ADX: a high field, high power density, advanced divertor and RF tokamak

The challenges that face a power-producing tokamak reactor can be best appreciated by considering the challenges that face ITER. In ITER's planned baseline scenario [15], its Q_DT = 10 burning core with 33% input power radiated would send 100 MW of heat exhaust via a narrow channel (with an assumed width, λ_q ∼ 5 mm, at outer midplane) towards its tungsten, vertical target plate divertor [16]—the most advanced divertor concept at the time of ITER's design, pioneered by Alcator C-Mod [17]. Assuming that disruptions can be avoided/mitigated and ELMs can be reduced or eliminated [16], this steady heat flow would be accommodated (i.e., reduced to less than 10 MW m⁻² surface heat flux) by producing a 'partially detached' divertor state, in which 70% of the scrape-off-layer (SOL) power is radiated, facilitated by impurity seeding (Ar, Ne, N). The projected value heat flux channel width (λ_q ∼ 5 mm) was highly uncertain, however, and a coordinated multi-machine investigation was initiated in 2010 to develop an empirical scaling law to ITER conditions. The results of this investigation do not support the assumption of λ_q ∼ 5 mm. In fact, the results clearly indicate [14] that λ_q in low-recycling H-modes is independent of machine size and scales inversely with poloidal magnetic field strength at the outer midplane, projecting to a λ_q in ITER that is the same as C-Mod, λ_q ∼ 1 mm (figure 2). Moreover, from simple geometry, this result means that the heat flux flowing along field lines into the divertor scales as q_∥ ∼ P_SOLB/R,⁷ with P_SOL being the total power entering the SOL. This is a serious challenge for successful ITER operation; experiments are now examining if this scaling for upstream λ_q holds true when the divertor is in a partially detached regime.

Zoom In Zoom Out Reset image size

A fallback position is to increase core radiation but at the penalty of degrading confinement. As shown in figure 3, experiments suggest that in order to maintain H₉₈ above 1, P_SOL must exceed the L–H power threshold, which is expected to be P_LH ∼ 50 MW for ITER. This implies that a doubling of the core radiated power fraction from 33% to 66% is possible, but not more. A factor of 5 reduction in 'planned λ_q' means that q_∥ at the entrance to the divertor may be 2.5 times the value originally anticipated.

Zoom In Zoom Out Reset image size

Experiments are presently using a combination of core and divertor seeding to explore the maximum P_SOLB/R that a conventional high-Z (W or Mo) vertical target plate can handle while maintaining good core confinement. ASDEX-Upgrade has obtained P_SOLB/R ∼ 18 [MW-T/m] [19] while C-Mod has obtained P_SOLB/R ∼ 25 [18]. Although experiments are continuing to optimize the tradeoff between power handling and confinement, this parameter is significantly short of P_LHB/R ∼ 45 required for ITER. As illustrated graphically in figure 4, the requirements for a DEMO (as embodied in ARIES ACT1 and ACT2 concepts [20]) or a FPP (ARC concept [11]) are much worse, even if large core radiation fractions (∼85%) are assumed such that these devices operate with P_SOL ∼ P_LH—a situation potentially at odds with avoiding fuel dilution and maintaining adequate core confinement. If ITER finds that the empirical projection for λ_q is somehow wrong and that λ_q is more in line with originally planned values, then the power handling challenge for ACT1/ACT2/ARC is still a factor of 2 to 3 times worse than ITER, simply because plasma exhaust power densities are 2 to 3 times higher.

Zoom In Zoom Out Reset image size

Long pulse length (∼5 × 10⁷ s) operation of FNSF/FPP/DEMO imposes an additional challenge on divertor operation, not encountered in present experiments or ITER—much more severe than the divertor power handling challenge: nearly complete suppression of material erosion and PMI damage at divertor target plates is required [21]. Steady-state heat removal of q_s ∼ 10 MW m⁻² on high-temperature solid targets requires thin refractory (W) plasma-facing components, on the order of ∼5 mm thick [22]. This restricts the allowed net surface erosion rate to be less than 1 mm per year, or a net W erosion flux of Γ_W ∼ 1.5 × 10¹⁸ m⁻² s⁻¹. At an electron temperature of 6 eV, plasma surface heat flux at q_s ∼ 10 MW m⁻² corresponds to a bombarding ion flux of Γ_ion ∼ 1.5 × 10²⁴ m⁻² s⁻¹. Thus net W erosion yield per incident ion must be suppressed to 10⁻⁶, even in the presence of impurity ions (Ar, Ne, N) used for seeding. Measurements [23] and models find W sputtering yields in the divertor to be on the order of 10⁻⁴ for typical impurity concentrations—a performance 'gap' of a factor of 100. It was thought that prompt-redeposition physics would help—sputtered W ionizing within a Larmor orbit that carries it back to the plate. But deposited material is not tightly bound to the bulk and becomes mixed with impurities; subsequent sputtering yields can be 10 times that of the original bulk material [24]. Helium ion bombardment at energies exceeding 20 eV also damages tungsten via implantation and, among other things, leads to the formation of 'nano-tendrils' at elevated surface temperatures [25]—a new mechanism to destroy a solid material surface. Thus, the only solution—apart from using self-annealing, liquid metal targets—is to operate in regimes in which the divertor plasma is fully detached, i.e., a state in which electron temperatures are well below 5 eV, impurity ion impact energies are below sputtering and damage thresholds, ion–neutral collisions remove plasma momentum, radiation/neutrals carry heat to surfaces and ion fluxes to target plates drop by more than two orders of magnitude. Unfortunately, present experiments (with conventional divertors) have not been able to attain a fully detached divertor (i.e., an order of magnitude reduction in ion flux at all divertor target surfaces) while at the same time maintaining good core energy confinement. The highest values of q_∥ ∼ P_SOLB/R that have been attained with acceptable heat loads and good confinement are with partially detached divertors. The difficulty in extending the detachment region over the entire divertor target comes from the tendency of the 'thermal front' to intrude into the confined plasma by forming an 'X-point MARFE' [26], thereby cooling the pedestal region and degrading core performance. Just as in the case of heat flux control, conventional divertors seem to impose an unavoidable tradeoff between maintaining good core confinement and protecting divertor targets from destruction. While specific details of the detachment behaviour differ among experiments (e.g., radiative species mix, divertor target plate material and orientation, extent of detachment front), the overall behaviour is the same. Divertor solutions that can decouple the detachment response from the core plasma response have not yet been attained—let alone demonstrated at the power and particle flux densities required for a DEMO.

Advanced magnetic divertors, such as the X-divertor (XD [27]), snowflake (SF [28]), Super-X divertor (SXD [29]) and X-point target divertor (XPT [9]), were conceived to address the extreme power exhaust challenge via a number of techniques (figure 5). Proof-of-principle experiments performed at low and moderate power fluxes are finding that the SF concept behaves roughly as anticipated: surface heat flux is reduced due to flux expansion, increased radiation and convective heat flows [30–34]. Further experiments are needed to compare performance improvement (i.e., heat flux reduction) of SF and XD concepts relative to an ITER-like vertical target plate divertor, which employs flux expansion via target tilt angle rather than poloidal field reduction.

Zoom In Zoom Out Reset image size

SXD and XPT concepts are long-legged divertors that attempt to solve the power handling challenge by placing targets at large major radius in a separate divertor chamber. This yields a corresponding reduction in q_∥ by total flux expansion, i.e., B ∼ 1/R. The XPT concept (shown in figure 5) starts with the SXD idea, but places an X-point in a divertor chamber as a 'virtual target', intercepting flux tubes that carry the highest q_∥. The local X-point in the divertor can be a first order or higher order null, taking advantage of a 'churning mode' identified for snowflake topologies [35] to activate multiple legs—a 'snowflake target divertor'. Most importantly, the XPT may produce a stable 'X-point MARFE' that is localized to the divertor chamber, ideally equidistance from divertor chamber walls, to spread heat uniformly over a large surface area. This is the ultimate goal for any advanced magnetic divertor—produce a fully detached divertor, reduce surface heat loads and erosion to acceptable levels while at the same time maintaining high core plasma confinement by avoiding pedestal cooling.

Analysis shows [36] that the position of the detachment 'thermal front' (i.e., a region where the temperature drops abruptly along a field line) is set by the balance between parallel heat inflow at the front's leading edge, q_∥, and the front's volumetric losses via a variety of means: intrinsic impurity radiation, extrinsic (seeded) impurity radiation, hydrogenic radiation, and charge exchange neutrals. A thermal front can fully accommodate the q_∥ up to maximum value, q_∥max, which, for a given impurity species radiator, is set by the plasma pressure and the impurity fraction. The lifetime of the impurity in the plasma also plays a role, affecting its average charge state and thus radiation level. When q_∥ exceeds q_∥max, the front gets 'pushed back' towards the target plate and/or the front gets 'burned though', i.e., the abrupt temperature drop along field lines is eliminated. Thus by placing the target plate or X-point target at large major radii, the detachment front may be stabilized; if the front attempts to move 'upstream' towards the confined plasma, it encounters a region of higher q_∥ which arrests its motion (bottom panel in figure 5). For these reasons, the SXD and XPT concepts are the ones targeted first for testing in ADX, which sets the initial poloidal field coil design. MAST is scheduled to perform proof-of-principle tests of the SXD concept for the first time in 2016/17 [37]. Although these tests will be at low power density, they should provide valuable information.

Liquid metal targets, despite their added complexity, may turn out to be the only means of handling the highly focused exhaust from a fusion reactor, as well as tolerating heat transients (ELMs, disruptions). Long-leg, highly baffled divertor geometries can work synergistically with liquid metal target concepts, which employ vapour shielding physics, e.g., as addressed in [38]. By operating divertor surfaces at temperatures lower than the main chamber walls, liquid metals and impurities may be controlled by preferential condensation in the divertor. A successful liquid metal target technology would also eliminate the need to operate with a fully detached divertor; the self-annealing surface would accommodate the bombarding ion flux without damage. As discussed in section 7.2, ADX will employ a modular vacuum vessel design which would enable the development and implementation of liquid metal divertor target ideas. The lower, outer divertor leg would be a natural test bed. Implementation of successful liquid metal target options in the upper, outer leg may require up–down asymmetric divertor magnetic topologies/geometries, different from that shown in figures 7 and 13. The ability to modify the vacuum vessel to accommodate such possibilities is an important asset.

The above considerations strongly motivate the need for an ADX and define its requirements: ADX must not only have the power density and magnetic field strength to simulate reactor-level conditions, it must also have the flexibility in its design to test the most promising ideas for solving heat flux and erosion challenges.

The electron temperature at the divertor target plate in a reactor will be in the range of ∼5 eV or less—this is dictated by atomic physics and the requirement that partial or full divertor detachment be attained. Stangeby and Leonard [21] argue that divertor target plate conditions of T_e,div ∼ 5 eV and n_div > 10²¹ m⁻³ must be attained in order to avoid rapid destruction of the divertor targets in a reactor by sputtering erosion. Having fixed the divertor electron temperature, the plasma density (and thus plasma pressure) in the divertor is determined by q_∥—either by the requirement of conducting q_∥ through the plasma sheath or by the requirement of dissipating q_∥ by impurity radiation in a thermal front [36]. In order to match key dimensionless parameters that control PMI physics (e.g., λ_debye/λ_ion, ρ_z/λ_ion, with λ_debye = Debye length, λ_ion = ionization mean free path, ρ_z = impurity Larmor radius), T_e,div, n_div and B must be made identical to a reactor [2]. Taken together, this means that B and q_∥ combined with magnetic/target plate geometry must be made identical to a reactor. Consistent with this picture is the requirement that reactor-level B−fields are needed to produce reactor-level plasma pressures in the boundary plasma, due to beta limits associated with the edge plasma pedestal [39]. As discussed by Hutchinson and Vlases [3], by matching B, n and q_∥ of a reactor it is also possible, at least in principle, to perform an approximate 'divertor similarity experiment' in which five key scale parameters (T_e, ν* = L_d/λ_ei, Δ_d/λ₀, ρ_i/Δ_d and β, with L_d = divertor field line length, λ_ei = electron–ion collisional mean free path, Δ_d = SOL thickness, λ₀ = neutral mean free path, ρ_i = ion Larmor radius, β = 2μ₀nT/B²) are identical to those of a reactor. The poloidal flux expansion and divertor leg length could be adjusted to do this. In any case, the overall message is clear: matching absolute reactor divertor parameters (B, q_∥, n_div, T_e,div) is necessary to study reactor-relevant physics regimes. Based on the recent multi-machine empirical scaling of parallel heat flux [14], the parallel heat flux scales as q_∥ ∼ P_SOLB/R. This means that B, n and P_SOLB/R in an ADX should be made similar to that in a reactor. From figure 6, it is clear that a compact, high-field tokamak is the natural choice for an ADX—it is the way to perform meaningful tests of advanced divertor prototypes without constructing a full-scale reactor; it removes the uncertainty and risk in trying to extrapolate divertor performance from extremely complex and uncertain computational models, developed and tested in regimes that are far from those expected in a reactor.

Zoom In Zoom Out Reset image size

It is recognized that RF current drive and heating technologies must evolve into primary actuators for fusion reactors, replacing the roles that neutral beams now play in most devices. As stated in a 2007 US DoE [12] report (addressing 'gap G-7' on page 190): 'The auxiliary systems typically used in current experiments, while extremely useful tools, are not generally suitable for a reactor. RF schemes are the most likely systems to be used and will require significant research to achieve the levels of reliability and predictability that are required.' For ICRF antennas and LHCD launchers, it means placing these actuators close to the plasma, where the challenge of efficient power coupling competes with minimizing PMI. For long pulse operation, the lifetime of the launcher will be determined by PMI. But most critically, the current drive system's efficiency (wall plug to plasma) and ability for current profile control are what matter the most; the viability of the tokamak concept as a steady-state fusion reactor depends on them. For this reason, LHCD is considered as a first-choice actuator; it is recognized as the most efficient RF current drive scheme. Unlike electron cyclotron current drive (ECCD) [42], the RF waves impart parallel momentum directly to electrons. On the other hand, ECCD may be a fallback option should the LHCD launcher PMI challenge prove impossible to solve.

5.2.1. Plasma–material interaction

From a PMI perspective, locating ICRF and LHCD antenna structures on the HFS of the torus with double-null magnetic equilibria is a potential game-changer (figure 7), yet it has never been exploited. Experiments clearly show that: (1) a low heat flux, quiescent boundary layer naturally forms on the HFS in double-null plasmas with: no heat/particle pulses reaching there from ELMs [45], essentially zero fluctuation-induced fluxes [40] and no 'blobs' (figure 8); (2) local plasma recycling fluxes are low, which minimizes neutral pressures in the vicinity of antenna/waveguide structures, leading to improved RF voltage handling [46]; (3) the flux of energetic ion orbit loss on the HFS is virtually non-existent; (4) energetic particles formed by RF fields have drift orbits that move away from the launch structure; (5) there is no impact from runaway electrons at this location; (6) any impurity ions that are produced by PMI on RF antenna structures would likely be very well screened, based on results from impurity transport experiments [43] (figure 9); (7) because of the steep profiles, the local density at HFS launch structures can be precisely controlled by adjusting the upper/lower X-point flux balance [47] and/or distance from the last-closed flux surface (LCFS) to launcher (figure 8).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

One might ask: can antennas be placed on the inner wall of a tokamak? Is there enough space? The answer is yes [11, 48]. The size of antenna/waveguide structures is dictated by RF wavelengths—not machine size—and these structures are already employed in compact tokamaks (e.g., C-Mod). Thus, inside-launch RF antennas can be designed to fit into the blanket of an FNSF/FPP/DEMO-sized device without compromising neutron shielding requirements while at the same time tapping into the local blanket cooling system. The RF 'plumbing' for an inside-launch LHCD system would not be much different than for coolant pipes in terms of size and robustness—metal waveguides, with no dielectrics, insulators or windows placed in the neutron environment. Potential designs for HFS LHCD and ICRF launchers in ADX are discussed in sections 7.3 and 7.4.

5.2.2. RF wave physics

HFS launch is a potential game-changer for LHCD wave physics, resulting in high efficiency current drive with profile control to mid minor radius [49–52] (figure 10). The increased magnetic field at this location enables waves to be launched with lower n_∥, penetrating deeper into the plasma before damping. This alone can lead to significant increases in current drive efficiency—more than 60% in the Vulcan study [49]. By launching off midplane near a region of reduced poloidal field (i.e., near an X-point), ray trajectories can also be 'steered' to propagate directly towards the plasma centre. These improvements have been exploited in both the Vulcan and ARC [11] designs. In fact, the favourable LHCD physics of high-field launch is central to making these concepts viable.

Zoom In Zoom Out Reset image size

Recent LHCD experiments in C-Mod have uncovered a clear, direct correlation between the local plasma density in the SOL and the onset of parasitic wave energy loss mechanisms (collisional damping [54], parametric decay instabilities [55, 56], wave scattering [57, 58] and full wave effects [59]). On the LFS, the SOL density profiles are broad, and, due to the rapid cross-field transport mechanisms that take place there, the density in the local SOL and at the launcher mouth is not controlled; the launcher must be placed very far from the LCFS to avoid PMI, particularly in a reactor. In contrast, the sharp density falloff in the HFS SOL (figure 8) allows the launcher to be located very close to the LCFS. This, combined with the ability for precise density control at the mouth may allow these SOL loss mechanisms to be virtually eliminated. Wave scattering from density blobs would also be non-existent.

HFS launch also leads to beneficial ICRF fast-wave physics [52]. With HFS launch, fast waves mode convert directly to ion Bernstein waves (IBW) and ion cyclotron waves (ICW), leading to strong single-pass absorption and direct electron heating. This avoids the generation of energetic ion tails and possible destabilization of energetic particle populations (including fusion alphas in a reactor), completely suppressing this mechanism for ion-induced sputtering of first-wall components.

In summary, we note that there is excellent three-way synergy among double-null plasmas, advanced divertor solutions and inside-launch RF systems—all working to focus the PMI onto systems that can handle it and away from systems that cannot—and at the same time providing unprecedented improvements in current drive and heating wave physics. These ground-breaking ideas, and the means to test and exploit them, are primary foci of the ADX conceptual design.

Results from a magnetic equilibrium scoping study, examining the ability to produce different divertor configurations using a fixed poloidal field coil set, are shown in figure 13. The cases shown correspond to a 1.0 MA, 5.4 T discharge. ADX is expected to have a similar power exhaust channel width as Alcator C-Mod, in the range of 1 to 3 mm. The first 10 mm of the SOL (highlighted in orange) is accommodated by envisioned divertor topologies. A set of 10 poloidal field coils (5 upper and 5 lower) is added to the C-Mod coil set to produce the geometries shown. The arrangement of coils shown provides for coaxial feed line connections to external power systems, but a wide range of other configurations and divertor leg shapes is also possible. Maximum currents in the poloidal field coils do not exceed typical values for a comparable C-Mod discharge. Magnetic forces are also similar and readily accommodated by the rigid vacuum vessel and mechanical supports placed at regular intervals in toroidal angle—the technique presently used successfully in Alcator C-Mod.

Zoom In Zoom Out Reset image size

Cryopumps are incorporated into the outer leg divertor chambers (small circle under the outer divertor leg, shown in figure 13), although their optimal placement and the optimum shape of the chambers has not yet been determined. It is envisioned that full or partial loop cryopumps would be used, employing an efficient 'pot' design developed for C-Mod [64], which does not require a flowing helium loop. Diagnosis of the divertor plasma, boundary plasma, and PMI at the divertor surfaces in the different configurations is crucial. Not shown are embedded gas feeds and divertor instrumentation that will be required to control and monitor divertor conditions. Optical access to the divertor volume is restricted compared to present-day tokamaks and will therefore require special radial ports that are included in the conceptual design.

One might ask: 'These poloidal field coil arrangements look like they will not have adequate room for neutron shielding. How can divertors like these be used in a fusion reactor?' A D–T neutron environment certainly places severe constraints on the placement of the coils and the design choices. Lackner and Zohm [65] considered the consequences of locating poloidal field coils far from the plasma and found that the coil currents required to produce a snowflake topology may become prohibitively high. But ADX is not constrained in these ways. Rather, ADX addresses a more fundamental question: given nearly complete freedom in coil arrangement, does there exist an advanced divertor scheme that can handle a fusion reactor's power exhaust density? Once having identified a successful scheme, there are several approaches that might be taken to implement it in a D–T environment. The first is to employ a coil arrangement that does have adequate shielding while retaining the desired physics effect. Kotschenreuther et al have found such schemes for the Super-X divertor [66]. The second is to employ non-superconducting poloidal field coils in critical locations that can withstand enhanced neutron damage while having sufficient, albeit reduced, operational lifetime. For example, single-turn copper control coils were proposed in the ARC concept [11]. These would survive for lifetimes of ∼2 years in the neutron environment and be replaced with the vacuum vessel. The use of a demountable, superconducting toroidal field magnet makes this possible. A combination of both schemes would probably be employed in a reactor—a set of radiation-shielded long-life superconducting coils to provide basic shaping in conjunction with replaceable shorter-lifetime coils that produce the fine structure and control of the divertor leg topology. In this sense, the development of demountable superconducting magnets and replaceable, radiation-tolerant divertor coils goes hand-in-hand with the development of advanced magnetic divertor schemes that can handle a reactor's power exhaust power density.

The demountable toroidal field magnet, with external superstructure, is an important feature of the ADX design, as well as being highly desirable for a reactor; (1) it facilitates the construction and maintenance of a very high field tokamak with internal poloidal field coils and (2) it serves as a 'shell' that can accommodate different vacuum vessel designs, poloidal field coil sets and/or divertor target components, facilitating relatively nimble configuration modifications. With the exception of RF launcher and antenna structures, it is envisioned that all internal vessel components (first-wall and divertor tiles, diagnostics, RF feeds, cryopumps) would be installed and tested off-line during assembly of an integrated vacuum vessel/poloidal field coil unit that is composed of modular sections (figure 14). One can also envision performing experiments on a specific divertor and/or RF concept in ADX while at the same time preparing new/modified vacuum vessel modules off-line, either on-site or off-site by collaborating institutions. This would be particularly valuable for implementing liquid metal divertor targets, where extensive off-line tests of the liquid handling system may be required. It should be noted that as part of C-Mod's routine maintenance programme, the sliding joints of the toroidal field magnets are inspected at regular intervals, requiring full disassembly of the superstructure—lifting the vacuum vessel with embedded coils to a temporary work location. The approach for ADX would be similar. A newly prepared vacuum vessel would be swapped out with the old one, or subsections of the modular vacuum vessel could be upgraded/replaced as needed to support the experimental programme.

Zoom In Zoom Out Reset image size

Using cryogenically cooled copper magnetics (LN₂), the overall plasma pulse length in ADX is projected to be ∼3 s at full performance and longer (∼5 s) at modestly reduced field and current. Based on preliminary estimates, the existing 225 MVA motor-generator at the PSFC combined with primary source power would allow 6.5 T, 1.5 MA operation for a 1 s flat-top. However, since the magnet systems and overall superstructure are based on the Alcator C-Mod design, it would support 8 T, 2 MA operation with an appropriate power systems upgrade.

One might ask: 'how is a pulse length of ∼3 s with ∼1 s flat-top relevant for developing steady-state power handling, PMI and RF solutions?' The answer is that a ∼ 3 or even ∼1 s pulse (τ_pulse) is actually ideal for ADX's mission. It is much longer than plasma energy (τ_E) and particle (τ_p) confinement time scales as well as PMI (τ_PMI) and current relaxation (τ_CR) time scales [48], but it is shorter than first-wall component thermal equilibration times (τ_thermal). Solutions identified in ADX can therefore be projected with confidence toward steady-state operation:

• τ_pulse ≫ τ_E, τ_p—steady power and particle exhaust is established in core and divertor; divertor performance, control methods, impurity behaviour, core performance can be assessed.
• τ_pulse ≫ τ_PMI—steady ion fluxes and material erosion/redeposition fluxes are established; erosion and redeposition, ion-induced damage rates (divertor and RF actuators) can be assessed. State-of-the art, in situ material surface diagnostics (e.g., [60]) can be employed to determine material response between discharges and to project their viability for long-pulse operation.
• τ_pulse ≫ τ_CR—fully relaxed current profiles are established; non-inductive current drive and current profile control for enhanced core plasma confinement scenarios can be assessed.
• τ_pulse < τ_thermal—first-wall components are 'inertially-cooled'. In this respect, ADX would not simulate steady-state reactor conditions. But τ_pulse < τ_thermal provides important practical and economic advantages: (1) surface heat fluxes of ∼40 MW m⁻² (∼1 s exposure) can be accommodated by refractory materials without surface melt damage; (2) the high cost and complexity of steady-state cooling is avoided, both for first-wall components and for RF systems; (3) simplified schemes for testing liquid metal divertor target ideas can be employed.

As stated in section 3, ADX's central mission is to identify reactor-relevant plasma physics solutions that extrapolate to steady state. The proposed pulse length is ideal for this purpose. A top priority for magnetic fusion energy development is to demonstrate solutions for divertor power and erosion handling that can operate successfully on these time scales and at the absolute performance parameters that are required (power exhaust density, magnetic field strength, plasma pressures) while at the same time producing excellent core plasma performance. Lacking that demonstration, any investment in next-step steady-state experiments carries a very high risk.

ADX would not address effects related to thermal equilibrated walls and the 'working' of material from long-time plasma (or neutron) exposure, e.g., film growth and/or surface modifications at very large fluences. In principle, high power, linear plasma-material test devices can be used to explore key aspects of this long-time material 'working' behaviour, once viable divertor solutions are found—i.e., the magnetic geometry, compatible materials, control schemes needed—and the resultant plasma parameters at the target plates (particle and energy fluxes, plasma densities and temperatures) are characterized in ADX.

Finally, while ADX would readily identify core plasma regimes that are time stationary (or not), over its 1–3 s pulse length, ADX would not fully resolve very long-time core plasma evolution (i.e., exceeding ∼10 to ∼30 energy confinement times), such as might occur from high-Z impurity accumulation inside an internal transport barrier.

The parameters shown in figures 6, 15 and 16 highlight ADX's unique ability to test divertor solutions at reactor-level conditions. As discussed in section 4.4, it is necessary to match absolute values of both B and q_∥ of a reactor in order to reproduce the divertor and PMI physics regimes of a reactor. Based on a multi-machine database [14], the parameter P_SOLB/R is the figure-of-merit that determines the relative magnitude of the 'upstream' q_∥ in the tokamak SOL. ADX's high field, high power and small major radius push P_SOL,MAXB/R, λ_q and q_∥ well beyond what is available from current or other planned experimental tokamaks, accessing values that are comparable to and go beyond ITER (see figure 15, and top panel in figure 16).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The magnitude of the upstream poloidal heat flux density, q_θ, can be scaled from another figure-of-merit parameter, P_SOLB_θ/R. The bottom panel in figure 16 compares P_SOLB_θ/R among world tokamaks. Again, ADX pushes this quantity far beyond what will be available from current or other planned experiments.

For ITER and devices such as ARC, ACT2 and ACT1, which are representative of FNSF/FPP/DEMO (orange bars in figure 16), it is highly desirable to operate with the lowest possible values of P_SOL compatible with obtaining good core plasma confinement. As discussed in section 4.1 (figure 3), data suggest that this lowest value is near the L–H power threshold. Thus the bottom of the orange vertical bars in figure 16 represent the minimum values of P_SOLB/R and P_SOLB_θ/R for which a divertor solution must be found (consistent with good core plasma confinement) before these reactor concepts can be considered viable. Existing experimental tokamaks can test divertor concepts, but only up to the maximum values of P_SOL available. Thus, in addition to having the required reactor-level magnetic field strengths, ADX is uniquely able to test divertor solutions at the heat flux densities needed to qualify them for FNSF/FPP/DEMO devices, as indicated by the grey shaded regions in figure 16.

For advanced divertors that attain a fully detached, dissipative plasma condition, P_SOLB_θ/R is important because it determines the required upstream-to-divertor heat flux footprint expansion factor, i.e., the required spreading of the upstream heat-channel width via radiation and plasma–neutral interactions onto a much larger 'wetted area' in the divertor chamber, as illustrated in the top panel of figure 5. For a maximum surface heat load, q_s,max, the expansion factor, S, must be large enough to satisfy the rough inequality, P_SOL/(4πRSλ_q) < q_s,max, which accounts for two active divertor chambers. With λ_q scaling as 1/B_θ, S scales as P_SOLB_θ/R.

Figure 15 (note 8) shows values of S that are needed to spread the tokamak's total source power uniformly over the surface of two divertor chambers, while keeping the surface heat flux less than q_s,max ∼ 5 MW m⁻². For ARC, ACT1 and ACT2, expansion factors on the order of ∼1000 are required. These numbers, combined with values of λ_q ∼ 1 mm from the multi-machine database scaling, determine the minimum size that an advanced divertor chamber must have in a reactor, i.e., the poloidal perimeter of its cross-section must be roughly on the order of 1 m. Thus, the divertor chamber in a reactor will have a cross-sectional size that is actually not much bigger than ADX (figure 13). This fact also tells us that, unlike ADX, advanced divertors in a fusion power plant may not need to occupy a large fraction of the poloidal cross-section of the device. In scaling to a reactor, the cross-section of the core plasma must grow by a factor of ∼5 (ARC) or more (ACT1, ACT2) in linear dimension compared to ADX. But the upstream heat flux width entering into the divertor chamber and the required cross-sectional size of the divertor chamber would remain similar.

For Super-X and X-point target divertors, reactor designs must accommodate the extended major radii of the outer divertor legs. However, if the legs do not extend beyond the outer midplane of the core plasma, the impact on toroidal field magnet design and size is minor [29], particularly in reactor designs that must have the outer toroidal field coil leg extended in major radius to accommodate sector maintenance [80].

Finally, it must be recognized that the physics that underlies the scaling of the heat flux width, λ_q, is not yet fully understood. The empirical parameterization of upstream λ_q was obtained for low-recycling divertor regimes in which the 'heat flux footprint' is imprinted on the divertor target and measured by thermography. Further research might find that upstream λ_q departs from the multi-machine H-mode scaling [14], particularly in dissipative divertor regimes, perhaps exhibiting a positive power law scaling exponent with respect to major radius. This scenario is considered in the last row of data in figure 15 in which λ_q is taken to scale with major radius. In this case, ADX would access divertor q_∥ values that exceed ACT1, ACT2 and all other world tokamaks. In this sense, divertor concepts that prove viable in ADX would be qualified as low-risk options for next-step devices.

Through decades of research on the Alcator devices, the MIT PSFC has extensive experience with high power ICRF and LHRF systems—pushing the envelope of high-power antenna technologies and leading the field in the area of RF wave physics [82]. Ion cyclotron range of frequencies (ICRF) hydrogen minority heating of deuterium plasmas has proven particularly effective, coupling up to 6 MW from three antennas on the LFS of Alcator C-Mod (two two-strap and one four-strap) with surface power densities exceeding 10 MW m⁻². Most recently, an advanced field-aligned 4-strap antenna [81] has demonstrated remarkable performance compared to conventional antennas: low impurity source rate, excellent load variation tolerance and operation at high neutral pressure. ADX will make use of this breakthrough and employ high power field-aligned antennas on the LFS (figure 17). The vacuum vessel will be specially designed to locate coaxial vacuum feedthroughs at optimal positions with respect to field-aligned antenna elements. RF power systems presently being used for C-Mod would be reused for ADX, upgraded to higher power and modified for higher source frequencies (90–120 MHz) in order to take advantage of the high single-pass absorption of minority hydrogen heating scenarios at toroidal fields in the 6.5 to 8 T range. These systems will deliver ∼8 MW to the plasma, producing power exhaust heat flux densities through the LCFS that are prototypical of a reactor, >1 MW m⁻², thus providing the intense auxiliary heating needed to test the advanced divertor concepts on ADX.

Zoom In Zoom Out Reset image size

In addition to accommodating advanced divertors, the ADX vacuum vessel will be specifically designed to support inside-launch LHCD and ICRF systems. A concept for an embedded RF feed system is shown in figure 18. Waveguide windows and RF feedthroughs are placed inside the toroidal field coils at a small major radius, which has the added benefit of locating cyclotron resonances in the pressurized section of the transmission line, reducing the risk of arcing at high power. These attach to radial feeders that pass between toroidal field coils. It is envisioned that this 'RF plumbing' would be a permanent part of the vacuum vessel infrastructure. LHCD launchers and ICRF antennas would be connected to these feed points and replaced and/or modified as needed via manned access inside the vessel.

Zoom In Zoom Out Reset image size

7.6.1. Inside-launch LHCD

The MIT PSFC presently has 4.6 GHz RF sources with a total power of 4 MW that are used to support the Alcator C-Mod research programme. ADX would utilize these systems for HFS launch LHCD. Figure 19 shows a conceptual design for a launch structure placed below the midplane on the HFS. In this design, the launcher is fed by four WR187 waveguides with each carrying the output from a single klystron. Each waveguide would pass vertically along the inner cylinder of the vacuum vessel and split into four waveguide rows for this conceptual design (similar to the 'LH2' launcher on C-Mod [83]), although space is available to include additional rows. Each vertical waveguide also employs an embedded phase shifter (similar to a bi-junction) to produce the toroidal phasing. Each assembly would form a 4 × 8 phased array, optimized to launch LH waves with n_∥ of approximately 1.6, and after waveguide losses, provide 0.5 MW of power to the plasma. A total of four launchers would be employed. The face of the waveguide array would be set slightly behind first-wall protection tiles on the inner wall to prevent damage to the launcher if the LCFS makes contact with the inner wall during startup, rampdown or off-normal events.

Zoom In Zoom Out Reset image size

Results from an initial scoping study using the GENRAY/CQL3D ray tracing Fokker–Planck code [84, 85] are shown in figure 19. Target parameters are taken from an Alcator C-Mod I-mode plasma: 5.6 T, 1.0 MA, central density of 1.8 × 10²⁰ m⁻³ and central temperature of 5.5 keV. Waves are launched with a unidirectional spectrum at n_∥ = 1.6. As anticipated, favourable trajectories with excellent radial penetration are obtained when launched from the HFS below the midplane of the device, producing a broad non-inductive current profile. The poloidal 'steering' technique shown in figure 19 takes advantage of the combined effects of toroidicity and poloidal magnetic field that characterize off midplane launch and that result in a steady increase of parallel wavenumber as the ray propagates to the plasma core. Furthermore, HFS launch takes advantage of the low PMI and good RF coupling properties of the local SOL, as discussed in section 5.2.

The predicted wave propagation in figure 19 is qualitatively different from what was observed in the poloidal 'steering' study of the LFS launcher on Alcator C-Mod, where the launcher poloidal angle was optimized to achieve high single-pass absorption. In order to illustrate the difference, figure 20 shows a comparison of the LH wave propagation characteristics and absorption profiles for the HFS launcher and a LFS launcher with launched n_∥ = 1.6 and 2.5, respectively. The LFS launcher is located at the same height as the launcher in [59]. The radial power deposition profile (and also the driven current density profile—not shown) for HFS launch is distributed through the core region from 0.2 ⩽ r/a ⩽ 0.65 with a peak at r/a ≈ 0.3. In contrast, the LFS case has a power deposition profile with a narrow peak at r/a ≈ 0.6; only small levels of LH power make it into the core region. The ability to couple faster phase velocity waves (i.e. lower n_∥) from the HFS makes it possible to drive current farther into the hot plasma core where high current drive efficiency is possible $(\eta \sim {\rm 1}/n_{\parallel}^{2} )$.

Zoom In Zoom Out Reset image size

The density range over which lower hybrid waves are accessible to ADX plasmas is illustrated in figure 21, comparing HFS and LFS launch cases. High magnetic field strength combined with HFS launch provides excellent LHCD access in ADX. For on-axis magnetic fields greater than 6.5 T, lower hybrid waves launched with n_∥ = 1.6 from the HFS would have access to plasma densities at r/a ∼ 0.75 approaching or exceeding n_e ∼ 2 × 10²⁰ m⁻³—a factor of ∼3 increase in plasma density compared to LFS launch.

Zoom In Zoom Out Reset image size

7.6.2. Inside-launch ICRF:potential elimination of boronization

It is anticipated that the primary plasma heating systems on ADX will be field-aligned ICRF antennas located on the LFS (section 7.5). This robust, proven technology can deliver the power needed for ADX's advanced divertor mission. However, as indicated in section 5.2, there are many compelling reasons to implement HFS launch for all RF systems. In addition to the beneficial RF physics, this configuration would essentially eliminate all impurity sources on the LFS that would otherwise be in close proximity to the plasma (launchers, protection limiters), potentially yielding an order of magnitude reduction in core plasma impurity level for sputtered and non-recycling impurities (see figure 9). Indeed, experiments with boronization on Alcator C-Mod found that the most important surfaces to boronize are the ones located on the LFS [5]. Thus, implementing HFS launch for all RF systems would go a long way towards eliminating the need to apply low-Z coatings for impurity control—a technique that does not scale well to a reactor. For this reason, the ADX vacuum vessel will be designed to accommodate HFS ICRF antennas (figure 18) with the intention of migrating all ICRF power to that location. In principle, four two-strap antenna systems (8 MW delivered) can be implemented on the inside wall, while still accommodating room for four LHCD launchers.

Figure 22 shows a conceptual design for a two-strap, HFS ICRF antenna option in ADX. Each strap would be centre-grounded and fed symmetrically by RF striplines that pass vertically along the inner wall to coaxial vacuum feedthroughs. The antenna straps and Faraday cage could also be tilted off vertical (∼6°) to align these components approximately with the local pitch of magnetic field lines—a field-aligned antenna for the HFS, although this is not depicted in the design drawing in figure 22.

Zoom In Zoom Out Reset image size

TORIC simulations have been used to explore HFS ICRF wave propagation physics in ADX, using representative C-Mod plasmas. As anticipated, incident fast waves mode convert directly to IBW and ICW [86]. This is because the fast-wave branch connects directly to IBW/ICW branches at the ion–ion hybrid layer. In contrast, fast waves launched from the LFS must tunnel past the hydrogen cyclotron resonance layer to get there. The mode converted waves damp directly on electrons via linear electron Landau damping and the relative balance between ion/electron absorption may be controlled via the hydrogen minority fraction, n_H/n_e. For n_H/n_e ∼ 0.15 (case shown in right panel of figure 22) the incident fast-wave power is absorbed nearly 100% via mode conversion. This is a very important result—complete absorption of the incoming fast wave is obtained without formation of energetic ion tails. This means that issues associated with fast ion loss or destabilization of energetic particle populations (including fusion alphas in a reactor) are eliminated and resultant fast-ion induced sputtering of first-wall components is suppressed. High single-pass absorption for the fast wave is also very important for suppressing RF-enhanced sputtering associated with fast waves impinging on first-wall surfaces, converting to slow waves and causing large sheath potentials due to sheath rectification effects. This mechanism was shown to be very important in C-Mod [87] and would be virtually eliminated in ADX with HFS launch. The fast-wave to IBW mode conversion scenario is also attractive as a physics tool because it has been found to produce poloidal flow drive in Alcator C-Mod [88] and TFTR [89]. With 100% of the fast-wave power undergoing mode conversion, HFS ICRF launch may provide an essential tool—perhaps the only tool available—to produce and control flow shear in the core of a reactor at the levels needed to suppress plasma turbulence. Experiments in C-Mod (with LFS launch) have already demonstrated toroidal velocity drive of up to ∼30 km s⁻¹ MW⁻¹, which is comparable to neutral beam injection in JT60-U and JET [90].

7.6.3. Current profile control in ADX

It is well recognized that core plasma energy confinement, as commonly characterized by H₉₈, is a critical parameter that determines the viability of next-step devices. Uncertainty in the H factor at the 10% level can make the difference between success and failure for a burning plasma experiment or a fusion power plant [92]. Moreover, enhanced levels of H₉₈ are often assumed in reactor designs. Although much progress has been made in core transport modelling, first-principles quantitative prediction is well beyond our current capabilities. Prediction of pedestal performance in ELM-suppressed and ELM-less regimes is even more daunting. We have no physics-based models that can make quantitative predictions for such discharges. To minimize the degree of extrapolation for future devices, experiments in the right regime (magnetic field, plasma density, with electron heating and equilibrated electrons and ions, and ELM-suppressed) and with the right heating and current drivers (i.e. with no particle or momentum source) and with the right divertor geometry and materials are crucial.

While general studies of transport continue to be important, research on ADX would focus on the ability to robustly and reliably attain H₉₈ above 1 using only the class of tools and wall materials that would be available in a reactor. The high-field launch LHCD system should allow attainment of low-shear and reversed-shear equilibria [6] that have been shown to lead to improved energy confinement [93]. At the same time, physics-based models would be tested and calibrated to allow application in reactor regimes with minimal extrapolation and high confidence. To properly characterize transport models, comparisons of experimental fluxes, profiles and fluctuations with those derived from simulations must be analysed with modern statistical tools [94].

Without a means to apply significant external torque—which is the case for ADX and the expectation for a fusion reactor—a main control knob for core transport is the current profile. This is predicted from linear gyrokinetic theory as well as nonlinear simulations that show a strong dependence of critical gradients on magnetic shear [93]. ADX would provide an ideal platform on which to study the effects of magnetic shear on the achieved gradients and overall confinement. A key goal would be the development of scenarios that feature enhanced core confinement in non-inductive operation, made possible with the proposed current drive tools.

The origins of intrinsic rotation and its role in setting the transport levels would also be a point of focus. Significant levels of flow and flow shear are commonly present in plasmas without external torques [95] and C-Mod has already operated in regimes where the flow shear from intrinsic rotation enables creation of ITBs [96]. Even with lower levels of rotation, the shear could have a quantitative effect that must be characterized and understood. ICRF mode conversion flow drive, enhanced by HFS launch, is expected to produce significant flow shear in the core plasma of ADX. This can be used for physics studies and for the enhancement of core plasma performance [88]. Most importantly, ICRF mode conversion flow drive may be a means to affect flow profiles in the core of a reactor. Its physics and the prospect for its application to reactor-scale devices is therefore an important area of study in ADX.

With 10 MW of auxiliary power delivered to the plasma, ADX is expected to attain values of β_N(= 〈β〉aB_T/I_p) in the range of ∼1.5%. This is well below the threshold for external kink modes but may be sufficient to trigger neoclassical tearing modes and, without mitigation, could lead to confinement degradation. Although such modes would not impact ADX's divertor mission, their presence would provide an opportunity to demonstrate techniques for control and suppression. ADX has a number of RF actuators for this task: elevate q_min and tailor current profile with application of LHCD, reduce/eliminate sawtooth triggers by application of ICRF mode conversion electron heating and current drive (as demonstrated in C-Mod [97]) and affect flow shear via application of ICRF mode conversion flow drive.

ADX would address the issue of electron–ion coupling and the effect that has on gradients relative to the marginal stability point for each species. Strong electron–ion equilibration can have deeper effects than simply T_i ∼ T_e (though this is an important consequence for micro-stability). For strongly equilibrated plasmas (as reactor-grade plasmas must be) both ion and electron temperatures will be close to marginal stability for the species with the lower critical gradient length. If the critical gradient length profiles for the two species are not identical—and in general they will not be—then the turbulence may be heavily influenced by the species providing the marginal stability. This is unlike the case of heating each species to approximately the same temperature in an uncoupled plasma, where each is independently driven to marginality. Thus the turbulence dynamics may be quite different for the two cases. This may be particularly important for 'other' transport channels—that is, it may be critical for predicting density and rotation profiles which are expected to respond quite differently in plasmas dominated by electron versus ion turbulent transport.

Enhanced core performance in present-day and future devices depends on the attainment of edge transport barriers; these are the 'pedestals' on which core profiles build. However current understanding of pedestal physics—the conditions for barrier formation, the transport and stability physics that regulate them—is far from complete, particularly in regimes relevant for reactors: good core confinement, ELM suppression, acceptable levels of divertor heat load and target erosion, and at the same time, dealing with reactor-level power exhaust densities. In this regard, ADX is an essential pedestal physics research tool for the world research programme. With the implementation of an advanced divertor that can successfully handle plasma exhaust surface power densities well above 1 MW m⁻² with minimal negative impact on the confined plasma, ADX would achieve unprecedented pedestal parameters. Based on extrapolation from C-Mod results [98], and using the EPED-like scaling p_ped ∼ B_TB_POL [39], the ultimate pressure pedestal achievable on ADX could push well past that of any existing machine, even topping expected values for ITER (∼0.1 MPa). H-mode pedestals with normalized power P_SOL/P_LH in the range of 2–3 would be attained, promoting access to enhanced H₉₈ (see figure 3) and low pedestal collisionality (ν* < 1). Access to these unique conditions will test existing pedestal models (e.g. EPED [39]) and new first-principles models as they emerge. A particular focus will be on the critical task of assembling the database required to project pedestal conditions in ELM-suppressed regimes.

At high plasma densities, ADX should trivially access enhanced D-alpha (EDA) H-mode, one of the most well understood variety of stationary ELM-suppressed high-confinement regimes [99]. EDA H-mode would be a good initial platform for developing advanced divertor solutions that can fully mitigate heat fluxes, suppress divertor target erosion, while maintaining a hot pedestal—extending beyond the dissipative divertor studies with ITER-like target geometries pioneered in Alcator C-Mod [18]. Ultimately, operation in stationary low-collisionality regimes must be demonstrated and found compatible with an advanced divertor concept.

A promising ELM-suppressed, high energy confinement regime is the I-mode [8]. I-modes have temperature pedestals without density pedestals, L-mode impurity confinement and energy confinement that exhibits relatively weak degradation with heating power, $W_{{\rm th}} \sim I_{{\rm p}} P_{{\rm heat}}^{0.7}$, which scales very favourably for reactor operation. Recent work on C-Mod and other tokamaks shows that the window for I-mode opens up substantially at high magnetic field [100]. Data from C-Mod (figure 23) delineate the I-mode window in terms of toroidal magnetic field and input power (P_loss), normalized by density and plasma surface area, P_loss/n_eS. These data are for single-null operation with unfavourable grad-B drift direction. The L- to I-mode power threshold (L–I) is independent of magnetic field while the I- to H-mode threshold (I–H) increases strongly with field, ∼B^0.78. The access region for I-mode forms a triangle in this parameter space. These data imply that I-mode could be readily accessed on ADX with only 1.6 MW at line average densities of 10²⁰ m⁻³, while by increasing the density to 2.5 × 10²⁰ m⁻³, it should be maintained up to 12 MW of power. I-mode has been shown experimentally to be compatible with near-DN configurations, although more studies are needed to examine the heat flux sharing in this configuration. ADX will have the flexibility in input power and magnetic topology to explore I-mode in a comprehensive way, examining synergies with advanced divertors and developing physics-based understandings of pedestal transport and stability to guide its development as a reactor regime.

Zoom In Zoom Out Reset image size

The attainment of low collisionality pedestals at high pressure and high power (made possible with advanced divertors) combined with the ability to modify edge current profiles (inside-launch LHCD) would allow ADX to explore regions in pedestal MHD stability space that would otherwise be inaccessible—shifting the balance between pressure-driven versus current-driven instabilities in determining the limits on the pressure pedestal. ADX will use these tools to seek out and evaluate additional regimes with continuous edge relaxation mechanisms, like QH-mode [101], or the recently predicted super-H-mode [102].

Naturally occurring ELM-suppressed regimes are the first choice for reactor operation and they would be most vigorously pursued in ADX. However, one can consider systems for active control of the pedestal in ADX (e.g., ELM suppression, enhanced impurity transport), such as via the application of resonant and non-resonant magnetic perturbations [103] or specially designed antennas to excite edge modes [104, 105]. Although not yet included in the conceptual design outlined above, ADX could readily accommodate such systems. With all RF systems moved to the HFS, there would be ample room for a set of low n/m in-vacuum coils or antenna structures to be placed on the LFS. However, the benefits of these systems, beyond being interesting physics tools, must be weighed against their complexity, risk and scalability to reactors.