion literature often used without units. 12 Chapter 2 Theory Another limit that needs to be respected when operating a tokamak concerns the electron density ne . Empirically it was established that above the Greenwald density (in units of 1020 m−3 ) [8, 9] IM , (2.5) πa2 a disruption, an event in which control over the plasma is lost and which can damage the reactor, becomes very hard to avoid. Advanced regimes allow operation up to ne ≈ 1.5nG [10] which results in approximately double fusion power output compared to operation at ne = nG . nG = 2.3 Transport and confinement The study of energy transport, and to a lesser extent, particle transport in magnetically confined plasmas has for a long time been a major part of fusion research. The first calculations in the 1950s only took classical transport (i.e. collisional diffusion across a straight B-field) into account. However, it was quickly realised that with the introduction of curved magnetic geometries, classical transport was greatly enhanced because of the drift motions and the trapping of particles (due to the variation in field strength along a field line), and this realisation led to the study of neoclassical transport. When the reactors became bigger en more advanced and temperatures increased, the predictions again turned out to be far off the mark and this time there was no easy explanation, hence the name ’anomalous transport’. Increasing diagnostic capabilities and physical understanding led to the insight that this was in fact turbulent transport, which to this day is not completely understood, although advanced numerical models are reaching the point where experimental results can be reproduced and predicted. This section will briefly introduce the three forms of transport and the way the resulting confinement is modelled at reactor level. 2.3.1 Classical transport Previously it was stated that plasma particles stick to the field lines like ’beads on a string’. This picture is not entirely accurate. In a homogeneous, straight magnetic field the particles gyrate around the field lines with the Lorentz force acting as the centripetal force, with the cyclotron frequency ωc and gyroradius ρg (also referred to as Larmor radius) given by qB m mv⊥ ρg = , qB ωc = (2.6) (2.7) 2.3 Transport and confinement 13 with q denoting the charge of the particle (e for an electron or hydrogen nucleus), v⊥ the velocity perpendicular to the magnetic field, m the mass of the particle and B the magnetic field strength. In case of a collision, particles can ’hop’ onto a different field line which is typically one gyroradius away. This results in classical diffusion with the diffusion coefficient of the order of χCL = νei ρg , where νei is the electron-ion collision rate. Assuming a Maxwellian velocity distribution it can be shown to be [7] νei = with ln Λ ≈ ln 3/2 12π0 Te3/2 1/2 3 e ne √ 2 e4 3/2 12π 3/2 20 m1/2 e T ln Λ, ≈ 15−20 the Coulomb logarithm. Classical transport 1/2 −3 n20 /B 2 Tk coefficients amount to χCL e = 4.8 × 10 1/2 0.10n20 /B 2 Tk ≈ 7.2 × 10−4 m2 /s. 2.3.2 (2.8) ≈ 3.4 × 10−5 m2 /s and χCL = i Neo-classical transport Because the magnetic field is curved, the field lines are ’compressed’ on the inside, and ’rarified’ on the outside of the torus. This creates a 1/R2 gradient in the field strength that points towards the center of the torus. The curvature and gradient give rise to the so-called curvature B drift and gradient B drift, respectively. Furthermore, if there is a (radial) electric field the particles will experience an E × B drift. Together these drifts change the trajectories of the particles in the plasma; the radius of their gyration changes periodically over an orbit and they no longer follow the field lines, resulting in a drift motion perpendicular to the field. A second effect originating from the gradient in the field, is that particles can 2 become ’trapped’. The magnetic moment µ = mv⊥ /2B of a particle, with v⊥ its velocity perpendicular to the magnetic field, is conserved. Because the magnetic field is stronger on the inside of the torus than on the outside and the magnetic field lines are twisted, a particle starting at the outside and following a field line will initially move along the gradient. To keep its magnetic moment constant, v⊥ has to increase and conservation of energy dictates that its parallel velocity vk decreases. If its initial parallel velocity was too small, at some point it will decrease to zero and it will reverse direction. It now follows a field line in the opposite direction, until it again runs up against the magnetic ’hill’ and reverses direction again. These particles are ’trapped’ and will bounce back and forth. Looking in the poloidal plane, the centre of mass of particles traveling around the torus in the direction of the magnetic field, describes a circle that is shifted inwards and is slightly larger than the flux surface its associated with. For particles traveling agains the magnetic field the opposite holds true: they are on a trajectory that is slightly smaller and shifted outwards. This means that a trapped particle Chapter 2 Theory 14 that reverses direction does not retrace its original path exactly, but follows a more or less parallel trajectory. The resulting orbit looks like a banana in the poloidal projection, and the step size for collisional transport of these particles is not their gyroradius, but the width of their so-called banana orbit. An approximate scaling for the resulting neoclassical diffusion coefficient is χNC ∝ q 2 ε−3/2 χCL , with ε = a/R0 the inverse aspect ratio. Neoclassical transport is rock bottom: a tokamak cannot do better than this and although the proportionality factor to classical transport looks inconspicuous, it turns out to be a factor of 100 larger for tokamaks with a large aspect ratio. 2.3.3 Anomalous or turbulent transport In most cases the transport is orders of magnitude larger still than neoclassical transport and this phenomenon is referred to as anomalous transport. It is caused by turbulence and a complete description is extremely complicated due to the nonlinear nature of the turbulence. Turbulent transport is convective, which sets it apart from classical and neo-classical transport, which are both diffusive. However, it turns out that for most purposes it works quite well to describe turbulent transport with an effective diffusion coefficient χT ∝ γmax L2c , where γmax is the growth rate of the fastest growing mode and Lc is the turbulence correlation length [11]. Although there are many forms of turbulence in a tokamak plasma, two electrostatic drift wave instabilities are the major drivers of turbulent transport under fusion conditions. For ion thermal transport this is believed to be the ion temperature gradient (ITG) instability [12, 13, 14], and the electron transport is dominated by trapped electron modes (TEM) [12, 13, 15]. For drift wave turbulence, the value of Lc scales with the gyroradius ρg in the limit of small ρ∗ = ρg /a, which results in so-called gyroBohm scaling with a diffusion coefficient χT ∝ ρ∗ T /eB. This in contrast to the Bohm scaling that applies to modes with a size comparable to the plasma dimensions (or minor radius), which follows χT ∝ T /eB [16, 11]. Often there is a threshold gradient above which the turbulence growth rate increases sharply. The result is a corresponding sharp increase in diffusion coefficient, an effect which is referred to as profile stiffness, because above the threshold the gradient responds much less to changes in the heat flux. This phenomenon is illustrated in figure 2.3, where the heat flux is plotted as a function of the dimensionless temperature gradient length R dTi R = . LTi Ti dr The diffusion coefficient in the stiff region of the plasma is often approximated 2.3 Transport and confinement 15 χ1 χ0 Threshold (κ) Figure 2.3: Schematic representation of energy transport in a tokamak as a function of the normalised temperature gradient. For low temperature gradients, the heat flux increases linearly with R/LT , until a critical value is reached. Beyond this value, the gradient becomes stiff, i.e. it hardly responds to changes in heat flux anymore. Figure adapted from [18]. with a critical gradient model [13] R R χT = χgB χs − κc H − κc + χ0 , LTi LTi (2.9) with χgB = q 3/2 T ρg /eBR the gyroBohm normalisation, χs the stiffness level, H the Heaviside step function, κc the threshold (with a value around 5 often found for reactor relevant tokamaks, although there are also parametrisations based on gyrokinetic simulations [17]) and χ0 the level of residual transport in the absence of turbulence. 2.3.4 L and H mode In the eighties a new regime of operation was discovered in the ASDEX tokamak [19]. ASDEX was one of the first tokamaks equipped with a divertor, and when enough heating power was supplied, the plasma would ’jump’ to a state in which the confinement was roughly a factor two better than before. The new Chapter 2 Theory 16 1 8 Experiment (Scans) Core Plasma Temperature 6 Tiexp (0.4) (keV) Internal transport barrier (ITB) Sawteeth Edge localized modes (ELMs) L—mode Edge transport barrier in H-mode JG98.483/35c 2 Neutrals 0 4 0 1 Normalised radius r/a Tiexp (0.8) (keV) Figure 1: Schematic view showing regions with different Figure 2: Link between central and edge ion tempera Figure 2.4: Typical radial temperature profiles in a tokamak for for different operating a series of ASDEX-Upgrade discharges [30]. transport characteristics in tokamak. regimes. When going from L to H mode, an edge transport barrier is created which results in very steep temperature and pressure gradients at the plasma edge, and elevates the core temperature. Figure courtesy of EFDA-JET. JG99.238/4bw Ti(0) and Te(0) (keV) Ion temperature (keV) regime was dubbed H-mode (high confinement) and the ’normal’ regime retroactively received the name L-mode (for low confinement). The improved confinement originates from a transport barrier at the plasma edge, where the pressure Pulse gradient creates a radial electric field that drives E × B No: 47543, 47545, 47546 15 = 8-18MW shear flows that locally10.0reduce the turbulent transport [20, 21]. ThisPNBIcan be seen 8.0 Type-I ELMs in a strong reduction in the balmer α radiation around the plasma [22], indicating 6.0 a reduction in outward particle flux. The results are steep temperature en density gradients at the plasma4.0edge, and because the core transport remains unaffected, 10 it looks like their respective profiles are placed on a pedestal, which is illustrated in figure 2.4. Because the pedestal raises the temperature and density over the 2.0 whole cross section of the plasma, it has a large contribution to the total stored Ions energy W and therefore the confinement time. 5 The H-mode comes1.0at a price though. The transport barrier at the plasma Electrons 0.8 that the pressure gradient keeps increasing until it hits a edge is usually so strong, 0.6 (local) stability limit, which triggers an edge localized mode (ELM) that ejects up to 10% of the stored energy from the plasma [23]. This energy (and the particles IP = 1.8MA / Bt =3. 0 0.4 0 1 3 that carry it) travel through layer the 3.7 divertor, where they hit the 2 3.5 to3.6 3.8 3.0 the 3.1 scrape-off 3.2 3.3 3.4 Tiped and Teped (keV) radius (m) wall. The short timespan (≈1 ms) ofMajor these events results in transient heat loads of up to 1 GW/m2 on the divertor surface [24], which may damage the divertor. Figure 3: Ion temperature profile for a series of JET shots Figure 4: Link between central and edge temperatur For this reason, a reactor will edge require form of ELM[32]. mitigation toJT60-U protect it. [33]. a series of plasmas with varying densitysome and edge ion temperature 18 2.3 Transport and confinement 17 Analogous to the pedestal, the plasma can also develop transport barriers in the core, often associated with a region of strong flow shear and the presence of flux surface with rational values of q [25]. These regions are referred to as internal transport barriers (ITBs) and might be used in advanced reactor scenarios. 2.3.5 Sawtooth crashes An important mechanism in (particle) and energy transport in the center of the plasma is the sawtooth mechanism. This takes its name from a sudden drop in temperature in the center of the plasma, followed by a gradual recovery, until the process repeats itself. When the central temperature is plotted as a function of time, the resulting graph has a distinct sawtooth shape. The sawtooth crash is caused by the central value of the safety factor q dropping below one. This triggers an MHD instability, in which the hottest, central part of the plasma is pushed outwards and replaced by cooler plasma. The result is an outward propagating heat flux and a flattening of the temperature, and to a lesser extend, density profiles. Because the fusion power Pfus ∝ p2 , the effect on the fusion power can be significant. Generally, sawteeth will not cause a disruption, but if they become too big they might destabilise other, more harmful, MHD modes, like neoclassical tearing modes. They can also play a role in flushing accumulating impurities from the plasma core, but this is a double-edged sword as they can also help impurities penetrate into the plasma center [26]. 2.3.6 Energy confinement time From a reactor point of view, the overall transport properties of the plasma are more relevant than the precise values of the transport coefficients at each radial (and poloidal) position. These global properties are reflected in the energy confinement time τE which is a measure of how long the plasma retains its energy τE = W Pcond − dW dt . (2.10) Here, W = 3/2ntot T is the total internal energy of the plasma, with ntot = ne + Σi ni Zi , and Pcond is the conducted power. In equilibrium dW/dt = 0, so the definition simply becomes τE = W . Pcond (2.11) Because Pcond is hard to measure, often a different version (with a slightly different notation) of the confinement time is used τ̃E = W , Ploss (2.12) Chapter 2 Theory 18 in which case Ploss = Prad + Pcond , the total loss power. In experiments with only external heating, this makes determining τ̃E simply a matter of measuring the temperature and density, because in equilibrium Ploss = Pext . 2.3.7 Scaling laws Because of the complicated nature of the transport processes the expected confinement time for a new experiment is often calculated using scaling laws, which provide τE as a function of engineering or physics The most common Q parameters. i approach is to fit a function of the form τE ∝ i pα , known as a power law, to i the data. This can be done either in engineering variables, like major and minor radius, plasma current, magnetic field, density, power, etc, or in physics variables like the Bohm time τB = a2 B ∝ ε2 R2 BT −1 , T (2.13) the normalised ion gyroradius ∗ ρ = 2eT Mi 1/2 √ Mi T Mi ∝ , eBa εRB (2.14) the ratio of plasma and magnetic pressure β ∝ nT B −2 , (2.15) the normalised collision frequency (collision frequency divided by the bounce frequency of trapped particles) ν ∗ = νii Mi eT 1/2 R a 3/2 qR ∝ nRT −2 qε−3/2 (2.16) and the cylindrical safety number qcyl = RB f (κ, δ) ∝ BRI −1 ε2 κ, ε2 I (2.17) with f (κ, δ) a function of the plasma triangularity δ, the elongation κ = b/a, (with b and a the diameter of the poloidal plasma cross section along the principal axes), and T the ion temperature in eV [26]. The values obviously vary over the profile, but can be approximated by their volume average for a global analysis. Using the above definitions, a linear transformation can be made between the engineering and physics variables and their respective exponents. The number of free parameters in the fit can be reduced by placing constraints on the exponents using the method developed by Kadomtsev [27] and Connor and 2.3 Transport and confinement 19 Taylor [28]. This relies on finding linear transformations of the physics variables under which the governing equations are invariant. Applying these transformations to the general form of the scaling law then constrains the exponents. For instance, the Kadomtsev, or high β, constraint demands that the exponents satisfy 4αR − 8αn − αI − 3αP − 5αB = 5. The Bohm and gyro Bohm constrained scalings add αR −7αn −4αI −7αP −5αB = 0 and 6αR −22αn −9αI −12αP −15αB = 0, respectively, to the high β constraint. One of the first concerted efforts in compiling a database with results from different tokamaks was made in the eighties, which resulted in the ITER89P Lmode scaling [29] 0.85 1.2 0.3 0.5 0.2 0.5 0.1 −0.5 τ̃E = 0.048IM R a κ B A n20 P , (2.18) where R is the reactor major radius, A the average ion mass in amu, n20 the electron density in 1020 m−3 and P the heating power in MW. Until the first Hmode scalings were published in the nineties, the L-mode scaling was also used for H-mode discharges by multiplying the predicted confinement time with an H-factor fH = 2. In reference [26] a comprehensive review of confinement data was made, resulting in a set of closely related forms of the τE -scaling. The IPB98(y,2) is the most commonly used scaling law for H-mode plasmas and also the recommended scaling law for reactor extrapolations [31]. The value of τE that it predicts, is plotted in figure 2.5 against the measured value of τE , for experiments in the confinement scaling database [32, 33]. Its parameters are given in table 2.1, together with the other H-mode scalings presented in the ITER physics basis [26]. These scalings differ in the definition of κ (κ = b/a vs κ = πa2 /area and the database that they are based on (the differences between the databases lie mainly in the type of external heating for the plasma discharges that they contain). 2.3.8 Particle transport and confinement For a burning plasma it is of importance how fast the ash is removed, relative to the burn rate, as this determines the burn-up fraction. This ratio is governed by the relation between particle and energy transport. Particle transport is somewhat different process from energy transport, but it works by the same mechanisms: diffusion and turbulence (convection). Experimental observations put the particle diffusion coefficient at the same order as the energy diffusion coefficient [34, 35, 36, 37, 38]. Dp ≈ χE , (2.19) which agrees with the transport being dominated by turbulence. This finding is confirmed by gyro-kinetic simulations, that also reveal a convection term, the sign of which depends on the ratio between electron and ion heat flux [39, 40]. Chapter 2 Theory 20 Plasma Phys. Control. Fusion 50 (2008) 043001 10 10.0 τEexp 1.01 0.1 0.1 Review Article COMPASS JT-60U PDX TFTR ASDEX DIII–D MAST START ITER AUG JET NSTX TCV C-Mod JFT-2M PBX-M TdeV 0.001 0.001 0.001 0.001 JG06.455-1c 0.01 0.01 0.01 0.01 0.1 0.1 98(y,2) (s) τ th,IPB98(y,2) τ 11 10 10 E Figure 34. Plot value of the measured H-mode thermal energy confinement versus that predicted by Figure 2.5: The predicted of τE from the IPB98(y,2) scaling plottedtime against the IPB98(y, 2) scaling expression. The symbols indicate data from different tokamaks as noted the measured values [30]. Figure courtesy of EFDA-JET. in the legend. where ε is the inverse aspect ratio (ε ≡ a/R), B is the toroidal magnetic field in T, and n is the line-averaged density in 1019 m3 . However, the anticipated operating regime for ITER is not L-mode, but ELMing H-mode. Therefore, an H-mode database was formed [132] containing both ELM-free and ELMing H-mode data. A full description of the H-mode database and its variables can be found in [133, 134] and a description of the L-mode database in [21]. Both the L-mode and H-mode databases have been routinely updated. The current recommended expressions to be used for the scaling of energy confinement time with selected dimensional variables are ITER97-L [21] for the Lmode and IPB98(y) and IPB98(y, 2) [20] for the H-mode. Unlike the Goldston and ITER89-P expressions, both of these scaling expressions pertain to the thermal energy confinement time (τth ) rather than the global energy confinement time (τE ), which includes the energy content in fast ions from auxiliary heating. Scaling IPB98(y) IPB98(y,1) IPB98(y,2) IPB98(y,3) IPB98(y,4) C (10−2 ) 3.65 5.03 5.62 5.64 5.87 I 0.97 0.91 0.93 0.88 0.85 B 0.08 0.15 0.15 0.07 0.29 n 0.41 0.44 0.41 0.40 0.39 P -0.63 -0.65 -0.69 -0.69 -0.70 R 1.93 2.05 1.97 2.15 2.08 κ 0.67 0.72 0.78 0.78 0.76 ε 0.23 0.57 0.58 0.64 0.69 A 0.20 0.13 0.19 0.20 0.17 N 1398 1398 1310 1273 714 RMSE (%) 15.8 15.3 14.5 14.2 14.1 ITER τE (s) 6.0 5.9 4.9 5.0 5.1 Table 2.1: The exponents of the different parameters in the IPB98(y) and IPB98(y,1–4) scaling laws for τE in H-mode plasmas [26]. The IPB98(y) and (y,1) scalings are based on the ITERH.DB3 dataset for ELMy H-modes, the first using κ = b/a and the latter using κ = area/πa2 . Scalings IPB98(y,2–4) also use κ = area/πa2 , but are based on a restricted dataset: (y,2) on ITERH.DB3 restricted to NBI heated discharges, (y,3) on the same as (y,2) but without the Alcator C-MOD data and (y,4) on the same as (y,2) but using only data from the five ITER similar devices. All scalings meet the Kadomtsev constraint, except for (y,3) which is just a free fit to the data. 2.3 Transport and confinement 21 22 Chapter 2 Theory For our purposes a detailed treatment of particle transport is not practical. Because we are most interested in the global particle transport, it seems reasonable to introduce a particle confinement time, similar to the energy confinement time. The definition of τp is then, analogous to the definition of τE , the total particle content N divided by the net flux of particles Γp , leaving the plasma through the separatrix: τp,core = N . Γp (2.20) Although the definition looks innocuous, there is a caveat which complicates matters somewhat. When a particle exits the plasma and hits the wall, it loses its energy and electric charge. The, now cold and neutral, particle either enters the pumping duct that leads to the pumps that maintain the vacuum, or it reenters the plasma and is ionised again. The latter process is often referred to as recycling. Depending on the value of the recycling coefficient Rcyc , τp can be a lot longer than the primary particle confinement time τp,core . The value of Rcyc , defined as the ratio of the recycled particle flux and the total particle flux, depends on the wall and plasma conditions, but can reach values up to 0.95 or even higher. Hence, a distinction needs to be made between core and edge particle transport. A detailed discussion and two alternative descriptions can be found in [41], but for simplicity we will stick to the more general definition of the global particle confinement time τp = τp,core /(1 − Rcyc ). Experimental observations indicate that the ratio ρ between τE and τp is approximately the same for different species, which makes it a suitable figure of merit for helium transport [41, 24]. The energy and particle confinement times are then related; τp = ρτE (2.21) and, depending on the plasma conditions, ρ varies from ± 5 up to around 30, with higher values for L-mode and ITB plasmas and lower values for elmy H-mode plasmas [42, 37, 43, 44]. For the remainder of this thesis, we will write τp for the particle confinement time as defined in equation (2.21), unless explicitly stated otherwise. Although we assume the same confinement time for all particle species, there can be a difference due to the different transport and edge recycling coefficients, and different pumping efficiencies. All of these depend on the atomic mass and the effective ion charge Zeff = Σj nj Zj2 , nj Zj with nj the density of particle species j, and Zj its atomic number. (2.22) 2.4 Helium transport 23 The figure of merit for the relative efficiency of the helium exhaust, is the enrichment ratio η, defined as the ratio of the helium concentration at the separatrix to the helium concentration in the exhaust gases (that are removed from the system) ηHe = ΓHe ncore DT . 2ΓDT ncore He (2.23) Values of ηHe = 0.3 have been found in AUG with the original divertor [45], while after the upgrade to the ITER like divertor values up to 1 were reported [46]. Similar results (ηHe between 0.1 and 1) were obtained at DIII-D [37], JT60-U [43] and JET [47]. The ITER physics basis states that ηHe ≥ 0.2 is required for ITER to be successful [48], but future reactors might require higher values to meet the tritium breeding and recycling targets. 2.4 Helium transport Helium ash removal is of critical importance to the success of a fusion reactor. Remove the ash too quickly and it will not be able to transfer its energy to the plasma to heat it. Remove it too slowly, and it will dilute the fuel too much, resulting a lower fusion power and possibly extinguishing of the reactor. Besides the volume averaged helium content, it is also the spatial distribution that matters. Helium in the plasma center will have a much larger effect on reactor performance than helium in the plasma edge, because of the peaked fusion power profile that is expected. The issue of ash removal from future fusion machines is a rather complicated problem that is difficult to investigate in present machines, due to the fact that there is no significant production of helium in the core. Furthermore, in contrast to present day machines, the sawtooth period in ITER is expected to be significantly longer than the energy confinement time, resulting in a different effect on the transport of impurities from the core towards the edge of the plasma. Predictions for the helium transport and profile in future reactors therefore rely heavily on numerical simulations. 2.4.1 Helium profile The helium profile in the plasma is determined by three things: the source profile, the transport of the helium produced in the plasma core towards the edge and the boundary condition at the edge, set by the recycling (which in itself is determined largely by the pumping efficiency). The source profile can easily be determined if the fuel density and temperature profiles are known. The alpha particles are created with an energy of 3.52 MeV and it takes some time before they have thermalised, during which they do 24 Chapter 2 Theory not necessarily stay on the same flux surface. The thermal helium source profile therefore is not exactly the same as the fusion power profile. Since this thesis only deals with volume averaged values for the temperature and particle densities, this is of little interest at this point, but needs to be considered for a detailed reactor design. Already in the 1960s the issue of helium transport was investigated, but only in last decade of the century did people start to self-consistently study the effect of helium in burning plasmas [49, 50, 51, 52, 53]. The most popular approach was to use a zero-dimensional (0D) model that only included the power balance and the alpha particle balance, resulting in a cubic equation for the helium fraction in the plasma, with the ratio between particle and energy confinement time ρ as a free parameter, which will be discussed in more detail in section 2.7. Currently, a full understanding of helium (or impurity) transport in fusion plasmas is still lacking. It can either be described as a fully diffusive process, or as a combination of an effective diffusion and an inward pinch, resulting in a transport equation of the form ΓHe = −DHe ∇nHe + VHe nHe , (2.24) with DHe and VHe the flux surface averaged diffusion coefficient and pinch velocity respectively. Both approaches have been used for ITER modeling. In [54], only the diffusive term is taken into account, using the 1.5D BALDUR code with both the MMM95 and the Mixed Bohm/gyro-Bohm transport models to obtain the diffusion coefficients for the core transport. Three different pedestal models provided the boundary conditions at the top of the pedestal, while the helium density at the edge was calculated using Zeff and a specified impurity (Beryllium) content. The resulting helium profiles are only slightly peaked, and the effect of sawtooth oscillations on the central helium density is rather limited. The sensitivity study shows that the central helium density is strongly dependent on the helium fraction at the edge. The approach of combining diffusion with a pinch has been taken in [55]. Using PTRANSP (predictive TRANSP [56]) and calculating the helium profile for different values of DHe and VHe , it was found that the dependance of the central helium density on the recycling increases with the inward pinch velocity, as would be expected for a boundary value problem. The opposite holds for an outward pinch velocity of course, in which case hollow helium profiles are expected. The resulting predicted power for ITER ranges from 320 MW for the outward pinch (for both large and small diffusivity, independent of the recycling) and 170-240 MW for the case of an inward pinch. This would give values of Q ranging from 4.6 to 6.5. The most advanced predictions of the helium transport in ITER come from (non-) linear gyrokinetic simulations. The results presented in [39] show that the 2.5 Tritium breeding and burn-up fraction 25 prospects for ITER look rather promising. In this case, the helium transport equation reads ∂nα nα + ∇ (−DHe ∇nHe + VHe nHe ) = ∗ , ∂t τsd (2.25) ∗ with τsd the thermalization time for fast alpha particles. The helium concentration in the plasma center was found to be of 5% for a recycling factor Rcyc = 0.97, and for the ITER specification Rcyc = 0.9 this decreased to 1.7%. In case of the ITER reference scenario, a rather flat density profile (i.e. fast core particle transport compared to particle removal from the edge) is expected for global helium concentrations of 2% and higher. In that case the shape of the helium profile is expected to almost exactly match the electron density profile [39]. The cause of the relatively flat helium profile is the fast core transport of helium compared to the time it takes a helium particle to enter the pumping system from the plasma edge. Hence, the total helium concentration is mostly determined by the efficiency of the pumping system [37, 45]. The precise value of ρ that still allows ignition is usually found to range from 5 - 10 [53, 57], depending on the precise model used. Early simulation of ITER performance with the BALDUR 1.5D code give similar results, but also put requirements on the recycling coefficient to obtain the desired values of ρ [50, 52]. Although the precise knowledge of the transport processes involved has greatly increased over the past two decades, the fundamental requirements have not changed much. The ITER physics database still lists a value of ρ between 5 and 10 [24]. 2.5 Tritium breeding and burn-up fraction Tritium is an unstable isotope of hydrogen, with a half life of 12.3 years. Consequently, tritium does not occur naturally on earth, but it can be made by irradiating lithium with the neutrons from the DT fusion reaction. The possible reactions with both lithium isotopes, 6 Li and 7 Li, are given by: 6 7 Li + n → 4 He + T + 4.8 MeV, 4 0 Li + n + 2.5 MeV → He + T + n , (2.26) (2.27) where the neutron n0 released by the 7 Li reaction has less energy than the original neutron. The 6 Li reaction has a large cross section (940 barn) for thermal neutrons, resulting in nearly all of them being captured and breeding a tritium atom once they have slowed down enough. Because of the 4.8 MeV released in this reaction, this can have a substantial positive contribution to the overall power balance of a fusion reactor. The 7 Li reaction is endotherm and has a much lower cross section 26 Chapter 2 Theory (order 0.3 barn for neutron energies above 5 MeV), but does offer the possibility of breeding more than one tritium atom per neutron. Because of the inevitable losses associated with tritium handling and the natural decay, the tritium breeding ratio TBR = tritium bred tritium burnt (2.28) needs to be above one to compensate for this. On top of that, a small surplus is required to obtain a starting inventory for new reactors when they are under construction. The current world production of tritium (mostly from CANDU fission reactors) amounts to several kilograms per year [58], which is at best of the same order as the start-up inventory off a single fusion power plant, depending on the time it takes to extract the tritium from the breeding blanket a Figure 2.6: The cross sections of the relevant tritium breeding and neutron multiplication reactions [59]. The tritium breeding in a fusion reactor is foreseen to take place in a blanket surrounding the plasma, between the plasma facing first wall and the vacuum vessel. Even so, not all neutrons released in the fusion reaction will end up reacting with lithium, due to absorption by structural materials and openings in the blanket for the divertor, heating, diagnostics and fuelling systems. To obtain a TBR > 1, 2.6 Power balance 27 a neutron multiplier is integrated in the blanket, most likely beryllium or lead [59] n + 9 Be + 3 MeV → 2He + n0 + n00 , 0 00 n + Pb + 10 MeV → Pb + n + n . (2.29) (2.30) The cross sections for the lithium, beryllium and lead reactions are plotted in figure 2.6. The achievable TBR depends on the exact design of the reactor and blanket, and on technology that is still under development (ITER will be the first reactor to test tritium breeding blanket modules), but is expected to lie between 1.08 and 1.15 [60, 61, 62, 63]. Whether the achievable TBR values are sufficient depends strongly on the tritium burn-up fraction fb , which is the probability that a tritium atom injected into the plasma will actually fuse. This is defined as the fusion rate (since every fusion reaction consumes one tritium atom) divided by the rate at which tritium is lost from the plasma, which is the sum of the fusion and the loss rates: fb = nD nT hσvi , nD nT hσvi + nT /τp∗ (2.31) with nD and nT the deuterium and tritium densities. The burn-up fraction has implications for the required tritium breeding ratio, because 1/fb is the average number of cycles a tritium atom needs to make through the system before it fuses. The time it takes for a given tritium atom to fuse is the product of cycle time and number of cycles, and if one cycle takes one day, decay losses cannot be neglected because they amount to about 1% over 50 cycles. Of course the 3 He produced in by the decay of tritium can be converted back into tritium by letting it absorb a neutron in the blanket, but this is still a net loss because that neutron can no longer breed tritium from lithium (in other words: it takes two neutrons to make one tritium atom in that case). Part of the tritium might adsorb onto the surfaces of tritium handling systems in places where it would be difficult to recover. This needs to be taken into account in a reactor design to prevent the buildup of a large tritium inventory, which would be unacceptable both from a safety and a TBR point of view. Modelling of the tritium circulation puts the minimum at fb = 0.02 and above 0.05 if the reactor also needs to breed the startup tritium inventory for a new reactor every 3 to 5 years [64]. 2.6 Power balance The following section provides a brief overview of the energy balance of a fusion plasma, the derivation of the burn criterion and stable and unstable operating points, where we will follow the approach taken by Freidberg [7]. Subsequently 28 Chapter 2 Theory the energy balance will be complemented with the helium particle balance, which will lead to a modified version of the burn criterion and closed ignition contours, which is based on work presented in refs. [41, 49, 53, 65]. In order to keep the plasma in a fusion reactor at the required temperature, the heat losses from the plasma need to be balanced by a heat source, a requirement which can be written down in the form of an energy balance Sα + SΩ + Sext = Srad + Sκ , (2.32) with Sα the heating power density from alpha particles, Sext the external heating power density, SΩ the Ohmic heating power density, Srad the power density of radiation losses and Sκ the power density of the conduction losses. Assuming no fast particle losses3 , the alpha power density delivered to the plasma is given by Sα = nD nT hσviEα , (2.33) for a plasma with nD and nT the deuterium and tritium densities, hσvi the reactivity of the plasma and Eα = 3.52MeV the alpha particle energy. In case nD = nT and no impurities, this reduces to Sα = 1 2 n hσviEα . 4 e (2.34) The Ohmic heating power plays a role in case there runs a (large) current in the plasma, for instance in a tokamak. The power density is given by SΩ = ηJ 2 , (2.35) with η the electrical resistivity, which scales with T −3/2 , and J the current density. Since the Ohmic power density is comparatively low at fusion relevant temperatures, we will neglect it from now on. Note however, that Ohmic heating plays an important role during the first phase of the discharge when the plasma temperature is still quite low. The first term on the right hand side of the energy balance is the radiation losses. In the core of the plasma these consist of two parts: Bremsstrahlung and synchrotron radiation. The Bremsstrahlung radiation power density is given by ! X 21/2 e6 Z 2 nj ne Te1/2 , (2.36) SB = 5/2 3/2 3 3 3π ε0 c hme j 3 Some fast particle losses are expected due to imperfections in the magnetic field and interactions with MHD instabilities (mostly Alfvénic modes), but the magnitude of these losses is expected to be a few percent at most [66]. 2.6 Power balance 29 with Z the ion charge, nj the density of ion species j and Te the electron temperature. For future references it is convenient to define ! 1/2 e6 2 CB = . 3/2 3π 5/2 ε30 c3 hme Synchrotron radiation is emitted by charged particles because of the acceleration associated with their gyration around the magnetic field lines. Because of re-absorption and reflection on metallic surfaces, quantification of net radiated synchrotron power requires an involved calculation, although simpler fitting formulas have been derived [67]. The emitted synchrotron power in ITER is negligible compared to the Bremsstrahlung everywhere except in the core of the plasma, where they are approximately equal [68]. However, synchrotron radiation losses increase rapidly with temperature and at higher plasma temperatures can increase to about 20% of the total energy losses from the plasma (or roughly twice the Bremsstrahlung losses) [67]. Nevertheless, because synchrotron losses also depend strongly on the temperature profile, first wall reflectivity and reactor geometry, we have choses to neglect them in our analysis. At the plasma edge (or in case high Z impurities are present) line radiation, either from charge exchange or non-fully ionised atoms, also forms a loss mechanism. However, line radiation from the plasma core is only important during the so-called burn through phase at the beginning of the discharge. Once the plasma reaches fusion relevant temperatures line radiation from the core can be safely neglected for our purposes. Combining equations (2.34), (2.36) and (2.11), and setting Pext = 0, a criterion for ignition can be derived for a 50:50 DT plasma: ne τE = 3T , − CB T 1/2 1 4 hσviEα (2.37) which shows that the product of density and confinement time required for ignition can be expressed as a function of T . In fusion research it is common to multiply both sides by T , to obtain the triple product ne τE T on the left. The triple product is a convenient figure of merit for a reactor because generally speaking the pressure (product of density and temperature) scales with the magnetic field and the energy confinement time scales with machine size, both of which have a big share in the cost of the reactor. Increasing this triple product for a given reactor is therefore a good measure of progress. Figure 2.7 plots the minimum value of the triple product as a function of T , where it can be seen that the Bremsstrahlung has the effect of increasing the minimal ignition temperature, but doesn’t affect the burn equilibria at higher temperature. The curves plotted in figure 2.7 are contours of Ṫ = 0 with the dot denoting differentiation with respect to time (t), but the graph provides no information on the stability of these equilibria. In this case an equilibrium is stable if the second Chapter 2 Theory 30 60 ne τE T (atm s) With Bremsstrahlung ahlung 30 Bremsstr 40 Without 50 20 10 0 0 10 101 T (keV) 102 Figure 2.7: The triple product at ignition plotted against T . The dashed line is the ignition curve without radiation losses, which only deviates significantly at temperatures below 10 keV. 2.7 Burn equilibria 31 derivate of T is negative, so T̈ < 0, and unstable if T̈ > 0, which can be seen in a graph that plots Ṫ as a function of T . To do this, the unknown confinement time needs to be eliminated from the left hand side of the burn equation, to which end a scaling law, such as the IPB98(y,2) scaling [26], introduced in section 2.3.7, can be used. Figure 2.7 plots Ṫ as a function of T for a hypothetical fusion reactor (the ignition experiment presented by Freidberg in table 14.3 on page 520 in ref. [7]), at a density of 1.1 × 1020 m−3 . While it is necessary to choose specific values for the engineering parameters, the graph would look similar for any reactor capable of reaching ignition. For temperatures between 0 and roughly 5 keV Ṫ < 0, meaning that external heating is required to keep the plasma stable at a temperature in that range. Between 5 and 17 keV Ṫ > 0, so once the plasma enters this range the temperature will increase by itself until it reaches 17 keV, since above 17 keV Ṫ < 0 again (without external heating). This also means that the equilibrium at 5 keV is unstable and the one at 17 keV is stable (albeit at a higher fusion power than at 5 keV. Given the characteristics of the equilibria it seems logical to refer to the unstable equilibrium at lower temperature as the ignition point, since from that point onward the plasma will sustain itself and the external heating can be switched off. The second, stable, equilibrium at higher temperature can be seen as the burn point, since this is the operating point the plasma will converge to in the absence of external control. 2.7 Burn equilibria The assumption of a pure DT plasma is useful to understand the power balance in a fusion plasma and how this translates to ignition and stable burn, but is not self-consistent because it neglects the alpha particles produced by the fusion reaction. Since these form the energy source that keeps the plasma at the required temperature, a proper treatment of fusion plasmas needs to include the helium ash. In this section we will introduce the model used by Reiter et al. and Rebhan et al. [41, 49, 53, 65], which includes the helium concentration self-consistently and also allows a (fixed) concentration of impurity ions. Using the particle confinement time as defined above, the helium balance in the plasma reads nD nT hσvi = nα , τp (2.38) with nα the helium density. Given the fact that the densities of the different ion species are coupled through the electron density, it is convenient to define a Chapter 2 Theory 32 dilution parameter fi = ni = 1 − ZfZ − 2fα , ne (2.39) ni = nD + nT being the fuel density, fα = nα /ne and fZ = nZ /ne where nZ is the impurity density. The total particle density ntot then is ntot = ne + ni + nα + nZ = ne ftot , (2.40) with ftot = 1 + fi + fα + fZ = 2 − (Z − 1)fZ − fα being the total particle fraction. Unless stated otherwise, we will assume nD = nT from this point onwards, so ne = ni + 2nα + ZnZ . (2.41) Using these definitions, equation (2.38) can be written as ne τE = 4fα ρfi2 hσvi (2.42) 6ftot T , − 4Rrad (T ) (2.43) and equation (2.37) changes to ne τE = fi2 hσviEα with Rrad = fi RB,1 + fα RB,2 + fZ RB,Z 2 √ 1 4 Z = CB T fi gff + 4fα gff + Z 2 fZ . T T T (2.44) (2.45) These are the Bremsstrahlung losses written in a slightly different form, with gff (Z 2 /T ) the Gaunt factor. √ With respect to equation 2.4 in [41], we have used the approximation gff = 2 3/π from Wesson [69] and neglected any line radiation losses from impurities (which make only a minor contribution for low Z impurities with temperatures between 5 and 100 keV). We have now two expressions for ne τE , so combining equations (2.42) and (2.43) and substituting fi = 1 − ZfZ − 2fα results in a cubic expression for fα : a0 + a1 fα + a2 fα2 + a3 fα3 = 0, (2.46) 2.8 Reactor studies 33 with 3 a0 = − T fZ3 (Z 2 − Z 3 ) + fZ2 (4Z 2 − 2Z) + fZ (1 − 5Z) + 2 , 2 3 2 a1 = − T fZ (4Z − 5Z 2 ) + fZ (14Z − 4) − 9 2 Eα 2 2 + fZ Z − 2fZ Z + 1 ρ 4 + [(fZ Z − 1)RB,1 (T ) − fZ RB,Z (T )] , ρhσvi 4Eα 3 [fZ Z − 1] a2 = − T [fZ (4 − 8Z) + 12] + 2 ρ 4 + [2RB,1 (T ) − RB,2 (T )] ρhσvi (2.47a) (2.47b) (2.47c) and a3 = 6T + 4Eα . ρ (2.47d) Equation 2.46 can be solved to obtain the helium concentration as a function of T. This result can then be inserted into equation 2.42 to plot the burn contours in the ne τE , T -plane, something that we will come back to in section 3.2.3. 2.8 Reactor studies Even though construction of ITER, the first fusion reactor with Q > 10, has not yet been finished, most ITER partners are already developing preliminary reactor designs for a commercial fusion power plant. For the European countries this is the power plant conceptual study (PPCS) [70, 71]. The PPCS includes five designs for a demonstration reactor, labelled PPCS A, AB, B, C and D. They differ from each other in the maturity of their technology, and are anticipated to be representative of the first three or four generation power plants. Their main design parameters are summarised in table 2.2 The PPCS A, AB and B designs rely on materials and technology that is currently being developed, and expect an improvement in plasma parameters, mainly density and pressure, of about 20% over the ITER values. For PPCS models C and D, advanced materials need to be developed and another gain of about 20% in plasma performance is required. In the remainder of this thesis the reactor designs from the PPCS will be used to investigate the burn equilibria and their stability. 34 Chapter 2 Theory Table 2.2: Proposed parameters for the PPCS reactors. The plant efficiency is defined as the ratio between net power and fusion power, with net power the electric power delivered to the grid. Adapted from [70]. Parameter Major radius (m) Minor radius (m) Aspect ratio Bφ (T) Ip (MA) Avg. ne (1020 m−3 ) ne /nG βN (thermal/total) H98 Zeff Pfus (GW) Blanket gain Pnet (GW) Padd (MW) Plant efficiency Model A 9.55 3.18 3 7.0 30.5 1.1 1.2 2.8/3.5 1.2 2.5 5.00 1.15 1.55 246 0.31 Model AB 9.56 3.19 3 6.7 30.0 1.05 1.2 2.7/3.5 1.2 2.6 4.29 1.18 1.50 257 0.35 Model B 8.6 2.87 3 6.9 28.0 1.2 1.2 2.7/3.4 1.2 2.7 3.60 1.39 1.33 270 0.37 Model C 7.5 2.5 3 6.0 20.1 1.2 1.5 3.4/4.0 1.3 2.2 3.41 1.17 1.45 112 0.42 Model D 6.1 2.03 3 5.6 14.1 1.4 1.5 3.7/4.5 1.2 1.6 2.53 1.17 1.53 71 0.6 2.9 Stellarators 2.9 35 Stellarators The stellarator is an alternative type of fusion reactor, which, like the tokamak, relies on magnetic confinement in a toroidal geometry. Contrary to tokamaks, which are (quasi-) axisymmetric, stellarators generally have a complicated magnetic geometry. As a measure of the helicity, or pitch, of the field lines, the rotational transform ι is used for stellarators, which is related to the safety factor q in a tokamak in the following way: ι= q . 2π (2.48) The magnetic field in a stellarator is completely generated by external coils and it therefore has no plasma current. This makes it an inherently steady state design, with the added benefit that it is immune to disruptions. The absence of disruptions also means that there are no hard operational limits like the β and density limits in a tokamak. Although there are (MHD) instabilities in stellarators, they result in a degrading of the confinement, which brings the plasma back to a stable regime. Due to the complicated structure of the magnetic field, stellarators always suffered from worse confinement than tokamaks. However, recently the construction of the first optimised, large scale, stellarator has been completed in Greifswald, Germany [72]. If successful, this might lead to a stellarator reactor design, dubbed HELIAS [73]. The transport processes in stellarators are the same as in tokamaks and consequently, the same approach using a scaling law for the energy confinement time is used [74]: τE = 0.134a2.28 R0.64 B 0.84 ι0.41 ne0.54 P −0.61 , (2.49) where ι takes the place of the plasma current that is present in the scaling laws for tokamaks. Apart from being inherently stable, stellarators have the benefit of requiring no external heating for current drive, thus having a lower recirculating power fraction than a tokamak of comparable size. Although stellarators will not be treated explicitly in the remainder of this thesis, the analyses performed are also valid for stellarators since the scaling law for τE has the same form. 36 37 Chapter 3 Burn Equilibria 3.1 Introduction When designing a power-producing fusion reactor, one needs to know how it will operate exactly. Necessarily, future fusion reactors are thought to operate at, or very close to, ignition, where the alpha particles provide essentially all heating power, otherwise they would never succeed in generating electricity at a competitive cost. However obvious this may be, it doesn’t say anything about the composition and temperature of the plasma in operating point. Yet this is crucial information for maximising performance within the operational and material limits of the reactor. Furthermore, knowledge of the stability and sensitivity to changes in plasma and machine parameters of the operating point is important for burn control purposes. A good starting point for the analysis of the operating space in a reactor is a global 0D model based on the energy balance of the plasma. The model can then be expanded by including the particle balances of deuterium, tritium and helium to investigate the effects of fuel burn up and ash accumulation. This has been done ad hoc, by assuming the helium density is a certain fraction of the fuel density, and self-consistently where the helium density is calculated using a particle confinement time which is proportional to the energy confinement time, a method which was first introduced by Reiter et al. [41]. Furthermore the effect of profile shaping was investigated in a follow-up paper by Reiter et al. [49], in which the shaping factors were defined by the volume averaged value of the corresponding variables, and it was shown that a change in one of the profiles is equivalent to a translation of the system in the ne , T plane. Later, Rebhan et al. resolved the discrepancy introduced by the different definitions of the confinement time (including or excluding radiation losses) [53], and investigated the burn stability of the old ITER design using the ITER89P L-mode scaling to eliminate the confinement time from the equations [65]. 38 Chapter 3 Burn equilibria Since an analysis based on the ITER98 H-mode scaling is lacking and this scaling forms the basis for most reactor studies, it seems prudent to carry one out. Building on the work of Rebhan et al. [65] we will derive an analytical relationship between the density and temperature in a burning plasma and compare the obtained burn contours to those based on the ITER89P L-mode scaling. Furthermore, we will use this expression to investigate and comment on the sensitivity of the system with respect to changes in energy and particle confinement. Due to the nature of the scaling laws, the solutions that we find for the operating contours extend over many orders of magnitude in electron density. Obviously, solutions outside the density range on which the scaling law is based should be treated with extreme caution, and solutions far outside this range have no physical meaning whatsoever. In a real tokamak there are several mechanisms that will limit the density of the operating contours to more reasonable levels. The high density regime is not accessible because of the Greenwald and Troyon limits. On the low density side, there are at least three things to consider. At high temperatures and low densities, synchrotron radiation will become the dominant loss mechanism, because it scales linearly with density, as opposed to the Bremsstrahlung and fusion power which have a quadratic density dependence. We have neglected synchrotron radiation losses because at power plant relevant densities they have only a minor effect on the energy balance. Then there is the LH-transition, which has a power threshold which scales roughly linearly in ne [75], meaning that for very low fusion power the plasma will not enter H-mode. And finally, the alpha slowing down time depends on the density, and cannot become too high, otherwise the alpha particles will be lost before they can transfer their energy. The relevant time scale here is τE and not τp , because alpha particles cannot be recycled at the edge while maintaining their energy. For consistency with earlier work and because we are interested in the shape of the solutions, we have chosen to plot the full operating contours in many cases. However, when discussing the effect of different parameters on fusion power, we have chosen an electron density ne = 1 × 1020 m−3 , or selected a reactor relevant density range. Because all current reactor designs feature some amount of external heating, we will investigate the effect this has on the operating contours. We will then extend this analyses to include the effect of impurities and look at the sensitivity of the net electric power to the external heating power. The chapter will conclude with the effect of uncertainties in the confinement scaling on the operating contours and fusion power. This chapter is a synthesis of two papers, which are complementary to each other. Section 3.2 is included here as published in Nuclear Fusion, while an adapted version of section 3.3 will be submitted to Nuclear Fusion. 3.2 Burning plasmas 3.2 Burning plasmas1 ; 3.2.1 Introduction 39 In a fusion reactor of the type tokamak, a plasma of the hydrogen isotopes deuterium and tritium is kept at a temperature of hundreds of millions Kelvin, confined in a toroidal geometry by means of magnetic fields. To start the reactor, external heating is applied to bring the plasma to the ‘ignition’ point: a combination of sufficiently high temperature and density at which the heating power delivered by the fusion reactions balances the heat loss. Once the plasma is ignited its temperature – and thereby the fusion power – increases autonomously until the stable ‘burn temperature’ (Tburn ) is reached. Above this temperature the heat losses increases faster than the fusion power. For a given Tburn , the electron density ne is therefore determined by the reaction rate [4] and heat loss rate, which is the sum of the radiation and conduction losses. The latter are conveniently expressed by the energy W , defined as the ratio of the kinetic energy content W confinement time τE = Pcond of the hot plasma and the conductive power losses Pcond . As the heat loss is the result of complex turbulent processes, empirical scaling laws are used which express τE as a function of operational parameters such as the geometry of the reactor, ne and heating power. There are but a few global parameters under operator control that influence Tburn and might be used for burn control. Important ones are ne , the mixing ratio of the two fuel components deuterium (D) and tritium(T) and the quality of confinement, expressed by the H-factor H98 = τE /τIPB98(y,2) , i.e. the value of τE compared to the scaling law prediction. A fourth and less obvious burn control parameter is the ratio of particle and energy transport τp ρ= (3.1) τE In tokamak reactors, particle confinement is much better than energy confinement, with ρ typically between 5 and 10 [34], with values of 10-30 also reported [76]. The paradox of the fusion reactor is that whereas good energy confinement is essential to reduce power losses, good particle confinement makes the reaction choke on its own ash. The effect of particle confinement on the burn equilibrium is evident from the contours in the ne τE , T -plane (assuming T = Te = Ti ) for which the fusion power heating balances the losses, an analysis already performed by Reiter et al. [41]. Note that whereas the contours are open towards high energy confinement when the choking effect of particle confinement is neglected (ρ = 0), taking this effect into account closes and constricts the contours. For ρ > 14.7 no sustained burn is possible. To complicate matters, a further constraint comes from the fuel cycle, which requires ρ to be sufficiently high as was shown by Sawan et al. [64]. 1 Published as: Jakobs, M.A., Lopes Cardozo, N.L.C. and Jaspers, R.J.E., Fusion burn equilibria sensitive to the ratio between energy and helium transport, Nucl. Fusion 54 (2014) 122005 Chapter 3 Burn equilibria 40 1017 1016 ρ= 3 ne τE (s/cm ) ρ=0 ρ=1 ρ= ρ= 5 9 13 1015 1014 10 100 T (keV) Figure 3.1: The plasma operating contours (POPCON [77] in the ne τE , T -plane for different values of ρ. For ρ = 0 the curve is open because there is no helium to choke the reaction. For increasing values of ρ the operating range in T and ne τE becomes more and more limited, until it vanishes for ρ = 14.71. 3.2 Burning plasmas 41 For these reasons, we address the question how the Tburn and Pburn change under variation of H98 and ρ, as well as ne , while assuming that the fuel mix is 50-50. We first introduce the basic equations and definitions of the 0D-model, following the work of Freidberg [7], Reiter et al. [41] and Rebhan et al. [53, 65]. Although inclusion of profile effects will quantitatively change the analysis, Reiter et al. [49] showed that the qualitative properties of the system remain the same. We then present an expression for the ne (T ), construct universal burn contours and derive two new results for the influence of energy and particle confinement on the burn equilibrium. 3.2.2 Theory The energy balance of a burning fusion plasma is approximated by Sα = Srad + Sκ , (3.2) with Sα the alpha particle heating and Srad and Sκ the losses due to radiation and conduction, respectively. External and Ohmic heating are neglected as they have a minor influence on the burn equilibrium. The alpha power density is given by Sα = nD nT hσviEα , where hσvi is the DT-reactivity [4], Eα = 3.52 MeV the energy of the alpha particle that is produced in the DT-reation, while nj denotes the number density of species j in units of m−3 . The dominant radiation loss P 1/2 is due to the Bremsstrahlung, given by SB = j CB Z 2 nj ne gff Te,keV with CB = 5.35×10−37 Wm3 , Te,keV the electron temperature√in keV, Z the ion charge, gff the Gaunt factor which we have approximated with 2 3/π and the summation is over all ion species. To account for the helium density (nα ) resulting from the fusion reactions we write nD nT hσvi = nα /τp , thereby assuming that the confinement of alpha particles is the same as that of other species. We further introduce the fuel dilution parameter fi = (nD + nT )/ne and the alpha fraction fα = nα /ne . Using these notations, ref [41] finds ne τE ne τE = = 4fα ρfi2 hσvi 6ftot T , fi2 hσviEα −4Rrad (T ) (3.3) (3.4) and by solving these equations arrives at burn contours, i.e. the contours in the ne τE , T -plane for which the alpha heating balances the losses. 3.2.3 Results We have used the same formalism to produce the burn contours in shown in figure 3.1. Note that ne τE must exceed a critical value for burn to occur. For given ne τE there are two solutions: the unstable ignition temperature (left hand branch) Chapter 3 Burn equilibria 42 and the stable burn temperature (right hand branch), in agreement with the intuitive picture of ignition and burn. In this calculation τE is an independent parameter, whereas in fact it depends on plasma parameters. Rebhan et. al [65] proposed a self-consistent analysis by using a scaling law which expresses τ̃E = W/(Pcond + Prad ) as a function of plasma parameters. They used the ITER89 L-mode scaling [29] to find burn contours in the ne , T -plane for this specific scaling, for a specific choice of reactor parameters. We follow this approach, using the now more relevant scaling for Hmode confinement, the IPB98(y,2) -scaling [26], which is commonly used to predict the performance of future fusion devices. Since the radiation losses are not included in IPB98(y,2) , the method applied in ref [65] cannot be used. Instead, we inserted the expression for the alpha heating power to eliminate the confinement time and derived an expression for ne as a function of T which is valid for all scaling laws of the form τE = KAk nle P −m : ne = 4fα Ak ρK 1 1−2m+l 1 2 f hσvi 4 i m−1 1−2m+l m (Eα V ) 1−2m+l . (3.5) Here K is a constant that depends on the engineering parameters of the reactor, A is the average ion mass in amu and P the power deposited in the plasma (by the alpha particle or external sources). Since K and the plasma volume V are the only reactor specific parameters in this equation, ne (V −m K)1/(1−2m+l) represents a normalised density that is the same for all fusion reactors that follow the same scaling law. Figure 3.2 shows the POCPONs for ITER ID [78] and 3 conceptual reactor designs PPCS models A - C as described in the conceptual power plant study [70]. There is a large difference between the IPB98(y,2) and ITER89P scaling for the ITER ID reactor. The solid curves for ITER ID and PPCS models A - C are isomorphic, which can be shown by applying the normalisation described above. It is important to note that while the formalism using the scaling laws leads to burn equilibria at values of ne and T that are far from the normal operating conditions of a fusion reactor, these are probably artefacts due to the mathematical form of the scaling laws. Reliable extrapolations can only be made in the parameter range where the database on which the scaling laws are based is well populated, i.e. with ne in the range 1019 to 1021 m−3 . To elucidate the role of particle confinement, while zooming in on the reactor relevant ne range, figure 3.3 shows the burn contours of PPCS model A in the ne , T plane. The plot shows clearly how, at constant density, the reactor will move from ignition at temperature of 5 to 8 keV to burn at a temperature around 30 keV, while the fusion power at the same time increases by an order of magnitude. The fusion power at ignition and burn depends quadratically on ne , and therefore the Greenwald density limit nG = Ip /πa2 is of fundamental importance. For all but the lowest ρ-values this limit is more restrictive than the Troyon pressure limit 3.2 Burning plasmas 43 1020 Density limit ne (m −3 ) 1021 PPCS A PPCS B PPCS C ITE R ID 1022 β li mit ITER ID 89L 1019 4 5 6 7 8 9 10 20 30 40 50 T (keV) Figure 3.2: The POPCONS for ρ = 5 for the ITER ID [78] and the PPCS A, B and C designs [70]. The contours for the PPCS reactors are made with the IPB98(y,2) scaling, as is the solid blue ITER ID curve, whereas the dashed ITER ID contour is created using the 89L-scaling following the procedure developed in [65]. The Greenwald density limit and the Troyon β limit for the ITER ID design are plotted. They also give a good indication of these limits for the PPCS designs, although the precise position is reactor specific. 2G W 10 ρ=0 nG 1G 20 W -3 ne (m ) 500 200 100 1019 ρ=5 W ρ = 10 5G ρ = 14.71 Chapter 3 Burn equilibria 44 4 5 β lim it MW MW MW 6 7 8 9 10 20 30 40 50 T (keV) Figure 3.3: The operating contours for ρ = 0, 5, 10 and 14.71 for the PPCS A design made with the IPB98(y,2) scaling. The colour indicates the fusion power in each operating point and the dashed dotted lines indicate lines of constant fusion power. The Greenwald density limit and the Troyon β limit are also indicated. 3.2 Burning plasmas I 45 p given by βmax = gT aB with Ip in MA and the Troyon factor gT = 0.03. Calculating the β limit requires knowledge about the composition of the plasma and that is only available on the equilibrium contours, so plotting it is not straightforward. We have taken the following approach: for figs. 3.2 and 3.3 we have calculated the β limit for a pure DT plasma, which results in a underestimation of ne of at most 20%. For figs. 3.4 and 3.5 we have taken the value of fα and T at the equilibrium and used that to calculate the value of ne and subsequently the fusion power at the β limit. This results in a β limit that has two values a one value of ρ and H98 (because every point on an equilibrium contour is associated with a point on a β limit contour). Figure 3.4 shows that the Tburn , and therefore Pburn , will change under variation of ρ. In other words, if by some process in the plasma or in the exhaust the ratio between particle and energy confinement changes, this will significantly affect the output power of the reactor. This may be a point of concern as it may lead to unwanted excursions of Pburn , but may also have potential as actuator for burn control. The dependence of fusion power (at burn equilibrium) on ρ is depicted in figure 3.4 for PPCS models A, B and C, for ne = 1 × 1020 m−3 . These curves show a lower (ignition) and upper (burn) branch that meet at ρmax , the highest value of ρ that can be tolerated at this particular density. For PPCS model A ρmax ≈ ρcrit , the fundamental limit on ρ set by the Bremsstrahlung and fusion cross section for the DT-reaction. For PPCS models B and C ρmax < ρcrit . Along most of the upper branches, i.e. in the burn equilibria for 5 < ρ < 10, Pburn is approximately inversely proportional to ρ. Note that part of the high power branch is not accessible because it exceeds the β limit, but the expected impact for future reactor designs is minimal because constraints on the achievable tritium breeding ratio will most likely set a lower limit of ρ = 5 [64]. In another projection of the parameter space, the influence of H98 on Pburn can be analyzed. Figure 3.5 displays Pburn as a function of H98 for ρ = 5, 10 and ne = 1 × 1020 m−3 , for PPCS model A. We see that below H98 = 0.73 and 0.83 for ρ = 5 and 10 , respectively, there is no ignition because the confinement is too low. For the stable burn branch (top half of the contour), increasing H98 first results in a steeply increasing Pburn until a maximum is reached at H98 = 1.1 to 1.3, depending on ρ. A further increase in H98 will lower Pburn because the increased helium content due to the better confinement chokes the fusion reaction. For ρ . 9, some parts of the high power branch exceed the β limit, which needs to be taken into account when choosing the operating point for a reactor. On the unstable ignition branch (bottom half of the contour), an increase in H98 beyond its minimum value initially results in ignition at lower temperature and power. Both branches meet again at the maximum H98 = 6.38 and 2.79 (ρ = 5 and 10) and beyond that there are no more burn equilibria, i.e. the fuel has become so diluted that the fusion power can no longer balance the conduction and radiation losses. Chapter 3 Burn equilibria 46 104 Fusion power (MW) β limit β limit PP C β limit PP CS PP CS C SA B 103 5 10 15 ρ Figure 3.4: The fusion power at a constant electron density ne = 1 × 1020 m−3 as a function of ρ for PPCS models A, B and C. It can be clearly seen that there above the critical value ρ = 14.7 there are no burn equilibria for PPCS model A. For PPCS models B and C, the maximum value of ρ is lower because the burn contours are shifted towards higher densities with respect to PPCS model A. Below the maximum value of ρ there are two equilibria: the unstable ignition branch at lower fusion power and the stable burn branch at higher power. The latter is especially sensitive to changes in ρ, but can exceed the β limit (dash dotted lines) for low value of ρ (the dotted part of the curves). The PPCS D design does not ignite in the current model below ne ≈ 1 × 1020 m−3 . 3.2 Burning plasmas 47 Fusion power (MW) 104 it im βl 103 β lim it ρ = 10 ρ=5 0.5 1 1.5 2 2.5 3 3.5 4 H98 Figure 3.5: The fusion power as a function of the H-factor H98 = τE /τE,98 for PPCS model A at a density of ne = 1 × 1020 m−3 . While it is not surprising that a too large reduction in the H-factor will lead to a loss of burn, the result when H98 increases deserves more attention. The fusion power initially increases but reaches a maximum around H98 = 1.3 for ρ = 5 and at H98 = 1.14 for ρ = 10, beyond which it drops until at some point the burn equilibria vanish (H98 = 6.4 and H98 = 2.8 respectively. The dash-dotted lines indicate the β limit for the two contours, and at low values of ρ this is the limiting factor of the fusion power (dotted part of the contour). Chapter 3 Burn equilibria 48 In short, for a given reactor there is no gain to be expected from improvement of energy confinement. Either the plasma exceeds the β limit, or the power output decreases. Rather, the reactor should be designed in such a way that its operating point is at Hmax , provided it does not conflict with the β limit. Of course, better confinement does allow one to reach ignition in a reactor with smaller dimensions and lower Pburn . 3.2.4 Conclusions We have derived an analytical expression (3.5) relating T and ne in a fusion reactor with self-consistent treatment of fuel burn up and helium accumulation, using the IPB98(y,2) scaling law for confinement time. This expression is valid for all fusion reactors that obey the same energy confinement scaling if one takes into account a scale factor that depends on the reactor parameters. Using these results we have plotted the burn contours of the PPCS A design in the ne , T plane, including the curves of constant fusion power and the Greenwald and Troyon limits. The fusion power at these equilibria was found to be very sensitive to changes in ne , ρ and H98 . The fusion power scales quadratically with the density around the Greenwald density, although this will be different for reactors that have a minimum density for ignition that is close to this limit. The dependence on ρ is especially strong for intermediate to high values of ρ, and since the value of ρ is to a large extent determined by the helium exhaust at the plasma edge, this offers possibilities for burn control using helium pumping [79, 80]. The effect of H98 on the fusion power could have implications for advanced tokamak scenarios where values of H98 well above 1 are expected. 3.3 Burn equilibria with impurities and Pext 3.3 Burn equilibria with impurities and Pext 3.3.1 Introduction 49 The analysis in section 3.2 deals exclusively with equilibria in burning plasmas without external heating, which is unlikely to be achieved in a tokamak because of the need for non-inductive current drive. Furthermore, we have only looked at ’pure’ plasmas so far, containing only deuterium, tritium and helium. A real fusion plasma always contains a non-zero amount of impurities, be it beryllium or tungsten from the reactor wall, or for example neon or nitrogen to increase the radiated power in the divertor. This section therefore investigates the burn equilibria in PPCS model A with external heating. First we derive a way to determine the minimum and maximum temperature on a POPCON, after which we present a procedure to determine the helium fraction in a plasma with external heating. Then we analyse the operating contours with external heating, and describe the effect of impurities on the operating contours and the power output at a given operating point. Finally, we study the effect that external heating has on the net electrical power delivered to the grid. 3.3.2 Temperature domain of a burning plasma It turns out that we do not need to solve equation (2.46) completely to be able to say something about the solution. In fact, we can already determine the allowed temperature domain of a burning plasma by looking only at the determinant ∆ = 18a3 a2 a1 a0 − 4a32 a0 + a22 a21 − 4a3 a31 − 27a23 a20 , (3.6) which is only a function of ρ and T . Equation (2.46) has three real solutions if ∆ > 0, two real solutions of which one is a multiple root if ∆ = 0 and one real and two imaginary roots for ∆ < 0. The solutions we are looking for have to satisfy 0 < fα < 0.5, because fα = 0.5 corresponds to a pure helium plasma and fα = 0 is only possible in case τp = ρτE = 0, which for finite ρ would mean that ne τE = 0 which has no physical relevance. It turns out that there is a real root fα > 0.5 for 1 keV < T < 1000 keV that we can discard. Because we are looking for physically meaningful solutions (fα has to be real), this restricts us to the temperature domain where ∆ ≥ 0. In figure 3.6 the discriminant is plotted for several values of ρ in the domain 1 keV < T < 1000 keV. Below 4 keV ∆ > 0, but the two real roots here are both negative and thus have no physical meaning. This leaves us with ∆ > 0 in a temperature window ranging from 5 to several hundred keV, depending on the value of ρ, where the solutions Chapter 3 Burn equilibria 50 2 ∆/∆peak 1 0 102 ρ = 0.1 101 ρ=1 −2 100 ρ=5 9 ρ= 3 ρ=1 −1 103 T (keV) Figure 3.6: The discriminant ∆ of equation (2.46) plotted as a function of T for different values of ρ. The values are normalised to the peaks around 10 keV. 3.3 Burn equilibria with impurities and Pext 51 satisfy 0 < fα < 0.5. Thus, the second and third zero crossing in figure 3.6 are the lower limit Tmin and upper limit Tmax of the accessible temperature window, and by definition the two roots coincide at these points. 3.3.3 Helium fraction with external heating The derivation of a cubic equation for the helium fraction by means of equations (2.42) and (2.43) no longer works for a plasma with external heating because the product of ne and Pext shows up in the resulting equation. Instead we can divide both equation (2.42) and (2.43) by ne and equate the two expressions for τE that we obtain that way. Solving for fα again yields a cubic equation 0 = a0 + a1 fα + a2 fα2 + a3 fα3 , (3.7) with coefficients 3 3 2 3 2 2 a0 = − T fZ (Z − Z ) + fZ (4Z − 2Z) + fZ (1 − 5Z) + 2 n2e , 2 3 a1 = (− T fZ2 (4Z − 5Z 2 ) + fZ (14Z − 4) − 9 2 Eα 2 2 + fZ Z − 2fZ Z + 1 ρ 4 Sext + (fZ Z − 1)RB,1 (T ) − fZ RB,Z (T ) + 2 )n2e , ρhσvi ne 4Eα 8RB,1 (T ) 1 3 (fZ Z − 1) − , a2 = 2 − T (fZ (4 − 8Z) + 12) + ne 2 ρ ρhσvi Eα n2e , a3 = 6T + 4 ρ (3.8a) (3.8b) (3.8c) (3.8d) with Sext the external power density. These differ from the coefficients in equation (2.47) by a factor of n2e and have an extra term 4Sext /ρhσvi in a1 . In this case only one of the roots satisfies 0 ≤ fα ≤ 0.5, and the solution is fα = − a2 + C + 3a3 ∆0 C (3.9) , with the discriminant ∆ defined by equation (3.6) and ∆0 = a22 − 3a1 a3 , ∆1 = C= 2a32 (3.10) 27a0 a23 , !(1/3) − 9a1 a2 a3 + p ∆1 + −27a23 ∆ 2 (3.11) . (3.12) 52 Chapter 3 Burn equilibria All terms on the right hand side of equation (3.9) are either a free parameter or a function of T , so we have the desired expression for fα . The catch is that from equation (3.9) alone we cannot tell what the equilibrium temperature is. One possible solution is to use the obtained value for fα to compute Pfus and insert this in the scaling law to compute τE . This value of τE can then be equated to the value of τE obtained from equation (2.42). Solving for T then yields the desired equilibrium temperature. This cannot be done analytically, but is easily achieved numerically and this procedure is much faster than finding the equilibrium by solving the energy and particle balance simultaneously for fα and T . 3.3.4 Burn equilibria with external heating The operating contours presented so far have been obtained for a plasma without external heating. This provides insight in the operational space of an ignited plasma, but future reactors are designed to operate with some external heating for current drive, and possibly control purposes. This leaves us with density, temperature, ρ and Pext as parameters, of which three can be chosen ’freely’ (respecting machine and plasma limits of course). First, we want to know how the addition of external heating changes the operating contours that were obtained previously. Figure 3.7 shows this for the PPCS A reactor design with different levels of external heating. The solid lines represent the operating contours for ρ = 5 (blue) and ρ = 10 (red), and the dashed lines represent operating contours with different levels of external heating. As expected, the equilibria with external heating almost coincide with the contours obtained previously at high densities, because in this region the external heating is insignificant compared to the fusion power. These equilibria are well above the Greenwald density or the β limit and therefore of little meaning for reactor design, except for a reactor so large that the operational contours are shifted towards lower densities. The equilibria at low density and intermediate temperature (between 5 and 20 keV) have disappeared for reasonable amounts of external heating for PPCS model A, because there the external heating outweighs the power losses from the plasma. This is of little consequence, since these equilibria are mainly a mathematical artefact and are not relevant for realistic reactor designs because of the (extremely) low densities. When following a burn contour toward lower temperatures, the curves with external heating start to deviate and they converge on a minimum density that is only a function of external heating and independent of ρ. This density is determined by the balance between radiation losses and external heating because there is little alpha heating at these low temperatures. After reaching this (local) minimum in the density, the curve will turn upwards again when the temperature decreases even further. Since our model does not 53 ρ=5 ρ=5 ρ = 10 1022 ρ = 10 3.3 Burn equilibria with impurities and Pext 1021 ne (m−3 ) 200 100 MW MW nG 50 MW 1020 β lim it 10 MW 20 0M 10 1019 W 0M 50 M W W 10 1018 4 5 6 7 8 9 10 20 T (keV) 30 40 M W 50 Figure 3.7: POPCON plot for PPCS model A with fixed external heating power for ρ = 5 ( ) and ρ = 10 ( ). The solid lines represent contours without external heating. For high densities, the curves almost coincide with the burn curves because the external heating is only an insignificant fraction of the fusion power. The equilibria at low densities have disappeared because in that case both the conduction and radiation losses are very small and outweighed by the external heating. On the low temperature side, the equilibrium is mainly determined by the radiation losses because there is hardly any fusion power due to the low cross section at these temperatures. On the high temperature side, the equilibrium is also set by the balance between radiation losses and external heating, but the curve has a downward slope because an increase in temperature has to be compensated by a reduction in density to maintain equilibrium. Chapter 3 Burn equilibria 54 include (impurity) line radiation and Ohmic heating, the results for temperatures below a few keV should be treated with extreme caution. On the high temperature (stable) branch, the external heating curves also deviate from the burn contours for decreasing temperature. In this case the curves trend towards an asymptote determined by the degradation √ of confinement with increasing power and the increase in radiation losses with T . In this temperature range, around 70 keV, the reactivity is almost temperature independent and the fusion power is only sensitive to fuel density. From a reactor perspective, the interesting area are the stable operating points around the Greenwald density and just below the β-limit. For a given electron density, adding external heating shifts the operating points to a higher temperature (and consequently higher fusion power). The shift becomes larger for increasing values of ρ, and causes the curves of different values of ρ to converge on the same asymptote. This means that depending on the precise design of the reactor and the corresponding positioning of the operating contours in the ne , T -plane, either ρ, the external heating power Pext , or a combination of both could serve as an actuator for burn control. Finally, adding external heating changes the sign of dne /dT : when the curve starts to deviate from the operating contour without external heating, it enters a regime where the temperature increases with decreasing density on the stable branch (and vice versa on the unstable branch). This needs to be taken into account when designing a controller for the reactor. In contrast to the PPCS A, B and C designs, the ITER reactor is not anticipated to ignited. Hence there is always a non-zero amount of external heating required, and the operating contours look very different from those for the PPCS A reactor, which is illustrated in figure 3.8. 3.3.5 Impurities Although all expressions presented so far allow for the presence of impurities in the plasma, the previous sections have neglected their effect on the operating contours. This is of course not realistic and therefore this section presents operating contours for several impurities at different concentrations in the plasma. It turns out that the main effect of impurities is a contraction of the operating contours for the reactor, and therefore a reduction in fusion power on the stable burn branch, and an increase in fusion power on the unstable ignition branch. The maximum allowable value of ρ is lowered. Every impurity has a critical concentration above which the burn can no longer sustain itself. The fusion power at this critical concentration is not zero however, because at a given temperature and density, the radiation losses are not zero. Low and high Z impurities have a different effect, because the Bremsstrahlung losses scale with Zeff which is a quadratic function of Z, whereas the fuel dilution 3.3 Burn equilibria with impurities and Pext 2 55 ·1020 nG 100 70 50 30 ne (m−3 ) 1.5 β -l im it 1 20 0.5 10 30 5 5 50 70 100 20 10 15 T (keV) 20 25 30 100 Figure 3.8: POPCON plot for ITER at ρ = 5 with constant levels of external heating (labels are in MW). There are no equilibria for Pext = 0, i.e. ITER does not ignite in our model. Chapter 3 Burn equilibria 56 scales linearly in Z. However, this is only a minor effect compared to the overal contraction of the burn contours. 1045 ρ=5 1% Be 1040 W 0.5% Ne ρ = 10 1% Be 1030 0. 0 1% 2% Be 1025 W 1% N ne (m−3 ) 2% Be % 1035 N 01 0. 1% 0.5% Ne nG 1020 β limit 1015 1010 4 5 6 7 8 9 10 20 T (keV) 30 40 50 Figure 3.9: PPCS model A POPCON plots for plasmas containing different impurities (beryllium, nitrogen, neon and tungsten) at different concentrations. The presence of impurities dilutes the fuel and enhances the radiation losses by increasing Zeff . The result is a large reduction in the accessible density and temperature range for a given value of ρ, and consequently the maximum value of ρ is lowered. Note that although low Z and high Z impurities affect the plasma in a different way (because the dilution is linear in Z whereas Zeff scales with Z 2 ), this is only a minor effect compared to the reduction in operating range. 3.3 Burn equilibria with impurities and Pext 57 Of more importance to a particular reactor design is the effect of impurities in the plasma on the fusion power and energy gain factor Q. Figure 3.10 shows the fusion power and helium fraction (left plot), and Q and Zeff as a function of the beryllium fraction fBe = nBe /ne in the plasma, for ρ = 5 and 10 in PPCS model A at a density of ne = 1 × 1020 m−3 with 246 MW of external heating. The fusion power drops rapidly with increasing values of fBe , as do fα and Q. Although the curves for different ρ values of converge on each other, the relative effect is fairly similar, with a decrease 22% and 23% at Zeff = 1.5 respectively. It is clear that the PPCS estimate of Zeff ≈ 2.5 yields unacceptable results for the fusion power if helium and beryllium are the only impurities. 10 ρ ff, Ze 10 = 5 ρ= ff, Ze Qρ 30 2.5 10 us ,ρ 20 2 = ,ρ 2.5 5 fα Pf 0.1 ,ρ= Zeff fα Q 5 fα 5 = 5 = s,ρ Pfus (GW) 3 40 0.15 P fu 7.5 0.2 = 10 0.05 1.5 10 Q ρ= 10 0 0 0.05 0.1 fBe 0.15 0 0 0.05 0.1 0.15 fBe Figure 3.10: The fusion power and helium concentration (left plot), Q and Zeff (right plot) for PPCS model A, with 246 MW of external heating, as a function of ) and 10 ( ) for the high temperature beryllium concentration fBe for ρ = 5 ( equilibrium at ne = 1 × 1020 m−3 . The relative decrease of fusion power is more or less independent of the value of ρ. Even a moderate value of Zeff = 1.5 already leads to a decrease in fusion power of 22% (ρ = 5) and 23% (ρ = 10) . The impurity content in ITER or future reactor plasmas cannot be know precisely of course, but JET reported Zeff = 1.2 in the core plasma during divertor operation with the ITER like wall [81]. This was in a divertor plasma without significant fusion power and therefore no helium content. Neglecting the tungsten contribution (so assuming a ’pure’ D-Be plasma), this amounts to a beryllium concentration fBe = 0.017, which corresponds to a 17% decrease in fusion power for PPCS model A relative to a plasma without beryllium pollution. The erosion rates reported in [81] are probably on the low side for a fusion power plant, because of the presence of fast He particles impacting on the first wall. Chapter 3 Burn equilibria 58 The effect of tungsten in the plasma on the fusion power is much weaker than that of beryllium for similar values of Zeff , as can be seen in figure 3.11, which displays the fusion power and helium fraction fα on the left, and Q and Zeff on the right, as a function of the tungsten concentration in the plasma. Both JET and ASDEX report tungsten concentrations up to 10−4 , but we have taken fW = 3 × 10−4 as the upper limit to obtain Zeff values that are foreseen in the PPCS. 0.2 10 40 3 Pfus,ρ= 5 30 Pfus,ρ=10 5 0.15 2.5 20 10 2.5 10 fα,ρ=5 0 1 2 fW 3 −4 ·10 0.1 0 2 Qρ=10 ρ= ff, Ze 5 1.5 ρ= ff, Ze 0 Zeff 10 fα Q Pfus (GW) 7.5 0 Qρ=5 fα,ρ= 1 2 fW 3 −4 ·10 Figure 3.11: The fusion power and helium concentration (left plot), Q and Zeff (right plot) for PPCS model A, with 246 MW of external heating, as a function of tungsten concentration fW for ρ = 5 ( ) and 10 ( ). Also for tungsten, the relative decrease of fusion power is more or less independent of the value of ρ, but the effect is much weaker for the same value of Zeff than for beryllium. Whereas for beryllium the relative decrease in fusion power is 22% at Zeff = 1.5, the corresponding reduction in power for tungsten is only 2%. This is caused by the quadratic dependence of Zeff on Z, and the quadratic dependence of the fusion power on the fuel density, an effect which was already reported in [82]. The atomic number of beryllium is 4, and it takes quite a lot of it to obtain the same value of Zeff compared to tungsten with an atomic number of 74, for which a small concentration already leads to a rather large change in Zeff . Please note that the effect of high Z impurities is not limited to fuel dilution and an increase in Zeff which results in higher Bremsstrahlung losses. They will most likely not ionise completely and therefore emit line radiation, which can have a non-negligible effect on the (local) energy balance in the plasma. The resulting temperature decrease might very well result in a significant loss in fusion power, but a complete treatment is outside the scope of this thesis. 3.3 Burn equilibria with impurities and Pext 59 So far we have looked at the effect of impurities on the fusion power in isolation, but most likely a change in particle confinement time because of a change in ρ will be accompanied by a change in impurity content (assuming the impurity source stays the same of course). Figure 3.12 shows contours of constant Pfus in the fBe , ρ-plane, as well as contours of constant beryllium influx SBe (dark blue lines). 10 5,0 00 4,0 00 6,0 8 00 5,0 00 7,0 00 02 6 6,0 0.0 ρ 8,0 0.001 00 4 9,0 0.0 00 00 05 7,0 00 8,0 00 11, 000 12, 000 1 10, 0.0 000 9,0 00 2 0.02 0 0.5 1 1.5 2 2.5 fBe 3 3.5 0.05 4 4.5 5 −2 ·10 Figure 3.12: Contour plot of the fusion power (coloured lines) in MW for PPCS model A with Pext = 246 MW as a function of the beryllium concentration fBe and ρ, at a density of ne = 1 × 1020 . The impurity concentration is of course also a function of ρ, and to indicate the effect of this coupling the contours of constant beryllium influx are also plotted (dark blue lines). The labels denote particles per second per m3 in units of 1020 . If the impurity source stays constant during a change in ρ, the plasma will move parallel to one the SBe = constant contours. This would have a significant effect on the fusion power: for instance, if ρ increases from 8 to 9, and the original beryllium concentration was 2%, the fusion power decreases with roughly 1 GW, or 18%. Of course it is unlikely that the impurity influx will stay constant when the fusion power changes by such a large fraction, but the example nevertheless shows Chapter 3 Burn equilibria 60 just how sensitive the fusion power output is to small changes in ρ. 3.3.6 Power output with external heating and impurities In section 3.3.4 we analysed the effect of external heating power on the position of the burn equilibria in operating space. Obviously, when the equilibrium changes, so does the fusion power, and figure 3.13 displays Pfus (left y-axis) and the derivative dPfus /dPext (right y-axis) as a function of Pext for ρ = 5 and 10 (blue and red curves respectively) for fBe = 0 and fBe = 0.025 for the PPCS A design at ne = 1020 m−3 . ·104 1 20 Pfus,fBe =0 s f dP fu t, dP ex 15 0. 0 25 0.5 dP dP fus , ext f Be= 0.25 dPfus dPext , fBe=0 0 0 50 10 Pfus,fBe =0 dPfus dPext Pfus,fBe =0.025 Be= Pfus (MW) 0.75 Pfus,fBe =0.025 0 5 dPfus dPext , fBe=0.025 100 150 Pext (MW) 200 0 Figure 3.13: The fusion power Pfus (left axis) and its derivative with respect to the external heating power dPfus /dPext (right axis) for PPCS model A at ne = ) and ρ = 10 ( ), with fBe = 0 and 0.025. The fusion 1×1020 m−3 for ρ = 5 ( power increases with Pext , but this effect decreases for increasing Pext . Equilibria at higher ρ and fZ are more sensitive to Pext than those are lower ρ and fZ values, but this difference also decreases for higher values of Pext , both in absolute and in relative terms. With the addition of Pext the fusion power increases, and this effect is stronger for higher values of ρ and for higher impurity concentrations. For ρ = 10 and fBe = 0.025 the fusion power increases by 1.8 GW when Pext goes from 0 to 246 MW, while for ρ = 5 and FBe = 0 the change is only 878 MW. The sensitivity of Pfus to Pext decreases for higher values of Pext , and this effect 3.3 Burn equilibria with impurities and Pext 61 is stronger, both in relative and absolute terms, for higher values of ρ and fZ : the value of dPfus /dPext for ρ = 10 and fBe = 0.025 at Pext = 246 MW is only 16% of what it is at Pext = 0 MW, whereas for ρ = 5 and FBe = 0 the corresponding ratio is 70%. So for higher values of ρ and fZ external heating becomes a more effective tool to increase the fusion power output of the reactor. But the ultimate goal is not the fusion power, but the power delivered to the grid. 3.3.7 The effect of Pext on net electric output The net (electric) power delivered to the grid, Pnet , is determined by the (overal) thermal efficiency η of the plant, and the conversion efficiency ξ of the external heating power. (Obviously the plant also uses a significant amount of power for the cryostat, coolant pumps, and a host of other auxiliary systems, but their power consumption is relatively insensitive to the fusion power and we will neglect it for now.) For the net power we can write Pnet = PE − Pext , ξ (3.13) with the electrical power PE given by PE = η (Pfus + Pext ) . (3.14) Here we have assumed that the external heating power is delivered to the plasma with 100% efficiency. Combining these two expressions, we get Pnet = ηPfus − (1 − ξ) Pext . ξ (3.15) On the face of it, any amount of external heating will reduce the power delivered to the grid, but this is only true if Pfus is independent of Pext . This is not the case, and figure 3.13 shows that dPfus /dPext > 0 for reactor relevant levels of Pext , so an increase in Pext also results in an increase in Pfus . To determine what happens to Pnet if Pext changes, we take the derivative of equation (3.15) with respect to Pext dPnet dPfus 1−ξ =η − . dPext dPext ξ (3.16) If dPnet /dPext is greater than zero, an increase in Pext will result in an increase in Pnet . Some algebra allows us to transform this into a condition on dPfus /dPext : dPfus 1−ξ > . dPext ηξ (3.17) 62 Chapter 3 Burn equilibria Using η = 0.31 as anticipated for PPCS model A [71], and ξ = 0.35 (reference [71] gives a value of ξ = 0.6, but this is based on negative ion source neutral beam injection, which seems highly unlikely for a commercial fusion reactor.) the condition is dPfus /dPext > 6.0. In figure 3.13 it can be seen that for ρ = 5 this means that the addition of any amount of external heating leads to a reduction in Pnet , while for ρ = 10 the limit lies somewhat above 100 MW, depending on the impurity concentration in the plasma. In the analysis above we have neglected the power gain from tritium breeding in the blanket, which changes the numbers but not the argument. If we assume all tritium breeding is done by the 6 Li + n → 4 He + T + 4.8 MeV reaction, the total thermal power Pth is given by 4.8 Pth = 1 + TBR Pfus , (3.18) 17.6 with TBR the tritium breeding ratio, which has a maximum value of approximately 1.15 [60, 61, 62, 63]. We can absorb this power multiplication factor in η and see that it relaxes the requirement on dPfus /dPext by roughly 30%. So depending on the values of η and ξ that can be achieved, adding external heating may be positive or detrimental to the overall performance of the reactor. Having said that, it seems unlikely that the foreseen level of external heating for the first generation of fusion power plants is optimal from a cost of electricity point of view. 3.3.8 Uncertainties in scaling laws The scaling laws for the energy confinement time are based on a database containing several thousand experiments, and a scaling law is just a fit through the datapoints, not an exact representation. The coefficients of the different parameters in the scaling law come with an uncertainty, which is also apparent in the differences between the respective scaling laws in the ITER physics basis. These are all based on different subsets of the same data, sometimes with different physics restrictions. The variation in predicted energy confinement time is of the order of one second, or roughly 25%, and the 95% log-nonlinear confidence interval for the IPB98(y,2) scaling law is 3.5 - 8.0 s for ITER [26]. It stands to reason therefore, that a small change in one of the exponents should only have a small effect on the burn contours. This is indeed the case for changes to the exponents of all parameters, except ne and P . The burn contours are very sensitive to variations in the scaling of τE with ne and P , as is shown in figure 3.14. Reducing the P exponent compresses the burn contours along the density axis, and increasing the exponent of P results in a stretching. The reverse is true for changes in the exponent of ne . Already changes as small as one percent result in a change of several decades in the predicted density range for the burn. The temperature range remains unaffected as explained before. 3.3 Burn equilibria with impurities and Pext 63 1050 9 05 0.4 ne 1045 P −0.6969 ρ= 1040 5 ne0.414 1 1 P P −0.6969 ne (m−3 ) 1035 1030 683 −0 . 40 0. ne 59 ρ= 5 ne0.4141 P −0.6831 1025 nG 1020 β limit 1015 4 5 6 7 8 9 10 20 T (keV) 30 40 50 Figure 3.14: Operating contours for PPCS model A for ρ = 5 and ρ = 10 applying the normal IPB98(y,2) scaling, and the IPB98(y,2) scaling with a ± 1% change in the exponents of ne and P respectively. These relatively small changes to the scaling law result in comparatively large changes in the operating contours. The accessible temperature range stays the same, since this is only a function of the fusion and radiation cross sections combined with the transport modelling using a confinement time, but the density range is heavily affected. 64 Chapter 3 Burn equilibria This strong sensitivity of the operating contours to minor variations in the energy confinement time scaling are clearly undesirable, because the exact position of the operating point and the corresponding fusion power output have a major impact on reactor design. Further investigation of this phenomenon is outside of the scope of this section and will be addressed in chapter 5. 3.4 Discussion and conclusions Future fusion reactors will to a large extent rely on alpha particles to provide the necessary heating. If a fixed ratio between the energy and particle confinement time is assumed, it can be shown that the resulting equilibria form closed contours in the ne τE , T -plane [41]. We have derived an analytical expression for the temperature range that is spanned by such contours, based on the discriminant of the cubic equation for the helium fraction in the plasma. This result is valid for all reactors (not only tokamaks), for which the energy and particle confinement time have a fixed ratio ρ. Following the approach presented in [65], we also derived an expression for the electron density as a function of temperature for reactors that follow the IPB98(y,2) scaling law. Compared to the results presented in [65], the operating contours obtained with the IPB98(y,2) scaling are shifted towards higher densities in the ne , T -plane. This is because the electron density of the equilibria is extremely sensitive to changes in the exponents of ne and P in the scaling law, something that we will come back to in chapter 5. Please note that to mimic H-mode confinement, the confinement time predicted by the ITER89P scaling is modified by an H-mode factor fH , which in this case is taken to be fH = 2. The contours obtained using the IPB98(y,2) scaling also have a shape that differs from the ones found using the ITER89P L-mode scaling. This is because the radiation losses are not included in the IPB98(y,2) scaling, as opposed to the ITER89P L-mode scaling that does include them. An explicit treatment of the radiation losses is thus required, and this does away with the artificial ’radiation limit’ [53], eliminating the ’bump’ that is present on contours obtained using a scaling law for τ̃E . The use of scaling laws to eliminate τE from the burn equations results in burn contours that span many orders of magnitude in density. Care has to be taken when interpreting equilibria outside of the density range on which the scaling laws are based, and often the contours extend to densities that have no physical meaning. This is purely a mathematical artefact, which originates in the form of the equations. Burn contours in the ne τE , T -plane are universal since they only depend on the reactivity and the radiation cross sections, but this property appears to be lost when they are transformed to the ne , T -plane. This can be resolved by dividing the expression for ne by the reactor specific terms in the scaling law for τE , resulting in 3.4 Discussion and conclusions 65 isomorphic burn contours that are valid for all reactors. The normalisation factor is arbitrary, and the resulting isomorphic burn contours still extend over several orders of magnitude in normalised electron density, depending on the scaling law for τE . The main benefit of this discovery lies in the fact that a sensitivity analysis of the burn equilibria is also valid for all reactors: the results can simply be scaled by the ratio of the reactor specific part of τE . We performed a sensitivity analysis of the fusion power with respect to the three parameters that are under operator control in a burning plasma ne , ρ and H98 . The fusion power scales quadratically with electron density, apart from the region around the minimum and maximum operating density on a contour. This strong dependence makes the density a powerful actuator for control of the fusion power. However, due to the desired operation close to the density limit the actual achievable variation in density might be too small for practical use. The fusion power is also very sensitive to H98 around the minimum and maximum values of H98 . In between, the power output reaches a maximum (and minimum) and the different PPCS designs project H98 right at the value where these extrema lie, effectively disqualifying the energy confinement as a control tool. Also the precise value of ρ has a strong influence on the fusion power on the burn branch, the sensitivity on the ignition branch is rather weak. Although acting on the slowest time scale of the three parameters that were investigated, ρ might be the most suitable actuator for power control in a fusion reactor operating on the burn branch since it can be varied over a wide range without the risk of crossing a stability limit. One needs to take the tritium breeding requirements into account when considering this approach however, as values of ρ < 5 might result in a too low tritium burn up fraction [64]. When including external heating in the model, several things change. For reactors capable of ignition and high (≈ 25 keV) or low (<5 keV) temperatures, the operating contours converge to a single contour that depends only on the amount of external heating and not on the value of ρ. For intermediate temperatures, the low density solutions have disappeared, because they feature such low fusion powers that for realistic values of Pext the system is completely dominated by the external heating. Only the high density solutions remain, dominated by the alpha heating and consequently the curves with external heating almost coincide with the original operating contours. In between the intermediate and the low and high temperature regimes there is a smooth transition. The fact that contours with different values of ρ but the same level of Pext converge to the same curve means that depending on the reactor design and choice of operating point, either ρ, Pext or a combination of both could be used as actuators for control. The inclusion of impurities in the system results in reduction of accessible temperature and density range. On top of that, the maximum allowable value of ρ decreases for increasing impurity concentrations. There is a small difference 66 Chapter 3 Burn equilibria between low and high Z impurities, due to the different scaling of fuel dilution and conduction and radiation losses with Z. This changes the shape of the operating contours, but it is insignificant compared to the contraction of the contours. When looking at the effect of impurities on fusion power, the difference between high and low Z impurities becomes more pronounced. Because the required concentration of low Z impurities is much higher for a given Zeff , they dilute the fuel much more than high Z impurities. Since Pfus scales with the fuel density squared, the corresponding effect on Pfus is even larger. The effect of increasing fZ is larger for lower values of ρ, although this difference is smaller for high Z than it is for low Z impurities. Most likely a change in ρ will be accompanied by a change in the impurity content of the plasma, and the fusion power is very sensitive to this. This sensitivity increases for higher impurity concentrations. The addition of external heating has a positive effect on the fusion power output of the reactor, and Pfus is more sensitive to Pext for higher values of ρ and higher impurity fractions. Whether the addition of Pext is beneficial for the net electricity production of the plant depends on the thermal and heating efficiency of the reactor, but a reduction in external heating power would most likely result in a larger Pnet for the PPCS A design. Finally, we looked at the sensitivity of the operating contours to small changes in the scaling laws for the energy confinement time. The predicted value of τE at the ITER operating points varies relatively little between the different scalings in the ITER physics basis and the predictions are robust against small changes in the exponents of the individual scaling laws [26], but this is not the case for the operating contours when small changes are made to the exponents of ne and P . Even changes of 1% already result in the contours being stretched or compressed along the density axis by many orders of magnitude. Although the corresponding shifts in operating points around the Greenwald density and just below the β limit are far less severe, we consider this an unphysical and also unwanted effect. A more detailed investigation of this phenomenon can be found in chapter 5. 67 Chapter 4 Burn stability 4.1 Introduction When designing a fusion reactor that is capable of (assisted) ignition, it is essential to know the nature of the operating point. What will the plasma do when left alone. Will it wander off to some unknown destination in phase space? Or will it remain happily where it is. And if shaken or rattled by some external event, how will it respond? In other words: is the operating point stable and if so, what is the stability radius? In the current understanding of a burning plasma two equilibria are identified: an unstable one at low temperature and a stable equillibrium at high temperature [7, 65]. For a simple DT plasma without conduction losses this follows from the fact that the reactivity has a maximum, whereas the radiation losses increase monotonically with temperature. There are therefore two temperatures at which the radiation losses are equal to the alpha heating power from the fusion reaction. Because the energy balance contains only cross sections that follow from atomic physics, the temperatures at these equilibria are independent of engineering parameters and can be determined from √ hσvi = CB T . (4.1) The addition of conduction losses to the energy balance introduces a reactor dependence in the system, but in general two equilibria remain. Again the one at low temperature is unstable and the other is stable. Irrespective of the reactor design, some general observations can be made. For instance, the temperature of the stable equilibrium decreases drastically because at high temperatures conduction is the dominant energy loss mechanism (because of turbulent and, to a lesser extent, neoclassical transport), reducing the importance of radiation losses. 68 Chapter 4 Burn stability The unstable (low) temperature equilibrium is far less affected because in that temperature range the radiation losses dominate. If the conduction losses are higher than the alpha heating power for all temperatures (for instance because the reactor is too small), there will be no equilibria. For a reactor with critical size, the alpha heating power will balance the losses at only one point, and this equilibrium necessarily will be unstable. More precisely: it will be a saddle point, with dT /dt < 0 on both sides of the equilibrium. In a pure DT plasma, assuming fD = fT , the temperature is the only variable (with the density being the free parameter). It is trivial to determine the stability of the system, because the only eigenvalue of such a system is −J, the Jacobian. While easy to analyse, it does not accurately represent a burning plasma because the fusion reaction produces helium which has a finite residence time in the plasma. The helium particles change the equilibrium by diluting the fuel, increasing radiation losses and even affecting the conduction losses. So any realistic description of a burning plasma needs to take this into account. Linear stability of the burn point is not the only consideration for the design of a future fusion reactor. Of even greater importance is the convergence radius of the equilibrium, or the size of the basin of attraction. The burn equations are highly non-linear and although the Poincare-Lyapunov theorem states that a nonlinear system is stable in a region around an equilibrium of the linearised system, the theorem does not say anything about the size of this region. The questions this chapter aims to answer are the following: how does the plasma respond to a disturbance of the burn equilibrium? Does it return to the original equilibrium, or does it find a new one (possibly at T = 0)? And in doing so, does it cross any operational limits? And what happens to the fusion power during these excursions? Does the system have bifurcations and associated limit cycles? In short: can we rely on the plasma to regulate the burn by itself, or does it need to be controlled by external means? First we will present the system of differential equations that describe a burning plasma, and subsequently introduce a reduced system (with only two variables) which is able to reproduce the most important dynamics. We will study the stability of this system as a function of the free parameters by deriving an analytical expression for the Jacobian and applying planar bifurcation theory [83]. Using these results we will investigate the effect of changes in the τE scaling on the stability and compare different PPCS designs, both with and without external heating and reflect on the implications for operating point selection and control requirements. Then we will return to the full system, derive the Jacobian, present the stability of the system, and compare the respective reactor designs. 4.2 Theory 4.2 Theory 4.2.1 Burn equations 69 The burn dynamics can be described by four coupled, non-linear ordinary differential equations (ODEs). Rebhan and Vieth [65] performed a first analysis of this system and determined the linear stability by deriving the Jacobian and evaluating its eigenvalues for the equilibria that they found using the ITER98P L-mode scaling. In their analysis the effect of helium ash accumulation and fuel dilution on the ion mass was omitted, resulting in a symmetry in the fD and fT terms in the Jacobian matrix. This symmetry enables a reduction of the number of dimensions by developing the Jacobian with respect to the first row or column, which yields ñD = −ñT to be an eigenfunction of the system. However, since a complete analysis should take the mass dependence into account we did include this term, breaking the symmetry and requiring a numerical treatment of the system. The 0D model of a burning plasma features the variables nD , nT , nα and T [65], and their evolution in time is governed by 4 differential equations: dnD dt dnT dt dnα dt dT dt nD , τp nT = sT − nD nT hσvi − , τp nα = nD nT hσvi − , τp 3 2(sD + sT )T nD nT T T Srad hσvi Eα + T − − = 3 + − 3 . 2 τ τ ntot E p 2 ntot 2 ntot = sD − nD nT hσvi − (4.2a) (4.2b) (4.2c) (4.2d) Here hσvi is the reactivity of the plasma as a function of T , for which we have taken the Bosch and Hale parametrisation [4], Eα = 3.52 MeV the energy of an alpha particle, Srad denotes the radiation losses due to Bremsstrahlung and the particle and energy confinement time are related through τp = ρτE , with ρ a free parameter with a typical value in the range 5-10, although values up to 30 have been reported [42, 37, 43, 44]. Equation (4.2d) was first derived from the energy balance by Rebhan et. al. [65] using dW/dt = d/dt(3/2ntot T ) and solving for dT /dt. Consequently, equation (4.2d) contains some terms with an interpretation that may not be immediately obvious. Their meaning is most easily understood when considering that a change in internal energy can be brought about by a change of temperature (at constant density) or a change of density (at constant temperature). A change in temperature is then equal to the total change in internal energy minus the change in internal energy due to a change in density, divided by 3/2 times the total density (note that the Boltzmann constant is absorbed into the temperature definition). Chapter 4 Burn stability 70 The nD nT hσviT /ntot term denotes the temperature change due to the change in density when two particles fuse. Since a fusion reaction doesn’t exchange energy with the environment, the process has to be adiabatic (apart from the energy gained from the change in binding energy obviously), so the kinetic energy of the reacting particles has to be retained in the plasma. The T /τp term compensates for the difference between particle and energy transport. In case τp = τE particles are lost at the same rate as their energy, so there is no effect on temperature. In other cases the change in temperature is equal to the difference between particle and energy transport, with τp > τE in general. Of course the use of energy and particle confinement times neglects the close relationship between particle and energy transport, but this is unavoidable in such a simple model. Finally the 2(sD + sT )T /ntot term represents the cooling effect of refuelling. In this case the fuel is assumed to be at zero kelvin, which is a good approximation in case of pellet fuelling and gas puffing. Only in the case of NBI heating does the new fuel have significant energy, but we will not include this effect in our analysis. The electron density in the plasma is given by ne = nD + nT + 2nα + ZnZ , (4.3) ntot = 2nD + 2nT + 3nα + (Z + 1)nZ , (4.4) the total particle density and the average ion mass A= 2nD + 3nT + 4nα + mZ nZ . nD + nT + nα + nZ (4.5) This burning plasma model implicitly assumes that dnZ /dt = 0, which is of course most likely not the case in a fusion reactor. For our purposes is suffices though, since we are mainly interested in the burn dynamics of the system, and not in plasma wall interaction or impurity seeding, which are expected to be the main sources of impurities in future reactors. From here onwards we will assume nZ = 0 for simplicity unless explicitly stated otherwise. However, relaxing this assumption does neither change the analysis nor the conclusions. The above system is four-dimensional, and its properties are determined by the three free parameters sD , sT and ρ. Although the stability of the system can easily be determined by calculating the eigenvalues of the linearised system, it is complicated to investigate the behaviour of a four-dimensional system around the equilibria and the dynamics of the system in general. In the next section we will therefore present a reduced system that has only two degrees of freedom and is determined by two parameters. This will allow the application of planar stability theory, and we can study the transition between different stability regions using bifurcation theory. 4.2 Theory 4.2.2 71 Stability of a two-dimensional system Determining the stability of a non-linear system is generally not trivial. However, the Poincare-Lyapunov theorem says that if a linearised system is asymptotically stable in a certain point, the non-linear system will be stable in that point too. Hence we will focus on the stability of the linearised system. Determining the stability of a linear system is most easily achieved by determining the Jacobian matrix J, and subsequently calculating its eigenvalues at the equilibria. Positive eigenvalues correspond to an unstable equilibrium and negative eigenvalues to a stable point. The eigenvalues of any (linear) system of equations can be determined by solving det(J − λI) = 0. (4.6) In the two-dimensional case this can be written as (j11 − λ) (j22 − λ) − j12 j21 = 0. (4.7) Defining p = j11 + j22 and q = j11 j22 − j12 j21 transforms this to λ2 − pλ + q = 0, (4.8) so that the solutions are λ1,2 = √ 1 p± ∆ , 2 (4.9) with ∆ = p2 − 4q [84]. Depending on the value of ∆ the following cases can be distinguished: i ∆ > 0: λ1 and λ2 are real and distinct, ii ∆ = 0: λ1 and λ2 are real and equal, and iii ∆ < 0: λ1 and λ2 are complex conjugates. We can use p and q to determine the signs of λi , i = 1, 2, and thus the stability of the equilibrium. The different possibilities, unstable (US), stable (S) and asymptotically stable (AS) are summarised in table 4.1. In case both p and q are zero, λ1 = λ2 = 0 and the stability of the system is unknown (indicated with UK in table 4.1). However, by looking at the different elements of J we can still obtain information about the stability. If J = ( 00 00 ) is the system stable, and if J 6= ( 00 00 ) it is unstable. 72 Chapter 4 Burn stability Table 4.1: Stability properties of a two-dimensional system as a function of the elements of the Jacobian matrix J, with p = j11 + j22 and q = j11 j22 − j12 j21 . p>0 p=0 p<0 4.2.3 q>0 q=0 q<0 US S AS US UK S US US US Bifurcation theory In order to fully appreciate the changes in behaviour between different regions in the stability diagram of a burning plasma, we will briefly introduce the accompanying bifurcations. We will encounter them again in section 4.3.4. The definition of a bifurcation is ẗhe division of something into two branches or parts,̈ or, in the study of dynamical systems, a sudden change in the qualitative or topological structure of the system. This abrupt change is brought about by a small, continuous change in the bifurcation parameter(s) of the system. From bifurcation theory we know that there can be local bifurcations in the system at points where one or both eigenvalues has a real part that is equal to zero [83]. This property can be exploited when looking for bifurcations: we will find a single zero eigenvalue at points where q = 0 (λ1 = p and λ2 = 0), and a double zero eigenvalue when p = q = 0 (λ1,2 = 0). In case p = 0 and q > 0 there are two completely imaginary eigenvalues, and if p = 0 and q < 0 both eigenvalues are real. The type of bifurcation depends on the nature of the eigenvalues at the bifurcation point. For a single zero eigenvalue, the result is a saddle-node bifurcation, also known as a fold, or limit point bifurcation [83, 85]. On one side of the bifurcation there are two equilibria: a saddle point (two real eigenvalues: one positive and one negative), and either a source (an unstable equilibrium) or a sink (a stable equilibrium). At the bifurcation, these two equilibria meet and annihilate each other, and on the other side of the bifurcation they have disappeared. At the point where both eigenvalues are completely imaginary, the system goes through a (Poincaré–Andronov–) Hopf bifurcation [83, 86], which signifies the birth of a limit cycle. The bifurcation can be either supercritical, in which case it is an attracting (stable) limit cycle, or subcritical if the limit cycle is unstable (repelling). A Bogdanov-Takens bifurcation occurs at the point where λ1,2 = 0 [83, 87]. Near this point the system has two equilibria, a saddle and a non-saddle (a source or a sink), which annihilate in a saddle-node bifurcation. An Andronov-Hopf bifurcation generates a limit cycle at the non-saddle equilibrium and finally a 4.3 Reduced system 73 non-local bifurcation called a saddle-homoclinic bifurcation also originates at this point. The saddle-homoclinic bifurcation occurs when a limit cycle collides with a saddle point and connects it with itself, which does not depend on the local value of the eigenvalues, but instead is determined by the global properties of the system [83]. There are several other bifurcations that can occur in systems with two or more dimensions, but they are of little relevance in our case. 4.3 Reduced system 4.3.1 Derivation There are several ways of reducing the dimensionality of the system. The most logical reduction follows from the observation that the fusion power output has a maximum at, or very close to, nD = nT , so it is reasonable to take this as a property of the system. Because the transport of deuterium and tritium in our model is exactly the same, this also means that sD = sT , hence this assumption reduces the number of parameters by one. A further reduction in dimensionality is facilitated by a coupling between ni and nα , which can be achieved by fixing the electron density at a constant value. Although this is not possible in reality because it would require an instantaneous feedback system on the particle sources sD and sT , it enables us to isolate the effect of helium accumulation on the burn dynamics at a constant density and it has the added effect of making the system a bit less complicated. The other possible relation between ni and nα would be through Zeff , but this would also imply a coupling with T because in equilibrium fα = f (T ). Combining the condition nD = nT with equation (4.3), allows us to express the fuel density ni = nD + nT as ni = ne − 2nα , (4.10) which means we can combine equations (4.2a) and (4.2b) into an expression that contains only one unknown, nα : dni 1 ni = si − n2i hσvi − dt 2 τp 1 ne − 2nα dnα 2 = si − (ne − 2nα ) hσvi − = −2 , 2 τp dt (4.11) where si = 2sD = 2sT . We can derive an expression for the refuelling rate si = −2 dnα 1 ne − 2nα 2 + (ne − 2nα ) hσvi + , dt 2 τp (4.12) Chapter 4 Burn stability 74 using equation (4.2c), which results in si = ne . τp (4.13) This result could have been anticipated by realising that in order to have a constant electron density, every particle that leaves the plasma needs to be replaced. Since the fusion process only affects the number of ions, and leaves the number of electrons unchanged, the only way an electron can leave the plasma is by transport to the wall. The refuelling only needs to compensate for these losses, hence the intuitive form of expression (4.13). Also the expressions for the total particle density and the average ion mass ntot = 2ne − nα , (4.14) 2.5ne − nα ne − nα (4.15) A= take on a simpler form, as do the radiation losses Srad = ne [(ne − 2nα ) Rrad (T, 1) + nα Rrad (T, 2)] = ne (ne + 2nα )Rrad (T, 1) (4.16) where we have made use of the fact that Rrad (T, 2) = 4Rrad (T, 1) Using these results we end up with two equations for the variables nα and T dnα 1 nα 2 = (ne − 2nα ) hσvi − , dt 4 τp 2 (ne − 2nα ) 3 nα T T Srad dT = hσvi Eα + T − − − 3 , dt 6ntot 2 ntot τp τE 2 ntot (4.17) (4.18) where in the latter the refuelling term has been combined with the particle transport term. The properties of the system are determined by the choice of the free parameters ne and ρ. 4.3.2 Jacobian matrix of the reduced system Writing xi = {nα , T }, the burn equations (4.17) and (4.18) can be expressed as dxi = fi (x1 , x2 ) i = 1, 2 dt We can linearise the system around an equilibrium point xj0 as 2 X dxi ∂fi = (xj − xj0 ) , i = 1, 2. dt ∂xj xj0 j=1 (4.19) (4.20) 4.3 Reduced system 75 Determining the linear stability of the system requires that we find the eigenvalues (and eigenvectors) of the Jacobian matrix J, whose elements jij = ∂fi /∂xj are evaluated at the equilibrium points xj0 . Determining J comes with some extensive algebra, and the full derivation can be found in section A.1. Here we will restrict ourselves to the final result: nα ∂τE 1 + 2 , ρτE ρτE ∂nα nα ∂τE (ne − 2nα )2 dhσvi + 2 , = 4 dT ρτE ∂T 3 (ne − 2nα )(2nα − 7ne ) hσvi Eα + T = 6n2tot 2 T 2ne τE nα ∂τE 5n2 Rrad (T, 1) − 2 , − 1 + − e 3 2 2 τE ρntot ρntot ∂nα 2 ntot dhσvi 3 (ne − 2nα )2 3 = + hσvi Eα + T 6ntot 2 dT 2 nα Srad 1 T ∂τE − − +1 − 2 . ρntot τE τE ∂T 3T ntot j11 = (2nα − ne )hσvi − (4.21) j12 (4.22) j21 j22 (4.23) (4.24) Although consisting of long expressions, numerical evaluation of the Jacobian matrix is straightforward and very quick, because of the analytical expressions for all derivatives which makes computationally expensive numerical derivates unnecessary. 4.3.3 Normalisation The density spans a range of several orders of magnitude, and consequently a numerical treatment of the problem is prone to precision and rounding errors. This can be solved by normalising the density, which we have done by expressing the helium density as a fraction of the electron density, fα = nα /ne for the numerical calculations. To a lesser extent the same holds true for the temperature, which can be normalised to the equilibrium temperature Teq to obtain T ∗ . So for the different terms of the Jacobian, we need to make the following Chapter 4 Burn stability 76 substitutions: n α = n e fα , (4.25) ∗ (4.26) T dnα dt ∂ ∂nα dT dt ∂ ∂T = Teq T , dfα = ne , dt 1 ∂ = , ne ∂fα dT ∗ = Teq , dt 1 ∂ = . Teq ∂T ∗ (4.27) (4.28) (4.29) (4.30) With these transformations, the relations between the elements of the Jacobian J and the normalised Jacobian J ∗ take the form ∗ j11 = j11 , ne ∗ j12 = j , Teq 12 Teq ∗ j21 = j , nα 21 ∗ j22 = j22 . 4.3.4 (4.31) (4.32) (4.33) (4.34) Reduced system stability For each combination of ρ and ne there are two equilibria (which possibly coincide), and the elements of the Jacobian matrix depend on the local values of fα and T . If we want to find the zero eigenvalues in the system, we have to solve q = 0 (or p = q = 0), dfα /dt = 0 and dT ∗ /dt = 0 simultaneously. In reference [41] an expression relating fα and T is presented, which can be used to express ne as a function of T for different forms of the τE scaling law [65, 88]. Although an analytic solution to the problem might exist if the inverse function of the relationship ne (T ) can be found, the resulting expression will probably not provide a lot of insight into the physics behind the solution. Alternatively, we can look for eigenvalues equal to zero on an equilibrium plane [T ∗ , fα ](ne , ρ), which is the approach that we have taken. Looking at the shape of the equilibrium contours presented in references [65] and [88], we see that the equilibria form nested, closed contours in the ne , T -plane, with increasing values of ρ resulting in smaller (contracted) curves. The question now is what happens to the stability when we change either ne or ρ. Changing ρ at constant ne means moving parallel to the T -axis, whereas varying ne at constant ρ means moving along a burn contour (black curves in figure 4.1). 4.3 Reduced system 77 In doing so, the system will pass through regions of different stability, which we will discuss in the following paragraphs. Figure 4.1 shows the nature of the equilibria in the reduced system in the ne , T -plane for the PPCS A reactor design [70]. On the low temperature side, the equilibrium is a saddle point having both a stable (green in figure 4.1) and an unstable (red) eigenvector. Moving towards higher temperatures, both eigenvectors first become unstable and subsequently acquire an imaginary part (blue), which introduces oscillatory behaviour. After crossing into a stable oscillatory region (magenta), the system becomes asymptotically stable (two stable eigenvectors), until at the high temperature side a small region of oscillatory stable behaviour is again encountered. λ1 10 λ2 70 ne (m−3 ) 1060 1050 1040 1030 1020 1010 5 10 20 50 T (keV) 100 200 5 10 20 50 T (keV) 100 200 Figure 4.1: The stability of the two eigenvectors of the reduced burn system for PPCS model A [70]. The black lines indicate contours of constant ρ, starting with ρ = 1 on the outside and ending with ρ = 14 for the innermost curve. Unstable means oscillatory unstable behaviour for a given eigenvector is denoted , behaviour, oscillatory stable behaviour and stable behaviour. Starting at the lowest temperature on a contour of constant ρ, we find that the equilibria are saddle points, with a stable and an unstable eigenvector. Following the contour in the clockwise direction, the stable eigenvector also becomes unstable, which happens exactly at the maximum density of the contour. This transition is accompanied by a saddle-node bifurcation, which occurs when a system has one eigenvalue equal to zero. In this case the source point meets the saddle point and they annihilate each other. For constant ρ, the bifurcation parameter is ne and we can understand the 78 Chapter 4 Burn stability physical reason for the bifurcation as follows. When increasing the density, the temperature difference between the stable an unstable equilibria is reduced, until at the maximum density both equilibria coincide. A further increase in ne is not possible without also changing the value of ρ. Further along the contour, the eigenvalues acquire an imaginary part, resulting in unstable oscillating behaviour of the equilibrium (blue points in figure 4.1). The real parts of the eigenvalues now become increasingly smaller, until at some point they become negative, corresponding to a stable oscillatory equilibrium (green points). At the change from unstable to stable oscillations, both eigenvalues are purely imaginary, which means that the system goes through a (Poincaré– Andronov–) Hopf bifurcation. In this case the bifurcation is subcritical, resulting in the birth of an unstable limit cycle. Moving along, the limit cycle increases in size until it degenerates into a homoclinic orbit to the saddle equilibrium at the same density, but lower temperature, where it disappears in a (saddle-)homoclinic bifurcation. This is a global bifurcation, meaning that it does not depend on the local parameters of the system, but instead arises from the properties of the system at different points, contrary to the fold and Hopf bifurcations, which are local bifurcations. Traversing the contour further in the clockwise direction, for ρ ≤ 10.5, the imaginary part of the eigenvalues can decrease to zero yielding a stable equilibrium. Approaching the low density side of the contour, the equilibrium once more enters the stable oscillatory part, before moving into the saddle region at the minimum density on the contour by transitioning either through the unstable oscillatory (ρ ≥ 8) or a stable region (ρ ≤ 8). For 10.5 ≤ ρ ≤ 14.5, the system always has an oscillatory behaviour on the high temperature side, but for ρ ≥ 14 there are no stable equilibria anymore, effectively lowering the upper limit on the accessible value of ρ for a fusion reactor if stable burn without external control is a requirement. At the point where the fold, Hopf and saddle-homoclinic bifurcations meet (close to the ρ = 8 contour at the minimum electron density), the system has a Bogdanov-Takens bifurcation [87]. A Bogdanov-Takens bifurcation occurs when the system has a zero eigenvalue of multiplicity two. There are two nearby equilibria: a saddle point and a node (sink or source), which annihilate via a saddle node bifurcation. The non-saddle equilibrium undergoes a Hopf bifurcation that generates a limit cycle that connects to the saddle point in a saddle-homoclinic bifurcation (which we encountered a little while back). 4.3.5 Physical interpretation Besides looking at the stability properties by themselves, we can also try to understand the physical mechanisms behind the stability or instability of the different eigenvectors. This section will interpret the local stability properties in terms of the physical mechanisms that drive them. Because the parameter space is too 4.3 Reduced system 79 large to be covered in detail, we will discuss the eigenvectors and eigenvalues at different relevant points, all at a density of ne = 1 × 1020 m−3 for the PPCS A design with the IPB98(y,2) scaling for τE . The Jacobian matrix consists of the partial derivatives of the burn equations with respect to the different variables. Evaluating it at an equilibrium point tells us something about the sensitivity of the equilibrium to changes in both variables. Looking at equation (4.7), it is clear that ∗ ∗ j11 < −j22 (4.35) ∗ ∗ ∗ ∗ j11 j22 ≥ j12 j21 . (4.36) is a necessary, but not sufficient condition for stability. The other requirement is So the stability of the system depends on all four elements in J ∗ . ∗ Looking at the physical meaning, the top left term j11 represents the derivative of dfα /dt with respect to fα . A negative value here means that an increase in fα will result in a negative value for dfα /dt, driving the system back to equilibrium. ∗ The bottom right term j22 corresponds to the sensitivity of the rate of change of temperature to changes in temperature, and again, a negative value has a stabilising effect. When taking the top right and bottom left terms into account, the picture becomes more complicated. They describe the change in dfα /dt and dT ∗ /dt, caused by variations in T ∗ and fα respectively. Their effect is stabilising if they have opposite signs, and destabilising in case they have the same sign, but the overall stability is determined by the signs and values of all four elements of J ∗ . Say an increase in helium content causes the temperature to rise. Even though ∗ j11 could be negative and drive the helium concentration back to the equilibrium level, the temperature increase might promote an increase in helium content. If this latter effect is stronger than the former, the system will be unstable. Notice that ∗ this effect can take place even when j22 < 0, it really depends on the magnitude of the different terms. We will take a closer look at the different terms that make up the different elements of the Jacobian matrix, and try to determine on physical grounds whether they will be positive or negative. This will provide insight in the driving mechanisms behind the instabilities that are present in the system. The first element, j11 , consists of three parts. The first (2nα − ne ) hσvi is always negative, because ne ≥ 2nα , and corresponds to the reduction in reaction rate when the fuel dilution increases. The second term 1 − ρτE Chapter 4 Burn stability 80 is always negative by definition. This leaves us with the last part nα ∂τE , ρτE2 ∂nα which is always positive. This can be seen by looking at equation (A.6) and realising that nα < 0.5ne . The second element, j12 , has only two terms. The first can be positive or negative, depending on the sign of dhσvi/dT . Because the maximum of hσvi lies at 67 keV, this term will be positive under reactor relevant conditions. This represents the increase in helium production because of a higher reactivity if the temperature rises. The sign of second term of j12 is also determined by dhσvi/dT (equation (A.7)), but has the opposite sign to the first term, because a higher temperature results in a higher fusion power, which decreases the (particle) confinement time. The third element j21 , turns out to be the term with the largest magnitude for all temperatures at ne = 1 × 1020 m−3 . It also has a negative sign everywhere, meaning that an increase in helium content will result in a decrease of temperature (all else remaining constant). Equation (4.23) has three terms, the first of which is (ne − 2nα )(2nα − 7ne ) 3 hσvi E + T . α 6n2tot 2 This is the derivative of heating power per particle with respect to nα . An increase in helium content leads to lower fuel concentration and hence lower fusion power. The ntot term in the denominator becomes slightly smaller for increasing nα , but since nα < 0.5ne it can easily be seen that this term is always negative by substituting ne − 2nα > ne − ne = 0 and realising that 2nα − 7ne < 0. The second part reads − T ∂τE 2ne T ∂τE 2ne T T ∂τE + − − , ρτE2 ∂nα ntot ρτE2 ∂nα τp n2tot τE2 ∂nα | {z } | {z } | {z } | {z } 1 2 3 4 where we have separated several terms to make their physical meaning clearer. Here part 1 stems from the changes in particle losses due to a change in helium content. This affects the particle losses because the average ion mass is present in the IPB98(y,2) scaling, where τE ∝ A0.19 , so in this case this term is negative. Parts 2 and 3 are the derivative of the cooling term from refuelling with respect to the helium content. Part 2 represents the changes in refuelling that accompany a change in particle confinement, which itself is caused by a change in helium content (which is positive), and part 3 the changes due to a change in ntot when the helium content changes. In the latter case the refuelling itself is not directly affected, but the cooling effect is divided among more or less particles. This term has a minus sign because a higher helium content means a lower number of particles. 4.3 Reduced system 81 Part 4 is due to a change in energy confinement caused by a change in helium content, which is negative because a larger helium content leads to a longer confinement time and hence lower energy losses. Finally, the third term in equation (4.23), − 5n2e Rrad (T, 1) , 3 2 2 ntot represents temperature decrease due to the change in radiation losses, which in2 crease when nα increases due to the Zeff term in the Bremsstrahlung. Summarising j21 consists of three terms, of which the first and the last have a cooling effect. Because the first term is the largest for all equilibria, the fact that the sign of the second term depends on position of the equilibrium doesn’t matter. Finally, j22 is made up of three terms. The first, (ne − 2nα )2 dhσvi 3 3 + hσvi , Eα + T 6ntot 2 dT 2 is positive and corresponds to the temperature increase due to the increased alpha heating for a rising temperature. The second term nα 1 T ∂τE − +1 − 2 ρntot τE τE ∂T looks complicated, but it represents the energy losses because of lower confinement, and a cooling effect associated with the refuelling with cold particles. Determining the sign of this term is not difficult, because equation (A.7) tells us that ∂τE /∂T is negative for reactor relevant temperatures, so this term of j22 is negative. The final term represents the increased radiation losses for higher temperatures, and this obviously has a negative contribution to the derivative. The overall sign of j22 thus depends on whether the increased fusion power outweighs the increased losses and this changes with temperature. 4.3.6 Low temperature stability We can apply the results from the previous section to different equilibria in the phase plane. Starting with the low temperature equilibrium at ρ = 5, which has a temperature of 6.24 keV, we can evaluate the normalised Jacobian matrix: −0.011693 0.000094 ∗ J = . −2.542608 0.158415 The (normalised) eigenvectors for the low temperature equilibrium are v1 = (0.066202, 0.997806) 82 Chapter 4 Burn stability and v2 = (0.000556, 1.000000), with v1 stable and v2 unstable. Although in both eigenvectors the alpha en temperature components have the same sign, the v2 has a much larger temperature component than v1 1 . The temperature component in both v1 and v2 is destabilising, but it is the larger helium component that stabilises v1 . 4.3.7 High temperature stability For the high temperature equilibrium at ρ = 5 and T = 29 keV, the Jacobian matrix is −0.086266 0.000082 J∗ = . −17.932381 −0.255725, Compared to the low temperature equilibrium, we can observe that the only qualitative difference can be found in the bottom right term. This term has become negative, reflecting the fact that at the high temperature equilibrium the losses increase faster with temperature than the fusion power. The derivative of the change in helium density with respect to the temperature is still positive, which is caused by the fact that dhσvi/dT , which is always positive for reactor relevant temperatures, is larger than ∂τE /∂T , which is negative because the increase of power with increasing temperature results in a lower value of τE . The major quantitative difference is in the bottom left element, which shows that the sensitivity of the temperature with respect to changes in the helium concentration at constant ρ increases with temperature, which finds its origin in the strong temperature dependence of hσvi. In this case the eigenvectors are and v1 = (0.066202, −0.997806) v2 = (0.000556, −1.000000), which are both stable. Notice that again both vectors are nearly parallel. From figure 4.1 we know that in between the unstable and stable equilibria there is transition region where the eigenvalues acquire an imaginary component, giving rise to oscillatory behaviour. Also, the values of the different elements of the Jacobian change when going to higher or lower densities, or when changing the value of ρ, but the overal picture stays the same. 1 Even with normalisation, J ∗ is highly asymmetrical and the eigenvectors are nearly parallel (their inner product is 0.997843), so special caution is warranted when calculating the eigenvalues, because the results are sensitive to small (rounding) errors in the elements of the Jacobian. Since a 2x2 matrix is always in Hessenberg form, balancing does not help in this case (in fact balancing a matrix in Hessenberg form changes the eigenvalues and vectors). 4.3 Reduced system 4.3.8 83 Phase portrait We can divide the ne , T -plane into regions with different behaviour. Figure 4.2 shows the phase portraits for PPCS model A for different values of ρ at a density of ne = 1020 m−3 . The red markers indicate the equilibria, whose position in the ne , T -plane is indicated in the top middle graph. The red lines indicate the separatrix between regions that converge towards a stable equilibrium and regions that do not (i.e. that will lead to a distinguishing of the burn if no action is taken). Starting from the top left image and moving anti-clockwise, the first image has ρ = 5 and we can see that the unstable point on the left is indeed a saddle point, with a stable eigenvector dominated by fα and an unstable one dominated by a the temperature. The stable equilibrium at higher temperature is a stable improper node, meaning that all orbits approach the node from opposite directions along the same line, except for two orbits which come in from opposite directions with a certain angle to the above mentioned line. Both axis can be found by finding the matrix T ∈ R2×2 that satisfies T AT −1 = J, with J the Jordan canonical form of A. Then define y(t) to satisfy the system y 0 = Jy (4.37) and all orbits come in along the direction of the y1 vector, except for the two orbits that come in along the y2 direction. The separatrix between the stable and unstable region starts on the T -axis a little above T = 5 keV in this case and increases more or less linearly until T = 10 keV and fα = 0.2, at which point the curve starts to flatten and disappears towards T = ∞. Lowering the value of ρ will lift the asymptote of the separatrix towards fα = 0.5, but it always stays below the physical limit fα = 0.5, which corresponds to a complete helium plasma. In the scope of our model this would be a an unrecoverable scenario without external heating, because a pure helium plasma means no fusion power and consequently infinite energy and particle confinement times. If the value of ρ is increased to eleven, we see that the separatrix starts to flatten a bit more strongly, but still approaches an asymptote that extends towards T = ∞, albeit at a lower value of fα . The unstable equilibrium has moved up a bit and the stable equilibrium is closer to the separatrix. At this point the eigenvalues of the system at the stable equilibrium have acquired an imaginary part and consequently this has become a stable spiral point. The pitch of the spiral depends on the ratio of the real and imaginary parts of the eigenvalues. Increasing the value of ρ until we hit the saddle-homoclinic bifurcation results in the separatrix revolving around the stable equilibrium and connecting with itself in the unstable equilibrium. This also means that the lower part of the separatrix has disappeared from the system and the stable region of the phase space has been reduced to the area within the homoclinc orbit. Chapter 4 Burn stability 84 0.3 ρ=5 ρ = 11 ρ = 13.8375 ρ = 14.4 fα 0.2 0.1 0 0.3 fα 0.2 0.1 0 10 ρ = 14.7 0.3 20 30 40 ne fα 0.2 0.1 0 10 20 30 T (keV) 40 5 10 20 T (keV) 50 Figure 4.2: Phase portraits for PPCS model A in the fα , T -plane of the reduced burn system for ne = 1020 m−3 for different values of ρ. The equilibria are indicated by corresponding red markers in the phase portrait and in the ne , T -plane (bottom right), and the red line indicates the separatrix between stable and unstable regions in the phase plane. The top left image at ρ = 5 shows a saddle point at low temperature and a stable equilibrium (sink) at high temperature. When the value of ρ is increased, an imaginary component is introduced in the system (top left). A further increase in ρ introduces a limit cycle, that grows until it becomes a homoclinic orbit that connects the saddle point with itself, orbiting the (high temperature) stable equilibrium. For even higher values of ρ the stability of the high temperature equilibrium changes: it becomes unstable with an imaginary part, until for very high values of ρ the high temperature equilibrium also becomes a source. Another effect of increasing ρ is that the two equilibria approach each other, until they coincide at the maximum allowed value of ρ. 4.3 Reduced system 85 The homoclinic orbit becomes an unstable limit cycle which shrinks in size with a further increase of ρ, until the system hits the Hopf bifurcation where the stable equilibrium changes to an unstable one. Note that right at the Hopf bifurcation there is an infinite number of limit cycles around the equilibrium, that at that point is stable, but not asymptotically stable. Towards the upper limit of ρ the imaginary part of the eigenvalues at the high temperature equilibrium disappears again, changing this point into an improper node, and the system is left with two unstable equilibria. An interesting observation that can be made when comparing the 5 phase portraits in figure is that the overall picture looks remarkably similar, with most orbits converging towards a curve that resembles a skewed parabole. The equilibria are located somewhere along this curve, and depending on the value of ρ, the separatrix partially runs along this curve as well. This observation could have implications for reactor start up or the design of burn control systems, because some orbits might be highly undesirable as they will cross the β limit, or put too much heat load on the first wall. 4.3.9 Stability for different scaling laws The ITER physics basis contains five different scaling laws, (IPB98(y) and IPB98(y,i), with i = 1, 2, 3, 4 [26]), and although their predictions for τE in ITER do not differ much, it has already been shown that the operating contours they predict for a burning plasma show large differences in density range [88]. In the following section we will investigate the changes in stability for the equilibria between the different scalings in the ITER physics basis. Figure 4.3 shows the different stability regions in the ne , T -plane for the five different energy confinement time scalings in the ITER physics basis. The global picture looks similar for the first four scalings: y, and y(1,2,3), with an unstable saddle point at low temperature and a second equilibrium at higher temperature, the stability of which is determined by ρ and ne . All plots feature the same stability regions, and they have a similar shape and position in the plot. At first glance the IPB98(y,4) scaling shows a completely different picture, but closer inspection learns that flipping the plot upside down makes it look very similar to the other four. The reason for this is that the value of 1 − 2m + l, with m and l being the exponents of the power and electron density in the scaling law, has a different sign for the (y,4) scaling compared to the other four scalings. The fact that this value is very close to zero for all scalings explains the large spread in density range between the different scalings, as was explained in chapter 4. Ignoring the fact that the plot for the last scaling is ’upside down’, the main differences in stability can be found when looking at the intersections of the curves of constant ρ with the stability boundaries. For the IPB98(y) scaling, the ρ = 14 contour extends to about two/thirds of the width of the unstable (dark blue) region on the high temperature side. The same contour for subsequent scalings reaches Chapter 4 Burn stability 86 98(y) ne (m−3 ) 1030 1025 1020 1015 1028 1055 1024 1040 1020 1025 1010 1016 98(y,4) ne (m−3 ) 98(y,3) 10100 1070 1075 1020 1050 10−30 1025 10−80 100 5 10 20 50 T (keV) 100 200 5 10 20 50 T (keV) 100 ne (m−3 ) ne (m−3 ) 1070 ne (m−3 ) 98(y,2) 98(y,1) 1032 10−130 200 Figure 4.3: Stability regions for the PPCS A design for the different τE scalings in the ITER physics basis [26]. The area is asymptotically stable, asymptotically stable with an oscillation, is unstable with a oscillation, is unstable and is a saddle point (unstable). 4.3 Reduced system 87 even further, and for the (y,2), (y,3) and (y,4) scalings extends into the stable (purple). Similarly, the ρ = 10 contour doesn’t quite extend into the green area for the (y) and (y,1) scalings, whereas for the other three scalings there is a significant part of this curve that traverses the asymptotically stable (green) area of the plot. The scaling laws for τE contain three parameters that change when moving through the phase space of a burning plasma; the average ion mass A, the electron density ne and the heating power P . Looking at the exponents k, l and m of these parameters in the scaling law, we see that l decreases and m increases for the subsequent scalings, whereas there is no clear pattern for k. The expansion of the unstable and oscillatory region therefore correlate with an increase in density and a decrease in power dependence, but we have not been able to identify a physical mechanism for this. For all scalings the point where both eigenvalues are zero lies roughly on the ρ = 8 contour, but the density for this point varies greatly. Depending on the sign of 1 − 2m + l, it can be found below or above the Greenwald density, but it is never in the reactor relevant density range. Figure 4.4 zooms in on the reactor relevant density and temperature range for the PPCS A design, plotting the stability of the equilibria on curves of constant ρ as well as the Greenwald density and β limit. In all subplots we see an unstable saddle point on the low temperature side, and a, mostly stable, equilibrium at the high temperature side. Only for ρ = 14 does this equilibrium become unstable, but it already acquires an imaginary eigenvalue part above ρ ≈ 10, depending on the scaling. In the two plots on the top row, the ρ = 2 curve shows an oscillatory region (in the bottom right of the plot). This is the intersection of this particular contour with the long ’tail’ of the stable oscillating region, which could already be seen in figure 4.1. The fact that it is not visible in the other plots is an artefact from plotting only integer values of ρ. Concluding we can say that for the scaling laws in the ITER physics basis there are only minor differences in stability, both on a global level and in the reactor relevant domain. Those differences manifest themselves at the high temperature equilibrium, and correlate with the values of the power en density exponents in the scaling law: a weaker power dependence and a stronger density dependence result in an expansion of the unstable area to lower values of ρ. However, based on the results of present day tokamaks [34], it seems unlikely that this will be an issue for future reactors, although much higher values of ρ have also been reported in limiter plasmas [76]. 4.3.10 Stability with external heating So far we have only considered ignited plasmas without external heating. This approach allows for an analytical treatment of the system, but it is an unrealistic Chapter 4 Burn stability 88 IPB98(y) lim it ρ= 5 ρ=9 ρ = 13 1 ρ = 13 β ρ=1 ne (1020 m−3 ) 2 nG IPB98(y,1) IPB98(y,2) IPB98(y,3) IPB98(y,4) ne (1020 m−3 ) 2 1 ne (1020 m−3 ) 2 1 5 10 15 20 25 T (keV) 30 35 40 5 10 15 20 25 T (keV) 30 35 40 Figure 4.4: Stability for the reduced system burn contours of PPCS model A at different values of ρ for the different τE scalings in the ITER physics basis. The asymptotically stable part of the contours is denoted , whereas is asymptotically stable with an oscillation, is unstable with an oscillation, is unstable and is a saddle point (unstable). 4.3 Reduced system 89 scenario for a tokamak reactor design, if only because some form of non-inductive current drive will be needed. The external heating power required for this will also affect the stability of the equilibria. In the case of external heating, equation (4.18) changes to 2 (ne − 2nα ) nα T T Sext − Srad dT 3 = hσvi Eα + T − − + , (4.38) 3 dt 6ntot 2 ntot τp τE 2 ntot where Sext is the external heating power density in keV/m3 . Since Sext is independent of nα or T , it does not show up in the Jacobian, which consequently has the same eigenvalues as before. Adding external heating does therefore not affect the stability of the system directly. MW 0M W 15 0M W 20 0M W 25 0M W 10 50 0.2 0M 50 W 10 MW 0 15 MW 20 0 MW 25 0 M 0 MW W fα (%) 0M W 0.25 0.15 0.1 22 24 26 28 T (keV) 30 32 Figure 4.5: The position of the high temperature equilibrium in the fα , T -plane as a function of Pext , at a density of ne = 1 × 1020 m−3 for the PPCS A design for ρ = 5 ( ) and ρ = 10 ( ). For increased levels of external heating, the equilibrium temperature increases and the helium fraction decreases. Whereas the relative decrease in helium fraction is the same at roughly 5% at both ρ values, the relative increase in temperature is significantly larger at ρ = 10 (9% versus 24%). Indirectly there is an effect because the equilibria are shifted in phase space to different values of fα and T , and the eigenvalues of the Jacobian evaluated at these points are different. The shift in fα and T is illustrated in figure 4.5, which shows the equilibrium position in the fα , T -plane as a function of Pext for the PPCS A design at ρ = 5 and ρ = 10. Increasing Pext results in a shift of the equilibrium to higher temperatures and lower helium fraction, which can be understood as follows. Adding external heating increases the temperature and consequently the reactivity of the plasma. This increase, together with the resulting increase in fusion power, lowers the energy confinement time, which results in a faster exhaust of helium ash. 90 Chapter 4 Burn stability The relative shift in helium fraction is roughly 5%, and this value is more or less independent of the value of ρ. The relative temperature shift, on the other hand, increases for higher values of ρ. The reason this effect is stronger for higher values of ρ lies in the ratio between alpha heating power and external heating. At higher values of ρ the equilibrium temperature is lower and the helium fraction higher, which implies a lower fusion power output. Adding a certain amount of external heating will therefore have a bigger effect at high ρ values. An interesting observation is that the effect on the fusion power output of the reactor, is almost linear in T , with the gradient dPfus /dT a function of ρ. For high values of ρ it might therefore be interesting to consider external heating to increase the power output of the reactor. As long as the gain in fusion power output is larger than the additional external heating divided by the plant efficiency times the heating efficiency, the net effect is positive. To determine the stability of the system with Pext 6= 0, we first solve the burn equations (4.17) and (4.38) to obtain the equilibrium values of nα and T . These are used to obtain the Jacobian of the system at the equilibrium and subsequently we determine the eigenvalues of the system. Figure 4.6 plots the stability of the system in the ne , T plane for the PPCS A reactor with different levels of external heating, at different values of ρ. The plot on the top left is identical to the bottom left in figure 4.4. It is immediately apparent that the green area at high temperature increases for increasing levels of external heating and that the blue segment on the ρ = 14 curve disappears. At the same time the addition of Pext results in a change from closed contours to open contours, which manifests itself in a change in the sign of the slope of the high temperature equilibrium curves. Instead of having a positive value of dne /dT , this now becomes negative. On the low temperature side the equilibrium curves also change. Additional heating first of all changes their slope from negative to positive, and secondly causes them to curve upwards after hitting a local minimum in density, which introduces a second stable equilibrium at low temperatures. Around the minimum density point, the system goes through a transition from unstable to stable, passing through an oscillating area, with the same bifurcations as described in section 4.3.4. The stability of the equilibrium to the right of the local density minimum is unaffected by the amount of external heating, it remains a saddle point (which is always unstable). Increasing the additional power has the effect of shifting the minimum density on the low temperature side upwards, and at some point they disappear from the plot (bottom middle plot). The relevance of these equilibria for reactor purposes is minimal because of the very low fusion power output, but it basically rules out a low power startup scenario at high density. Concluding we can say that the addition of external heating has a beneficial effect on the stability of the high temperature equilibrium, for which the stable 4.3 Reduced system 91 Pext = 0MW Pext = 50MW 2 lim it ρ=1 ρ=5 ρ=9 ρ = 13 ρ=1 1 ρ = 13 ne (1020 m−3 ) β nG Pext = 100MW Pext = 150MW Pext = 200MW Pext = 250MW ne (1020 m−3 ) 2 1 ne (1020 m−3 ) 2 1 5 10 15 20 25 T (keV) 30 35 40 5 10 15 20 25 T (keV) 30 35 40 Figure 4.6: Stability for the reduced system burn contours of PPCS model A at different values of ρ for different levels of external heating. The asymptotically stable part of the contours is denoted , whereas is asymptotically stable with an oscillation, is unstable with an oscillation, is unstable and is a saddle point (unstable). Chapter 4 Burn stability 92 area increases for increasing Pext . The stability of the low temperature equilibrium is unaffected for higher densities, and a stable, third equilibrium is introduced at an even lower temperature. The transition between the stable and unstable equilibrium at low temperature occurs at the local density minimum. 4.3.11 Reactor comparison There are currently four different prototype reactor designs under consideration in Europe [71, 70], with PPCS models A and B being more conservative and PPCS models C and D applying riskier technology extrapolations. This section analyses and compares the stability for the four different designs at their designed level of external heating. Figure 4.7 plots the stability of the burn contours for the PPCS model A, B, C and D designs for different values of ρ with the amount of external heating specified in reference [71], which is 246 MW, 270 MW, 112 MW and 71 MW respectively. We take the same approach as in section 4.3.10, first calculating the equilibrium values of nα and T for the given values of ρ and Pext and subsequently determining the eigenvalues of the Jacobian. The resulting stability curves are plotted in figure 4.7, in which the top left plot closely resembles the bottom right plot in figure 4.6, since the level of external heating differs by only 2% between them. All four reactor designs are stable in their complete operating range for their designated level of external heating, but the type of stability differs between the different reactors at different operating points. The operating space for the PPCS A and B designs are almost completely asymptotically stable, the PPCS C and D designs show considerable areas with oscillatory stable behaviour. For the PPCS C design, this only occurs for ρ ≈ 2 or well above nG for high values of ρ. 4.4 Full system Having studied the reduced system in detail, we will now return to the full system of burn equations, as presented in section 4.2.1. First we will linearise the system around the equilibria and derive the Jacobian matrix, and use this result to investigate the stability of the system in the ne , T -plane for different reactor designs and energy confinement scaling laws. Finally, we will discuss the stability of the system with external heating. 4.4.1 Jacobian matrix of the full system Writing xi = {nD , nT , nα , T }, the burn equations (4.2) can be expressed as dxi = fi (x1 , x2 , x3 , x4 ) dt i = 1, 2, 3, 4 (4.39) 4.4 Full system PPCS A, Pext = 246MW 93 PPCS B, Pext = 270MW 2 lim it ρ=1 ρ=5 1 ρ=9 13 ρ= ne (1020 m−3 ) β nG PPCS C, Pext = 112MW PPCS D, Pext = 71MW ne (1020 m−3 ) 2 1 5 10 15 20 25 T (keV) 30 35 40 5 10 15 20 25 T (keV) 30 35 40 Figure 4.7: Stability for the reduced system burn contours of PPCS models A, B, C and D at different values of ρ for the design value of external heating as specified in [71]. The asymptotically stable part of the contours is denoted , whereas is asymptotically stable with an oscillation. Chapter 4 Burn stability 94 We can linearise the system around an equilibrium point xj0 as 4 X dxi ∂fi (xj − xj0 ) , = dt ∂xj xj0 j=1 i = 1, 2, 3, 4. (4.40) To determine the linear stability of the system requires determining the eigenvalues (and eigenvectors) of the Jacobian matrix jij = ∂fi /∂xj evaluated at the equilibrium points xj0 . As with the reduced system, determining the derivatives used in the Jacobian is not difficult, but involves some extensive algebra which can be found in section A.2. Here we simply present the Jacobian of the full system: 1 nD ∂τE + 2 , ρτE ρτE ∂nD nD ∂τE = −nD hσvi + 2 , ρτE ∂nT nD ∂τE = 2 , ρτE ∂nα dhσvi nD ∂τE = −nD nT + 2 , dT ρτE ∂T nT ∂τE = −nT hσvi + 2 , ρτE ∂nD nT ∂τE 1 = −nD hσvi − + 2 , ρτE ρτE ∂nT nT ∂τE = 2 , ρτE ∂nα dhσvi nT ∂τE = −nD nT + 2 , dT ρτE ∂T nα ∂τE = nT hσvi + 2 , ρτE ∂nD nα ∂τE = nD hσvi + 2 , ρτE ∂nT 1 nα ∂τE + 2 , =− ρτE ρτE ∂nα dhσvi nα ∂τE = nD nT + 2 , dT ρτE ∂T j11 = −nT hσvi − (4.41) j12 (4.42) j13 j14 j21 j22 j23 j24 j31 j32 j33 j34 (4.43) (4.44) (4.45) (4.46) (4.47) (4.48) (4.49) (4.50) (4.51) (4.52) 4.4 Full system j41 j42 ntot nT − 2nD nT 3 = hσvi Eα + T 3 2 2 2 ntot 4(sD + sT )T (ρ − 1)T ∂τE 1 ∂Srad 2Srad + − 3 + 3 2 + , 2 ρτE ∂nD ∂nD n2tot 2 ntot 2 ntot 3 ntot nD − 2nD nT hσvi T E + = α 3 2 2 2 ntot (ρ − 1)T ∂τE 4(sD + sT )T 1 ∂Srad 2Srad , − 3 + 3 2 + ρτE2 ∂nT ∂n n2tot T 2 ntot 2 ntot 2nD nT 3 = − 3 2 hσvi Eα + T 2 2 ntot (ρ − 1)T ∂τE 6(sD + sT )T 1 ∂Srad 3Srad + , − 3 + 3 2 + ρτE2 ∂nα ∂n n2tot n n α tot 2 2 tot 3 ∂hσvi nD nT 3 = 3 hσvi + Eα + T 2 2 ∂T 2 ntot 1 Srad 2(sD + sT ) T ∂τE 1−ρ 1 − 3 − − 2 . + ρ τE τE ∂T 2T ntot 2 ntot + j43 j44 95 (4.53) (4.54) (4.55) (4.56) This result differs from the Jacobian presented by Rebhan and Vieth [65] for two reasons. Firstly, our result is valid for scaling laws for the energy confinement time that do not include radiation losses, whereas ref [65] uses the ITER89P scaling [29], which includes the radiation losses in the energy confinement time. This introduces some extra terms in our Jacobian to account for this. The second difference arises because ref [65] assumes a constant ion mass of 2.5 amu, which is only correct in case of a pure DT plasma. A burning plasma by definition contains helium and this needs to be taken into account in the average ion mass, which consequently will be ≥ 2.5 amu. 4.4.2 Full system stability Using the expression for the Jacobian presented in section 4.4.1, we can determine the stability of the different eigenvectors of the system for each equilibrium by determining their respective eigenvalues. In figure 4.8 the stability of the different eigenvectors of the system is plotted. Again, red is unstable, blue is oscillatory unstable, purple is oscillatory stable and green is stable. We can see that on the high temperature side the system is stable, whereas on the low temperature side there is one eigenvector that turns the system unstable. In the central region of the operating space their are several transitions between stable and unstable behaviour for the different eigenvectors. A transition between different stability regimes means that either the real or the imaginary part of the eigenvalues has a zero crossing. There is only one point (near Chapter 4 Burn stability 96 1070 ne (m −3 ) 1060 1050 1040 1030 1020 1010 1070 ne (m −3 ) 1060 1050 1040 1030 1020 1010 5 10 20 50 T (keV) 100 200 5 10 20 50 T (keV) 100 200 Figure 4.8: The stability of the different eigenvectors for a burning plasma in the PPCS A reactor using the IPB98(y,2) scaling for τE . Unstable behaviour is indicated by , signifies oscillatory unstable behaviour, oscillatory stable behaviour and stable behaviour. The irregularities in the boundaries between the different colours, for instance on the green-purple boundary in the two plots on the right is caused by a an interchange of two or more eigenvectors.Please note that the density and β limits are not taken into account in this plot. 4.4 Full system 97 the bottom of the plot where the blue parts end) where the absolute value of both eigenvalues goes to zero, which corresponds to the Bogdanov-Takens bifurcation described in section 4.2.3. Hence the fact that blue always borders red and purple and never borders green. Similarly, purple is never adjacent to red. Note that although it appears otherwise, there is actually a small slither of red between the green and blue areas in the bottom right plot. The irregularities on the border between purple and green that can be seen in the two plots on the right of figure 4.8 are caused by a numerical difficulty sorting the eigenvectors. When determining the eigenvectors, most algorithms will determine the largest eigenvalue and corresponding eigenvector first, then reduce the dimension of the matrix and repeat. If two (or more) eigenvalues happen to approach each other closely enough, both in argument and absolute value, it can be difficult to know which is which at the next point in parameter space. This can of course be solved by looking closely at the trajectory of the different eigenvalues and vectors, but developing a fail safe algorithm is rather involved and doesn’t lead to new physical insight. Although this plot allows us to investigate the stability of each eigenvector in a particular equilibrium, it is hard to obtain the overal stability at a glance. Figure 4.9 plots the stability areas of the complete system in a single figure, with red the unstable and green the stable part. The system is unstable on the low, and stable on the high temperature side. The transition between the two regions lies at the minimum density of the burn contours below the Bogdanov-Takens bifurcation (which occurs at ρ ≈ 8) for the low density part of a burn contour. For higher values of ρ and at the high density part of the contour, the transition occurs at a higher temperature and a higher respectively lower density. Comparing this picture to figure 4.1 the overal stability of the four- and twodimensional systems looks very similar, lending greater credibility to the claim that the reduced system captures the essential physics. 4.4.3 Eigenvectors and eigenvalues Apart from looking at the stability of the individual eigenvectors and of the system as a whole, we can also try and interpret the eigenvectors of the system in a physical sense. Although new for the ITER physics basis scalings, a similar approach was taken by Rebhan and Vieth for the ITER89P L-mode scaling [65]. They investigated three cases: one where they assumed that τE and τp kept their equilibrium values during a perturbation, one where τE and τp followed the ITER89P scaling law but the heating power was equal to the loss power, and one where they equated the heating power to Sα = nD nT hσviEα 2π 2 κa2 R, again with τE and τp following the scaling law. We have only investigated the latter case, of course using the IPB98(y,2) scaling, since this is of most relevance for future reactors. Besides the fact that the ITER89P scaling includes radiation losses and the IPB scalings do not, we have Chapter 4 Burn stability 98 also included the presence of helium in the plasma on the average ion mass, something which was absent in the analysis in [65]. 4.4.4 Low temperature stability We will discuss the eigenvectors for the full system at the same equilibria as for the reduced system, at a density of ne = 1020 m−3 and ρ = 5 for the PPCS A reactor with IPB98(y,2) scaling. The normalised eigenvectors for the unstable, low temperature equilibrium at T = 6.2 keV are displayed in table 4.2. Table 4.2: The normalised eigenvectors and eigenvalues (λ) at the low temperature equilibrium for a burning plasma in the PPCS A reactor using the IPB98(y,2) scaling at a density of ne = 1020 m−3 and ρ = 5. The first eigenvector also has a very weak fα dependence which doesn’t show up at this level of accuracy. v1 v2 v3 v4 fD 0.01 0.21 −0.14 −0.48 fT 0.01 0.21 −0.14 0.29 fα 0 −0.03 0.05 0.05 T∗ −1.00 −40.96 0.98 0.83 λ 0.195 −0.018 −0.014 −0.012 For the low temperature equilibrium all four eigenvectors have four non-zero elements, which agrees with our finding that the dimensionality of the system cannot be reduced without discarding physics information, as was done in ref [65] . The first eigenvector v1 is unstable and is predominantly a temperature perturbation. This corresponds to the temperature instability that is present in the simplest model for a burning plasma that contains only Bremsstrahlung losses and alpha particle heating. At the low temperature equilibrium the reactivity of the plasma has a stronger temperature dependence than the radiation losses. Therefore a temperature perturbation will quickly grow until the plasma either extinguishes (in case of a negative perturbation) or the temperature reaches the stable equilibrium. Because a temperature perturbation also has an effect on the plasma composition, v1 also has non-zero deuterium, tritium and helium components. The second and third eigenvectors differ from the first in that they show much larger deuterium, tritium and helium components. They both correspond to an in-phase deuterium and tritium density fluctuation coupled to a temperature and helium density fluctuation. The fourth eigenvector v4 differs from the previous three in that the deuterium and tritium fluctuation have an opposite sign. The temperature component is 4.4 Full system 99 the smallest for this eigenvector. This eigenvector probably corresponds to the eigenvector (ñD , −ñT , 0, 0) that was found in reference [65]. The eigenvalue of v4 is indeed λ4 = −1/τp , and the reason the eigenvector looks different is caused by the fact that we have included the ion mass dependence in the τE scaling law, which causes a coupling between the temperature and particle densities that cannot be removed. Concluding we can say that it is the thermal part that causes the linear instability of the low temperature equilibrium in a burning plasma. The particle dominated perturbations are all stable. 4.4.5 High temperature stability The eigenvectors at the high temperature equilibrium (T = 21.5 keV) look a little different and are listed in table 4.3. The most obvious difference is the fact that the first two eigenvectors have acquired an imaginary part. Secondly the particle components are a lot smaller (for almost the same normalisation: the particle densities for both the high and low equilibrium are normalised to ne = 1020 m−3 and the temperature is normalised to the equilibrium temperature of 6.22 and 21.5 keV respectively). Table 4.3: The normalised eigenvectors and eigenvalues (λ) at the high temperature equilibrium for a burning plasma in PPCS model A using the IPB98(y,2) scaling at a density of ne = 1020 m−3 and ρ = 5. The first eigenvector also has a very weak fα dependence which doesn’t show up at this level of accuracy. v1 v2 v3 v4 fD (10−4 ) 2 + 68i 2 − 68i −1747 116 fT (10−4 ) 2 + 68i 2 − 68i 1642 116 fα (10−4 ) 2 − 9i 2 +9i -102 328 T∗ 1.0000 1.0000 0.9708 0.9993 λ −0.210 − 0.120i −0.210 + 0.120i −0.074 −0.060 The unstable temperature dominated fluctuation at the low temperature equilibrium, is stable at the high temperature equilibrium, but has acquired an imaginary part which causes an oscillating behaviour. The magnitude of the imaginary part relative to the real part of the eigenvalue increases with ρ, giving rise to higher ratio of oscillation period to damping time. The third eigenvector corresponds to the (ñD , −ñT , 0, 0) eigenvector found by Rebhan and Vieth with eigenvalue λ3 = −1/τp , but in our case this is again coupled to a temperature and (weak) helium fluctuation. The fourth and final eigenvector is a combination of temperature en density fluctuations, but it stabilises slower than the other three eigenmodes. 100 Chapter 4 Burn stability Because of the lack of symmetry in the Jacobian, both at the high and low temperature equilibria all eigenvectors are coupled density and temperature fluctuations. This makes a simple interpretation in terms of a pure temperature or density perturbations difficult, but nevertheless we can identify the underlying mechanisms of several eigenvectors. One eigenvector is a density dominated fluctuation, with eigenvalue λ = −1/τp , which is expected since this is the particle transport timescale. Another eigenvector corresponds to the thermal mode that is present in a pure DT plasma. The eigenvalue of this mode changes with T : for low values of T it is small compared to τE , for higher values of T the ratio increases to close to one. This can be understood by realising that at low temperature the radiation losses account for a large fraction of the total energy losses and these are not included in τE , which is consequently much larger than the timescale of the fluctuation. At higher temperature the importance of the radiation losses decreases, bringing the energy confinement time and the eigenvalue of the mode closer together. The other two eigenvectors are hybrid modes that consequently have timescales somewhere in between −1/τE and −1/τp . Their eigenvalues also depend on the precise location of the equilibrium. We can plot the stability of the system in the ne , T -plane like in figures 4.8 and 4.1, or look at the composition of the eigenvectors in individual equilibria, but this doesn’t tell us how the stability is affected by changing the parameters of the system that are under control of the operator. In the model that we used, the properties of the system are determined by three parameters which, within limits, can be chosen freely: sD , sT and ρ. In our calculations so far we have made the assumption that sD = sT , effectively giving us two inputs that we can adjust. The standard approach in this case would be to plot the stability boundary (or the eigenvalues) as a function of sD and ρ-plane. However, because for each value of ρ there are two equilibria that have the same refuelling rate and the boundary of stability is near the maximum or minimum density on a ρ = const contour, such a plot would not be very instructive. Instead, we have plotted the lines of constant ρ and constant sD (or sT for that matter) in the stability overview for PPCS model A with the IPB98(y,2) scaling in figure 4.9. At low density the stability boundary intersects the contours of constant ρ at the minimum density for ρ < 9. For higher values of ρ the stability boundary deviates towards higher temperatures and consequently higher density at the intersection with the ρ iso-contours. This trend continues at the high density side of the contours, where the intersection lies below the maximum density. The dashed lines that indicate constant fuelling rate are slightly curved downward and are just a bit lower at the high temperature side. The reason for this it that in the center the value of ρ increases, which means a better particle confinement and consequently a less need for refuelling. At the low and high temperature 4.4 Full system 101 1070 1060 -3 ne (m ) 1050 1040 1030 1020 1010 5 10 20 50 T (keV) 100 200 Figure 4.9: Stability plot for the PPCS A reactor with IPB98(y,2) scaling, including contours of constant ρ (solid lines) and constant refuelling rate (dashed lines). Please note that the density and β limits are not taken into account in this plot. 102 Chapter 4 Burn stability sides the particle confinement is lower, so for a given density there is a higher particle transport. Finally the reason for the slope in the lines lies in the fact that the fusion power increases with temperature, so the confinement time decreases towards higher temperatures. This effect weakens when the temperature approaches the maximum in the reactivity around 70 keV, which can be seen in the lines of constant refuelling rate at high density and temperature, which run almost horizontally at high temperature. Looking at the intersections between the lines of constant refuelling rate and the stability boundary, it is obvious that it is only possible to cross the stability boundary by changing the refuelling rate while keeping ρ constant for high densities or high values of ρ. For ρ < 10 at low densities the stability boundary intersects the constant ρ contours at the point where they are tangent to the lines of constant refuelling rate. 4.4.6 Stability for different scaling laws The ITER physics basis contains five different scaling laws, each based on a fit to a different subset of the database, or using slightly different fit restrictions. The effect of using a different scaling on the stability boundary in a burning plasma is illustrated in figure 4.10 for the PPCS A reactor design. Again green indicates stable behaviour and red unstable. The stability boundary hardly changes between the different scalings, except for the region around the minimum density. There the position of the critical point, which corresponds to the position of the Bogdanov-Takens bifurcation in the reduced system, varies. Zooming in on the reactor relevant density area, as is done in figure 4.11, we see that only at very high values of ρ there is a difference in stability between the different scalings at the high temperature equilibrium. It is interesting to note however, that although basically all high temperature equilibria are stable, they all have at least two imaginary eigenvalues. A perturbation of these equilibria will result in damped oscillations of the plasma parameters. From these plots there is no way of telling how strong the damping is compared to the oscillatory part, but in general higher values of ρ exhibit weaker damping (which makes sense because these equilibria are closer to the unstable equilibria at low temperature). 4.4.7 Reactor stability comparison with external heating Most reactor designs, even though capable of ignition, still employ some level of external heating, mostly for current drive and control purposes [70]. Figure 4.12 shows the stability of the PPCS A, B, C and D reactor designs with their respective levels of external heating, for ρ ranging from 1 to 14. Obviously, the shape of the curves is exactly the same as in figure 4.7, but the stability properties are somewhat different. The addition of two extra degrees of freedom has resulted in the addition of an imaginary component to the eigenvalues 4.4 Full system 103 98(y) ne (m−3 ) 1030 1025 1020 1015 1028 1055 1024 1040 1020 1025 1010 1016 98(y,4) ne (m−3 ) 98(y,3) 10100 1070 1075 1020 1050 10−30 1025 10−80 100 5 10 20 50 T (keV) 100 200 5 10 20 50 T (keV) 100 ne (m−3 ) ne (m−3 ) 1070 ne (m−3 ) 98(y,2) 98(y,1) 1032 10−130 200 Figure 4.10: The stability regions for a burning plasma in the PPCS A reactor for the five scalings in the ITER physics basis using the full system. Please note the change in density range between the different plots, and be aware that the density and β limits are not taken into account. Chapter 4 Burn stability 104 IPB98(y) it ρ= 5 ρ = 13 ρ = 13 1 lim ρ=9 β ρ=1 ne (1020 m−3 ) 2 nG IPB98(y,1) IPB98(y,2) IPB98(y,3) IPB98(y,4) ne (1020 m−3 ) 2 1 ne (1020 m−3 ) 2 1 5 10 15 20 25 T (keV) 30 35 40 5 10 15 20 25 T (keV) 30 35 40 Figure 4.11: Stability for the full system burn contours of PPCS model A at different values of ρ for the different τE scalings in the ITER physics basis. The asymptotically stable part of the contours is denoted , whereas is asymptotically stable with an oscillation, is unstable with an oscillation and is unstable. 4.4 Full system PPCS A, Pext = 246MW 105 PPCS B, Pext = 270MW 2 lim it ρ=1 ρ=5 1 ρ=9 13 ρ= ne (1020 m−3 ) β nG PPCS C, Pext = 112MW PPCS D, Pext = 71MW ne (1020 m−3 ) 2 1 5 10 15 20 25 T (keV) 30 35 40 5 10 15 20 25 T (keV) 30 35 40 Figure 4.12: Stability of the full system burn contours of the PPCS A, B, C and D designs [70], for ρ ranging from 1 to 14. At low density and high temperatures, most ). At higher densities, they are stable, but the solutions in equilibria are stable ( this region will oscillate ( ), and the PPCS D design has a region with unstable oscillating behaviour ( ). Finally, at high density and low temperatures, there are unstable equilibria ( ). 106 Chapter 4 Burn stability for most of the high temperature equilibria, compared to the stability of the reduced system. Only at lower densities do equilibria without oscillatory behaviour still exist. Again, the PPCS D design is the odd one out, and in this case the system is unstable for higher densities at low temperatures, which were stable for the reduced system with external heating. Unfortunately, the anticipated operating point for PPCS model D (T = 12 keV and ne = 1.4 × 1020 m−3 ) lies within the unstable density range, although the equilibrium temperature that our simulations predict is somewhat lower than the value from the PPCS. 4.5 Discussion and conclusions We have derived a simple two-dimensional system to study the operating point stability of a burning plasma, by assuming a constant electron density and equal deuterium and tritium concentrations. This allowed us to use planar bifurcation theory to describe the transitions between regions with different stability in the phase plane. Furthermore, we have analysed the physical mechanism behind the stabilising or destabilising effect of the different elements of the Jacobian matrix. In general, a burning plasma has two equilibria at a given density, one at a lower, and one at a higher temperature. The low temperature equilibrium is always unstable, and the stability of the high temperature equilibrium depends on the density and ρ. For low values of ρ, the high temperature equilibrium is stable for all but the highest densities (which are inaccessible anyway because they are far above the Greenwald and Troyon limits). For high values of ρ the high temperature equilibrium is stable for intermediate densities, and the stable range shrinks for increasing values of ρ and it disappears completely when approaching ρcrit . At the boundaries of the different stability regions, the system features different bifurcations. In the two dimensional system these can be easily distinguished by looking at the eigenvalues. Generally speaking, there is a local bifurcation in the system when either one, or both, of the eigenvalues is zero or has a real part equal to zero. These zeros occur at the boundary between different stability regions. It turns out that the reduced system contains five bifurcations: two saddle node bifurcations at the maximum and minimum density on a contour respectively, a sub-critical Hopf bifurcation at the stable-unstable transition which results in the birth of a limit cycle and a saddle homoclinic bifurcation where the homoclinic orbit collides with the saddle point at the unstable equilibrium and disappears. The final bifurcation is a Bogdanov-Takens bifurcation which occurs at the point where both eigenvalues are equal to zero. This is a point where two fold bifurcations, a Hopf bifurcation and a saddle-homoclinic bifurcation meet. The Hopf bifurcation creates a limit cycle, which grows and collides with the low temperature saddle point in the saddle-homoclinic bifurcation. For low values of ρ both the reduced and the full system show linearly stable 4.5 Discussion and conclusions 107 behaviour at high temperature and linearly unstable behaviour at low temperatures, while the density has negligible influence on the stability properties. For intermediate temperatures, which are only accessible at higher values of ρ for intermediate densities, the system goes through a transitional phase, where the eigenvalues acquire an imaginary part which gives rise to oscillatory behaviour. The linear stability properties of a system only provide (very) limited information about its non-linear stability. For the reduced system we made streamline plots that show the temporal evolution of the system in the ne , T -plane which can be used to identify safe operating regimes, where a perturbation of the system will not grow to violate the β-limit. The Greenwald limit cannot be exceeded in the reduced system because we assume the density to be fixed. For the four dimensional system it is not possible to make such streamline plots. Also there is no guarantee that the orbits in phase space map out contiguous volumes; there is a real possibility that some of the trajectories of the system pass through a saddle point or even form interlocking loops. While we haven’t found any trajectories starting at a stable, high temperature equilibrium that cross the β or density limit, we cannot exclude that such orbits exist. A perturbation of an unstable equilibrium at low temperature on the other hand will almost certainly cross the β-limit in case the trajectory converges to the high temperature equilibrium. In case the trajectory leads to an extinguishing of the plasma there is the risk of crossing the density limit, since the particle confinement increases and the refuelling rate stays the same. In the intermediate temperature range with imaginary eigenvalues there is also the possibility that a perturbation of a stable equilibrium leads to a trajectory that ends up in an unstable part of phase space, which can lead to either an extinguishing of the plasma and crossing the density limit, or an increase in temperature which can lead to a violation of the β-limit. The addition of external heating power lifts the low temperature equilibria to inaccessible densities, and has a stabilising effect on the high temperature equilibria at high values of ρ. Only for high levels of Pext , ρ close to ρcrit and ne ≈ 2nG does external heating have a destabilising effect. When comparing the stability of the four different reactor designs in the PPCS, models A, B and C show similar stability characteristics. Because PPCS model D is not capable of ignition and requires external heating to maintain the required temperature, the position of the operating points deviates significantly from those of the other three reactors. However, also for PPCS model D the operating points are stable. Although the dynamics of the four-dimensional system of burn equations are much more complex than those of the two-dimensional system, the overal stability looks very similar. The main difference being that all operating points for the PPCS A, B and C designs lie in the stable region of phase space where at least one of the eigenvalues has an imaginary component. In most cases the imaginary 108 Chapter 4 Burn stability part is small compared to the real part, so possible oscillations will experience a strong damping, but nevertheless this might need to be taken into account in a detailed reactor design. The inclusion of the ion mass dependence in our derivation breaks the symmetry between deuterium and tritium that was present in the results presented in reference [65]. Consequently, the eigenvectors that describe a pure density and a pure temperature perturbation that they identified do not exist anymore. Nevertheless, we were able to identify the corresponding thermal and particle dominated eigenvectors in the full system. The other eigenvectors are hybrid modes for which we have not been able to find a simple physical interpretation. Concluding we can say that the stability properties of a two-dimensional burn system with constant ne and D/T-ratio are in good agreement with those of the full four-dimensional system. Operating points in the reactor relevant density and temperature range are mostly stable, with the exception of those of the PPCS D design. The use of different scaling laws yields only slightly different stability properties at the operating points, and the addition of external heating has a stabilising effect in all reactor relevant scenarios. One point of concern that remains is the temperature evolution during the start-up of the reactor. After heating the plasma to the ignition point, the temperature and helium content will evolve until they settle at their stable values. However, when starting with a pure DT plasma, the temperature can overshoot the equilibrium temperature by more than 10 keV, which will most likely violate the β-limit. To prevent this from happening the plasma could be started with a non-zero helium concentration, but this would require a higher heating power. A better solution might be to start with a pure DT plasma, but replace (part of) the fuelling by helium injection once ignition has been reached. That way the reactor could be started with a minimum amount of external heating while still preventing the dangerous temperature overshoot. 109 Chapter 5 Sensitivity of burn contours to form of scaling laws 5.1 Introduction Unlike present day experiments, an economically viable fusion reactor cannot rely on external heating to keep the plasma at the required temperature. While some level of external heating might still be required, for example for non-inductive current drive, the main source of power has to come from the alpha particles produced by the fusion reaction. This constitutes a radical change to the dynamics of the plasma, because it creates a strong link between the density n, temperature T and fusion power Pfus . The common method for evaluating reactor designs is to make use of a scaling law for the energy confinement time τE , because a complete description of the transport properties of the plasma is too complicated for this purpose. For lack of a better alternative, this is also the common approach for reactors capable of ignition (or in any case have predominant alpha heating). The implicit assumption is that the plasma doesn’t know what form of heating is applied. From a transport point of view, this seems reasonable, because the main drive for turbulent transport is the temperature gradient which should be independent of the precise heating method. From the Lawson criterion [89] for ignition follows that to good approximation the triple-product nτE T must exceed a critical value, (which is why the triple product is commonly used as a metric of progress in fusion research). In a burning plasma Pfus ∝ n2 , and the commonly used scaling law IPB98(y,2) [26] predicts an energy confinement time τE ∝ n0.41 P −0.69 . Combining these two proportionalities, we find that τE ∝ n−0.97 , according to which the triple product for an ignited plasma is (almost) independent of density. 110 Chapter 5 Sensitivity of burn contours to form of scaling laws This suggests that a fusion power plant could ignite at arbitrarily low density, and consequently, low fusion power. Of course there are physical mechanisms such as the alpha slowing down time, synchrotron radiation losses or the power threshold for the LH-transition (all linearly dependent on density), which will put a lower limit on the density. Nevertheless, these only become a factor below electron densities ne < 1019 m−3 . So either ignition at such low densities is possible, or there is a problem with the application of the current scaling laws to burning plasmas. Scaling laws for τE have been used from the very beginning of fusion research in an attempt to compare the results of different reactors and develop a basis on which to design new (and better) experiments. In this regard they have been highly successful, but care has to be taken when applying them outside the range of plasma parameters present in the dataset on which they are based. This is especially important in the case of burning plasmas, as they will explore parameter regimes that are currently inaccessible. A second problem arises from the assumption underlying all current τE scaling laws: that the electron density ne and heating power P are independent of each other. However, in a burning plasma density and power are strongly coupled and consequently a scaling law should not treat them as such. In this chapter we investigate the application of current scaling laws to burning plasmas, and show that there is a singularity in the system. We will explore the consequences of this singularity for the predicted operating contours and conclude with some suggestions that might help resolve the issue. 5.2 Theory In a burning plasma, the only parameters that are under direct control of the operator are the electron density and fuel ratio. Given these two, the plasma might find an equilibrium. The existence, stability and position in phase space has been the topic of several studies. Early work on burning plasmas was done by Kolesnichenko et al. [90], and Houlberg et al. [77] were the first to introduce the plasma operation contour (POPCON) plot, that displays the operating contours in the ne , T -plane. Other studies looked at the required power for ignition, for instance Mitarai et al. [91], and Reiter et al. [41] looked at the influence of helium on the burn equilibria when it is taken into account self-consistently. Rebhan et al. studied the stability of the operating points for ITER ID ([78]) with self-consistent helium treatment and the ITER89P L-mode scaling ([29] for the energy confinement time τE (including an H-mode factor fH = 2 to mimic H-mode behaviour). All these studies apply a scaling law for τE in their calculation of the operating points (dx/dt = 0, with x = [nj , T ] and j running over the different ion species), either implicitly by solving the equations numerically, or explicitly by eliminating τE from the equations by means of the scaling law. 5.2 Theory 111 We will follow the approach presented by Reiter et al. in [41], which was expanded by Rebhan et al. in [53, 65], and largely adopt the notation introduced therein. The power balance in a burning plasma is Pα = nD nT Eα hσvi = W + n2e Rrad = Pcond + Prad , τE (5.1) where nD and nT are the deuterium and tritium density, Eα is the alpha particle energy, hσvi the reactivity, ne the electron density, W = 3/2ne ftot T the internal energy of the plasma and Rrad can be interpreted as the Bremsstrahlung reactivity multiplied with the energy per interaction, defined by 2 1 4 Z 1 Rrad = CB T 2 fi gff + 4fα gff + ZfZ gff . (5.2) T T T Here fi = ni /ne = (nD + nT )/ne , fα = nα /ne , fZ = nZ /ne , T is the √ temperature and gff the Gaunt factor, which can be approximated by gff ≈ 2 3/π for fusion plasmas [69]. The quantitative error introduced by this approximation is of the order of 10%, but does not affect the stability or the dynamics of the system. From equation (5.1) the burn criterion can be derived: ne τE = 3 2 ftot T 1 2 4 fi Eα hσvi − Rrad , (5.3) with ftot = 1 (ne + ni + nα + nZ ) = 1 + fi + fα + fZ . ne (5.4) Note that ftot = 2 in case of a pure hydrogen plasma and limZ→∞ ftot = 1. A similar condition can be derived for the alpha particle fraction fα . Since fi = 1 − ZfZ − 2fα , the alpha particle balance is n e fα 1 2 2 ne fi hσvi = , 4 ρτE (5.5) which translates to ne τE = 4fα , ρfi2 hσvi (5.6) where we used the definition τp = ρτE . Equating (5.3) and (5.6) results in a cubic equation for fα , which can be solved to obtain fα as a function of T . The result is plotted in the left half of figure 5.1 for different values of ρ. Substituting the result in equation (5.6) yields a self-consistent expression of the burn criterion and the value of ne τE can now be plotted as a function of T Chapter 5 Sensitivity of burn contours to form of scaling laws ρ= 13 ρ ρ= = 9 13 5 = 9 0.2 ρ= ρ= fα 0.3 ρ 0 1 1016 1015 5 ρ 0.1 1017 3 ρ=0 0.4 ne τE (s/cm ) 112 = 1 ρ= 10 T (keV) 100 10 T (keV) 1014 100 Figure 5.1: Helium fraction (left) and ne τE plotted as a function of T for different values of ρ. to obtain so-called plasma operating point contour (POPCON) plots, as shown in the right half of figure 5.1. For reactor design purposes, it is desirable to plot the operating points in the ne , T -plane instead of the ne τE , T -plane. This requires elimination of τE from the burn criterion and for want of a good description of the transport in a tokamak this can only be done by means of a scaling law. Most scaling laws for the confinement time in a tokamak take the general form τE = Knle P −m , (5.7) with P the external heating power delivered to the plasma and K a factor which is obtained from several machine parameters. For the well-known ITER89P and IPB98(y,2) scaling laws K is given by 0.85 1.2 0.3 0.5 0.2 0.5 K89 = 0.048IM R a κ B A K98 = 0.93 1.39 0.58 0.78 0.15 0.19 0.145IM R a κ B A . (5.8) (5.9) Here R is the major radius, a the minor radius, IM the plasma current in MA, κ the plasma elongation, B the applied toroidal field on axis and A the plasma ion mass in amu. The exponents of density and power are l = 0.1 and m = 0.5 for the ITER89 scaling and l = 0.41 and m = 0.69 for the IPB98(y,2) scaling. Note that the average ion mass A depends on the plasma composition, and is therefore not constant. 5.3 Results 113 There are two significant differences between the ITER89 and IPB98(y,2) scaling laws. Firstly, the former is based on L-mode plasmas, whereas latter is developed for H-mode. However, the most important difference for the problem at hand is that the IPB98(y,2) scaling law does not include radiation losses in the definition of the confinement time, where the ITER89 scaling does include these. This has implications for the form the power balance takes, and consequently for the calculation of the alpha particle content and the determination of the operating points. Henceforth we will use τ̃E to denote the confinement time including radiation losses. In [88] we presented an analytical expression for ne as a function of T using the IPB98(y,2) scaling: ne = fα ρK 1 1−2m+l m−1 1−2m+l m 1 2 (1 − 2fα − ZfZ ) hσvi (Eα V ) 1−2m+l , 4 (5.10) but we did not include the derivation, which can be found in appendix B. 5.3 Results 5.3.1 Operating contours Using equation 5.10 we can plot operating contours in the ne , T -plane, as is done in figure 5.2 for the PPCS A design at ρ = 5 and 10. Simply following the math has resulted in nicely closed operating contours, but they extend to either very high or very low densities. The high density points can be discarded on the grounds of being above the Greenwald density, the β-limit or both, but this is not the case for the low density points. Looking at equation (B.9), it is apparent that the confinement time scales with nel−2m which is n−0.97 for the IPB98(y,2) scaling law. Of course there is e some effect from the variation in helium concentration over a burn contour, but the main trend is determined by the density. Lower densities therefore result in longer confinement times which, in combination with the alpha concentration at intermediate temperatures, results in burn equilibria that extend to extremely low densities. The obvious thing to try to put a lower limit on the accessible density is to look at the neo-classical confinement time, because this is the absolute upper limit on confinement in a tokamak. So if with decreasing density the value of τE predicted by the scaling laws at some point exceeds the value of τENC , this puts a lower limit on the density of the operating contours. However, as shown in section C, τENC also scales linearly with density and exceeds τE for all densities. A second thing to look at is the LH-transition power threshold: when the heating power becomes too low, the plasma will lose the H-mode confinement. 114 Chapter 5 Sensitivity of burn contours to form of scaling laws 1045 ρ 1040 = 5 ne (m−3 ) 1035 10 ρ= 30 10 1025 nG 1020 β-limit 1015 4 5 6 7 8 910 20 T (keV) 30 40 50 70 Figure 5.2: The operating contours for the PPCS A design [70, 71], for ρ = 5 and 10. The contours extend to extremely high and low densities. The solution might be trusted in the the density range on which the scaling law was based, i.e. around the Greenwald limit and about a decade below. But also in this range the near-degeneration of the solution leads to virtually vertical contours, and there obvious way to tell where the solutions are no longer valid. In other words, from the point of view of the scaling law, there is no good reason why burn could not be achieved at densities of 1019 m−3 or even lower. 5.3 Results 115 This threshold has a roughly linear dependence on density [75], as opposed to the alpha power which scales with density squared. Hence this puts a lower limit on the accessible density range. Another factor to take into account is the alpha slowing down time, which also depends on density. At first glance, the relevant time scale appears to be τp , but because τp is mainly determined by the edge recycling and alpha particles cannot be recycled without losing their energy, it is actually τE that matters. In other words: the alpha particles need to transfer their energy to the plasma before hitting the wall. 5.3.2 Density and power coupling The more fundamental problem stems from the coupling between power en density in a burning plasma. In present day fusion reactors, the heating power and density can be chosen independently. Consequently, the confinement database is populated with shots for which there is no coupling between density and power. In a burning plasma, this is not the case. In the absence of external heating power the temperature cannot be influenced directly, only the density is under control of the operator. When using the expression for the alpha power to eliminate the energy confinement time from the power balance (with the purpose of expressing ne as a function of T ), the fraction 1/1 − m2 + l shows up on the temperature side of the equation. Here m and l are respectively the exponents of the power and electron density in the scaling law for τE . Looking at equation (B.9), it is immediately obvious that there is a singularity at z = 1 − 2m + l = 0. At this point equation (5.10) is no longer valid, and has to be replaced by an expression from which the density dependence has disappeared: K 1 2 f hσviEα V 4 i −m = 4fα . ρfi2 hσvi (5.11) In this expression, fi , fα and hσvi are all functions of T , so this is one equation for one variable, T . It can be rearranged to K Ak 4 V Eα −m = m−1 4fα 2 f hσvi , ρAk i (5.12) where the left hand side is constant (the factor A−k gets rid of the ion mass dependence which was included in K) and determined by the reactor parameters. The right hand side is reactor independent and only a function of T . Depending on the reactor properties, this equation has two, one or no solutions, because changing the reactor parameters changes the value of V and K, which determines the intersections with the closed contours described by the expression on the right 116 Chapter 5 Sensitivity of burn contours to form of scaling laws hand side. If the solutions exist, these are independent of density, meaning that the burn contours have degenerated into vertical lines in the ne , T -plane. A similar exercise can be performed for a scaling law including radiation losses. First an expression for ne similar to equation (5.10) can be derived using the ITER89P scaling law (see Appendix B), yielding: ne = Em K 1 1−2m+l 4fα ρfi2 hσvi 1−m 1−2m+l (5.13) , with E = 32 ftot T 2π 2 κa2 R = 32 ftot T V . In this case the expression at the singularity reads K Em 1 1−m = 4fα , ρfi2 hσvi (5.14) which can also be rewritten to have only reactor dependent, constant term on the left and reactor independent terms on the right hand side: K Vm 1 1−m = 3 ftot T 2 m 4fα ρfi2 hσvi 1−m . (5.15) It turns out that the different scaling laws in the ITER physics basis are all close to the singularity. As a matter of fact, the IPB98(y,4) scaling is located on the other side of the singularity than the other four scaling laws, which results in a burn contour that is ’mirrored’ along the density axis compared to the other (flipped up-down around approximately the Greenwald density). Table 5.1: Predicted values of τE for ITER [26] Scaling IPB98(y) IPB98(y,1) IPB98(y,2) IPB98(y,3) IPB98(y,4) τE 6.0 5.9 4.9 5.0 5.1 The predicted confinement times from the different τE scalings in the ITER physics basis are very similar (see table 5.1). Yet the fact that the scaling laws are so close to this singularity means that small variations in the exponents l or m have a major impact on the operating contours. Approaching the singularity results in a stretching of the burn contours along the density axis, until at the singularity the operating points no longer form contours but instead degenerate into two 5.3 Results 117 vertical lines, or isothermals. The positions of these lines on the temperature axis is determined by the solutions to equation (5.12) or (5.15). Because the right hand side of these equations is reactor independent, the positions of the solutions is determined by the left hand side, i.e. the reactor parameters. For large enough reactors there will be two solutions and for reactors that are not capable of ignition there are none. In between there are reactors that have a ’critical size’, where both solutions coincide and there is only one operating temperature. Ergo, for a large enough reactor, a hypothetical scaling law for τE with the values of l and m such that z = 0 would provide reasonable predictions for τE . Yet the burn contours would consist of two vertical lines in the ne , T -plane, and ignition could be achieved at any power and density between the LH-transition threshold, and the Troyon and Greenwald limits. While the scaling laws yield similar values for the operating temperature around the Greenwald density, which is to be expected because the data they are based on contain mostly points in this region. However, the value of dne /dT varies greatly between the different scalings and changes sign when crossing the singularity. This is an issue because the required response of a control system depends on the value of dne /dT : in the case of IPB98(y,4) an increase in density will result in a decrease in temperature on the stable burn branch, which is the opposite of what is currently expected. A further illustration of the problem can be seen in figure 5.3, which displays the burn contours of the ITER ID design for ρ = 3 using the IPB98(y,2) scaling law, and but for two different exponents l of the electron density: the original value n0.41 e . We deliberately chose the ITER ID design because a slight different value n0.35 e the IPB98(y,2) scaling law predicts that it will not ignite. However, choosing a different reactor design does not change the analysis below in any meaningful way. The figure also shows the Greenwald density limit of nG = Ip /πa2 and the Troyon pressure limit βmax ≡ 0.072ε(1 + κ2 )/2 for a pure hydrogen plasma.1 For the original value of l, the operating contour looks like we expect it to look. The ITER ID design is slightly too small to achieve ignition, so the minimum density on the contour lies above the Greenwald and Troyon limits. However, for l = 0.35 the situation looks completely different: all of a sudden the ITER ID design does ignite, and the maximum density at which it ignites lies well below the Greenwald and Troyon limits. And the change in predicted confinement (for the same density and power of course) is just a few percent, depending on the exact value of the density. To demonstrate the effect of variations in l or m, figure 5.4 shows the minimal density at which the ITER ID design ignites for the IPB98(y,2) scaling, as a 1 The exact plasma β at a given electron density depends on the corresponding ion density, which can only be determined on a burn contour. For all other points in the ne , T -plane this depends on the trajectory in phase space taken by the plasma. 118 Chapter 5 Sensitivity of burn contours to form of scaling laws 1065 1050 1035 ) τE τE ∝ n0 1020 .69 1P 0.4 e ∝n −3 ne (m −0 nG β-limit . e 35 105 P −0. 69 10−10 10−25 5 10 20 T (keV) 50 100 Figure 5.3: Burn contours for the ITER ID design for ρ=3, using the IPB98(y,2) scaling but with varying exponents l for the density (l = 0.41 and l = 0.35). This small change in density dependence τE has a dramatic effect on the predicted burn equilibria. 5.4 Discussion and conclusions 119 function of the density exponent l (left plot) and power exponent m (right plot). m = 0.705 l = 0.38 ne 1025 1020 1015 0.3 0.35 0.4 l 0.45 0.5 0.6 0.65 0.7 m 0.75 0.8 Figure 5.4: The maximum (red curves) and minimum (blue curves) density at ignition as function of l (left plot) and m (right plot). Small changes in the values of l and m lead to large changes in the predicted density at ignition. The red and blue curves represent the maximum and minimum of the dashed and solid contours in figure 5.3 respectively. The left and right plot are not quite mirror images because of the appearance of m in the numerator of the exponent in expression 5.10. Since there is no obvious physical reason why 2m − l > 1 is not allowed, the solutions on the ’other side’ of the asymptote cannot be disqualified at this point. Yet it seems unlikely that such small changes to the scaling laws, which are well within the error margins, can have such enormous effects on the operating points in a burning plasma. 5.4 Discussion and conclusions When applying the current τE scaling laws to burning plasmas, this leads to predictions for the operating points at extremely high and low densities. The high density operating points can be discarded because they lie above the Greenwald and Troyon limits, but this is not the case for the low density points. Of course, the LH-transition power threshold, synchrotron radiation losses and alpha slowing down time will put a lower limit on the density, but this only happens below ne ≈ 1019 m−3 . We suspected that including neoclassical transport explicitly might solve this problem, but this is not the case as can be seen in section C. While it can be argued that this a mathematical artefact that is of little consequence for real world applications, the predictions for the operating points are also extremely sensitive to small variations in the exponents l and m of ne and P in 120 Chapter 5 Sensitivity of burn contours to form of scaling laws the scaling law. We have used the expression for ne as a function of T in a burning plasma to show that small changes in l and m lead to big changes in the predicted minimum (or maximum) density on a burn contour. Moreover, for l + 1 = 2m, the operating contours degenerate into two vertical lines in the ne , T -plane, which means that the operating points have become independent of density. In other words, ignition is possible at arbitrarily low densities and fusion power. Unfortunately, we have not been able to identify a physical mechanism that could be the reason for this problem. In a burning plasma, the density and temperature (and therefore the fusion power) are coupled, and one cannot be changed without affecting the other. In fact, al else staying constant, the operator can only change the temperature in a burning plasma by changing the density. This coupling is of course absent in the confinement database. In recent years considerable effort has been spent on the development of two term scaling laws [92], combining a core and pedestal scaling. However, these still treat density and power as independent parameters and will therefore fundamentally suffer from the same problems, although the precise value of the density and power exponents might be further removed from the singularity than is the case for the ITER scaling laws. For a more detailed treatment of these two term scaling laws see section D. We have therefore, unfortunately not been able to find a solution to this problem. It would be interesting to have shots in the confinement database that have densities, temperatures and heating levels that are expected in a burning plasma. Since these do not exist by definition, one could look for shots that would be operating points if a hypothetical fusion reaction was used that delivers more energy to the plasma. For instance, if the alpha particles in the DT reaction had an energy of 5 MeV, or even 10 MeV, some present day devices would be capable of achieving ignition. Looking at the energy confinement scaling of shots that would have burned if that were the case might shed some light on the expected confinement scaling in burning plasmas. A second suggestion is to mimic a burning plasma in, for instance, JET, by coupling the heating systems to the temperature and density in a feedback loop, with a gain factor to compensate for the fact that JET does not ignite. 121 Chapter 6 Discussion and conclusions At the beginning of this thesis several research questions were formulated. This chapter will provide an answer to these questions using the results presented in the previous chapters, and subsequently discuss these answers in the broader context of developing electricity producing fusion reactors. The overarching research question of the thesis was What are the properties of burn equilibria in fusion reactors? Because of the general nature of this question, several sub questions were raised whose answers, when combined, provide a good overview of the properties of burn equilibria. What parameters determine the position of the burn equilibria in operating space and how sensitive is the system with respect to these parameters? The burn criterion for a pure DT plasma is a well known result in literature, but little work has been done on burning plasmas with a self consistent helium treatment. Taking helium accumulation into account leads to closed burn contours in the ne τE , T -plane which are completely determined by the temperature [41], and these contours can be translated to the ne , T -plane [65]. The work presented in chapter 3 shows that burn contours in the ne , T -plane are exactly the same for all reactors that obey the same τE scaling law, apart from a scaling factor that is a function of the engineering parameters of the reactor. Consequently, the burn equilibria in different reactors will coincide when plotted on a normalised density scale. Only the position relative to the Greenwald en β limits will differ because these depend on the engineering parameters. Figure 6.1 plots two such universal operating contours in the normalised density and T -plane. Note that we cannot indicate the Greenwald density and β-limit in this plot, since these are reactor specific. Also, the contours extend over many orders of magnitude in the density, which is an artefact of the mathematical form of the scaling law. 122 Chapter 6 Discussion and conclusions = ρ = 10 1030 1020 ρ normalised density 5 1040 1010 1 1 10 T (keV) 100 Figure 6.1: The generic operating contours for any fusion reactor that follows the IPB98(y,2) scaling for τE for ρ = 5 and 10. How does the power output of a burning plasma respond to changes in energy confinement or particle transport? There are several parameters that affect the position of the equilibria and the fusion power at these points. The most interesting parameters from a reactor design perspective are the H-factor and ρ = τp /τE . The density also plays a role of course, but since the fusion power scales quadratically with density over most of the operating range, it is desirable to choose a point close to the density limit. For most of the past sixty years the fusion community has focussed on increasing the energy confinement time, and considerable gains have been made. In fact, for a given reactor design, the confinement time predicted by the IPB98(y,2) scaling is only 10 to 30 percent below the value at which maximum fusion power is achieved. Increasing it beyond that value will result in lower power output for a given reactor design, although further improvements in energy confinement will allow the construction of smaller reactors. The power output of a reactor also depends strongly on the helium accumulation in the plasma, which depends on both the temperature and ρ. For a given density, the output power on the high temperature side of a burn contour scales roughly inversely with ρ, creating an incentive to keep ρ as low as possible. Since the value of ρ is mainly determined by the helium recycling at the plasma edge, increasing the pumping capacity at the divertor might offer a possibility for burn control through the particle confinement time τp . Minimising τp might seem desirable from a power balance point of view, but it comes at a price. The tritium burn up fraction is a critical parameter in reactor 123 design and this is closely linked to the tritium confinement time. When this becomes too low, the tritium has to be recycled too often, resulting in unacceptably high tritium losses which cannot be replaced because there is an upper limit on the tritium breeding ratio that can be achieved. In our model we have assumed that the confinement time for all particle species is the same, which is not necessarily true. As already mentioned, a major factor is the recycling at the plasma edge and this might well be different for different particle species. Identifying or creating a mechanism that allows a controlled increase in tritium recycling (or decrease in helium recycling) would offer interesting possibilities to increase the power production of a fusion reactor without impeding its tritium breeding capacity. What are the stability properties of the operating points? In chapter 4 we have derived an analytical expression for the Jacobian of the system of burn equations for a plasma that obeys a scaling law for the energy confinement time of the form τE = KAk nle P m , which is the form of all scalings in the ITER physics basis. We did this for both the full system of burn equations, and a reduced system that contains only expressions for the helium density and the temperature. This requires the assumption of a constant electron density and nD = nT . The properties of the reduced system are governed by ne and ρ, whereas for the full system it is sD , sT and ρ. While the system exhibits a complex stability diagram and features interesting transitions and bifurcations, the reactor relevant operating points are stable, except for very high values of ρ. The addition of a significant amount of external heating, as foreseen for the reactors in the PPCS, stabilises also these operating points. Only for PPCS model D does the stability of the operating point remain a concern. Using a different scaling law has a negligible effect on the stability properties of the operating points. It seems reasonable to assume that some form of burn control will be necessary, if only to maintain the fusion power at the desired level. In that case, most likely a stable and robust operating point can be found for all values of ρ, where even a sizeable disturbance of the equilibrium does not lead to a violation of an operational limit on a timescale that cannot be dealt with by the control systems. A point of consideration is the start-up of the plasma. To minimise the required amount of external heating, a pure DT start is desirable, but this will result in an overshoot of the temperature that could cause the plasma to exceed the Troyon limit. This might be circumvented by injecting helium or changing the DT fuelling ratio once the ignition temperature has been reached. How sensitive are the burn equilibria to errors in the energy confinement time scaling laws? As is shown in chapter 5, great care has to be exercised in using the current τE scaling laws for burning plasmas. The expression for the electron density as a function of T that can be derived from the scaling laws contains the possibility for 124 Chapter 6 Discussion and conclusions a singularity, which depends on the precise value of the exponents of ne and P . All five scaling laws in the ITER physics basis happen to have their exponents very close to the critical values, and the singularity lies well within the error margin of the fit. The singularity arises from the coupling of density and power in a burning plasma, which is not present in the τE database on which the scaling laws are based. The physical interpretation of the singularity is a decoupling of density and temperature, meaning that the burn contours degenerate into two vertical lines in the ne , T -plane. In other words: a reactor capable of ignition would either burn at constant temperature, regardless of the plasma density. Given the hypersensitive response of the operating points to changes in the scaling law, it seems prudent to investigate this issue more thoroughly and put the energy confinement scaling laws for burning plasmas on a firmer basis. However, the database has to contain shots where the heating power, temperature and density correspond to an equilibrium in a burning plasma, even though the absolute value of the confinement time is too low for real burn to occur should the experiment have been carried out in a DT plasma. Because the energy confinement scaling fits the database rather well, it should also be able to describe these shots. With a new DT-campaign in JET in the works, it might be possible to perform some simulated burn experiments, by coupling the heating power to the temperature and density, possibly with a gain to compensate for the fact that JET is too small to achieve ignition. By performing such experiments at a range of densities, a few data points could be acquired to investigate the effect of the coupling between density and heating power in burning plasmas on the τE scaling in more detail. 125 Chapter 7 Outlook and recommendations From this work follow a few points of attention for reactor design. Looking at the reactor and the burn conditions from an integral systems perspective, it became clear in the course of writing this thesis that a successful fusion reactor needs to meet a number of conflicting requirements, which on their own might seem reasonable enough. For instance, the need for a high tritium burn up fraction, which arises from the need to keep the required tritium breeding ratio and the total tritium inventory as low as possible, calls for a long particle confinement time. Maximum power output, on the other hand, benefits from a low particle confinement time because it reduces the helium accumulation in the plasma. A high particle confinement time will also result in lower stability margins for the operating point, increasing the need for control. Similarly, we want a high power density, because the capital costs of a fusion reactor scale with the plasma volume. This is in direct conflict with the need to minimise the heat load on the divertor. In a commercial power plant the recirculated power must be kept as low as possible. Yet the power plant concepts foresee significant amounts of external heating, typically 5-10% of the fusion power. This is primarily needed to drive the plasma current, but does help to stabilize the burn and provide the operator with some level of control. Still, unless radical improvements in the current drive efficiency and/or the wall-plug efficiency of the current drive systems are made, such a level of external power necessarily leads to a recirculated power of tens of percents of the gross electric output power, to which all other power consumption – such as the power needed to pump the coolant and run the cryo-plant – still must be added. 126 Chapter 7 Outlook and recommendations Simple models, such as the zero-dimensional one used in this thesis, are very powerful in identifying such discrepancies, because they can easily be combined and run for a large range of input parameters. The results can then be used to target specific issues that are critical to the future success of fusion. Furthermore, the detailed physics models that physicists develop to describe more detailed phenomena that occur in fusion plasmas, are often of little use for the engineers that are tasked with developing control systems to regulate the power output or to stabilise the plasma. They are often looking for so-called ’OK’ models, which are comparatively easy to understand, run fast and that capture just enough of the physics to implement reliable model-based control systems. From that perspective, an effort to model the effects of self-heating and helium accumulation on reactor performance and energy confinement scaling, would be well spent, in preparation for the ITER DT campaign. A start could be made by developing a scaling law based on the points in the database that feature selfconsistent values of heating, density and plasma composition and see whether it suffers from the same issues as the current ITER scalings. The next step could be burning plasma simulation experiments in a JET DT-campaign, or the development of a scaling law from first principles, or maybe based on empirical relations from gyro-kinetic simulations. Beyond ITER we need to address the conflicting requirements that arise from the different challenges that a successful fusion reactor needs to overcome, using an integral systems perspective to find the compromise that is optimal. 127 Appendix A Partial derivatives for the Jacobian A.1 Reduced system To write down an expression for the Jacobian, we need to know the partial derivatives of the confinement time with respect to nα and T . For the energy confinement time we assume a power law of the form τE = K ∗ Ak nle P −m , (A.1) with K ∗ a constant depending on machine parameters. For the heating power P we substitute the alpha power Pα = 1 (ne − 2nα )2 hσviEα V, 4 (A.2) where V is the plasma volume. First of all we take the partial derivatives of the different terms in (A.1) and (A.2) with respect to nα , which we subsequently use for the derivative of the whole system. Then the process is repeated for the ∂/∂T terms. In the reduced system the electron density ne is kept constant, yielding the trivial result ∂ne /∂nα = 0. (A.3) The average ion mass A changes with the composition of the plasma (equation (4.15)) and its derivative with respect to nα is ∂A 1.5ne = . ∂nα (ne − nα )2 (A.4) 128 Appendix A Partial derivatives for the Jacobian For the fuel density it reads ∂nD nT 1 ∂ = (ne − 2nα )2 = −(ne − 2nα ). ∂nα 4 ∂nα (A.5) The reactivity only depends on the temperature and does not contribute to the ∂/∂nα terms. Using the above expressions, the derivative of the confinement time with respect to nα can now be written down explicitly 1.5ne 4m K ∗ Ak nle ∂τE k m + . (A.6) = 1 2 ∂nα A (ne − nα )2 (ne − 2nα ) 4 (ne − 2nα ) hσviEα V The partial derivative of the confinement time with respect to T is less complicated ∂τE = ∂T 1 4 (ne K ∗ Ak nle −m dhσvi m , hσvi dT − 2nα )2 hσviEα V (A.7) however, this includes the derivative of hσvi with respect to T . The often taken approximation hσvi ∝ T 2 is not valid in our case because we want to cover the complete temperature axis. Instead for hσvi we use the Bosch and Hale parametrization [4], which is valid from 0.5 keV up to 550 keV. r ξ hσvi = C1 θ e−3ξ , (A.8) mr c 2 T 3 2 1/3 BG ξ= , (A.9) 4θ T . (A.10) θ= T (C2 +T (C4 +T C6 )) 1 − 1+T (C3 +T (C5 +T C7 )) Taking the derivative involves applying the chain rule a few times: r dhσvi 3 ξ d ln θ 1 − 6ξ dξ −3ξ = C1 θ + − , e dT mr c 2 T 3 dT 2ξ dT 2T 1/3 dξ dξ dθ 1 BG dθ ξ dθ = =− =− dT dθ dT 3θ 4θ2 dT 3θ dT (A.11) (A.12) and 1 + 2C3 T + C23 − C2 C3 + C4 + 2C5 T 2 +2(C3 C5 − C2 C5 + C6 + C7 )T 3 + C25 − C4 C5 + C3 C6 − 3C2 C7 + 2C3 C7 T 4 dθ + 2(C5 − C4 )C7 T 5 + C7 (C7 − C6 )T 6 = . dT (−1 + T (C2 − C3 + T (C4 − C5 + C6 T − C7 T )))2 (A.13) A.2 Full system 129 For the radiation losses the partial derivatives with respect to T and nα are given by ∂Srad Srad = , ∂T 2T ∂Srad 5n2 Rrad (T, 1) . = e 3 2 ∂nα 2 ntot A.2 (A.14) (A.15) Full system The derivatives in the Jacobian with respect to T are of course (almost) the same as for the reduced system, but the derivatives with respect to nj are different. The derivatives of the energy confinement time now read ∂τE k ∂A K ∗ Ak nle l ∂ne m ∂nD nT = + − (A.16) m ∂nj A ∂nj ne ∂nj nD nT ∂nj (nD nT hσviEα V ) and (j − 1)nD + (j − 2)nT + (j − 3)nα + (j + 1 − mZ )nZ ∂A = , ∂nj (nD + nT + nα + nZ )2 2 ∂ne j − 3j + 4 , = ∂nj 2 j = 1; nT ∂nD nT = nD j = 2; ∂nj 0 j = 3. (A.17) (A.18) (A.19) The expression for the alpha heating power has changed to Pα = nD nT hσviEα V. Also the expression for the radiation losses needs to be modified: X Srad = ne nj Rradj , (A.20) (A.21) j √ Rradj = gff CB Zj2 T , (A.22) and consequently its derivatives have taken a slightly different form: ∂Srad Srad ∂ne = + ne Rradj . ∂nj ne ∂nj (A.23) 130 131 Appendix B Derivation of ne as function of T The idea of eliminating the confinement time from the power balance by means of a scaling law was presented and carried out for the ITER89 scaling by Rebhan et. al. [65], which yielded the following result: m τ̃E = Knl+1 e T . (B.1) It is possible to take the approach presented in [65] one step further and derive an analytical expression for ne as a function of T . Starting with the definition of τ̃E and equating that to the scaling law: τ̃E = W = P 3 2 ftot ne T P 2π 2 κa2 R = Knle P −m , (B.2) = Knle . (B.3) which can be written as W P 1−m Raising both sides to the power 1/1 − m and using the definition of τ̃E again m 1 l τ̃E W 1−m = K 1−m ne1−m , (B.4) and solving for τ̃E , the following expression is found τ̃E = K Em 1 1−m l−m ne1−m , (B.5) where we have defined E = 23 ftot T 2π 2 κa2 R = 32 ftot T V , with V the plasma volume. Appendix B Derivation of ne as function of T 132 Using expression (B.5), we can eliminate τ̃E from the burn criterion in the case the confinement time includes radiation losses, and be left with an expression for ne (T ). Using the burn criterion derived from the helium balance, equation (2.42), and substituting the expression for τ̃E yields # " 1 l−m 4fα K 1−m 1−m . (B.6) ne = ne Em ρfi2 hσvi From this it follows that ne = Em K 1 1−2m+l 4fα ρfi2 hσvi 1−m 1−2m+l (B.7) . The above approach only works in case the definition of the confinement time includes the radiation losses. If not, these need to be taken into account explicitly, which makes it impossible to eliminate the heating power using the definition of the confinement time analytically. Instead, for the IPB98(y,2) scaling, it can be done by inserting the expression for the alpha particle heating (plus any external heating if appropriate) into the scaling law. The expression for the alpha power reads Pα = 1 2 2 1 2 n f hσviEα = n2e (1 − 2fα − ZfZ ) hσviEα V, 4 e i 4 (B.8) with fi = (nD + nT )/ne = 1 − 2fα − ZfZ the fraction of fuel ions in the plasma and V the plasma volume. Inserting this in the equation for the confinement time yields τE = Knl−2m e 1 2 (1 − 2fα − ZfZ ) hσviEα V 4 −m . (B.9) 4fα . ρfi2 hσvi (B.10) Again inserting this expression into equation (2.42), we find ne τE = Kn1+l e 1 2 2 n (1 − 2fα − ZfZ ) hσviEα V 4 e −m = Some rearranging and eliminating fi from the last term, the gives the desired expression for ne (T ) ne = fα ρK 1 1−2m+l 1 2 (1 − 2fα − ZfZ ) hσvi 4 m−1 1−2m+l m (Eα V ) 1−2m+l . (B.11) 133 Appendix C Neoclassical confinement time Since the scaling law for the energy confinement time does not include an explicit description of neoclassical transport, we expected that setting a lower limit to the confinement time equal to the value predicted by neoclassical transport would increase the minimum density on a given burn contour. A simple estimate for the neoclassical energy confinement time is [7] τENC ≈ a2 , χNC i (C.1) with a the plasma minor radius and χNC the neoclassical temperature diffusion i coefficient for ions, which is given by [93] χNC = 0.68q 2 i R r 3/2 χCL i , (C.2) with R and r the major and minor radius, χCL = 0.10n20 /B 2 T 1/2 [7] and q the i safety factor. Since we use a 0D-model, we evaluate χNC at the plasma edge, so i we adopt the definition of q95 from the ITER physics basis [94]: q95 = 5a2 BT f, RIM with f a form factor to account for the shaping of the plasma 1 + κ2 1 + 2δ 2 + 1.2δ 3 1.17 − 0.65ε−1 f= . 2 2 (1 − ε−2 ) (C.3) (C.4) 134 Appendix C Neoclassical confinement time Here κ and δ are the plasma elongation and triangularity at q95 and ε = R/a is the aspect ratio. From equation (C.2) and the definition of χi CL we find that neoclassical transport scales linearly with ne , and therefore τENC ∝ n−1 e . Coincidentally, this is very close to the density dependence of τE ∝ n−0.97 in a burning plasma according to e the IPB98(y,2) scaling law (from equation(B.9) with l = 0.41 and m = 0.69). Because of this, τENC exceeds τE for every point on a burn contour and implementing neoclassical transport to be the lower limit for transport doesn’t resolve the issue. 135 Appendix D Alternative scaling for the confinement time The coupling between density and power in a burning plasma makes it desirable to have a scaling law for the confinement time that includes only one of these two, preferably the density. Such a scaling law would have to be relatively simple, but would ideally be based on a deeper understanding of the transport. Given the fact that future reactors are foreseen to operate in H-mode, we will focus our attention on these types of plasmas. In H-mode plasmas, there is a distinct difference between the edge transport barrier that is responsible for the pedestal and the transport in the core (be it with or without ITB). It has therefore been attempted to separate the pedestal from the core, by developing a two term scaling model that includes expressions for the energy content of the pedestal and of the core plasma. Cordey et. al. [92] have presented two different models for the thermal energy content of both the pedestal and the core (making 2 x 2 different combinations between them). The thermal conduction model, where it is assumed that the dominant loss term in the pedestal is heat conduction down the gradient, provides the following parametrisation for the pedestal energy: Wped = 0.000643I 1.58 R1.08 P 0.42 n−0.08 B 0.06 κ1.81 ε−2.13 A0.2 Fq2.09 , (D.1) where Fq ≡ q95 /qcyl and qcyl = 5κa2 B/RI. Even though in this model the density and power still appear together, the dependence on the density is rather weak. The other pedestal model presented in [92] considers the MHD stability limits to be the limiting factor for the pressure gradient in the edge, which has the same coefficient for the density, but exchanges power dependence for temperature dependence: ∗−0.08 0.2 2.29 −2.56 2.48 βped = 0.000833ρ∗0.27 A Fq ε κ , ped ν (D.2) 136 Appendix D Alternative scaling for the confinement time 1/2 2 with βped = Wped /RI 2 , ρ∗ped = Tpav /I, ν ∗ = nped R/Tpav , Cv = 0.92 the fraction of the total volume occupied by the pedestal (taking into account the pedestal width), Tpav = 2 × 102 Wped /Cv Vnped , and V the device volume. However, dropping ν ∗ from the scaling doesn’t significantly affect the results for reactor relevant machines and yields an root mean square error value of 25%, which is comparable to the 24% of the conduction model [92], and for our case has the benefit of removing the density dependence from the scaling altogether. We can therefore use the following scaling law for the pedestal pressure 0.2 2.18 −2.67 2.27 βped = 0.000643ρ∗0.3 Fq ε κ . ped A (D.3) Defining γ = 6.43 × 10−4 A0.2 Fq2.18 Rε−2.67 κ2.27 and using the definitions for ρ∗ped , Tpav and βped this can be written as 2 0.15 1.7 Wped = γρ∗0.3 =γ ped I = γTpav I 2 × 102 Wped Cv Vnped 0.15 I 1.7 . (D.4) Solving for Wped yields 1 Wped = γ 0.85 2 × 102 Cv Vnped 0.15 0.85 I2 (D.5) and using Wped = 3nped Tped Cv V as defined by Thomsen et. al. [95], this gives the following scaling for the pedestal temperature Tped = 0.15 1 1 − 0.6 −1.85 . γ 0.85 2 × 102 0.85 (Cv V) 0.85 I 2 nped 3 (D.6) Using this result, we can write down an expression for the energy confinement time in the pedestal that depends only on the density, and not on the power. While this is possible for the pedestal scaling, the core scalings presented in [92] do not allow this and therefore suffer from the same problem as the IPB98(y,2) scaling law, even though the exact value of the coefficients might be a bit further removed from the singularity. In order to completely remove this issue from the scaling law, a different approach to confinement in the core needs to be taken. 137 Bibliography [1] H. A. Bethe, “Energy production in stars,” Physical Review, vol. 55, no. 5, p. 434, 1939. [2] F. S. Troyon, R. Gruber, H. Saurenmann, S. Semenzato, and S. Succi, “MHD-Limits to Plasma Confinement,” Plasma Physics and Controlled Fusion, vol. 26, no. 1A, p. 209, 1984. [3] A. Einstein, “Ist die Trägheit eines Körpers von seinem Energieinhalt abhängig?,” Annalen der Physik, vol. 323, no. 13, pp. 639–641, 1905. [4] H.-S. Bosch and G. M. Hale, “Improved formulas for fusion cross-sections and thermal reactivities,” Nuclear Fusion, vol. 32, p. 611, 1992. [5] I. A. Crawford, “Lunar resources: A review,” Progress in Physical Geography, vol. 39, pp. 137–167, 2015. [6] G. G. Dolgov-saveliev, D. P. Ivanov, V. S. Mukhovatov, K. A. Razumova, V. S. Strelkov, M. N. Shepelyev, and N. A. Yavlinsky, “Investigations of the stability and heating of plasmas in toroidal chambers,” in Proceedings of the 2nd Geneva Conference on the Peaceful Uses of Atomic Energy, p. P/2527, 1958. [7] J. Freidberg, Plasma Physics and Fusion Energy. Cambridge: Cambridge University Press, 2008. [8] M. J. Greenwald, J. L. Terry, S. M. Wolfe, S. Ejima, M. G. Bell, S. M. Kaye, and G. H. Neilson, “A new look at density limits in tokamaks,” Nuclear Fusion, vol. 28, no. 12, pp. 2199–2207, 1988. [9] M. J. Greenwald, “Density limits in toroidal plasmas,” Plasma Physics and Controlled Fusion, vol. 44, pp. R27–R53, aug 2002. 138 Bibliography [10] M. A. Mahdavi, T. Osborne, A. W. Leonard, M. Chu, E. Doyle, M. Fenstermacher, G. McKee, G. Staebler, T. Petrie, M. R. Wade, S. Allen, J. Boedo, N. Brooks, R. Colchin, T. Evans, C. M. Greenfield, G. Porter, R. Isler, R. L. Haye, C. Lasnier, R. Maingi, R. Moyer, M. Schaffer, P. Stangeby, J. Watkins, W. P. West, D. G. Whyte, and N. Wolf, “High performance H mode plasmas at densities above the Greenwald limit,” Nuclear Fusion, vol. 42, no. 1, pp. 52–58, 2002. [11] X. Garbet, “Introduction to drift wave turbulence modelling,” Fusion Science and Technology, vol. 57, no. 2T, pp. 364–371, 2010. [12] W. Horton, “Drift waves and transport,” Reviews of Modern Physics, vol. 71, pp. 735–778, apr 1999. [13] X. Garbet, P. Mantica, C. Angioni, E. Asp, Y. Baranov, C. Bourdelle, R. V. Budny, F. Crisanti, J. G. Cordey, L. Garzotti, N. Kirneva, G. M. D. Hogeweij, T. Hoang, F. Imbeaux, E. Joffrin, X. Litaudon, A. Manini, D. C. McDonald, H. Nordman, V. Parail, A. G. Peeters, F. Ryter, C. Sozzi, M. Valovič, T. J. J. Tala, A. Thyagaraja, I. Voitsekhovitch, J. Weiland, H. Weisen, A. Zabolotsky, and the JET EFDA Contributors, “Physics of transport in tokamaks,” Plasma Physics and Controlled Fusion, vol. 46, pp. B557–B574, dec 2004. [14] A. Kendl, “Two-dimensional turbulence in magnetized plasmas,” European Journal of Physics, vol. 29, pp. 911–926, sep 2008. [15] Y. Xiao and Z. Lin, “Turbulent Transport of Trapped-Electron Modes in Collisionless Plasmas,” Physical Review Letters, vol. 103, p. 085004, aug 2009. [16] J. Wesson, The science of JET. Culham: JET Joint Undertaking, 2006. [17] F. Jenko, W. Dorland, and G. W. Hammett, “Critical gradient formula for toroidal electron temperature gradient modes,” Physics of Plasmas, vol. 8, no. 9, pp. 4096 – 4104, 2001. [18] T. W. Versloot, Edge rotation and momentum transport in JET fusion plasmas. PhD thesis, Eindhoven University of Technology, 2011. [19] F. Wagner, G. Becker, K. Behringer, D. J. Campbell, A. Eberhagen, W. Engelhardt, G. Fussmann, O. Gehre, J. Gernhardt, G. v. Gierke, G. Haas, M. Huang, F. Karger, M. Keilhacker, O. Klüber, M. Kornherr, K. Lackner, G. Lisitano, G. G. Lister, H. M. Mayer, D. Meisel, E. R. Müller, H. Murmann, H. Niedermeyer, W. Poschenrieder, H. Rapp, H. Röhr, F. Schneider, G. Siller, E. Speth, A. Stäbler, K.-H. Steuer, G. Venus, O. Vollmer, and Z. Yü, “Regime of Improved Confinement and High Beta in Neutral-BeamHeated Divertor Discharges of the ASDEX Tokamak,” Physical Review Letters, vol. 49, pp. 1408–1412, nov 1982. Bibliography 139 [20] K. H. Burrell, “Effects of E x B velocity shear and magnetic shear on turbulence and transport in magnetic confinement devices,” Physics of Plasmas, vol. 4, no. 5, pp. 1499–1518, 1997. [21] P. W. Terry, “Suppression of turbulence and transport by sheared flow,” Reviews of Modern Physics, vol. 72, pp. 109–165, jan 2000. [22] F. Wagner, G. Fussmann, T. Grave, M. Keilhacker, M. Kornherr, K. Lackner, K. McCormick, E. R. Müller, A. Stäbler, G. Becker, K. Bernhardi, U. Ditte, A. Eberhagen, O. Gehre, J. Gernhardt, G. v. Gierke, E. Glock, O. Gruber, G. Haas, M. Hesse, G. Janeschitz, F. Karger, S. Kissel, O. Klüber, G. Lisitano, H. M. Mayer, D. Meisel, V. Mertens, H. Murmann, W. Poschenrieder, H. Rapp, H. Röhr, F. Ryter, F. Schneider, G. Siller, P. Smeulders, F. X. Söldner, E. Speth, K.-H. Steuer, Z. Szymanski, and O. Vollmer, “Development of an Edge Transport Barrier at the H-Mode Transition of ASDEX,” Physical Review Letters, vol. 53, pp. 1453–1456, oct 1984. [23] H. Zohm, “Edge localized modes (ELMs),” Plasma Physics and Controlled Fusion, vol. 38, pp. 105–128, 1996. [24] A. Loarte, B. Lipschultz, A. S. Kukushkin, G. F. Matthews, P. C. Stangeby, N. Asakura, G. F. Counsell, G. Federici, A. Kallenbach, K. Krieger, M. A. Mahdavi, V. Philipps, D. Reiter, J. Roth, J. Strachan, D. G. Whyte, R. P. Doerner, T. Eich, W. Fundamenski, A. Herrmann, M. E. Fenstermacher, P. Ghendrih, M. Groth, A. Kirschner, S. Konoshima, B. LaBombard, P. Lang, A. W. Leonard, P. Monier-Garbet, R. Neu, H. Pacher, B. Pegourie, R. A. Pitts, S. Takamura, J. Terry, E. Tsitrone, and the ITPA Scrape-off Layer and Divertor Physics Topical Group, “Chapter 4: Power and particle control,” Nuclear Fusion, vol. 47, pp. S203–S263, jun 2007. [25] P. C. de Vries, E. Joffrin, M. Brix, C. D. Challis, K. Crombé, B. Esposito, N. C. Hawkes, C. Giroud, J. Hobirk, J. Lönnroth, P. Mantica, D. Strintzi, T. J. J. Tala, and I. Voitsekhovitch, “Internal transport barrier dynamics with plasma rotation in JET,” Nuclear Fusion, vol. 49, p. 075007, jul 2009. [26] ITER Physics Expert Groups on Confinement and Transport and Confinement Modelling and Database, ITER Physics Basis Editors, and ITER EDA, “Chapter 2: Plasma confinement and transport,” Nuclear Fusion, vol. 39, pp. 2175–2249, dec 1999. [27] B. B. Kadomtsev, “Tokamaks and dimensional analysis,” Soviet Journal of Plasma Physics, vol. 1, p. 295, 1975. [28] J. W. Connor and J. B. Taylor, “Scaling laws for plasma confinement,” Nuclear Fusion, vol. 17, no. 5, pp. 1047–1055, 1977. 140 Bibliography [29] P. N. Yushmanov, T. Takizuka, K. S. Riedel, O. J. W. F. Kardaun, J. G. Cordey, S. M. Kaye, and D. E. Post, “Scalings for tokamak energy confinement,” Nuclear Fusion, vol. 30, no. 10, 1990. [30] T. C. Luce, C. C. Petty, and J. G. Cordey, “Application of dimensionless parameter scaling techniques to the design and interpretation of magnetic fusion experiments,” Plasma Physics and Controlled Fusion, vol. 50, p. 043001, apr 2008. [31] E. J. Doyle, W. A. Houlberg, Y. Kamada, V. S. Mukhovatov, T. H. Osborne, A. R. Polevoi, G. Bateman, J. W. Connor, J. G. Cordey, T. Fujita, X. Garbet, T. S. Hahm, L. D. Horton, A. E. Hubbard, F. Imbeaux, F. Jenko, J. E. Kinsey, Y. Kishimoto, J. Q. Li, T. C. Luce, Y. R. Martin, M. Ossipenko, V. Parail, A. G. Peeters, T. L. Rhodes, J. E. Rice, C. M. Roach, V. Rozhansky, F. Ryter, G. Saibene, R. Sartori, A. C. C. Sips, J. A. Snipes, M. Sugihara, E. J. Synakowski, H. Takenaga, T. Takizuka, K. Thomsen, M. R. Wade, and H. R. Wilson, “Chapter 2: Plasma confinement and transport,” Nuclear Fusion, vol. 47, pp. S18–S127, jun 2007. [32] J. P. Christiansen, J. G. Cordey, K. Thomsen, A. Tanga, J. C. DeBoo, D. P. Schissel, T. S. Taylor, O. J. W. F. Kardaun, F. Wagner, F. Ryter, S. M. Kaye, Y. Miura, N. Suzuki, M. Mori, T. Matsuda, H. Tamai, T. Takizuka, S.I. Itoh, and K. Itoh, “Global energy confinement H-mode database for ITER,” Nuclear Fusion, vol. 32, no. 2, pp. 291–338, 1992. [33] K. Thomsen, D. J. Campbell, J. G. Cordey, O. J. W. F. Kardaun, F. Ryter, U. Stroth, A. Kus, J. C. DeBoo, D. P. Schissel, Y. Miura, N. Suzuki, M. Mori, T. Matsuda, H. Tamai, T. Takizuka, S.-I. Itoh, K. Itoh, and S. M. Kaye, “ITER H mode confinement database update,” Nuclear Fusion, vol. 34, no. 1, pp. 131–167, 1994. [34] G. Becker, “Empirical scaling laws for local transport in neutral beam heated plasmas,” Nuclear fusion, vol. 28, no. 8, pp. 1458–1459, 1988. [35] E. J. Synakowski, B. C. Stratton, P. C. Efthimion, R. J. Fonck, R. a. Hulse, D. W. Johnson, D. K. Mansfield, H. Park, S. D. Scott, and G. Taylor, “Measurements of radial profiles of He2+ transport coefficients on the TFTR Tokamak,” Physical Review Letters, vol. 65, no. 18, pp. 2255–2258, 1990. [36] E. J. Synakowski, P. C. Efthimion, G. Rewoldt, B. C. Stratton, W. M. Tang, B. Grek, K. W. Hill, R. A. Hulse, D. W. Johnson, M. W. Kissick, D. K. Mansfield, D. McCune, D. R. Mikkelsen, H. K. Park, a. T. Ramsey, M. H. Redi, S. D. Scott, G. Taylor, J. Timberlake, and M. C. Zarnstorff, “Helium, iron, and electron particle transport and energy transport studies on the Tokamak Fusion Test Reactor,” Physics of Fluids B: Plasma Physics, vol. 5, no. 7, pp. 2215–2228, 1993. Bibliography 141 [37] M. R. Wade, D. L. Hillis, J. T. Hogan, R. Maingi, M. M. Menon, M. A. Mahdavi, W. P. West, K. H. Burrell, P. Gohil, R. J. Groebner, R.-M. Hong, D. H. Kellman, J. C. Phillips, R. P. Seraydarian, the DIII-D Team, and D. F. Finkenthal, “Helium transport and exhaust studies in enhanced confinement regimes in DIII-D,” Physics of Plasmas, vol. 2, no. 6, pp. 2357–2365, 1995. [38] M. R. Wade, T. C. Luce, and C. C. Petty, “Gyroradius Scaling of Helium Transport,” Physical Review Letters, vol. 79, no. 3, pp. 419–422, 1997. [39] C. Angioni, a. G. Peeters, G. V. Pereverzev, A. Bottino, J. Candy, R. Dux, E. Fable, T. Hein, and R. E. Waltz, “Gyrokinetic simulations of impurity, He ash and α particle transport and consequences on ITER transport modelling,” Nuclear Fusion, vol. 49, p. 055013, may 2009. [40] S. S. Henderson, L. Garzotti, F. J. Casson, D. Dickinson, M. F. J. Fox, M. O’Mullane, A. Patel, C. M. Roach, H. P. Summers, and M. Valovič, “Neoclassical and gyrokinetic analysis of time-dependent helium transport experiments on MAST,” Nuclear Fusion, vol. 54, no. 9, p. 093013, 2014. [41] D. Reiter, G. H. Wolf, and H. Kever, “Burn condition, helium particle confinement and exhaust efficiency,” Nuclear Fusion, vol. 30, pp. 2141–2155, oct 1990. [42] K. H. Finken, “Particle, in particular helium removal experiments optimized by the dynamic ergodic divertor,” Fusion Engineering and Design, vol. 37, pp. 445–448, 1997. [43] A. Sakasai, H. Takenaga, N. Hosogane, S. Sakurai, N. Akino, H. Kubo, S. Higashijima, H. Tamai, N. Asakura, K. Itami, and K. Shimizu, “Helium exhaust in ELMy H-mode plasmas with W-shaped pumped divertor of JT-60U,” Journal of Nuclear Materials, vol. 266, pp. 312–317, 1999. [44] K.-D. Zastrow, S. J. Cox, M. G. von Hellermann, M. G. O’Mullane, D. Stork, M. Brix, C. D. Challis, I. H. Coffey, R. Dux, K. H. Finken, C. Giroud, D. L. Hillis, J. T. Hogan, K. D. Lawson, T. Loarer, A. G. Meigs, P. D. Morgan, M. F. Stamp, A. D. Whiteford, and JET EFDA Contributors, “Helium exhaust experiments on JET with Type I ELMs in H-mode and with Type III ELMs in ITB discharges,” Nuclear Fusion, vol. 45, no. 3, pp. 163–175, 2005. [45] H.-S. Bosch, D. P. Coster, R. Dux, G. Haas, A. Kallenbach, M. Kaufmann, K. Lackner, J. Neuhauser, S. D. P. Hempel, W. Poschenrieder, R. Schneider, the ASDEX Upgrade Team, and NI- ICRH- and ECRH Teams, “Particle exhaust studies in ASDEX Upgrade,” Plasma Physics and Controlled Fusion, vol. 39, pp. 1771–1792, 1997. 142 Bibliography [46] H.-S. Bosch, W. Ullrich, D. P. Coster, O. Gruber, G. Haas, J. Neuhauser, R. Schneider, and R. Wolf, “Helium transport and exhaust with an ITER-like divertor in ASDEX Upgrade,” Journal of Nuclear Materials, vol. 290-293, pp. 836–839, 2001. [47] M. Groth, P. Andrew, W. Fundamenski, H. Y. Guo, D. L. Hillis, J. T. Hogan, L. D. Horton, G. F. Matthews, A. G. Meigs, P. D. Morgan, M. F. Stamp, D. Stork, and M. G. von Hellermann, “Helium and neon enrichment studies in the JET Mark IIAP and Mark IIGB divertors,” Plasma Physics, vol. 42, pp. 591–600, 2002. [48] ITER Physics Expert Group on Divertor, ITER Physics Expert Group on Divertor Modelling and Database, ITER Physics Basis Editors, and J. ITER EDA, Naka Joint Work Site, Mukouyama, Naka-machi, Naka-gun, Ibarakiken, “Chapter 4 : Power and particle control,” Nuclear Fusion, vol. 39, no. 12, pp. 2391–2469, 1999. [49] D. Reiter, G. H. Wolf, and H. Kever, “Stationary burning: Analysis of profile effects on the required exhaust efficiency and the permitted helium particle confinement time,” Journal of Nuclear Materials, vol. 176-177, pp. 756–762, dec 1990. [50] M. H. Redi and S. A. Cohen, “Effects of particle transport on Helium ash accumulation and sustained ignition in the ITER design,” tech. rep., Princeton Plasma Physics Laboratory, Princeton, NJ, 1990. [51] D. Reiter, H. Kever, G. H. Wolf, M. Baelmans, R. Behrisch, and R. Schneider, “Helium removal from tokamaks,” Plasma Physics and Controlled Fusion, vol. 33, no. 13, pp. 1579–1600, 1991. [52] M. H. Redi, S. A. Cohen, and E. J. Synakowski, “Transport simulations of helium exhaust in ITER using recent data from TFTR, TEXTOR and JT-60,” Nuclear Fusion, vol. 31, no. 9, p. 1689, 1991. [53] E. Rebhan, U. Vieth, D. Reiter, and G. H. Wolf, “Effect of helium concentration on ignition curves with energy confinement time including radiation losses,” Nuclear fusion, vol. 36, pp. 264–269, feb 1996. [54] P. Leekhaphan, “Impurity Transport and Helium Accumulation in ITER,” Thammasat Int. J. Sc. Tech, vol. 15, no. 3, 2010. [55] R. V. Budny, R. Andre, G. Bateman, F. D. Halpern, C. E. Kessel, A. H. Kritz, and D. McCune, “Predictions of H-mode performance in ITER,” Nuclear Fusion, vol. 48, p. 075005, jul 2008. [56] A. H. Kritz, “Final Report for National Transport Code Collaboration PTRANSP,” tech. rep., LeHigh University, 2011. Bibliography 143 [57] D. Reiter and S. Wiesen, “Helium removal and recycling,” Transactions of fusion science and technology, vol. 49, pp. 248–257, 2006. [58] M. B. Kalinowski and L. C. Colschen, “International Control of Tritium to Prevent Horizontal Proliferation and to Foster Nuclear Disarmament,” Science and global security, vol. 5, pp. 131–203, 1995. [59] D. Mcmorrow, “Tritium,” Tech. Rep. November, The MITRE Corporation, 2011. [60] F. Wasastjerna, Using MCNP for fusion neutronics. PhD thesis, Helsinki University of Technology, 2008. [61] L. A. El-Guebaly and S. Malang, “Toward the ultimate goal of tritium selfsufficiency: Technical issues and requirements imposed on ARIES advanced power plants,” Fusion Engineering and Design, vol. 84, pp. 2072–2083, dec 2009. [62] U. Fischer, D. Große, P. Pereslavtsev, S. Stickel, H. Tsige-Tamirat, and V. Weber, “Neutronics design analyses of fusion power reactors based on a novel integral approach,” Fusion Engineering and Design, vol. 84, pp. 323–328, jun 2009. [63] U. Fischer, P. Pereslavtsev, D. Große, V. Weber, A. Li Puma, and F. Gabriel, “Nuclear design analyses of the helium cooled lithium lead blanket for a fusion power demonstration reactor,” Fusion Engineering and Design, vol. 85, pp. 1133–1138, dec 2010. [64] M. E. Sawan and M. A. Abdou, “Physics and technology conditions for attaining tritium self-sufficiency for the DT fuel cycle,” Fusion Engineering and Design, vol. 81, pp. 1131–1144, 2006. [65] E. Rebhan and U. Vieth, “Burn stability and a safe operating regime of a tokamak reactor with ITER scaling,” Nuclear Fusion, vol. 37, pp. 251–270, feb 1997. [66] A. Fasoli, C. Gormezano, H. L. Berk, B. N. Breizman, S. Briguglio, D. S. Darrow, N. Gorelenkov, W. W. Heidbrink, A. Jaun, S. V. Konovalov, R. Nazikian, J.-M. Noterdaeme, S. E. Sharapov, K. Shinohara, D. Testa, K. Tobita, Y. Todo, G. Vlad, and F. Zonca, “Chapter 5: Physics of energetic ions,” Nuclear Fusion, vol. 47, pp. S264–S284, jun 2007. [67] F. Albajar, J. Johner, and G. Granata, “Improved calculation of synchrotron radiation losses in realistic tokamak plasmas,” Nuclear Fusion, vol. 41, no. 6, pp. 665–678, 2002. 144 Bibliography [68] J. Garcia, G. Giruzzi, J. F. Artaud, V. Basiuk, J. Decker, F. Imbeaux, Y. Peysson, and M. Schneider, “Analysis of DEMO scenarios with the CRONOS suite of codes,” Nuclear Fusion, vol. 48, p. 75007, jul 2008. [69] J. Wesson, Tokamaks. Oxford: Oxford University Press, 2nd ed., 1997. [70] D. Maisonnier, I. Cook, S. Pierre, B. Lorenzo, D. P. Luigi, G. Luciano, N. Prachai, and P. Aldo, “DEMO and fusion power plant conceptual studies in Europe,” Fusion Engineering and Design, vol. 81, pp. 1123–1130, feb 2006. [71] D. Maisonnier, D. J. Campbell, I. Cook, L. Di Pace, L. Giancarli, J. Hayward, A. Li Puma, M. Medrano, P. Norajitra, M. Roccella, P. Sardain, M. Tran, and D. J. Ward, “Power plant conceptual studies in Europe,” Nuclear Fusion, vol. 47, pp. 1524–1532, nov 2007. [72] M. Wanner and the W7-X Team, “Design goals and status of the WENDELSTEIN 7-X project,” Plasma Physics and Controlled Fusion, vol. 42, no. 11, pp. 1179–1186, 2000. [73] Y. Igitkhanov, T. Andreeva, C. D. Beidler, E. Harmeyer, F. Herrnegger, J. Kisslinger, F. Wagner, and H. Wobig, “Status of HELIAS reactor studies,” Fusion Engineering and Design, vol. 81, pp. 2695–2702, nov 2006. [74] A. Dinklage, E. Ascasibar, C. D. Beidler, R. Brakel, J. Geiger, J. H. Harris, A. Kus, S. Murakami, S. Okamura, R. Preuss, F. Sano, U. Stroth, Y. Suzuki, J. Talmadge, V. Tribaldos, K. Y. Watanabe, A. Weller, H. Yamada, and M. Yokoyama, “Assessment of global stellarator confinement: Status of the international stellarator confinement database,” Fusion Science and Technology, vol. 51, pp. 1–7, 2007. [75] F. Ryter and the H-mode Threshold Database Group, “Progress of the international H-mode power threshold database activity,” Plasma Physics and Controlled Fusion, vol. 44, pp. A415–A421, 2002. [76] D. L. Hillis, K. H. Finken, J. T. Hogan, K. H. Dippel, R. A. Moyer, A. Pospieszczyk, D. Rusbüldt, K. Akaishi, R. W. Conn, H. Euringer, D. S. Gray, L. D. Horton, R. A. Hulse, R. C. Isler, C. C. Klepper, P. K. Mioduszewski, A. Miyahara, and G. H. Wolf, “Helium exhaust and transport studies with the ALT-II pump limiter in the TEXTOR tokamak,” Physical Review Letters, vol. 65, pp. 2382–2385, jun 1990. [77] W. A. Houlberg, S. Attenberger, and L. Hively, “Contour analysis of fusion reactor plasma performance,” Nuclear Fusion, vol. 22, pp. 935–945, jul 1982. [78] ITER-JCT and Home Teams (presented by G. Janeschitz), “Status of ITER,” Plasma Physics and Controlled Fusion, vol. 37, no. 11A, p. A19, 1995. Bibliography 145 [79] M. Fichtmüller, G. Corrigan, L. Lauro-Taroni, R. Simonini, J. Spence, E. Springmann, and A. Taroni, “Simulation of helium exhaust in JET and ITER,” Journal of Nuclear Materials, vol. 266-269, pp. 330–336, mar 1999. [80] S. J. Zweben, R. V. Budny, D. S. Darrow, S. S. Medley, R. Nazikian, B. C. Stratton, E. J. Synakowski, and G. Taylor for the TFTR Group, “Alpha particle physics experiments in the Tokamak Fusion Test Reactor,” Nuclear Fusion, vol. 40, pp. 91–149, jan 2000. [81] S. Brezinsek, A. Widdowson, M. Mayer, V. Philipps, P. Baron-Wiechec, J. Coenen, K. Heinola, A. Huber, J. Likonen, P. Petersson, M. Rubel, M. Stamp, D. Borodin, J. Coad, A. Carrasco, A. Kirschner, S. Krat, K. Krieger, B. Lipschultz, C. Linsmeier, G. Matthews, and K. Schmid, “Beryllium migration in JET ITER-like wall plasmas,” Nuclear Fusion, vol. 55, no. 6, p. 063021, 2015. [82] D. J. Ward, “The physics of DEMO,” Plasma Physics and Controlled Fusion, vol. 52, p. 124033, dec 2010. [83] Y. A. Kuznetsov, Elements of Applied Bifurcation Theory. Springer-Verlag, 2nd ed., 1998. New York: [84] Q. Kong, A Short Course in Ordinary Differential Equations. Springer, 2014. [85] Y. A. Kuznetsov, “Saddle-node bifurcation,” Scholarpedia, vol. 1, no. 10, p. 1859, 2006. [86] Y. A. Kuznetsov, “Andronov-Hopf bifurcation,” Scholarpedia, vol. 1, no. 10, p. 1858, 2006. [87] J. Guckenheimer and Y. A. Kuznetsov, “Bogdanov-Takens bifurcation,” Scholarpedia, vol. 2, no. 1, p. 1854, 2007. [88] M. Jakobs, N. Lopes Cardozo, and R. J. E. Jaspers, “Fusion burn equilibria sensitive to the ratio between energy and helium transport,” Nuclear Fusion, vol. 54, p. 122005, dec 2014. [89] J. D. Lawson, “Some Criteria for a Power Producing Thermonuclear Reactor,” Proceedings of the Physical Society. Section B, vol. 70, no. 1, p. 6, 1957. [90] Y. I. Kolesnichenko, S. N. Reznik, and V. A. Yavorskij, “Stability of the steady-state course of a thermonuclear reaction in an inhomogeneous plasma,” Nuclear Fusion, vol. 16, no. 1, pp. 105–111, 1976. [91] O. Mitarai and K. Muraoka, “Ignition analysis with the H-mode power threshold scaling in a DT tokamak reactor,” Plasma physics and controlled fusion, vol. 38, pp. 551–569, 1996. 146 Bibliography [92] J. G. Cordey, “A two-term model of the confinement in Elmy H-modes using the global confinement and pedestal databases,” Nuclear Fusion, vol. 43, pp. 670–674, 2003. [93] M. N. Rosenbluth, R. D. Hazeltine, and F. L. Hinton, “Plasma Transport in Toroidal Confinement Systems,” Physics of Fluids, vol. 15, no. 1, pp. 116–140, 1972. [94] ITER Physics Basis Editors, ITER Physics Expert Group Chairs and CoChairs, ITER Joint Central Team and Physics Unit, and ITER EDA, Naka Joint Work Site, Mukouyama, Naka-machi, Naka-gun, Ibaraki-ken, Japan, “Chapter 1: Overview and summary,” Nuclear Fusion, vol. 39, pp. 2137–2174, dec 1999. [95] K. Thomsen, J. G. Cordey, the H-Mode Database Working Group, and the Pedestal Database Working Group, “Development of a two-term model for the confinement in ELMy H-modes using the global confinement and pedestal databases,” Plasma Physics and Controlled Fusion, vol. 44, pp. A429–A435, may 2002. 147 Acknowledgements Writing a PhD thesis is a long and arduous road, which I most likely wouldn’t have completed without the support of colleagues, friends and family. Although it is impossible to list everyone who has made a contribution in one form or another, there are some people who I think deserve a special mention here. First of all I would like to express my gratitude and great appreciation to my promotor, Niek Lopes Cardozo. Niek, throughout my professional, and also personal, struggles of the last few years, you always remained positive and supportive. I especially enjoyed the ’thinking out loud’ sessions we had in front of a white board, and the many discussions on issues that were not necessarily related to fusion. I also want to thank my co-promotor Roger Jaspers, who has supervised me since I did my internship in China nine years ago and taught me many of the things I know about fusion. Roger, you have been a constant factor throughout this long journey, and were always prepared to make time to help me, even though I had long since abandoned my original topic. Leon Kamp, my other co-promotor, always found time for my questions and greatly helped me during my brief foray into liquid metal flows. Leon, I greatly appreciate your unwavering focus on the physics questions and your rigorous approach for every problem. A great thank you to all my (former) colleagues who made my time at Eindhoven University of Technology a pleasant one: Clazien and Hélène for their helpful support in all practical matters, Herman for his assistance with the liquid metal experiment and his limitless supply of interesting facts, Hans for his encouraging enthusiasm, Maarten for sharing his extensive knowledge on plasma physics in general and plasma rotation in particular, and Mark for his probing questions and unfailing ICT support. I want to express my gratitude towards the PhD students, both at DIFFER and at the TU/e, that I had the privilege of calling my peers: Ephrem, Geert Willem, Menno, Willem, Wolf, Bram, Rianne, Pieter Willem, Matthijs, Vitor and others, thank you for your help and the constructive discussions we had. The 148 Acknowledgements same goes for the staff at DIFFER, or FOM Rijnhuizen as it was called when I started: Tony, Marco, Hugo, Peter, Egbert and Dick, who helped sharpen my ideas and understanding of fusion. At the TU/e my appreciation also extends to Maarten Steinbuch for coaching me when needed. Furthermore, I want to thank my students Stefan, Laszlo, Kevin, Benjamin, Selwyn, Wouter and Peter for all their hard work, and all other students who shared my office for their company and entertainment. Then there are the people I am lucky to call my friends. Athina, Jonathan and Thijs, thank you for all the professional and personal talks we shared. Gillis, thank you for the many lunches, bike rides and holidays, and for taking the initiative to meet when I didn’t. Gar, you made some of the lonely working hours less lonely. Hjalmar, I enjoyed the many talks on topics that no one else seems to care about. Margit, I took great pleasure in discussing the big question of life with you. Eveline, your encouragement was of great help in the final push to complete this thesis. Saskia, Leon and Jisse, you have been great friends despite the long periods of, sometimes one-sided, radio silence. Jeroen, Erik, Eric, Toine: the shared holidays, bike rides and discussions are unforgettable. Maaike, you have taught me many things with your energy and lust for life, and I cherish fond memories of our shared experiences. Mirja, of all my friends you probably understand me best. Your continuous support, both close and from further away, means a lot to me and I admire your caring and selfless attitude. Finally, I want to thank my parents Anthonie and Marleen, and my sisters Arwen en Niniane for their unwavering support. Arthur, thank you for being not only my brother, but also my best friend. The many hours we spent on our bikes, the many holidays together, but especially the many deep conversations we had have helped me reach this point. 149 Curriculum Vitae I was born on the 28th of April 1982 in the city of Eindhoven, the Netherlands, and moved to Bakel shortly after my fourth birthday. I attended primary school at the Vrije School Peelland in Helmond, and completed my Waldorf education at the Vrije School Brabant in Eindhoven and, after moving to Terhorst, at the Bernard Lievegoed School in Maastricht. After obtaining my VWO diploma at the Montessori College in Maastricht, I started studying Applied Physics and Applied Mathematics at Eindhoven University of Technology, but quit the latter after completing the propedeuse (first year). In 2008 I obtained my BSc in Applied Physics, for which I made sequential images on the breakdown of electric discharges in different gases at different conditions in a non-planar geometry. For my Master in Applied Physics I chose the plasma and radiation technology track and did an internship at the South Western Institute of Physics in Chengdu, China, where I investigated the hydrogen/deuterium ratio in the HL-2A tokamak. Other work included the design of optics for the charge exchange recombination spectroscopy system and assisting in installing the neutral beam injection system. My graduation project, on the interaction between magnetosonic Whistler waves and runaway electrons during disruptions in tokamaks, I did at Chalmers University in Gothenburg, Sweden. Upon obtaining my master’s degree I started my PhD project at Eindhoven University. After initial forays into the interaction between plasma shear flow and turbulence, and free surface liquid metal MHD flows, this culminated in the thesis you are currently reading, which focusses on equilibria in burning fusion plasmas. 150