Why carbon dioxide makes stellarators so important

Allen H. Boozer Columbia University, New York, NY 10027 [email protected] (Dated: March 9, 2020) The increasing level of atmospheric carbon dioxide has driven public discourse throughout the world. An immediate implementation of carbon-free energy sources is demanded with little discus- sion of costs, technical constraints on the sources, or implications of high residual levels of carbon dioxide. Residual carbon-dioxide can be removed from the air, but the cost to remove the carbon- dioxide produced by human activity during a year is thought to be trillions of dollars—otherwise it remains in the atmosphere for centuries. Economic considerations may limit wind and solar sources to less than 40% of the electricity production. Fission or fusion may be the only choice for most of the rest. Development costs are orders of magnitude smaller than implementation costs, which are tens of trillions of dollars for fission. A needless delay in the development of fusion has enormous financial implications. As will be shown stellarators are better positioned than any other concept for a fast path to fusion. A computationally derived conceptual design for a stellarator reactor may allow final design and construction to be initiated without the delay of intermediate generations of experiments. The most urgent issue is the development of conceptual designs.

I. INTRODUCTION enormous financial costs while exposing the world to the risks of elevated carbon-dioxide levels. Such options can have a large moral hazard by blocking Societal risks associated with the increase in at- the rapid development of more effective options. mospheric carbon dioxide make the rapid develop- ment of fusion energy compelling. This is empha- The prominently discussed options are not solu- sized by a 2019 report of the Organization for Eco- tions in the sense of returning carbon-dioxide to a nomic Co-operation and Development (OECD) [1], pre-industrial-revolution level. They seek to only which found a large cost penalty when wind and so- limit the carbon-dioxide increase to a tolerable level. A tolerable level [2] is often taken to be consistent lar exceed 30% to 40% of the electricity-generation o fraction. Of all fusion concepts the stellarator ap- with a temperature increase of 2 C. But, the tem- pears best poised for rapid development. perature increase that will occur under various sce- narios is uncertain as are the effects, detrimental and The increase in atmospheric carbon-dioxide is un- beneficial, that arise at various carbon-dioxide lev- leashing powerful political forces but has aroused els. The options for a rapid transition to carbon-free little interest in determining the options that sci- electricity production frequently ignore the OECD ence could offer—neither carbon-dioxide removal nor limit [1] on the fraction of the electricity-generation carbon-free energy sources, such as fusion. As will be that can be economically produced by solar and shown, the development of options costs several or- wind, 30% to 40%. Whatever strategy is adopted, it ders of magnitude less than their deployment. This should be consistent with an increasing use of energy. enormous cost ratio makes it irrational to implement For example, Table 49 in [4] shows an approximate carbon-dioxide mitigation without also having re- 4% annual increase in the electricity-generation ca- search focused on the fastest possible development pacity of the world. of better options. When elevated levels of atmospheric carbon diox- The risks associated with the increase in carbon- ide are perceived to have dangers, the cost of removal dioxide are sometimes described in apocalyptic of the carbon dioxide that humans place in the at- terms—neither limitations of finance nor on gov- mosphere [2] during a year, ∼ 50 Gt, defines the

arXiv:1912.06289v3 [physics.plasm-ph] 6 Mar 2020 ernmental power should be allowed to stand in the financial risk of not moving to carbon-free energy way of implementing emergency measures. The term sources. As discussed in Chapter 1 of [2], the natu- “moral hazard” means the encouragement of risk ral removal of carbon-dioxide from the atmosphere over responsibility. The development of science- requires centuries [2]. based options can be viewed as a moral hazard by The cost of removal is not the standard method removing the mandate for immediate implementa- of estimating the cost of carbon-dioxide emissions. tion. The standard method is the Social Cost of Carbon Nevertheless, ill-considered options can entail (SCC), which uses models to estimate financial im- plications: damages and benefits. A recent review less than seven. The Apollo program to land and of the literature [3] finds the mean value of the es- return a person from the moon was announced in timates of the social cost of carbon-dioxide emission May 1961 and reached its goal in July 1969, just over is ∼$50/t. When a discount rate is included that eight years. The time commonly envisioned for ad- cost is reduced to ∼$30/t. The range of estimates is dressing the carbon-dioxide problem is thirty years large, but using the $30/t estimate, the damage pro- [1, 2], which includes the time for deployment as well duced by emitting 50 Gt is approximately $1.5 tril- as for the development of a solution. lion. The OECD report [1] discussed carbon-dioxide What are the options for carbon-free energy pro- pricing at $50/t as a policy instrument to internalize duction? Energy from nuclear fission is carbon free, costs and achieve a low emission outcome. The cost but has proliferation, radioactive waste, and safety of 50 Gt at $50/t is $2.5 trillion. issues. Wind and solar provide cost-effective carbon- Removal is the only ensured way to avoid effects free options, but both suffer from site specificity, in- of carbon-dioxide emission. Direct air capture fa- termittency, and grid stability issues. Intermittency cilities could remove a year’s emission quickly but accounts in part for the discrepancy between wind at an annual operating cost that is projected [2] to and solar providing 32% of the electricity genera- lie between $5 trillion and $30 trillion. The capital tion capacity in 2016 but only 25% of the electricity cost, which was estimated in 2019 by Fasihi et al [5], [4]. The intermittency issue can be addressed by also has large uncertainties, but $200 to $300 per batteries, which must be large and expensive when t/yr may be credible. This estimate gives a cost of wind and solar dominate a grid, and by power trans- $10 to $15 trillion for facilities that would remove mission over a scale greater than that of weather 50 Gt/yr. If the capital costs were sufficiently low, patterns. A more subtle issue is the stability of these facilities could be operated intermittently us- the electrical grid. Wind and solar power eliminate ing extremely low-cost wind and solar power. the heavy rotating generators of conventional power Costs should be compared to the world economic plants, which have sufficient inertia to stabilize the output, which is estimated [6] to be $92 trillion in grid during changing power loads [7]. Chapter 5 of 2020, and the wholesale value of annual electricity the 2019 OECD report [1] stated that financial fea- production, approximately $2 trillion a year. This sibility limits the electricity-generation fraction that can be derived by multiplying the $80/MWh for the can be produced by solar and wind to 30% to 40%. wholesale price of electricity in Table 8 on p. 127 The cost of deploying carbon-free energy sources of [1] by the 2016 world electricity production of can be assessed by multiplying the overnight costs of 25×109 MWh given in Table 37 of [4]. electricity-generation technologies by the electricity The costs of carbon-dioxide removal have large un- generation capacity of the world, which is approx- certainties, and the research proposed in the 2019 imately 8000 GW in 2020 (Table 49 in [4]). The National Academy study [2], with a maximal an- overnight cost is the expense of constructing a gen- nual expenditure of less than $250 million, is clearly eration facility, ignoring the interest charges, and not designed to develop minimal cost options on the dividing by the number of Watts generated. The shortest possible time scale. Chapter five of the overnight cost [8] is $1.32/W for wind, $1.33/W study [2] said a facility that would remove 104 t/yr for tracking solar photovoltaic, and $6.32/W for ad- of carbon-dioxide could demonstrate the technology vanced nuclear fission. Replacing the present electric and would cost approximately $100 million. The generating capacity with new nuclear fission facilities ratio of the cost of deployment to the cost of devel- would cost approximately $50 trillion. Replacement opment of a carbon-dioxide removal option is mea- with wind and solar facilities would cost approxi- sured by the ratio of the $10 trillion estimated as mately $10 trillion, but this cost contains no provi- the capital cost to the $100 million required for a sion for addressing intermittency and grid-stability demonstration facility, a factor of a hundred thou- issues. Wind and solar appear to be obvious choices sand. to replace some fraction of the electricity generation, The time required to develop science-based op- but that fraction may be very limited [1] and is an tions can be studied but not definitively answered important topic for additional research. without their development. Nevertheless, science Fusion energy has fundamental advantages com- can move with remarkable speed in periods of soci- pared to alternative carbon-free energy sources— etal crisis. The splitting of the nucleus in December especially when most of the electricity-generation ca- 1938 to the launching of the first nuclear-powered pacity should come from sources other than wind or submarine in January 1954 was approximately fif- solar. The development costs of fusion energy are teen years; nuclear weapon development required low compared to deployment costs. The ratio of de-

2 ployment to development costs is largely determined not be the fastest possible path to a demonstration by the ratio of the 8000 GW of total generation ca- of fusion energy. It was only fifteen years between pacity to the envisioned size of a demonstration mag- the splitting of the uranium nucleus and a fission- netic fusion reactor, approximately 1 GW. powered submarine. The low cost of developing a fusion option, com- A necessary step in stellarator development is the pared to the societal risks of not doing so, makes an formulation of conceptual designs for one or more assessment of rapid paths to fusion development im- attractive, low-risk, stellarator demonstration reac- perative. A needless delay imposes enormous costs tors. Major improvements in stellarator reactors can on society: the cost of carbon-dioxide removal— be made through computational design, which is fast trillions of dollars per year of delay—and the cost of and has a low cost. No reason is known why a con- a less than optimal replacement for a large fraction of ceptual design for a stellarator reactor cannot be suf- the electricity-generation capacity—ten of trillions ficiently attractive that its final design and construc- of dollars spread over a few decades. tion could be initiated without the delay of inter- When societal risks are considered, the case for mediate generations of experiments. Clarifying and stellarators is compelling. The stellarator, among confirming experiments that are constructed simul- all fusion concepts, has properties that best open taneously do not produce delays. Many such exper- a fast and low-risk path to reactors. The plasma iments were built while TFTR and JET were under in a stellarator is externally controlled, rather than construction, such as the tokamak now called DIII- self-organized, to a far greater extent than in any D. other fusion concept—magnetic or inertial. In ad- In 2018 the U.S. stellarator community published dition, the stellarator can make use of an order of an article Stellarator Research Opportunities [9], magnitude more distributions of external magnetic which reviewed the issues of stellarators and should fields. These differences lead to two distinct research be consulted for additional details and references. paradigms: optimization using computational de- The Stellarator Research Opportunities article also sign confirmed by experiments for stellarators and defined a research program, but that program did extrapolation from one generation of experiments to not consider the implications of societal risks. another as in tokamaks. Here societal risks are the focus, which makes the Section II compares stellarators and tokamaks, primary question whether the development time for which clarifies why with present knowledge the stel- fusion power could be shortened by eliminating a larator appears to offer a far better path for a fast generation or more of major stellarator experiments and low-risk development of a fusion reactor. The before beginning construction of a reactor. A major time required for development is essentially deter- experiment means with a time scale of a decade or mined by the number of consecutive generations of longer. In a development program defined by its ur- experiments that are needed. Relative risk is de- gency, that question may be best answered as part of termined by the issues requiring proof-of-principle the review process of conceptual designs for stellara- demonstration, such as disruption avoidance in toka- tor reactors. The question addressed in this paper maks. Using these criteria, the step to a power is the urgency of the development of conceptual de- plant from the ITER tokamak appears more difficult signs, which has a low total cost—presumably well than going directly using our present understand- under $100 million—compared to the overall cost of ing of stellarators. Tokamaks are a certain geomet- developing stellarator reactors. ric limit of quasi-axisymmetric stellarators, but with The sections following Section II provide addi- only 10% of the possible external magnetic field dis- tional details on important topics and areas in which tributions available to them. It should not be sur- major improvements can be quickly made through prising that an additional order of magnitude in de- computational design. Section III discusses coil is- sign freedom aids the achievement of fusion energy. sues including the importance of open access, which Agreement with the statement that stellarators of- has had almost no mention in the fusion literature, fer a far more likely path for a fast and low-risk de- possibly because the type of open access available in velopment of a fusion reactor does not imply ITER stellarators does not appear to be possible in toka- or the tokamak should be terminated. Deuterium- maks. Section IV discusses the space of stellarator tritium experiments on ITER, which are scheduled configurations, which is so large that special strate- to begin in fifteen years, followed by a tokamak gies are required to determine those most suitable demonstration reactor may provide a longterm op- for fusion reactors. Section V discusses strategies tion for carbon-free energy, and that may prove to for dealing with the implications of microturbulent be important. But, an ITER-centric time scale may transport on reactor design. A method of assess-

3 ing the implications is developed in the Appendix. not using the plasma itself to provide an Appendix A 2 shows that gyo-Bohm scaling fits the essential part of its confinement concept. overall scaling laws of both tokamaks and stellara- tors with remarkable accuracy and derives implica- tions for reactor design. Section VI considers di- This allows stellarators to be designed compu- vertors and the protection of the walls from alpha- tationally with far more reliability than any particle damage in stellarators. Section VII dis- other fusion concept. cusses technical developments in the areas of coils, The alternative to computational design is liquid films for the walls, solid walls, and breeding extrapolation from one generation of experi- blankets. ments to another, as is traditional in tokamaks. The abstract of the original paper on the ITER Physics Basis emphasized extrapolation [12]. II. STELLARATORS AS A PATH FOR THE The paper that introduced the scientific basis RAPID DEVELOPMENT OF FUSION of W7-X emphasized computational optimiza- tion of designs [13]. The stellarator, among all fusion concepts, has Four disadvantages of extrapolation in com- properties that best open a fast and reliable path parison to computational design are: to reactors. These properties can be illustrated by comparing stellarators with tokamaks. Far more (a) Experiments build in conservatism—even tokamak than stellarator experiments have been per- apparently minor changes in design are formed, but far more details about fusion plasmas not possible and therefore remain unstud- are required for the design of a tokamak than of a ied. stellarator reactor. (b) Experiments are built and operated over The extra information required for tokamak reac- long periods of time—often a number of tors makes the step from ITER to a demonstration decades. power plant (DEMO) appear more difficult than go- ing from our present understanding of stellarators to Multiple experiments carried out at the a stellarator DEMO. Open questions that ITER will same time do not delay development, but address are summarized in a 2019 article by Hawry- extrapolation using consecutive genera- luk and Zohm in Physics Today [10] and the issues tions of experiments does. The need for that are being considered for the European tokamak consecutive generations of experiments demonstration power plant are reviewed in a 2019 should be minimized. article in Nuclear Fusion [11]. (c) The cost of computational design is many 1. No proof-of-principle issue, such as dis- orders of magnitude smaller than building ruption avoidance in tokamaks, blocks a major experiment, as well as having a rapid development of stellarators. much faster time scale. Disruptions are an existential threat to (d) Extrapolations are dangerous when reactor-scale tokamaks, particularly the threat changing physics regimes. Examples are of strong currents of relativistic electrons. (i) plasma control in ignited versus non- Nevertheless, disruptions receive only cursory ignited plasmas and (ii) the formation of consideration in the papers on open questions a current of relativistic electrons during that will be addressed by ITER [10] and on a disruption. the European tokamak demonstration reactor In existing tokamaks, external heat- [11]. ing provides plasma control that is not present when the heating is dominated As discussed below, no solution that is gener- by DT fusion. As will be discussed, a ally perceived to be reliable is known for dis- stellarator has far more available degrees ruptions, which makes all tokamak planning of freedom for control but requires fewer problematic. A common presumption is that than a tokamak. the disruption problem will be solved because it must. Although relativistic electrons are an ex- istential threat to tokamak reactors, the 2. The stellarator is unique among all fu- danger is removed in standard stellara- sion concepts, magnetic and inertial, in tor designs. In stellarators, the plasma is

4 robustly centered within the confinement extremely limited and slow, seconds for chamber, and the effect of plasma cur- the temperature and density profiles and rents on the rotational transform, which minutes for the current profile. is the derivative of the poloidal relative Steering alone does not provide complete to the toroidal flux, is minimized. The protection against disruptions. A disrup- loop voltage, which accelerates electrons, tion can be initiated if a part of a wall is the rate of slippage of the poloidal rel- tile or even a tiny flake from the tung- ative to the toroidal magnetic flux. sten diverter targets were to enter the In tokamaks, a loss of position control ac- plasma [17]. But, in Alcator C-Mod dis- companies disruptions and each megaam- ruptions caused by tungsten flakes could pere of decay in the plasma current can be avoided when the poloidal magnetic increase the current in relativistic elec- field could be maintained [17]. Of course, trons by a factor of ten. As noted in a tile falling into a stellarator plasma Chapter 3, Table 5, of the ITER physics would cause a rapid drop in the plasma basis [14], this implies an amplification of pressure. Stellarators should be designed a current of relativistic electrons a tril- so currents associated with the diamag- lion times greater in ITER than in JET. netic effect of the plasma and the quick Recent theoretical work [15] indicates the loss of the plasma energy through radia- amplification may be far larger than for- tion would be tolerable. These require- merly expected. ments are far less demanding than those The danger of electrons running away to of a tokamak. relativistic energies has been prominent 3. Stellarator reactor designs are only in the literature for more than twenty weakly dependent on the plasma pres- years, but no method has yet been de- sure profile. vised of mitigating the danger in a way that is perceived to be reliable. The last (a) The sensitivity of tokamaks to the profile paragraph of the 2019 Nuclear Fusion re- of the current density makes them highly view of the physics of runaway electrons sensitive to the pressure profile, through (RE) [16] noted: With ITER construction the bootstrap current, and in non-steady- in progress, reliable means of RE miti- state reactors to the temperature profile, gation are yet to be developed. Sections through the resistivity. 10.8 and 11.3 of this review discussed the wall damage that runaway electrons can (b) Microturbulent transport is an issue for produce. A runaway current as small as all magnetic-fusion systems. The insensi- 300 kA could cause the melting limits of a tivity of stellarators to the pressure pro- wall panel roof to be exceeded. This is a file implies that only the overall level of worse-case number, and several megaam- the transport is of central importance. peres of relativistic electrons striking the For tokamaks, not only is the overall walls may be required for extreme dam- level important, but also the radial de- age. pendence of the transport. The severity of the damage that can be (c) The overall effective transport rate can be produced by even a single relativistic- normalized to gyro-Bohm transport by a electron incident implies: (i) The achieve- coefficient D, Appendix A. The H-factor ment of the ITER mission will be difficult of the tokamak literature scales as H ∝ when more than one such incident occurs 1/D2/5. in a year. (ii) The strategy for avoidance Too large a D implies that either the must be fundamentally based on theory power output of a single reactor or the and computation. magnetic field strength become exces- Tokamak disruptions are often said to re- sively large. Too small a D implies sult from exceeding operating boundaries the plasma radius is small compared to [10]. Unfortunately, methods of steer- the thickness of the blankets and shields, ing tokamak plasmas away from operat- which means the power production is too ing boundaries during a fusion burn are small compared to the reactor cost. The

5 problem of too small a D could be ad- as large as is consistent with a reasonable wall life- dressed by reducing the plasma reactiv- time to minimize the cost of fusion per kilowatt hour. ity by departing from a 50/50 deuterium- The total power output PT of a single fusion reac- tritium mixture. tor would ideally be small to maximize flexibility and D of order but somewhat smaller than minimize the capital required for the construction of 2 unity is optimal. a single unit. The obvious relation PT ∝ (R/a)a pw couples a high power density with a high total power (d) Stellarators do not have constraints such output. The aspect ratio R/a is determined by the as the Greenwald Limit on the plasma fundamental properties of a fusion concept, and the density or the high electron temperature aspect ratio of stellarators is several times larger required in tokamaks for current mainte- than that of tokamaks. The minor radius a cannot nance. The higher the electron tempera- be too small compared to the thickness of blankets ture the greater the number of energetic and shields around a fusion plasma, but otherwise it alpha particles, which increases the sen- is determined by transport. sitivity to energetic particle instabilities. Appendix A shows that the empirical energy- As discussed in Appendix A, the degra- confinement scaling of tokamak and stellarator ex- dation of confinement with power seen in periments closely match what would be expected empirical scaling laws implies a degrada- if the diffusion coefficient were a factor D times tion of confinement with temperature. gyro-Bohm transport. The required plasma radius squared scales with the quality of confinement 1/D, 4. Stellarators offer far more freedom of the magnetic field strength B, the power density on control than do tokamaks. the walls pw, and the central plasma temperature T0 Approximately fifty externally produced dis- as tributions of magnetic field are available for D4/5T 6/5 plasma control in stellarators; approximately a2 ∝ 0 . (1) 8/5 2/5 five are available in axisymmetric tokamaks. B pw Unlike in stellarators, these require careful time-dependent control. For a given quality of confinement, wall loading, and 2 aspect ratio, the total power output PT ∝ a can be The plasma profiles in tokamaks require far made smaller by using a larger magnetic field and a more control than in stellarators, but the avail- lower plasma temperature—as long a T0 > 10 keV. able degrees of freedom to provide that control The higher central temperature in tokamak reactors are far fewer. offsets the advantage of a smaller aspect ratio for allowing fusion power plants to have a smaller total 5. The coil systems in stellarators, unlike power output P . Tokamak reactor designs often those in tokamaks, can be designed for T require a significant fraction of this power output be open access to the plasma chamber, Sec- used for maintenance of the plasma current and for tion III A. control while stellarators do not. The larger recircu- lating power fraction is a significant burden on the If fusion is to be developed rapidly, a demon- economic viability of tokamak reactors. stration reactor (DEMO) must be designed Early operations of the W7-X stellarator achieved to allow first wall components to be changed D = 0.13 in ten-second steady-state plasma condi- quickly—too many uncertainties remain in tions, which yields attractive reactor designs, and first wall materials [18, 19], in concepts such D = 0.05 during short intervals. DIII-D has carried as walls being covered by liquids [20], and in out long-pulse tokamak experiments that achieved blankets for breeding tritium [21] for it to be D = 0.31. Both the stellarator and tokamak results otherwise. for D are discussed in Appendix A 4. Open access also shortens maintenance times in operating reactors. III. COILS FOR STELLARATOR Stellarators do have the disadvantage of a larger REACTORS aspect ratio, the ratio of the major to the minor ra- dius, R/a, of the torus. The power density on the Three properties of the coils that produce the ex- walls pw, megawatts per square meter, should be ternal magnetic field are of particular importance to

6 if the joints are not demountable. A demountable coil means parts can be repeatedly separated and re- joined. The provision of joints is much easier when the joints can be located in a low-field region as they can be in stellarators. Rapid changes in the components that surround the plasma are critical for the fast development of fu- sion energy and would minimize maintenance time in a reactor. Despite the obvious importance, coil FIG. 1: H. Yamaguchi has published a set of continuous concepts such as the one illustrated in Figure 1 re- helical coils that generate a magnetic field that approxi- main largely unexplored. mates a quasi-isodynamic stellarator. The full coil set is illustrated on the left and the outer magnetic surface to- gether with the primary coils are illustrated on the right. As discussed, the red coil is the only coil that need limit B. Efficient magnetic field distributions access to the plasma chamber. This is Figure 1 in H. Yamaguchi, Nucl. Fusion 59 104002 (2019). A curl-free magnetic field decays with the distance x from the coil that produces it as e−kx where k is the wavenumber of the field. All possible exter- a rapid and reliable development of fusion energy: nal magnetic field distributions can be ordered by (1) Coils that offer easy access to the plasma cham- their efficiency of production [22, 25]. The are ap- ber. (2) Coils that are relatively easy to manufac- proximately fifty distributions that have adequate ture because the magnetic fields that they produce efficiency [26]. Stellarator optimizations could be are not unnecessarily strong nor rapidly varying in constrained so only magnetic fields that can be pro- space. (3) Coil systems that maximize the flexibility duced efficiently at a distance are included, which of plasma control [22]. are the only magnetic fields that can be produced by practical coils. The benefits of limiting the design to the ef- A. Coils with easy chamber access ficiently produced external field distributions are largely unexplored. Coils systems for both tokamaks and stellarators have been designed in a manner that makes access to the plasma chamber extremely constrained. Coils C. Coils needed for plasma control can be designed with demountable joints, so parts of the coils can be removed to provide chamber access The important stellarator control parameters are [23]. For stellarators, but not for tokamaks, a far the efficient magnetic field distributions. It is known simpler and more robust method of providing access that the importance of the various magnetic field is available. The red coil in Figure 1, which is from distributions for plasma control varies widely [22]. [24], is the only coil that need encircle a stellarator What has not been done is to assess which of these plasma and limit plasma access. Mathematics en- distributions are the most important and how the sures the rest of the magnetic field could be produced control of these distributions can be incorporated in by coils, each shaped like a windowpane, with some coil design. embedded in the removable sections of the walls [22]. The speed and the completeness with which a There is no necessity for the removal of wall sections given machine allows fusion to be developed is to be more restricted than that produced by the red largely determined by its available control. coil of Figure 1. Discreteness in the coils that produce the toroidal magnetic field in tokamaks produces an unac- ceptable toroidal ripple unless the space between IV. STELLARATOR CONFIGURATIONS toroidal-field coils is small. Figure 1 shows how the toroidal ripple can be used in stellarators to provide The space in which stellarators are designed has the helical magnetic field that they require. about fifty degrees of freedom—far too many for an The winding and assembly of coils can be sim- optimization code to ensure that a global optimum plified if joins are possible during assembly, even has been found. Although a direct and complete

7 optimization is impossible, practical numerical opti- which can be ω(θ, ϕ), is required to obtain magnetic mizations can (1) refine an initial guess or (2) main- coordinates leaving R(θ, ϕ) and Z(θ, ϕ) free. Ob- tain the optimization of a curl-free magnetic field as taining well confined particle trajectories constrains the plasma pressure is increased. half of the freedom of another function of the two The large size of this space compared to what has angles. For example, quasiaxisymmetry is obtained been explored is illustrated by Figure 2. when the field strength B has the property that B(θ, ϕ) = H B(θ, ϕ)dϕ/2π. The curl-free solution can be extended throughout the volume enclosed by A. Identification of states for optimization the surface by choosing efficient magnetic field dis- tributions so the magnetic field perpendicular to the The attractiveness of an optimized stellarator is optimization surface is zero. Maximizing the coil ef- largely determined by the initial state used in the op- ficiency is equivalent to placing another constraint timization, which makes makes methods of choosing on a function of the two angles. There are only two- an initial state of great practical importance. Two and-a-half functions of constraints on the three func- concepts for finding advantageous initial states are tions of θ and ϕ. Consequently, there is additional (1) a Taylor expansion around the magnetic axis and freedom in the properties of the magnetic field. (2) the optimization of an outer magnetic surface of This method of defining curl-free states for opti- a curl-free magnetic field. mization is unexplored.

B. Annular Design 1. Expansion around the axis An optimal design for a stellarator may have low The original idea of defining equilibria using a plasma transport in the outer half of the minor ra- Taylor expansion around the central field line in a dius, but such rapid transport in the inner half that toroidal plasma, the magnetic axis, [27] is due to the pressure is essentially constant there, Appendix Mercier in 1964. Taylor expansion methods were A 3. The implications are essentially unexplored, found in 1991 to set important constraints on stel- but there are advantages to having the confinement larators [28], and recent advances have been made produced by an outer annulus. (1) A spatially con- [29]. Unlike axisymmetric systems, the achieve- stant pressure p maximizes R p2d3x for a fixed max- ment of adequate particle confinement is the primary imum pressure, which maximizes the fusion power. physics issue in stellarators. Methods of achieving The confinement time of the plasma is the ratio of particle confinement, quasisymmetry and omnigen- the total plasma volume to the volume of the annu- ity, are discussed in [30–32]. lus longer than the confinement time of the annulus. (2) Impurities tend to be flushed out more readily the narrower the confinement annulus compared to 2. Optimization of an outer surface the total confining volume. (3) The injection of fuel is easier. (4) The optimal 50/50 DT ratio can be An outer magnetic surface of a curl-free magnetic maintained in the fusing plasma, which is not triv- field can be found [33] that has desirable confinement ial when transport coefficients are small in core. (5) properties such as exact quasisymmetry. Quasisym- Control of the width of the annulus would provide metry gives tokamak-like confinement of individual an important control of the plasma. particles. This method is particularly important The situation in tokamaks with good confinement in conjunction with the concept of annular design, is related but different. The pressure drops across which is discussed in Section IV B and in Appendix the core of a tokamak, but there is a narrow re- A 3. gion right at the plasma edge, where a transport The shape of the outer magnetic surface is deter- barrier arises that creates a pedestal, which raises mined by three functions of two angles that must the plasma pressure everywhere inside. See Figure satisfy constraints. Using (R, ζ, Z) cylindrical coor- 2 in [10] and the related discussion. This annulus dinates, the three functions are R(θ, ϕ), ζ = ϕ + naturally has periodic instabilities called Edge Lo- ω(θ, ϕ), and Z(θ, ϕ), where θ and ϕ are the poloidal calized Modes (ELM’s), which must be mitigated to and toroidal angles in Boozer magnetic coordinates avoid unacceptable damage to the chamber walls, in which the constraint of exact quasi-symmetry is [14]. The extent to which such pedestals arise in easily specified [30]. One function of the two angles, stellarators, or whether it is even desirable that they

8 FIG. 2: Matt Landreman and his group have used analytic expansions around the magnetic axis to survey the landscape of possible quasisymmetric stellarators. The figure shows a database of 2.4 × 108 quasisymmetric stel- larator configurations. The few designated points on the left side of the figure indicate previously known stellarator configurations occur, is unclear. The self-organized state of an ax- confinement. Unlike the situation in tokamaks, the isymmetric tokamak plasma implies the control over details of the pressure profile in stellarators are of important features such as the pressure profile is little relevance. ITG turbulence need only be kept limited. Carefully designed non-axisymmetric per- at a level that is consistent with an adequate fu- turbations that preserve the quasisymmetry of the sion product, nτET , where n is the number density tokamak core could ameliorate this limitation [34]. of the deuterium and tritium ions, τE is the energy An example is the control of ELM’s by long wave- confinement time, and T is the temperature. length non-axisymmetric magnetic fields [35]. As discussed in Appendix A, power-law scaling re- lations hold with remarkable accuracy for tokamaks and stellarators. Nevertheless, scaling relations do not provide the certainty that is wanted for a reactor V. MICROTURBULENCE STRATEGIES design—even in stellarators. The effect of microtur- bulence is not well understood in either tokamaks or An uncertainty in the design of both stellarator stellarators. Non-linear calculations of microturbu- and tokamak fusion reactors is microturbulent trans- lence using the GENE code [36] show a W7-X case port. The physics of microturbulence in stellarators with a transport enhancement of twenty times the was reviewed in 2015 by Helander et al [36]. The characteristic gyro-Bohm value and a DIII-D case two most important types of microturbulence are with an enhancement of two-hundred times. the ion-temperature-gradient (ITG) mode and the In designing a stellarator reactor, the most im- trapped electron (TE) mode. The TE instability portant information on microturbulence is what fac- has much greater stability when the trapped elec- tors are beneficial in obtaining an adequate nτET . trons are primarily in a region of good magnetic A higher magnetic field strength B appears to be field line curvature as in W7-X. In tokamaks and clearly beneficial. The ion temperature gradient, in quasi-axisymmetric and quasi-helically symmet- d ln Ti/dr times a spatial scale is an instability fac- ric stellarators the trapped electrons are primarily tor, but what that spatial scale is is not agreed upon. in a region of bad curvature. It could be related to the average magnetic field- ITG microturbulence appears to have a beneficial line curvature, the local shear, or the global shear effect of expelling impurities, which implies some in the magnetic field. A density gradient is stabiliz- level is desirable. But, in a reactor ITG turbulence ing, so d ln T/d ln n should not be too large, but a can not be so large that it unacceptably degrades ion sufficiently weak temperature gradient can be stable

9 even when the density profile is flat. imply such explorations are probably not required to To avoid impurity accumulation, it is not clear build an attractive stellarator reactor other than to that stabilizing the trapped electron mode when ion increase the certainty that the transport in a given temperature gradient mode is unstable is beneficial. design is acceptable. This might make ion heat transport rapid compared There should be a focus on experiments and the- to particle transport, which is bad. What is needed ory that can contribute over a time scale of years, is a rapid transport of non-hydrogenic ions relative not decades, to clarify the constraints of microtur- to the heat transport in a plasma that primarily has bulence on reactor design. hydrogenic ions. What seems to be agreed upon is that linear insta- bility theory is a poor surrogate for relative levels of VI. EDGE CONTROL microturbulent transport [37]. The ITG mode can be stabilized by zonal flows [38], though in stellara- A. Divertors tors wave coupling in other forms than zonal flows may be more important [39]. The particle exhaust from plasmas should be con- Stellarators can be designed for optimal microtur- centrated to the location of pumps, but this concen- bulent transport [40], but such optimizations require tration makes the power loading on the walls intol- a surrogate. Full simulations of microturbulence are erably high unless a large fraction of the power is too time consuming to be practical. The reliabil- radiated away. ity of full gyrokinetic simulations is debated, but A divertor is a magnetic structure that directs the to the extent that they can be taken to be reliable plasma particles to the locations of pumps. A de- they could be used (1) to test whether a given stel- tached divertor means that radiation removes essen- larator configuration has acceptable transport prop- tially all of the energy from the plasma before it erties and (2) to determine which features of the contacts the wall. magnetic configuration have the greatest effect on Two types of magnetic structures are being con- the microturbulent transport. The microturbulence sidered for divertors in stellarator reactors: resonant codes GENE, XGC, and GTC have been primar- and non-resonant. ily developed for tokamaks but stellarator versions, such as XGC-S [41], are being developed.

Plasma confinement can be greatly enhanced by 1. Resonant divertors transport barriers. The best known is the H-mode enhancement of tokamak confinement by approxi- mately a factor of two by the formation of a narrow A resonant divertor locates a chain of islands at edge pedestal [10]. The formation and the stability the plasma edge, which requires extremely accurate of this transport barrier can be strongly influenced control of the edge rotational transform, ι = 1/q, by tokamak shaping and in particular by having neg- which is the twist of the magnetic field lines; q is the ative triangularity, which has the mid-plane point of safety factor. the triangle on the small major radius side of the W7-X has a resonant divertor [44, 45], so this con- plasma [42]. Transport barriers can form not only cept is being studied as part of the W7-X program. at the edge but also in the body of the plasma [43]. In particular, W7-X has demonstrated that a res- Internal transport barriers are associated with ra- onant divertor can maintain stable detachment and tional magnetic surfaces, low or negative magnetic radiate most of the plasma energy before the plasma shear, a strong local magnetic shear, such as that reaches the walls. produced by the Shafranov shift, and E~ × B~ flow shear, which has a far stronger effect on the ion than on the electron transport. 2. Non-resonant divertors Stellarators offer much more freedom to change the properties that control internal transport bar- Non-resonant divertors use the Hamiltonian me- riers than do tokamaks. Freedom from the details chanics concepts of Cantori and turnstiles. Mag- of the profile of the net plasma current, including netic field lines obey exactly the equations of one- the disruptions caused by that profile, imply trans- and-a-half degree of freedom Hamiltonian mechan- port barriers are an important area for exploration. ics, H(p, q, t), although for field lines the three vari- Nevertheless, existing W7-X results, Appendix A 4 ables of the Hamiltonian mechanics are three spatial

10 coordinates. Beyond the outermost confining mag- 1. Developments for coils netic surface, a double magnetic flux tube is formed in each period of the stellarator (1/2 the flux comes Technical developments in high-temperature su- in and 1/2 goes out). The two parts of these tubes perconducting coils for fusion applications was the strike the wall at remarkably robust locations. subject of a 2018 Nuclear Fusion review [50]. This Non-resonant divertors have been explored less review included a discussion of use of joints in coils. than resonant divertors, but there are several the- The construction of both tokamaks and stellarators oretical papers on non-resonant divertors [22, 46– could be faster and cheaper if coils could be delivered 49]. Unlike resonant divertors, non-resonant diver- in pieces that are joined during device construction. tors place no constraint on the edge rotational trans- Commonwealth Fusion Systems [51] has placed a form and the width of the escaping flux tube that strong focus on developing coils for fusion systems carries plasma to the pumps can be adjusted. that can operate at much higher magnetic fields than those in existing tokamaks. This work is important for stellarators as well as tokamaks. The required minor radius of a plasma will be found to scale as B. Protection of the walls from α particles a ∝ D2/5/B4/5 while the total power output of a 2 reactor PT for a given wall loading pw scales as a . Helium ions (alpha particles) produced by the nu- Higher magnetic fields allow power plants to be built clear reactions can become deeply embedded in the with a smaller unit size and allow compensation for walls if they strike while still energetic. The accumu- poor confinement, a large D. lation of helium gas in crystal lattices creates blisters and fuzzy regions, which destroys the structural in- tegrity of the walls. 2. Development of liquid films Three strategies have been proposed for address- ing this issue: (1) Apply whatever constraints are Even a thin layer of liquid on plasma-facing com- necessary on the variation of the magnetic field ponents can address four issues [20]. First, the layer strength on the magnetic surfaces to limit the loss of can eliminate the degradation of wall materials that alpha particles. (2) Design the edge magnetic field can be produced by fusing plasma plasmas. Exam- so the energetic trapped alpha particles, which are ples are alpha particle degradation and the sudden the problem, strike the wall in a location in which flash of radiative energy that would occur if a piece they harmlessly go into a liquid, such as lithium or of a tile fell into the plasma. Second, flowing liquids tin, not a solid wall. The feasibility of doing this can remove the surface heat load. Third, somewhat is essentially unexplored. (3) Avoid alpha-particle thicker liquid layers can reduce the nuclear damage. damage altogether by covering plasma facing com- Fourth, liquid layers can reduce gradients such as ponents with a thin liquid film, Section VII 2. temperature and stress.

3. Development of solid walls VII. IMPLICATIONS OF TECHNICAL DEVELOPMENTS Although liquids can mitigate issues associated with plasma-facing components, solid walls are re- Technical developments are of particular impor- quired even if there are liquids covering the walls. tance in four areas (1) coils, (2) liquid films for cov- Issues that must be addressed relative to materials ering first walls, (3) solid first wall materials, and (4) for the first wall are discussed in [18, 19] breeding blankets for tritium. The design of a stel- larator reactor that has open access to the plasma chamber requires a suitable choice for the coil sys- 4. Development of tritium breeding blankets tem. But, when this is done, a fast development of fusion requires only that an appropriate space allo- Major challenges and fundamentally different de- cation be made for the first wall, the blankets, and sign choices exist for the blankets that breed the shields. Several versions of these systems should be tritium burnt in fusion systems. These are reviewed made to test various designs. The replacement of in- in [21]. adequate components must be part of the research on the test reactor.

11 Acknowledgements 1. Deuterium-Tritium power density

This material is based upon work supported by the John Wesson [53] gave a convenient expression for U.S. Department of Energy, Office of Science, Office the power density of DT fusion, which holds with of Fusion Energy Sciences under Award DE-FG02- 10% accuracy for temperatures between 10 keV and 95ER54333 and by grant 601958 within the Simons 20 keV, Foundation collaboration ”Hidden Symmetries and 2 2 Fusion Energy.” pDT = 0.77n T . (A1) with 1/5 of the energy in alpha particles and 4/5 in neutrons. The power density in alpha particles is Appendix A: Fusion power and transport α 2 2 pDT = 0.154n T ; (A2) Freidberg, Mangiarotti, and Minervini have noted pα (0) c ≡ DT (A3) [52] “that the overall design of a tokamak fusion re- DT (nT )2 actor is determined almost entirely by the constraints ≈ 0.154. (A4) imposed by nuclear physics and fusion engineering.” Related constraints apply to stellarators and allow The derivation of the power density, megawatts a simplified determination of the requirements of a per meter cubed, begins with Equation (1.4.2) of stellarator reactor and how they depend on physics Wesson’s book Tokamaks [53]. The power den- properties of the plasma, which is the subject of this α 2 sity in alpha particle is pDT = n < σv > Eα/4. appendix. The energy released in alpha particles per reaction −19 A small unit size for fusion reactors, measured is Eα = (3.5 MeV) × (1.60 × 10 MJ/MeV) = by the total power output PT , is in conflict with 5.6 × 10−19 MJ. The velocity weighted cross section having a high power density on the walls pw since with Maxwellian ions is approximated within 10% PT ∝ Rapw. The basic fusion concept sets the as- accuracy for 10 keV< T <20 keV in Equation (1.5.4) pect ratio R/a, but the minor radius a is deter- as < σv >= 1.1 × 10−22T 2 when the units of T are mined by transport as long as the minor radius is α 10 keV. This calculation gives pDT , Equation (A2); sufficiently large compared to the thickness of the a multiplication by five gives the full power density, blankets and shields surrounding the plasma. When pDT , Equation (A1). transport would allow a minor radius smaller than The required energy confinement time to achieve this, the DT fuel mixture could be degraded from ignition is the optimal 50/50 mixture for the reactor design to be consistent with an adequate a. The diffusion co- 3kBnT α = pDT so (A5) efficient below which transport becomes too small τE is comparable to gyro-Bohm with an enhancement 3kB factor D ≈ 0.1. nT τE = ≈ 3.12. (A6) cDT The units that are used are 10 keV for temper- ature, 1020/m3 for number density, Tesla for mag- The minimum of nT τE is at T =1.4, which means at netic field, megajoules for energy, and seconds for 14 keV. What is precisely meant by nT is not clear time. In these units, the Boltzmann coefficient, since both n and T depend on radius. When central which converts 1020 particles/m3 times 10 keV into values are used in an analytic transport model, the 20 mega-Jules per cubic meter, is kB = 10 × 1.602 × constant 3.12 becomes 4.39, Equation (A33). 10−21 = 0.1602. The permeability of free space −7 µ0 = 4π × 10 in standard scientific units becomes 2. Transport model µ0 = 0.4π in the units that are used in this paper. The radial coordinate r is defined so the volume enclosed by a magnetic flux surface is (2πR)(πr2) The equilibrium between heat transport and fu- with R the major radius. The edge of the plasma is sion power in alpha particles is at r = a, which is the standard stellarator definition 1 d of the minor radius. The standard definition of the (rQ) = pα (0)f, where (A7) r dr DT minor radius of a tokamak, at has a plasma volume 2 κ (2πR)(πa2), where κ is the elongation. That is  nT  e √ t e f ≡ , (A8) at = a/ κe. n0T0

12 α and pDT (0) is the central power density provided by using Equation (A2). Equation (A22) gives the re- the fusion-produced alpha particles. The heat flux quired confinement for the power from alpha heating is to balance the thermal losses at the plasma edge. dnT The thermal energy in the plasma Wth is the in- Q(r) = −3k D , (A9) tegral of 3p/2 over the volume of the plasma; B dr Z a and D is the diffusion coefficient for plasma pressure. 2 3 p Wth = (2π) R p0 frdr (A23) An integration of the transport equation across 0 2 the plasma 0 < r < a gives 2 2 = 3p0π Ra m¯ 0 where (A24) Z a R λ0 p α 0 J0(x)xdx aQ(a) = pDT (0) frdr. (A10) m¯ 0 ≡ ≈ 0.608. (A25) 0 1 2 2 λ0

When performed numerically R λ0 pJ (x)xdx ≈ a. Analytic model 0 0 1.76. The energy confinement time is An analytic model is obtained for a diffusion co- efficient that is proportional to the plasma pressure, W τ ≡ th (A26) which is the case for gyro-Bohm diffusion when the E (2π)2RaQ(a) density is proportional to the square root of the tem- m¯ a2 perature. Let = 0 (A27) F0 D0 p D(r) = D f, so (A11) a2 0 ≈ 0.488 . (A28) 3 df D Q(r) = − p D ; (A12) 0 4 0 0 dr The requirement for alpha heating to balance the p ≡ 2k n T , (A13) 0 B 0 0 thermal losses is where D0 is a constant and p0 is the central plasma Z Wth pressure. = (2π)2R pα rdr or (A29) τ DT The solution to the equation for f given by Equa- E 3 R a tions (A7) and (A12) is 2 p0 0 prdr τE = (A30) pα (0) R a p2rdr f(x) = J (x), with (A14) DT 0 0 R a 1 3k p0 prdr x ≡ kr; (A15) = B 0 (A31) n T c R a p2rdr f(0) = 1, and f(ka) = 0. (A16) 0 0 DT 0 2 1 3kB m¯ 0λ0 J0(x) is the zeroth order Bessel function of the first = , so (A32) n0T0 cDT 2F kind, which has its first zero, J0(λ0) = 0, at λ0 = 2.405..., dJ /dx = −J (x), and d(xJ )/dx = xJ (x). 0 1 1 0 n0T0τE ≈ 3.12 × 1.409 ≈ 4.39. (A33) The boundary condition f(ka) = 0 implies ka = λ . (A17) 0 b. Gyro-Bohm diffusion The energy flux at the plasma edge is The implications of transport on designs of 3 p0D0 Q(a) = F0; (A18) toroidal magnetic fusion systems, stellarators and 4 a tokamaks, requires a normalizing transport model. F0 ≡ λ0J1(λ0), and (A19) Gyro-Bohm diffusion will be used for two reasons: α ISS04 2 pDT (0) (1) An empirical scaling law τE , Equation (A64), k = 3 , or (A20) p0D0 describes a broad range of stellarator and tokamak 4 experiments, Figure 3. This scaling law is accu- D pα (0) 0 = DT (A21) rately approximated by gyro-Bohm scaling, Equa- a2 3 2 4 p0λ0 tion (A44), with a dimensionless multiplying fac- 2cDT tor D. (2) Even non-turbulent transport models, = 2 n0T0. (A22) 3kBλ0 Appendix A 5, can differ by a number of orders of

13 magnitude from gyro-Bohm transport models, ei- gyroradius times the speed of sound with both cal- ther larger or smaller, but gyro-Bohm transport with culated using the temperature T . Therefore, one can D ≈ 1 gives optimal reactor designs. A heuristic let derivation of the gyro-Bohm diffusion coefficient is ρ C given in Appendix A 2 c. D = ∆ s s , (A39) The gyro-Bohm diffusion coefficient is a

C where the approximations are absorbed into the ra- D ≡ ρ2 s , (A34) dial scale size of the turbulence ∆. gB i a In gyro-Bohm diffusion, ∆ = ρs, which is a typical scale of fluctuations in ITG turbulence. In Bohm- where ρi ≡ Cs/ωci is the ion gyroradius using the speed of sound C ≡ pT/m , and a is the plasma like diffusion, ∆ ≈ a, which is as large as it can be. s i The enhancement factor of gyro-Bohm diffusion has minor radius. When DgB is evaluated at an ion mass of 2.5 times the proton mass, the interpretation ∆ c T 3/2 D = , (A40) D = gB ; (A35) ρ gB a B2 s cgB ≈ 162 (A36) but can also differ from unity because the plasma is √ not microturbulent, Appendix A 5, or turbulence is p 5 The speed of sound is Cs ≡ T/mi = 6.19×10 T present in only part of the plasma. and√ the ion gyroradius is ρi ≡ Cs/ωci = 1.62 × 10−2 T /B. The analytic model of Appendix A 2 a, is√ obtained d. Gyro-Bohm scaling of τE when√ the density profile has the form n ∝ T , then DgB/ f is constant. The scaling of the energy confinement time will be To study the effect of enhanced or reduced trans- studied using Equation (A27) with D0 replaced by port, a dimensionless coefficient D is introduced so the gyro-Bohm-scaled diffusion coefficient, Equation (A38), D0 = DDgB(0), or (A37) 2 3 3/2 m¯ 0 B a cgB T0 τ = . (A41) = D , (A38) E 3/2 2 cgBF0 a B DT0 where the constant cgB is given in Equation (A36). The convention is to replace the temperature de- pendence of τE with a thermal power Pth depen- dence. Since Pth = Wth/τE and the central pressure c. Heuristic derivation of gyro-Bohm diffusion is p0 = 2kBn0T0, Equation (A24) implies

2 The heuristic derivation of the gyro-Bohm diffu- 1 2 Ra = 6π kBm¯ 0 , and (A42) sion coefficient starts with general expression for a T0 Pth 2 2/5 radial diffusion coefficient, D ≈ ∆ /τco, where ∆ is   1 2
3/5 the radial scale of the microturbulence, and τco is the τE =m ¯ 0 6π kB F0cgB correlation time of the flow vr that gives the radial 3/5 scale; ∆ ≈ v τ . The radial velocity in electrostatic a12/5R3/5n B4/5 r co × 0 (A43) turbulence is vr = E˜θ/B. The radial motion pro- 2/5 3/5 D Pth duces a change in the electric potential φ˜ ≈ ∆(T/ea) a12/5R3/5n3/5B4/5 where T is the plasma temperature and the minor ra- ≈ 0.281 0 . (A44) dius a is the scale over which the temperature varies. 2/5 3/5 D Pth The poloidal variation in the potential is δ∆(T/ea), where δ∆ is the poloidal variation of the radial scale. The parameter dependencies is this formula repro- ˜ ISS04 Consequently, Eθ ≈ (δ∆/∆θ)(T/ea) ≈ T/ea; the duce those of the τE scaling law, Equation (A64) ISS04 radial scale varies by roughly the poloidal scale over with surprising accuracy. The τE scaling law rep- the poloidal scale of the turbulence, ∆θ. That is, resents both tokamak and stellarator experiments, D ≈ (T/eBa)∆, where T/eBa = ρiCs/a, the ion Figure 3.

14  α 0 0.5 0.6 0.7 0.8   λ(α) 2.405 2.554 2.703 2.963 3.455   m¯ (α) 0.6 0.7 0.8 0.8 0.9  F(α) 1.248 1.930 2.400 3.211 4.859

TABLE I: The argument λ of the Bessel functions at the plasma edge, r = a, the ratiom ¯ of the average to the central pressure, and the enhancement of the fusion power F are given as a function of α, which is the fraction of the plasma radius in which diffusion is assumed to go to infinity.

f. Gyro-Bohm scaling of ignited experiments

When the plasma is undergoing a steady fusion FIG. 3: The energy confinement times observed in both burn, Equation (A22) gives an expression for the re- stellarator and tokamak experiments are compared to 2 2 ISS04 quired D0/a . The expression obtained for D/a the stellarator scaling law, τE , Equation (A64). This figure was Figure 4 in the 2018 Nature Physics article on from gyro-Bohm scaling is given by Equation (A38). W7-X [55]. Equating these two expressions provides an expres- sion for the central density, n0 = nb, with √ 3k c λ2 D T e. Gyro-Bohm scaling of non-ignited experiments n = B gB 0 0 (A51) b 2c B2a3 DT √ The magnetic field that is required to reach a cen- 2 D T0 = 252.8λ0 2 3 . (A52) tral density n0 and temperature T0 can be calculated B a 2 in terms of the thermal power, Pth = (2π) RaQ(a), One less parameter is required to describe ignited supplied, the enhancement over gyro-Bohm diffu- than non-ignited experiments. sion, D, and the major R and minor radius, a. Equation (A18) for the edge heat flux, Q(a), p0 = Equation (A18) for the edge heat flux, Q(a), p = 0 2kBn0T0 for the central pressure, Equation (A38) for 2k n T for the central pressure, Equation (A38) B 0 0 the relation between D0 and gyro-Bohm diffusion, for the relation between D and gyro-Bohm diffu- 0 and n0 = nb using Equation (A52) imply that in an sion imply ignited plasma s r 5/2 2 2 !1/4 1/4 3 Dn T 9 c k  T 3  B = k c F 0 0 (A45) gB B 2 2 0 B gB 2 B = λ0F0 D 5 (A53) 2 a Q(a) 4 cDT a Q(a) s 5/2 2 3 1/4 Dn T 1/4  D T  ≈ 6.97 0 0 (A46) = 9.96 λ2F
0 . (A54) a2Q(a) 0 0 a5Q(a) s q 2 5/2 R/a = 3π kBcgBF Dn0T0 (A47) Pth 3. Transport with a confining annulus s 5/2 R/a A confining annulus means that the diffusion co- ≈ 31.0 Dn0T0 . (A48) Pth efficient D(r) is extremely large in the central part of the plasma 0 < r < αa, so f = 1 there, but Equivalently, within the confining annulus αa < r < a, the dif- fusion coefficient has the same form as in Appendix 2 √ 1 B Pth D = (A49) A 2, D(r) = D0 f. The solution for f in the con- 2 5/2 3π kBcgBF n0T0 R/a fining annulus αa < r < a is B2P = 1.043 × 10−3 th . (A50) Y0(λ)J0(x) − J0(λ)Y0(x) 5/2 f(x) = , (A55) n0T0 R/a J0(αλ)Y0(λ) − J0(λ)Y0(αλ)

15 The equation for energy balance R a pα rdr =  α = 0 α = 0.5 α = 0.7  0 DT r p d ln p p d ln p p d ln p aQ(a) can be used to define λ for any pressure profile  a p0 d ln r p0 d ln r p0 d ln r  that satisfies the transport equation as    0.5 0.818 0.448 1 0.408 1 0   0.6 0.737 0.729 0.907 0.695 1 0  α 2   2 pDT (0)a F  0.7 0.638 1.191 0.789 1.163 1 1.075  λ ≡ R a α . (A61)   p (r)rdr  0.8 0.518 2.084 0.642 2.063 0.822 2.000  0 DT   0.85 0.445 2.952 0.553 2.936 0.7105 2.887 Similarlym ¯ can be defined asm ¯ ≡ 0.9 0.361 4.66 0.448 4.647 0.577 4.613 R a 2 2 0 p(r)rdr/p0a . These definitions give the equations derived in Appendix A 2 general validity. TABLE II: The radial profiles of the pressure and the Plasmas are generally unstable to microturbulence logarithmic derivative of the pressure with respect to ra- when the logarithmic gradient of the pressure be- dius are gives for three values α, which is the fraction comes large compared to unity; of the plasma radius in which diffusion is assumed to go to infinity. As α becomes larger, the stability mea- d ln(p) x df/dx − = (A62) sure d ln p/d ln r becomes smaller at a given radius, but ln(r) 2 f the pressure at which d ln p/d ln r reaches a certain value x Y (λ)J (x) − J (λ)Y (x) becomes larger. = 0 1 0 1 . (A63) 2 J0(αλ)Y0(λ) − J0(λ)Y0(αλ)

2 α The pressure profile and the profile of the logarith- where k = 4pDT (0)/(3p0D0) as before, Equation (A20). The boundary conditions are f(αλ) = 1 and mic derivative of the pressure are given in Table II. f(λ) = 0. J0(x) and K0(x) are the Bessel functions of the first and second kind. Both obey the relations 4. Comparison with experiments dJ0/dx = −J1(x) and d(xJ1)/dx = xJ0(x). The function λ(α) is given implicitly by Equation (A10), which is obtained by equating the total exit- The observed global energy confinement in stel- ing heat flux 2πaQ(a) per unit length of the plasma larator experiments is summarized by the scaling in the toroidal direction with the total alpha-heating [54], power per unit length. a2.28R0.64 τ ISS04 = 0.134 n¯0.54B0.84ι0.41, (A64) Z λ (αλ)2 E P 0.61 e 2/3 fxdx = + F(λ) 0 2 where the energy confinement time is in seconds, the Y0(λ)J1(αλ) − J0(λ)Y1(αλ) minor a and the major radius R are in meters, the −αλ ;(A56) J0(αλ)Y0(λ) − J0(λ)Y0(αλ) volume averaged magnetic field B is in Tesla, the volume-averaged electron densityn ¯ is in 1020/m3, J1(λ)Y0(λ) − J0(λ)Y1(λ) F(α) ≡ λ ; (A57) the effective heating power P is in mega-Watts, and J0(αλ)Y0(λ) − J0(λ)Y0(αλ) the rotational transform ι2/3 is at a radius r = 2a/3.  df  ka = −F(α). (A58) The minor radius a is defined so the plasma volume 2 dx λ is (2πR)(πa ). ISS04 The τE scaling law represents both tokamak The implication is that Equation (A18) for Q(a) and stellarator experiments, Figure 3, though toka- holds when F0 is replaced by F(α). Indeed, F can mak H-mode experiments have up to a factor of two be defined by Equation (A18). better confinement than predicted. Paradoxically, Equation (A10) for energy balance is satisfied the radial dependence of the transport seen in W7- when X, as reported in the article from which this figure αλ Y (λ)J (αλ) − J (λ)Y (αλ) was taken [55], does not agree with that expected 0 = + 0 1 0 1 ; (A59) 2 J0(αλ)Y0(λ) − J0(λ)Y0(αλ) for gyro-Bohm transport. Nevertheless, the overall ISS04 dependencies of the τE scaling law are given by α λ aQ(a) = pDT 2 F(α). (A60) gyro-Bohm transport, Equation (A44). The coef- k ficient in the stellarator scaling is a factor of 2.09 Equatiion (A59) implicitly gives the function λ(α), times smaller than in gyro-Bohm scaling, which can Table I. In the absence of a region of rapid transport, be counterbalanced by D = 6.3. The rotational- ISS04 α = 0, the solution vanishes, f(λ) = 0 at λ = λ0 ≈ transform dependence of τE can be interpreted 2.405 and F = λ0J1(λ0). as D = ∆/ρs ∝ 1/ι, which may even be correct.

16 The detached divertor experiments in the Large This requires D = 0.11. The power loading is pw = 2 2 Helical Device (LHD) that were reported in 2018 [56] PT /((2π) κeRa) = 1.2 MW/m , where κe = 1.6 is had a = 0.55, R = 3.90, n0 = 0.7, B = 3, Pth = 9, the elongation. and a stored plasma energy Wth = 0.35. These re- When the confinement factor D, the magnetic field sults were said to be agreement with Equation (A64) strength B, and the wall loading pw are held con- for stellarator scaling. The central temperature is re- stant, the factor that determines the total power lated to the thermal energy content in the analytic 2 6/5 output PT scales as a ∝ T0 . The higher plasma model by temperature required even in a pulsed tokamak re- actor offsets its lower aspect ratio in comparison to W th a steady state stellarator reactor. For example using T0 = 2 2 (A65) 3π m¯ 0kBn0Ra stellarator definitions, the European Union DEMO ≈ 0.147. (A66) has an aspect ratio of 2.45 but a central temperature is 2.5 times greater than would probably be cho- Consistency with Equation (A48) is obtained for sen for a stellarator reactor, 2.45 × (2.5)6/5 = 7.364, D = 2.1. which is a reasonable aspect ratio for a stellarator If D were 2.1 for stellarators, but the energy con- reactor. finement time were factor of two longer, as in the case in H-mode tokamaks in Figure 3, then D would be 0.37 for H-mode tokamaks. Smaller values of D have been seen in tokamak and stellarator experi- 5. Non-turbulent transport ments. A study of long-pulse DIII-D results published in The characteristic diffusion coefficient for neoclas- 2018 [57] had T0 = (Te + Ti)/2 = 0.45 and n0 = 0.5, sical transport in which the particle drift trajectories B = 1.6, R = 1.7, at = 0.6, and Pth = 15.6. The make small excursions from the magnetic surfaces is elongation was κe = 2, which makes the stellarator √ 2 definition a = κeat = 0.849. A fit gives D = 0.31. Dnc = αncρi νi, and (A67) Early results from W7-X [44, 45] demonstrate that Dnc excellent confinement can be obtained, D = 0.05, Dnc ≡ (A68) Dgb though this confinement rapidly degrades, possibly a because continual pellet injection is not yet avail- = αnc , (A69) λi able. The central plasma has Ti = Te = 3.5 keV 20 3 and n0 = 0.8 × 10 /m , B = 2.5 T, R = 5.5 m, where αnc is a dimensionless coefficient, which can a = 0.5 m, and Pth = 5 MW. W7-X was able 2 be of order 10 and νi is the ion collision frequency. to maintain plasma parameters for ten seconds [58] 20 3 The mean free path λi ≡ Cs/νi ≈ 10.05 × with Ti = Te = 1.9 keV and n0 = 1.6 × 10 /m , 3 2 3 10 T /n, so λi/a ∼ 5 × 10 in a fusion reactor B = 2.5 T, R = 5.5 m, a = 0.5 m, and Pth = Non-turbulent transport scales differently when 5.9 MW. These results give D = 0.13, a value that the drift of some of the particles away from the mag- yields attractive reactor designs. netic surfaces is limited only by collisions. The char- Burning plasma experiments in ITER seem to re- acteristic transport coefficient for this type of trans- quire a value of D consistent with those seen in DIII- port is D. For example, the burning-plasma scenario out- lined in Table 1 of [59] for ITER had PT = 500,  2 ρi Cs at = 2, R = 6.2, κe = 1.8, B = 5.3, < n >= 1.1, D1/ν = α1/ν νi, and (A70) and < T >= 0.89. Assuming broad profiles so a νi D n0 = 1.18 and T0 = 1.1 gives D = 0.42. Similarly, 1/ν D1/ν ≡ (A71) the European Union design for a pulsed demonstra- Dgb tion (DEMO) tokamak reactor [11] has T0 = 25keV λi 20 3 = α . (A72) and n0 = 1.5 × 10 /m , B = 5.9 T, R = 9 m, 1/ν a at = 2.9 m, and PT = 2, 014 MW, assuming central values are twice their volume averages. The stellara- That is D can be orders of magnitude greater √ 1/ν tor equivalent minor radius is a = κeat = 3.67 m. than DgB.

17 [1] The Costs of Decarbonisation: System Costs with energy, Physics Today 72, issue 12, page 34 (De- High Shares of Nuclear and Renewables, OECD cember 2019). 2019 NEA No. 7299, (Nuclear Energy Agency, [11] G. Federici, C. Bachmann, L. Barucca, C. Baylard, Organization for Economic Co-operation and De- W. Biel, L.V. Boccaccini, C. Bustreo, S. Ciattaglia, velopment), http://www.oecd-nea.org/ndd/pubs/ F. Cismondi, V. Corato, C. Day, E. Diegele, T. 2019/7299-system-costs.pdf Franke, E. Gaio, C. Gliss, T. Haertl, A. Ibarra, J. [2] National Academies of Sciences, Engineering and Holden, G. Keech1, R. Kembleton, A. Loving, F. Medicine, Negative Emissions Technologies and Re- Maviglia, J. Morris, B. Meszaros, I. Moscato, G. liable Sequestration: A Research Agenda National Pintsuk, M. Siccinio, N. Taylor, M. Q. Tran, C. Academies Press, Washington, D.C. (2019). https: Vorpahl, H. Walden, and J.H. You, Overview of the //www.nap.edu/read/25259/chapter/1 DEMO staged design approach in Europe, Nucl. Fu- [3] P. Wang, X. Deng, H. Zhou, and S. Yu, Estimates sion 59, 066013 (2019). of the social cost of carbon: A review based on meta- [12] F. Perkins, D. E. Post, N. A. Uckan, M. Azumi, analysis, Journal of Cleaner Production 209, 1494 D. J. Campbell, N. Ivanov, N. R. Sauthoff, M. (2019). Wakatani, W. M. Nevins, M. Shimada, J. Van Dam, [4] 2019 Energy Statistics Pocketbook, United Nations D. Boucher, G. Cordey, A. Costley, J. Jacquinot, G. Department of Economic and Social Affairs Statis- Janeschitz, S. Mirnov, V. Mukhovatov, G. Porter, tics Division (United Nations New York, 2019), S. Putvinski, M. Shimada, R. Stambaugh, M. available online at https://doi.org/10.18356/ Wakatani, J. Wesley, K. Young, R. Aymar, Y. Shi- 1df8f86d-en. momura, D. Boucher,N. Fujisawa, Y. Igitkhanov, [5] M. Fasihi, O. Efimova, C. Breyer, Techno-economic A. Kukushkin, V. Mukhovatov, S. Putvinski, M. assessment of CO2 direct air capture plants, Journal Rosenbluth, and J. Wesley, Chapter 1: Overview of Cleaner Production 224, 957 (2019). and summary, Nucl. Fusion 39, 2137 (1999). [6] 2019 World Economic Outlook: Growth Slowdown, [13] G. Grieger, W. Lotz, P. Merkel, J. N¨uhrenberg, J. Precarious Recovery (International Monetary Fund, Sapper, E. Strumbeger, H. Wobig, R. Burhenn, V. Washington, DC, April 2018). Available on line at Erckmann, U. Gasparino, L. Giannoe, H. J. Hart- ISSN 1564-5215. fuss, R. Jaenicke, G. Kuhner, H. Ringler, A. Weller, [7] C. Seneviratne and C. Ozansoy, Frequency response and F. Wagner, Physics optimization of stellarators, due to a large generator loss with the increasing Phys. Fluids B 4, 2081 (1992). penetration of wind/PV generation—A literature re- [14] T.C. Hender, J.C Wesley, J. Bialek, A. Bon- view, Renewable and Sustainable Energy Reviews deson, A.H. Boozer, R.J. Buttery, A. Garofalo, 57 (2016) 659 (2016). T.P Goodman, R.S. Granetz, Y. Gribov, O. Gru- [8] U.S. Energy Information Administration, As- ber, M. Gryaznevich, G. Giruzzi, S. Gunter, N. sumptions to the Annual Energy Outlook Hayashi, P. Helander, C.C. Hegna, D.F. Howell, 2020: Electricity Market Module, https: D.A. Humphreys, G.T.A. Huysmans, A.W. Hyatt, //www.eia.gov/outlooks/aeo/assumptions/ A. Isayama, S.C. Jardin, Y. Kawano, A. Kellman, C. pdf/electricity.pdf, January 2020. Kessel, H.R. Koslowski, R.J. La Haye, E. Lazzaro, [9] D. A. Gates, D. Anderson, S. Anderson, M. Zarn- Y.Q. Liu, V. Lukash, J. Manickam, S. Medvedev, V. storff, D. A. Spong, H. Weitzner, G. H. Neilson, D. Mertens, S.V. Mirnov, Y. Nakamura, G. Navratil, Ruzic, D. Andruczyk, J. H. Harris, H. Mynick, C. M. Okabayashi, T. Ozeki, R. Paccagnella, G. Pau- C. Hegna, O. Schmitz, J. N. Talmadge, D. Curreli, tasso, F. Porcelli, V.D. Pustovitov, V. Riccardo, D. Maurer, A. H. Boozer, S. Knowlton, J. P. Al- M. Sato, O. Sauter, M.J. Schaffer, M. Shimada, P. lain, D. Ennis, G. Wurden, A. Reiman, J. D. Lore, Sonato, E.J. Strait, M. Sugihara, M. Takechi, A.D. M. Landreman, J. P. Freidberg, S. R. Hudson, M. Turnbull, E. Westerhof, D.G. Whyte, R. Yoshino, Porkolab, D. Demers, J. Terry, E. Edlund, S. A. H. Zohm and the ITPA MHD, Disruption and Mag- Lazerson, N. Pablant, R. Fonck, F. Volpe, J. Canik, netic Control Topical Group, Chapter 3: MHD sta- R. Granetz, A. Ware, J. D. Hanson, S. Kumar, C. bility, operational limits and disruptions, Nucl. Fu- Deng, K. Likin, A. Cerfon, A. Ram, A. Hassam, S. sion 47 S128 (2007). Prager, C. Paz-Soldan, M. J. Pueschel, I. Joseph, [15] L. Hesslow, O. Embreus, O. Vallhagen, and T. A. H. Glasser, Stellarator Research Opportunities: F¨ul¨op, Influence of massive material injection on A Report of the National Stellarator Coordinating avalanche runaway generation during tokamak dis- Committee, Journal of Fusion Energy 37, 51 (2018). ruptions, Nucl. Fusion 59, 084004 (2019). [10] R. Hawryluk and H. Zohm, The challenge and [16] B. N. Breizman, P. Aleynikov, E. M. Hollmann, and promise of studying burning plasmas: Answers to M. Lehnen, Review: Physics of runaway electrons in open questions that will be addressed by the ITER tokamaks, Nucl. Fusion 59 083001 (2019). experiment should enable the production of fusion [17] M. L. Reinke, S. Scott, R. Granetz, J.W. Hughes,

18 S.G. Baek, S. Shiraiwa, R. A. Tinguely, S. Wuk- Plasma Phys. 85, 905850103 (2019). itch, and The Alcator C-Mod Team, Avoidance of [30] A. H. Boozer, Transport and isomorphic equilibria, impurity-induced current quench using lower hybrid 26, 496 (1983). current drive, Nucl. Fusion 59, 066003 (2019). [31] J. R. Cary and S. G. Shasharina Omnigenity and [18] C. Linsmeier, M. Rieth, J. Aktaa, T. Chikada, A. quasihelicity in helical plasma confinement systems, Hoffmann, J. Hoffmann, A. Houben, H. Kurishita, Phys. Plasmas 4, 3323 (1997). X. Jin, M. Li, A. Litnovsky, S. Matsuo, A. von [32] M. Landreman and P. J. Catto, Omnigenity as gen- Muller, V. Nikolic, T. Palacios, R. Pippan, D. Qu, eralized quasisymmetry, Phys. Plasmas 19, 056103 J. Reiser, J. Riesch, T. Shikama, R. Stieglitz, T. (2012). Weber, S. Wurster, J.-H You, and Z. Zhou, Devel- [33] A. H. Boozer, Curl-free magnetic fields for stellara- opment of advanced high heat flux and plasma-facing tor optimization, Phys. Plasmas, 26, 102504 (2019) materials, Nuclear Fusion 57, 092007 (2017). [34] A. H. Boozer, Enhanced control, Nature Physics 14, [19] J.W. Coenena, Y. Maoa, S. Sistlac, A. V. M¨ullerb, 1157 (2018). G. Pintsuka, M. Wirtza, J. Rieschb, T. Hoeschenb, [35] J. K. Park, Y. Jeon, Y. In, J. W. Ahn, R. Nazikian, A. Terraa, J.-H. Youb, H. Greunerb, A. Kretera, G. Park, J. Kim, H. Lee, W. Ko, H. S. Kim, N. C. Ch. Broeckmannc, R. Neub, Ch. Linsmeiera, Mate- Logan, Z. R. Wang, E. A. Feibush, J. E. Menard, rials development for new high heat-flux component and M. C. Zarnstorff, 3D field phase-space control in mock-ups for DEMO, Fusion Engineering and De- tokamak plasmas, Nature Physics 14, 1223 (2018). sign 146, 1431 (2019). [36] P. Helander, T. Bird, F. Jenko, R. Kleiber, G. [20] C. E. Kessel, D. Andruczyk, J. P. Blanchard, T. G. Plunk, J. H. E. Proll, J Riemann, and P. Bohm, A. Davis, K. Hollis, P. W. Humrickhouse, Xanthopoulos, Advances in stellarator gyrokinetics, M. Hvasta, M. Jaworski, J. Jun, Y. Katoh, A. Kho- Nucl. Fusion 55, 053030 (2015). dak, J. Klein, E. Kolemen, G. Larsen, R. Majeski, [37] I. J. McKinney, M. J. Pueschel, B. J. Faber, B. J. Merrill, N. B. Morley, G. H. Neilson, B. Pint, C. C. Hegna, J. N. Talmadge, D. T. Anderson, M. E. Rensink, T. D. Rognlien, A. F. Rowcliffe, S. H. E. Mynick, and P. Xanthopoulos, A compari- Smolentsev, M. S. Tillack, L. M. Waganer, G. M. son of turbulent transport in a quasi-helical and a Wallace, P. Wilson, and S.-J. Yoon, Critical Explo- quasi-axisymmetric stellarator, J. Plasma Phys. 85, ration of Liquid Metal Plasma-Facing Components 905850503 (2019). in a Fusion Nuclear Science Facility, Fusion Science [38] P. H. Diamond, S. I. Itoh, K. Itoh, and T. S. Hahm, and Technology 75, 886 (2019). Zonal flows in plasma - a review, Plasma Phys. Con- [21] M. Abdou, N. B. Morley, S. Smolentsev, A. Ying, trol Fusion, 47, R35 (2005). S. Malang, A Rowcliffe, and M. Ulrickson, Blan- [39] C. C. Hegna, P. W. Terry, and B. J. Faber, Theory ket/first wall challenges and required R&D on the of ITG turbulent saturation in stellarators: Identify- pathway to DEMO, Fusion Engineering and Design ing mechanisms to reduce turbulent transport, Phys. 100, 2 (2015). Plasmas 25 022511 (2018). [22] A. H. Boozer, Stellarator design, J. Plasma Phys., [40] P. Xanthopoulos, H. E. Mynick, P. Helander, Y. 81, 515810606 (2015). Turkin, G. G. Plunk, F. Jenko, T. Goerler, D. Told, [23] F. J. Mangiarotti and J. V. Minervini, Advances T. Bird, and J. H. E. Proll, Controlling Turbulence on the Design of Demountable Toroidal Field Coils in Present and Future Stellarators, Phys. Rev. Lett. With REBCO Superconductors for an ARIES-I 113, 155001 (2014). Class Fusion Reactor, IEEE Transaction on Applied [41] M. D. J. Cole, R. Hager, T. Moritaka, J. Dominski, Superconductivity 25, 4201905 (2015). JR. Kleiber, S. Ku, S. Lazerson, J. Riemann, and [24] H. Yamaguchi, A quasi-isodynamic magnetic field C. S. Chang, Verification of the global gyrokinetic generated by helical coils, Nucl. Fusion 59 104002 stellarator code XGC-S for linear ion temperature (2019). gradient driven modes, Phys. Plasmas 26, 082501 [25] M. Landreman and A. H. Boozer, Efficient magnetic (2019). fields for supporting toroidal plasmas, Phys. Plasmas [42] M. E. Austin, A. Marinoni, M. L. Walker, M. W. 23, 032506 (2016). Brookman, J. S. deGrassie, A. W. Hyatt, G. R. Mc- [26] A. H. Boozer, Physics of magnetically confined plas- Kee, C. C. Petty, T. L. Rhodes, S. P. Smith, C. mas, 76, 1071 (2004). Sung, K. E. Thome, and A. D. Turnbull, Achieve- [27] C. Mercier, Equilibrium and stability of a toroidal ment of Reactor-Relevant Performance in Negative magnetohydrodynamic system in the neighbourhood Triangularity Shape in the DIII-D Tokamak, Phys. of a magnetic axis, Nucl. Fusion 4, 213 (1964). Rev. Lett. 122, 115001 (2019). [28] D. A. Garren and A. H. Boozer, Existence of quasi- [43] J. W. Connor, T. Fukuda, X. Garbet, C. helically symmetrical stellarators, Phys. Fluids B 3, Gormezano, V. Mukhovatov, M. Wakatani, the ITB 2822 (1991). Database Group and the ITPA Topical Group on [29] M. Landreman, W. Sengupta, and G. G. Plunk, Transport and Internal Barrier Physics, A review Direct construction of optimized stellarator shapes. of internal transport barrier physics for steady-state Part 2. Numerical quasisymmetric solutions, J. operation of tokamaks, Nucl. Fusion 44, R1 (2004).

19 [44] T. S. Pedersen, R. Konig, M. Jakubowski, M. Press, Oxford, 3rd edition, 2004. Krychowiak, D. Gradic, C. Killer, H. Niemann, T. [54] H. Yamada, J.H. Harris, A. Dinklage, E. Ascasibar, Szepesi, U. Wenzel, A. Ali, G. Anda, J. Baldzuhn, F. Sano, S. Okamura, J. Talmadge, U. Stroth, A. C. Biedermann, B.D. Blackwell, H.-S. Bosch, S. Kus, S. Murakami, M. Yokoyama, C.D. Beidler, V. Bozhenkov, R. Brakel, S. Brezinsek, J. Cai, B. Tribaldos, K.Y. Watanabe, and Y. Suzuki, Charac- Cannas, J.W. Coenen, J. Cosfeld, A. Dinklage, T. terization of energy confinement in net-current free Dittmar, P. Drewelow, P. Drews, D. Dunai, F. Ef- plasmas using the extended International Stellarator fenberg, M. Endler, Y. Feng, J. Fellinger, O. Ford, Database, Nucl. Fusion 45, 1684 (2005). H. Frerichs, G. Fuchert, Y. Gao, J. Geiger, A. Go- [55] A. Dinklage, C. D. Beidler, P. Helander, G. Fuchert, riaev, K. Hammond, J. Harris, D. Hathiramani, M. H. Maassberg, K. Rahbarnia, T. Sunn Pedersen, Y. Henkel, Ye. O. Kazakov, A. Kirschner, A. Knieps, Turkin, R. C. Wolf, A. Alonso, T. Andreeva, B. M. Kobayashi, G. Kocsis, P. Kornejew, T. Kre- Blackwell, S. Bozhenkov, B. Buttensch¨on,A. Czar- meyer, S. Lazerzon, A. LeViness, C. Li, Y. Li, Y. necka, F. Effenberg, Y. Feng, J. Geiger, M. Hirsch, Liang, S. Liu, J. Lore, S. Masuzaki, V. Moncada, U. H¨ofel,M. Jakubowski, T. Klinger, J. Knauer, G. O. Neubauer, T. T. Ngo, J. Oelmann, M. Otte, Kocsis, A. Kr¨amer-Flecken, M. Kubkowska, A. Lan- V. Perseo, F. Pisano, A. Puig Sitjes, M. Rack, genberg, H. P. Laqua, N. Marushchenko, A. Moll´en, M. Rasinski, J. Romazanov, L. Rudischhauser, G. U. Neuner, H. Niemann, E. Pasch, N. Pablant, L. Schlisio, J.C. Schmitt, O. Schmitz, B. Schweer, S. Rudischhauser, H. M. Smith, O. Schmitz, T. Stange, Sereda, M. Sleczka, Y. Suzuki, M. Vecsei, E. Wang, T. Szepesi, G. Weir, T. Windisch, G. A. Wurden, T. Wauters, S. Wiesen, V. Winters, G.A. Wurden, D. Zhang, and the W7-X Team, Magnetic config- D. Zhang, S. Zoletnik and the W7-X Team, First uration effects on the Wendelstein 7-X stellarator, divertor physics studies in Wendelstein 7-X, Nucl. Nature Physics 14, 855 (2018). Fusion 59, 096014 (2019). [56] M. Kobayashi, S. Masuzaki, K. Tanaka, T. [45] T. S. Pedersen, R. K¨onig, M. Krychowiak, M. Tokuzawa, M. Yokoyama, Y. Narushima, I. Ya- Jakubowski, J. Baldzuhn, S. Bozhenkov, G. mada, T. Ido, R. Sekiab, and The LHD Experi- Fuchert, A. Langenberg, H. Niemann, D. Zhang, K. mental Group, Core plasma confinement during de- Rahbarnia, H.-S. Bosch, Y. Kazakov, S. Brezinsek, tachment transition with RMP application in LHD, Y. Gao, N. Pablant, and the W7-X Team, First re- Nuclear Materials and Energy 17, 137 (2018). sults from divertor operation in Wendelstein 7-X, [57] J. M. Park, J. R. Ferron, C. T. Holcomb, R. J. But- Plasma Phys. Control. Fusion 61 014035 (2019). tery, W. M. Solomon, D. B. Batchelor, W. Elwasif, [46] A. Bader, A. H. Boozer, C. C. Hegna, S. A. Laz- D. L. Green, K. Kim, O. Meneghini, M. Murakami, erson, and J. C. Schmitt, HSX as an example of a and P. B. Snyder, Integrated modeling of high βN resilient non-resonant divertor, Phys. Plasmas 24, steady state scenario on DIII-D, Phys. Plasmas 25, 032506 (2017). 012506 (2018). [47] A. Bader, C. C. Hegna, M. Cianciosa, and G. J. [58] T. Klinger, T. Andreeva, S. Bozhenkov, C. Brandt, Hartwell, Minimum magnetic curvature for resilient R. Burhenn, B. Buttenschn, G. Fuchert, B. Geiger, divertors using Compact Toroidal Hybrid geometry, O. Grulke, H.P. Laqua, N. Pablant, K. Rahbarnia, Plasma Phys. Control. Fusion 60, 054003 (2018). T. Stange, A. von Stechow, N. Tamura, H. Thom- [48] A. H. Boozer and A. Punjabi, Simulation of stellara- sen, Y. Turkin, T. Wegner, I. Abramovic, S. kslom- tor divertors, Phys. Plasmas 25 , 092505 (2018). polo, J. Alcuson, P. Aleynikov, K. Aleynikova, A. [49] A. Punjabi and A. H. Boozer, Simulation of non- Ali, A. Alonso, G. Anda, E. Ascasibar, J.P. Bh- resonant stellarator divertor, Phys. Plasmas 27, ner, S.G. Baek, M. Balden, J. Baldzuhn, M. Ban- 012503 (2020). duch, T. Barbui, W. Behr, C. Beidler, A. Benndorf, [50] P. Bruzzone, W. H. Fietz, J. V. Minervini, M. C. Biedermann, W. Biel, B. Blackwell, E. Blanco, Novikov, N. Yanagi, Y. Zhai, and J. Zheng, Re- M. Blatzheim, S. Ballinger, T. Bluhm, D. Bcken- view: High temperature superconductors for fusion hoff, B. Bswirth, L.-G. Bttger, M. Borchardt, V. magnets, Nucl. Fusion 58 103001 (2018). Borsuk, J. Boscary, H.-S. Bosch, M. Beurskens, R. [51] D. Whyte, Small, modular and economically attrac- Brakel, H. Brand, T. Bruer, H. Braune, S. Brezin- tive fusion enabled by high temperature superconduc- sek, K.-J. Brunner, R. Bussiahn, V. Bykov, J. Cai, tors, Philosophical transactions of the Royal Society I. Calvo, B. Cannas, A. Cappa, A. Carls, D. Car- of London. Series A: Mathematical, physical, and ralero, L. Carraro, B. Carvalho, F. Castejon, A. engineering sciences 377, issue 2141, page 20180354 Charl, N. Chaudhary, D. Chauvin, F. Chernyshev, (2019). M. Cianciosa, R. Citarella, G. Claps, J. Coenen, [52] J. P. Freidberg, F. J. Mangiarotti, and J. Min- M. Cole, M.J. Cole, F. Cordella, G. Cseh, A. Czar- ervini, Designing a tokamak fusion reactor?How necka, K. Czerski, M. Czerwinski, G. Czymek, A. does plasma physics fit in?, Phys. Plasmas 22, da Molin, A. da Silva, H. Damm, A. de la Pena, 070901 (2015). S. Degenkolbe, C.P. Dhard, M. Dibon, A. Dinklage, [53] John Wesson, Tokamaks, International Series of T. Dittmar, M. Drevlak, P. Drewelow, P. Drews, Monographs on Physics 118, Oxford University F. Durodie, E. Edlund, P. van Eeten, F. Effenberg,

20 G. Ehrke, S. Elgeti, M. Endler, D. Ennis, H. Es- Papenfu, S. Paqay, E. Pasch, A. Pavone, E. Pawelec, teban, T. Estrada, J. Fellinger, Y. Feng, E. Flom, T.S. Pedersen, G. Pelka, V. Perseo, B. Peterson, D. H. Fernandes, W.H. Fietz, W. Figacz, J. Fontde- Pilopp, S. Pingel, F. Pisano, B. Plaum, G. Plunk, caba, O. Ford, T. Fornal, H. Frerichs, A. Freund, T. P. Plskei, M. Porkolab, J. Proll, M.-E. Puiatti, A. Funaba, A. Galkowski, G. Gantenbein, Y. Gao, J. Puig Sitjes, F. Purps, M. Rack, S. Rcsei, A. Reiman, Garca Regaa, D. Gates, J. Geiger, V. Giannella, A. F. Reimold, D. Reiter, F. Remppel, S. Renard, R. Gogoleva, B. Goncalves, A. Goriaev, D. Gradic, M. Riedl, J. Riemann, K. Risse, V. Rohde, H. Rhlinger, Grahl, J. Green, H. Greuner, A. Grosman, H. Grote, M. Rom, D. Rondeshagen, P. Rong, B. Roth, L. M. Gruca, C. Guerard, P. Hacker, X. Han, J.H. Rudischhauser, K. Rummel, T. Rummel, A. Runov, Harris, D. Hartmann, D. Hathiramani, B. Hein, B. N. Rust, L. Ryc, S. Ryosuke, R. Sakamoto, M. Heinemann, P. Helander, S. Henneberg, M. Henkel, Salewski, A. Samartsev, E. Sanchez, F. Sano, S. J. Hernandez Sanchez, C. Hidalgo, M. Hirsch, K.P. Satake, J. Schacht, G. Satheeswaran, F. Schauer, Hollfeld, U. Hfel, A. Hlting, D. Hschen, M. Houry, T. Scherer, J. Schilling, A. Schlaich, G. Schlisio, J. Howard, X. Huang, Z. Huang, M. Hubeny, M. F. Schluck, K.-H. Schl¨uter,J. Schmitt, H. Schmitz, Huber, H. Hunger, K. Ida, T. Ilkei, S. Illy, B. Is- O. Schmitz, S. Schmuck, M. Schneider, W. Schnei- raeli, S. Jablonski, M. Jakubowski, J. Jelonnek, H. der, P. Scholz, R. Schrittwieser, M. Schr¨oder,T. Jenzsch, T. Jesche, M. Jia, P. Junghanns, J. Kac- Schr¨oder, R. Schroeder, H. Schumacher, B. Schweer, marczyk, J.-P. Kallmeyer, U. Kamionka, H. Kasa- E. Scott, S. Sereda, B. Shanahan, M. Sibilia, P. hara, W. Kasparek, Y.O. Kazakov, N. Kenmochi, Sinha, S. Sipli, C. Slaby, M. Sleczka, H. Smith, W. C. Killer, A. Kirschner, R. Kleiber, J. Knauer, Spiess, D.A. Spong, A. Spring, R. Stadler, M. Ste- M. Knaup, A. Knieps, T. Kobarg, G. Kocsis, F. jner, L. Stephey, U. Stridde, C. Suzuki, J. Svensson, Kchl, Y. Kolesnichenko, A. Knies, R. Knig, P. Ko- V. Szab, T. Szabolics, T. Szepesi, Z. Szkefalvi-Nagy, rnejew, J.-P. Koschinsky, F. Kster, M. Krmer, R. A. Tancetti, J. Terry, J. Thomas, M. Thumm, J.M. Krampitz, A. Krmer-Flecken, N. Krawczyk, T. Kre- Travere, P. Traverso, J. Tretter, H. Trimino Mora, meyer, J. Krom, M. Krychowiak, I. Ksiazek, M. H. Tsuchiya, T. Tsujimura, S. Tulipn, B. Unter- Kubkowska, G. Khner, T. Kurki-Suonio, P.A. Kurz, berg, I. Vakulchyk, S. Valet, L. Vano, B. van Mil- S. Kwak, M. Landreman, P. Lang, R. Lang, A. ligen, A.J. van Vuuren, L. Vela, J.-L. Velasco, M. Langenberg, S. Langish, H. Laqua, R. Laube, S. Vergote, M. Vervier, N. Vianello, H. Viebke, R. Vil- Lazerson, C. Lechte, M. Lennartz, W. Leonhardt, brandt, A. Vorkper, S. Wadle, F. Wagner, E. Wang, C. Li, C. Li, Y. Li, Y. Liang, C. Linsmeier, S. N. Wang, Z. Wang, F. Warmer, T. Wauters, L. We- Liu, J.-F. Lobsien, D. Loesser, J. L. Cisquella, J. gener, J. Weggen, Y. Wei, G. Weir, J. Wendorf, U. Lore, A. Lorenz, M. Losert, A. Lcke, A. Lums- Wenzel, A. Werner, A. White, B. Wiegel, F. Wilde, daine, V. Lutsenko, H. Maaberg, O. Marchuk, J.H. T. Windisch, M. Winkler, A. Winter, V. Winters, Matthew, S. Marsen, M. Marushchenko, S. Ma- S. Wolf, R.C. Wolf, A. Wright, G. Wurden, P. Xan- suzaki, D. Maurer, M. Mayer, K. McCarthy, P. Mc- thopoulos, H. Yamada, I. Yamada, R. Yasuhara, M. Neely, A. Meier, D. Mellein, B. Mendelevitch, P. Yokoyama, M. Zanini, M. Zarnstorff, A. Zeitler, D. Mertens, D. Mikkelsen, A. Mishchenko, B. Missal, J. Zhang, H. Zhang, J. Zhu, M. Zilker, A. Zocco, S. Zo- Mittelstaedt, T. Mizuuchi, A. Mollen, V. Moncada, letnik and M. Zuin, Overview of first Wendelstein T. Mnnich, T. Morisaki, D. Moseev, S. Murakami, 7-X high-performance operation, Nucl. Fusion 59, G. Nfrdi, M. Nagel, D. Naujoks, H. Neilson, R. 112004 (2019). Neu, O. Neubauer, U. Neuner, T. Ngo, D. Nicolai, [59] B. J. Green for the ITER International Team and S.K. Nielsen, H. Niemann, T. Nishizawa, R. Nocen- Participant Teams, ITER: burning plasma physics tini, C. N¨uhrenberg, J. N¨uhrenberg, S. Obermayer, experiment Plasma Phys. Control. Fusion 45, 687 G. Offermanns, K. Ogawa, J. lmanns, J. Ongena, (2003). J.W. Oosterbeek, G. Orozco, M. Otte, L. Pacios Rodriguez, N. Panadero, N. Panadero Alvarez, D.

21