A New Optimized Quasihelically SymmetricStellarator
A. Bader, B.J. Faber, J.C. Schmitt, D.T. Anderson, M. Drevlak, J.M. Duff, H. Frerichs, C.C. Hegna, T.G. Kruger, M. Landreman, I.J. McKinney, L. Singh, J.M. Schroeder, P.W. Terry, A.S. Ware
UUnder consideration for publication in J. Plasma Phys. A New Optimized Quasihelically SymmetricStellarator
A. Bader † , B.J. Faber , J.C. Schmitt , D.T. Anderson ,M. Drevlak , J.M. Duff , H. Frerichs , C.C. Hegna , T.G. Kruger ,M. Landreman I.J. McKinney , L. Singh , J.M. Schroeder ,P.W. Terry , A.S. Ware University of Wisconsin-Madison, Wisconsin, USA Auburn University, Alabama, USA IPP-Greifswald, Germany University of Maryland-College Park, Maryland, USA University of Montana, Montana, USA(Received xx; revised xx; accepted xx)
A new optimized quasihelically symmetric configuration is described that has the desir-able properties of improved energetic particle confinement, reduced turbulent transportby 3D shaping, and non-resonant divertor capabilities. The configuration presented in thispaper is explicitly optimized for quasihelical symmetry, energetic particle confinement,neoclassical confinement, and stability near the axis. Post optimization, the configurationwas evaluated for its performance with regard to energetic particle transport, idealmagnetohydrodynamic (MHD) stability at various values of plasma pressure, and iontemperature gradient instability induced turbulent transport. The effect of discrete coilson various confinement figures of merit, including energetic particle confinement, aredetermined by generating single-filament coils for the configuration. Preliminary divertoranalysis shows that coils can be created that do not interfere with expansion of thevessel volume near the regions of outgoing heat flux, thus demonstrating the possibilityof operating a non-resonant divertor.
1. Introduction
This paper discusses results from optimizations to produce quasihelically symmetric(QHS) equilibria that simultaneously demonstrate multiple desirable properties for ad-vanced stellarators. These properties include excellent neoclassical and energetic particleconfinement, a reduction in turbulent transport, and a functional non-resonant divertor(Bader et al. et al. et al. et al. et al. et al. † Email address for correspondence: [email protected] a r X i v : . [ phy s i c s . p l a s m - ph ] A p r A. Bader et al reactor concepts. These include the ARIES-CS project (Ku et al. et al. et al. et al. b ). Advances in theoretical understanding have produced mechanismsfor optimization in the areas of turbulent transport (Mynick et al. et al. et al. et al. et al. a ), and non-resonant divertors (Bader et al. et al. et al.
2. Optimization
Typically, stellarators are optimized by representing the plasma boundary in a two-dimensional Fourier series and perturbing that boundary in an optimization scheme. Theboundaries for stellarator symmetric equilibria are given as, R = (cid:88) m,n R m,n cos ( mθ − nφ ) ; Z = (cid:88) m,n Z m,n sin ( mθ − nφ ) (2.1)Here ( R, Z, φ ) represent a cylindrical coordinate system, θ is a poloidal-like variable,and R m,n , Z m,n are the Fourier coefficients for the m th poloidal and n th toroidal modenumbers. This representation enforces stellarator symmetry, as is commonly used instellarator design.Using the boundary defined in 2.1 and profiles for the plasma pressure and current,the equilibrium can be solved for at all points inside the boundary. Equilibrium solutionsin this paper are calculated using the Variational Moments Equilibrium Code (VMEC)(Hirshman & Whitson 1983). The full equilibrium can then be evaluated for variousproperties of interest and the overall performance of the configuration can thus bedetermined.Optimization of the equilibrium is performed by the ROSE (Rose Optimizes StellaratorEquilibria) code (Drevlak et al. F ( R , Z ) = (cid:88) i (cid:0) f i ( R , Z ) − f target i (cid:1) w i ( f i ) σ i ( f i ) (2.2)Here, R and Z represent the arrays of the R m,n and Z m,n coefficients that define theboundary. The summation is over the different penalty functions, the f i s, chosen bythe user, each of which has a weight, w i , a target, f target i and a function σ i which New Optimized Quasihelically Symmetric Stellarator f i − f target i while a constraint only requires either f i < f target i or f i > f target i .The variation of the boundary coefficients, R and Z is performed through an op-timization algorithm. For the resulting configuration shown here, Brent’s algorithm isused (Brent 2013). Additional details of this optimization technique can be found in(Drevlak et al. et al. b ) and results for similar optimizations ofquasisymmetric stellarators can be found in (Ku et al. et al. • The deviation from quasisymmetry, where the quantity to be minimized is defined asthe energy in the non-symmetric modes normalized to some reference field, here the fieldon axis. In more detail, the configuration is converted into Boozer coordinates (Boozer1981) and the magnitude of the magnetic field strength is represented in a discrete Fourierseries with coefficients B m,n . Then the quasisymmetry deviation for a four field perioddevice is calculated as, P QH = (cid:88) n/m (cid:54) =4 B m,n /B , (2.3)where B , is the field on axis. • The Γ c metric is a proxy for energetic particle confinement (Nemov et al. J (cid:107) with fluxsurfaces. Its viability in producing configurations with excellent energetic ion confinementwas demonstrated in (Bader et al. • A magnetic well (Greene 1997), where present, provides stability against interchangemodes. Because finite β effects in quasihelical symmetry tend to deepen magnetic wells,it suffices to have a vacuum magnetic well of any strength. In this optimization, thevacuum well was required to exist at the magnetic axis, but was not optimized furtherbeyond that. • The rotational transform profile was chosen to avoid low order rational surfaces,and to be sufficiently high above the ι = 1 surface to avoid a resonance when nominalamounts of plasma current are added. • The aspect ratio of the configuration was fixed at the starting aspect ratio of 6.7.No deviations to the aspect ratio were allowed beyond that value. • The neoclassical transport metric, (cid:15) eff (Nemov et al. • Although not explicitly optimized for, it was desired that the configuration haslower turbulent transport for the same profile values relative to that in HSX. Explicitoptimization for turbulent transport is a subject of future work.2.1.
Configuration characteristics
A new optimized configuration, hereby termed the WISTELL-A configuration, hasbeen generated through the optimization procedures described above. Figure 1 shows
A. Bader et al some of the properties of the configuration. Contours of the vacuum magnetic fieldstrength on the boundary are shown in figure 1a. The helical nature of the magnetic fieldstrength, a feature of quasihelical symmetry, is clearly seen. Figure 1b shows toroidalcuts of the boundary surface at toroidal angles 0, π/ and π/ , colloquially referred to asthe "bean", "teardrop" and "triangle" surfaces respectively. Shown are both the vacuumboundary surfaces (red) and the surfaces with normalized plasma pressure β = 0 . .The surfaces with finite pressure are generated by a free boundary solution using coilsand self consistent bootstrap current. The procedure will be described in section 4.3.The finite pressure boundary is somewhat smaller than the vacuum boundary. Figure 1cshows the Boozer spectrum for the vacuum equilibrium where the n = 0 , m = 0 modehas been suppressed. The n = 4 , m = 1 mode is dominant, which is expected for a four-period quasihelically symmetric equilibrium. The largest nonsymmetric modes are themirror mode at n = 4 , m = 0 and the n = 8 , m = 3 mode. Note that the n = 8 , m = 2 mode is also symmetric. Figure 1d shows the rotational transform profile both for vacuumat β = 0 . %. The dashed black lines represent the major low order rational surfacesthat should be avoided, these are the ι = 1.0 and ι = 4/3 surfaces. The configurationpasses through the ι = 8/7 surface at the edge in the vacuum configuration. This vacuumsurface can possibly be used to test island divertor features, which will be discussed insection 4.4. The blue dotted line in figure 1d represents the rotational transform profileat β = 0 . %. As can be seen, the minimum of the rotational transform profile is justabove the ι = 1 surface.In addition, we note some derived features of the configuration. In figure 2, we showthe neoclassical transport, as quantified by the (cid:15) eff metric, the quasisymmetry deviation,as described in equation 2.3, and the Γ c metric for vacuum configurations. In order toprovide a baseline for comparison, we include the same quantities calculated for the HSXequilibrium (black). HSX has better values obtained for the quasisymmetry metric, butslightly worse values for (cid:15) eff and Γ c over the majority of the minor radius. In addition tothe optimized vacuum configuration, we also show the quantities for the vacuum fieldsproduced by filamentary coils. The results here show that the configuration produced withcoils does a very good job at reproducing the important qualities of the equilibrium.
3. Coil construction
Coils to reproduce the vacuum boundary were produced by the FOCUS code (Zhu et al. et al.
New Optimized Quasihelically Symmetric Stellarator Figure 1.
Contours of the magnetic field strength on the boundary are plotted in top left,(a). The top right plot, (b), shows boundary surfaces at toroidal cuts of toroidal angles 0, π/ and π/ for the vacuum configuration (red) and a configuration with 0.94% β (blue dot).The bottom left plot (c) shows the vacuum Boozer spectrum with the strengths of the 8 mostdominant modes as a function of normalized toroidal flux, s . The bottom right plot, (d), showsthe vacuum rotational transform profile (red) and the rotational transform profile at 0.94% β (blue dot). In (d) important rational surfaces are plotted with dashed and dotted black lines.
4. Performance evaluation
The performance of the configuration is evaluated in several topical areas. Theseinclude confinement of energetic particles evaluated by Monte Carlo analysis, turbulenttransport evaluated by non-linear
GENE simulations (Jenko et al. et al. et al.
Energetic particles
Energetic particle optimization was obtained in these configurations by targeting the Γ c metric that seeks to align contours of the second adiabatic invariant, J (cid:107) with flux A. Bader et al Q S d e v i a t i n (a)HSXWISTELL-AWISTELL-A (w. c ils) 0.2 0.4 0.6s0.0000.0010.0020.0030.0040.005 ε e ff (b) 0.2 0.4 0.6s0.0000.0020.0040.0060.0080.0100.012 Γ c (c) Figure 2.
Values of the quasisymmetry deviation (left, a), (cid:15) eff (middle, b) and Γ c (right,c) are plotted as a function of normalized toroidal flux, s for three configurations, the HSXconfiguration (black solid), the WISTELL-A configuration (red solid) and the WISTELL-Aconfiguration as produced by coils (blue dashed). Figure 3.
A representation of coils for the WISTELL-A configuration. Internal to the coils isa representation of the magnetic field strength on the boundary as produced by the coils. surfaces. The calculation for Γ c is given in Nemov et al. (2008, Eq. 61) as Γ c = π √ L s →∞ (cid:32)(cid:90) L s dsB (cid:33) − (cid:90) B max /B min db (cid:48) (cid:88) well j γ cj vτ b,j B min b (cid:48) . (4.1)where the electric field contribution is ignored and the arbitrary reference field B = B min . The quantity γ c is γ c = 2 π arctan (cid:18) v r v θ (cid:19) . (4.2)Here, v r is the bounce averaged radial drift, v θ is the bounce averaged poloidal drift. Theratio v r /v θ is the key quantity to minimize. A method to calculate v r /v θ from geometricalquantities of the magnetic field line is described in Nemov et al. (2008, Eq. 51). Thesummation in eq. 4.1 is taken over all the wells for a suitably long field-line. In ourcase between 60 and 100 toroidal transits were used. The calculation considers trapping New Optimized Quasihelically Symmetric Stellarator Figure 4.
Collisionless alpha particle losses for an ARIES-CS scale device. Alpha particle lossesare plotted as a function of time for three flux surfaces corresponding to normalized toroidalfluxes of 0.2 (left), 0.3 (center) and 0.4 (right). Blue represents the ideal optimized fixed boundaryconfiguration. Orange represents the vacuum field produced by filamentary coils. wells encountered by all possible trapped-particle pitch angles, with b (cid:48) representing anormalized value of the reflecting field. The bounce time for a particle in a specificmagnetic well is given by τ b,j . The parameters B max and B min are the maximum andminimum magnetic field strength on the flux surface.Previously, equilibria in ROSE were optimized by simultaneously minimizing Γ c andthe quasihelical symmetry deviation. This resulted in configurations with very lowcollisionless particle losses (Bader et al. ) and on-axisfield strength (5.7 T). We choose a flux surface and distribute the alpha particles onthe flux surface such that they properly resemble a distribution of alpha particles. Theevaluation is done with the ANTS code (Drevlak et al.
ANTS is well suited to evaluate both theideal equilibrium and the equilibrium produced by the filamentary coils. In both cases,5000 particles are included in the evaluation for each flux surface for 200 ms. Previouscalculations indicated that 5000 particles per flux surface were sufficient for Monte-Carlostatistical purposes (Bader et al. s = 0 . whereas in the ideal case, no particles are lost. Thelosses just outside the mid-radius, at s = 0 . increase from 1.2% in the ideal case to 1.7%in the configuration with coils.The lost particles can be shown as a function of starting pitch angle at s = 0 . in figure5. Here, the x-axis represents the field at which the particle reflects, B ref = E/µ where E and µ are the particle’s energy and magnetic moment. Low values of B ref representdeeply trapped particles, and high values of B ref represent particles near the trapped- A. Bader et al
Figure 5.
Alpha particle losses as a function of pitch angle for the ideal fixed boundaryconfiguration (blue) and the vacuum field from filamentary coils (orange). The dashed blackline represents the trapped passing boundary. passing boundary. The trapped passing boundary is indicated with a vertical dashed line.All passing particles are confined.As is clear from the results in figure 5, most of the particle losses are from parti-cles near the trapped-passing boundary. However, there are a few additional losses ofdeeply trapped particles, and from particles somewhat further from the trapped passingboundary, at
E/µ ≈ . . 4.2. Turbulent Transport
Improving turbulent transport is a key area for stellarator research. Recent resultsfrom W7-X indicate turbulent transport determines the overall energy and particleconfinement (Pablant et al. et al. et al. et al. et al. s = 0 . surface using the Gene code (Jenko et al. a/L
T i assumingadiabatic electrons. In figure 6, the heat flux for WISTELL-A are compared to twoother configurations, the HSX configuration which has been well-analyzed for turbulenttransport (Faber et al. et al. et al. et al. et al. Γ c ." It possessed favorableenergetic particle properties, but did not possess a vacuum magnetic well, and thuswas not considered as a viable configuration. Nevertheless, the turbulent properties ofthis configuration are of interest. The results show that WISTELL-A reduces turbulentheat flux relative to HSX in the low ion temperature scale length regime. However,for a/L T i (cid:62) , the heat flux is comparable to that of HSX. The turbulence-reduced New Optimized Quasihelically Symmetric Stellarator Figure 6.
Turbulent heat flux at the s = 0 . surface as a function of normalized ion temperaturescale length, a/L Ti for three different configurations, HSX (black dashed), WISTELL-A (red)and a turbulence reduced configuration (blue). configuration, on the other hand, demonstrates reduced heat flux across a range of a/L T i and a dramatic reduction in heat flux from HSX and WISTELL-A at a/L
T i (cid:62) .Analysis of the turbulence at a/L T i = 3 for each configuration indicates that thedifferences in the ion heat flux values at s = 0 . are not associated with changes in thelinear ITG instability spectrum. Figure 7 shows the heat flux at a/L T i = 3 for eachconfiguration as a function of k y ρ s . The HSX and WISTELL-A show similar heat fluxspectra, with the flux peaking at k y ρ s ≈ . . The prominent feature in the WISTELL-Aflux spectrum at k y ρ s = 0 . has been previously observed and analyzed in gyrokineticsimulations of HSX (Faber et al. k y ρ s from k y ρ s ≈ . to k y ρ s ≈ . . The linear growth rate spectra as a function of ( k x ρ s , k y ρ s ) for eachconfiguration at a/L T i = 3 is shown in figure 8. Visual inspection of the growth ratespectra indicates there is little difference in the dominant linear instability between eachconfiguration, and in fact, the turbulence-reduced configuration (figure 8c) has largergrowth rates than either HSX or WISTELL-A. This observation is supported moredirectly by using the eigenmode data from Fig. 8 in a quasilinear heat flux calculationusing the model described in Pueschel et al. (2016, Eq. 2), which is reproduced here: Q QL = aL T i C (cid:88) k x ,k y w i ( k x , k y ) γ ( k x , k y ) (cid:68) k i, ⊥ ( k x , k y ) (cid:69) , (cid:10) k ⊥ ( k x , k y ) (cid:11) = (cid:82) d z (cid:112) g ( z ) Φ ( k x , k y , z ) k ⊥ ( k x , k y , z ) (cid:82) d z (cid:112) g ( z ) Φ ( k x , k y , z ) . (4.3)The linear growth rate γ ( k x , k y ) is calculated from a linear Gene simulation at normal-ized perpendicular wavenumber ( k x , k y ) and produces an eigenmode Φ ( k x , k y , z ) , where z is the field-line-following coordinate. The Jacobian along the field line is given by (cid:112) g ( z ) and each contribution to the sum is weighted by w i = ˜ Q i / ˜ n i , where ˜ Q i and ˜ n i arethe calculated linear gyrokinetic heat flux and density perturbations from Gene . Thenormalizing coefficient C is fit to a nonlinear gyrokinetic heat flux calculation and a/L T i isthe normalized temperature gradient; only ratios of Q QL will be considered here to avoidmodel ambiguity. The quasilinear calculation predicts the heat flux for the turbulencereduced configuration should actually be larger than for HSX at a/L T i = 3 by a factor ofapproximately 1.1. This does not agree with the nonlinear gyrokinetic heat fluxes shownin Fig. 6. The discrepancy between the linear growth rates and the full nonlinear heat fluxis consistent with trends found in McKinney et al. (2019). The results presented here also0
A. Bader et al
Figure 7.
Turbulent heat flux spectrum at a/L Ti = 3 from Figure 6 as function of normalizedbinormal wavenumber, k y ρ s for HSX (black dashed), WISTELL-A (red) and a turbulencereduced configuration (blue). Figure 8.
ITG growth rate spectrum with a/L Ti = 3 as a function of normalized radialwavenumber k x ρ s and normalized binormal wavenumber k y ρ s (a) HSX, (b) WISTELL-A, and(c) turbulence-reduced configurations. indicate that linear growth rates can be a misleading indicator for stellarator turbulenceand turbulent transport. Furthermore, these results suggest that the turbulence-reducedconfiguration possesses enhanced turbulence saturation mechanisms.To make a preliminary assessment of the turbulence saturation characteristics, theturbulence saturation theory from Hegna et al. (2018) will be applied. A crucial aspectof the this theory is the supposition that the dominant nonlinear physics involves energytransfer from unstable to damped eigenmodes at comparable wavenumbers. This isaccomplished through a three-wave interaction quantified by a triplet correlation lifetimebetween unstable and stable ITG modes as defined in Hegna et al. (2018, Eq. 104) by τ pst ( k , k (cid:48) ) = − i ω t ( k (cid:48)(cid:48) ) + ω s ( k (cid:48) ) − ω ∗ p ( k ) ; k − k (cid:48) = k (cid:48)(cid:48) , (4.4)where ω ( k ) is the complex linear ITG frequency at normalized wavenumber k = ( k x ρ s , k y ρ s ) . Large values of the triplet lifetimes suggest energy can be veryeffectively transferred out of turbulent-transport-inducing instabilities into dampedeigenmodes that either dissipate energy or transfer it back to the bulk distributionfunction. High values of τ pst correspond to lowered turbulent fluctuation levels and New Optimized Quasihelically Symmetric Stellarator Figure 9.
Triplet correlation lifetimes on a log scale as a function of k for (a) HSX and (b)the turbulence-reduced configuration for a/L Ti = 3 . The value shown at any particular k isthe calculation of τ pst as defined by 4.4 where p is an unstable mode at that k , s is a stablemode at whatever wavenumber k (cid:48) such that Re ( τ pst ( k , k (cid:48) )) is maximized by a third mode t ,which can be unstable or stable. The triplet correlation lifetime value has been weighted bya fluctuation energy spectrum obtained from gyrokinetic calculations which emphasize tripletlifetimes involving energy containing scales. correspondingly reduced turbulent transport. In figure 9, the triplet correlation lifetimesare shown for the HSX and the turbulence optimized configuration. Importantly, theturbulence-reduced configuration shows larger triplet correlation lifetimes in the region k y ρ s (cid:46) . compared to HSX, where the larger correlation lifetimes are observed at higher k y ρ s . This is an important difference, as instabilities at larger scales (smaller | k ⊥ | ) canmore easily contribute to turbulent transport. Thus, larger triplet correlation lifetimesat smaller k y ρ s where at least one unstable and one stable mode are involved suggestsenergy is being transferred more efficiently from the modes driving the fluctuationspectrum to dissipation and thus lowering the contribution to turbulent transportat that k y ρ s . This may be contributing to both the decrease in overall transport infigure 6 and the downshift in heat flux spectrum in figure 7 between HSX and theturbulence-reduced configuration. The connections between nonlinear turbulent heatflux, quasilinear turbulent heat flux, and the triplet correlation lifetimes are summarizedin table 1. The triplet correlation lifetimes for each configuration are quantified bycomputing a spectral average defined as (cid:104) τ NZ (cid:105) = (cid:88) k x ,k y S G ( k x , k y ) Re ( τ NZ ( k x , k y )) . (4.5)The weighting factor S ( k x , k y ) is a turbulent fluctuation spectrum computed froma characteristic nonlinear gyrokinetic simulation that preferentially weights low | k ⊥ | contributions to provide consistency with the nonlinear simulations. The decrease innonlinear flux between HSX and the turbulence reduced configuration by a factor of 2correlates with an increase in triplet correlation lifetimes by more than a factor of two,while the increase in nonlinear flux between WISTELL-A and HSX correlates with adecrease in triplet correlation lifetimes. This will be explored in more detail in futurework, but this result already indicates that quasihelically symmetric configurations withlower ITG-driven transport can be obtained.2 A. Bader et al
Config. Q NL /Q NL,HSX Q QL /Q QL,HSX (cid:104) τ NZ (cid:105) / (cid:104) τ NZ,HSX (cid:105)
HSX 1 1 1WISTELL-A 1.1 1.05 0.76Turb. red. 0.5 1.1 2.85
Table 1.
Values of nonlinear gyrokinetic heat flux, quasi-linear heat flux, and spectral averagedtriplet correlation lifetimes at a/L Ti = 3 for the three configurations. All values have beennormalized to the corresponding HSX value. MHD Stability
The MHD properties of QH stellarators (Nührenberg & Zille 1988) are somewhatdistinct from other classes of optimized stellarators. The relatively reduced connectionlength (the distance along the field line between B max and B min ) implies that QH stel-larators have reduced banana widths, reduced orbit drifts for passing particles (Talmadge et al. et al. et al. ι eff = ( ι − N ) where N is the periodicity of the stellarator. Moreover,the bootstrap current in a QH stellarator is in the opposite direction relative to whatoccurs in a tokamak. This has the consequence of reducing the value of ι with risingplasma pressure and producing negative dι/ds in the core region. Negative values of dι/ds can have beneficial effects for both ideal ballooning (Hegna & Hudson 2001) andmagnetic island physics (Hegna & Callen 1994).The only stability quantity constrained in the ROSE optimization is the magnetic welldepth as described by d V /dΦ , the second derivative of volume with respect to toroidalflux, at the magnetic axis for the vacuum equilibrium. A magnetic well ( V (cid:48)(cid:48) < ) isnecessary for stability against interchange modes. The magnetic well depth at pointsaway from the magnetic axis was not explicitly optimized for, but can be quantified with W = (cid:32) dVdΦ (cid:12)(cid:12)(cid:12)(cid:12) ρ =0 − dVdΦ (cid:33) / dVdΦ (cid:12)(cid:12)(cid:12)(cid:12) ρ =0 (4.6)For calculations including finite pressure, a pressure profile is assumed where tem-perature is linear in normalized flux, T = T (1 − s ) and the density profile is broad, n = n (cid:0) − s (cid:1) . Here, T and n represent the temperature and density at the magneticaxis respectively. The pressure profiles as a function of normalized flux are given in figure10a. The pressure was varied by varying T at fixed n = 0.9 × m − , with T rangingfrom 1.3 keV to 3.5 keV. The free-boundary equilibrium was calculated with VMEC usingthe vacuum magnetic field given by the filamentary coils described in section 3. The self-consistent bootstrap current profiles are calculated using SFINCS (Landreman et al. T e = T i was assumed, and the bootstrap current was calculatedat the ambipolar radial electric field. The rotational transform profiles from the VMECequilibria evaluated for several different pressures are shown in figure 10b.As noted previously, the bootstrap current tends to lower the value of ι and producereversed magnetic shear in the core. As seen in figure 10b, the rotational transform profilecrosses ι = 1 around s ≈ . when the normalized pressure, β ≈ . Unless compensatedfor, this potentially sets an operational limit for this configuration. New Optimized Quasihelically Symmetric Stellarator Figure 10. (a) The pressure profiles versus normalized flux, s , for several values ofvolume-average β . The effects of finite- β are shown for the radial profiles (in s ) of the (b)rotational transform ( ι ≡ ι/ π ), (c) the magnetic well depth and (d) the Mercier stabilitycriterion. Positive values indicate Mercier stability. (inset shows more detail for . < s < . .)Vacuum quantites for the rotational transform and well depth are shown (dashed lines). With finite beta equilibria, relevant stability metrics can be calculated. The well depths,as given by equation 4.6 are shown in figure 10c. As seen in the figure, the vacuumconfiguration has a magnetic well, and the well depth gets larger as the pressure increases.However, a magnetic hill region remains near the plasma edge.The Mercier criterion is given by the sum (Bauer et al. et al. D Merc = D S + D W + D I + D G (cid:62) (4.7)where the individual terms in eq. 4.7 represent contributions (stabilizing or destabilizing)from the shear, magnetic well, current and geodesic curvature and are given by the4 A. Bader et al
Figure 11. (a) The radial profiles of the of the growth rates, as calculated by COBRAVMECfor various values of normalized pressure, β . (b) The value of the growth rate for the mostunstable ballooning mode is shown as a function of normalized pressure, β . The configurationat β =1.16% is stable to ballooning modes. following: D S = sι π ( Ψ (cid:48)(cid:48) Φ (cid:48) ) D W = sι π (cid:90) (cid:90) gdθdζ B g ss dpds × (cid:18) V (cid:48)(cid:48) − dpds (cid:90) (cid:90) g dθdζB (cid:19) D I = sι π (cid:90) (cid:90) gdθdζ B g ss Ψ (cid:48)(cid:48) I (cid:48) − ( Ψ (cid:48)(cid:48) Φ (cid:48) ) (cid:90) (cid:90) gdθdζ (cid:16) (cid:126)J · (cid:126)B (cid:17) g ss D G = sι π (cid:90) (cid:90) gdθdζ (cid:16) (cid:126)J · (cid:126)B (cid:17) g ss − (cid:90) (cid:90) gdθdζ (cid:16) (cid:126)J · (cid:126)B (cid:17) g ss B (cid:18)(cid:90) (cid:90) gdθdζ B g ss (cid:19) In the above expressions, Φ and Ψ are the toroidal and poloidal magnetic fluxes, g isthe Jacobian, p is the pressure, I is the net toroidal current enclosed within a magneticsurface, and the metric element g ss = |∇ s | .Figure 10d shows the Mercier stability criterion as given by eq. 4.7 and evaluated byVMEC. All configurations are Mercier stable ( D Merc > ) for s (cid:46) . . Calculationsof ballooning stability were obtained with COBRAVMEC (Sanchez et al. β (cid:54) . . Ballooning stability is violated at higher values of β with the specified pressureprofile shape. The region where ballooning instability tends to occur first is near s ≈ . .Higher critical β values for ideal ballooning instability can be obtained by tuning thepressure profile. This will be pursued in future work.While the configuration presented here has only modest MHD stability properties, itis anticipated these properties can be improved through further optimization. However,there is little evidence to support the notion that MHD stability provides any rigorous New Optimized Quasihelically Symmetric Stellarator et al. β .4.4. Divertor
Divertors for quasisymmetric stellarators require either resilience to changes in theplasma current and pressure profiles, or active control mechanisms to ensure properfunction of a resonant divertor, often referred to as an island divertor, from startup tothe operational point (König et al. et al. et al. et al.
FOCUS code is that it allows the coils to expandaway from the plasma in regions where they are not required to be close to the plasma.Fortuitously, the regions of high curvature where strike lines exit are also regions wherecoil expansion is possible. Therefore, a method for divertor construction is to adjust theuniform wall so it is expanded in these regions. This allows for both longer connectionlengths between the plasma and the wall, and some degree of divertor closure, allowingfor access to high neutral pressure.An initial attempt at such a construction is shown in figure 12. Here four plots of thedivertor structure at toroidal values of φ = 0 ◦ , 15 ◦ , 30 ◦ and 45 ◦ are shown. In addition, anEMC3-EIRENE simulation was carried out with nominal operating parameters for theupgraded scenario. The calculation from EMC3-EIRENE indicates that the heat flux isconcentrated in specific areas toroidally and poloidally near φ = 30 ◦ . A three dimensionalrepresentation of the divertor design is shown in figure 13Future iterations of the divertor structure design are necessary in order to smooth theheat flux deposition. The results presented here are therefore meant to indicate a firstattempt at how divertor design could proceed and not indicative of a final design.In addition, the configuration will have access to island divertor experiments byexploiting the n = 8 , m = 7 resonance. However, due to the presence of self-generatedplasma currents in quasisymmetric equilibria, ensuring the island position is maintainedthroughout the discharge to the operating point requires some external control, whetherby auxiliary coils or current drive. Designing such operational scenarios are beyond thescope of this current work.
5. Conclusion
A new optimized quasihelically symmetric stellarator is developed that has a numberof desirable features including improved energetic ion confinement, low neoclassicaltransport, reduced turbulent transport and non-resonant divertor capability. This con-figuration is made possible through the simultaneous improvements in optimization6
A. Bader et al
Figure 12.
A representation of a non-resonant divertor concept for the WISTELL-A device.Plots correspond to four toroidal positions at φ = 0 ◦ , 15 ◦ , 30 ◦ and 45 ◦ . The wall (solid blue)is expanded near regions of peak heat flux. The heat flux is calculated by EMC3-EIRENE andshown in color. Figure 13.
A three dimensional representation of a non-resonant divertor concept for theWISTELL-A device. The gray outer surface represents the wall. Temperature contours fromEMC3-EIRENE are presented at φ = ◦ , 22.5 ◦ and 45 ◦ , and the heat flux on the wall is alsorepresented on the boundary. (ROSE) and coil (FOCUS) tools as well as advances in physics understanding. Stellaratoroptimization is a rapidly developing field with new advancements in physics metrics,equilibrium solutions and optimization algorithms occurring at an impressive pace.Individual optimized equilibria represent markers of where progress is at a given pointin time. They highlight recent advancements and provide benchmarks for testing furtheroptimizations.A particularly significant advance is the development of configurations with excellentenergetic particle confinement. Importantly, this improved energetic ion confinement is New Optimized Quasihelically Symmetric Stellarator β .There is reason to believe this property can be improved.There is a need for a viable divertor option for stellarators with finite bootstrap current.Initial calculations show that non-resonant divertors, which do not rely on a low-orderresonance at the edge, may be a possible solution. The complex interaction betweenedge plasma, impurities, neutral gas, and plasma boundary surfaces in a stochastic edgeare only accessible numerically with the EMC3-EIRENE code. Validation of this codein different edge scenarios in experiments with relevant geometries and conditions isnecessary to predict the functionality of next step experiments.A final advantage of new equilibria is that they can be the foundation for new experi-mental designs. In the appendix of this paper, a conception of a midscale quasihelically-symmetric stellarator is given.
6. Acknowledgments
This work was supported by the University of Wisconsin, UW2020-135AAD3116 andthe US Department of energy grants DE-FG02-93ER54222 and DE-FG02-99ER54546.BJF is supported by the U.S. Department of Energy Fusion Energy Sciences PostdoctoralResearch Program administered by the Oak Ridge Institute for Science and Education(ORISE) for the DOE. ORISE is managed by Oak Ridge Associated Universities (ORAU)under DOE contract number de-sc0014664. All opinions expressed in this paper are theauthor’s and do not necessarily reflect the policies and views of DOE, ORAU, or ORISE.This research used resources of the National Energy Research Scientific ComputingCenter (NERSC), a U.S. Department of Energy Office of Science User Facility operatedunder Contract No. DE-AC02-05CH11231.
Appendix A.
This appendix details a possible realization of the new configuration as a midscaleexperiment. Such an experiment could significantly advance the quasisymmetric conceptand retire some of the risks related to energetic particle transport, turbulent transportand divertor operation in quasisymmetry.A 0-Dimensional analysis was carried out in order to determine a target equilibriumfor a realization of the WISTELL-A configuration for a midscale experiment. Details are8
A. Bader et al
Param. Initial Upg.( H =1) Upg.( H =1.5) Upg.( H =2) R (m) 2.0 a (m) 0.3 V (m ) 3.55 ι B (T) 1.25 2.5 2.5 2.5ECH (MW) 0.5 1.0 1.0 1.0NBI (MW) 0.0 1.0 1.0 1.0 H factor 1.5 1.0 1.5 2.0 n (10 m − ) 0.15 0.9 0.9 0.9 T e (keV) 1.2 0.8 1.3 1.7 T i (keV) 0.7 0.8 1.3 1.7 β % 0.36 0.49 0.73 0.98 ν ∗ i τ E (ms) 48 65 98 130 τ ie (ms) 35 5 9 14 Table 2.
Parameters and derived quantities (using 0-D analysis) for the WISTELL-Astellarator for the initial operational phase and 3 scenarios for the full operational phase presented in table 2. The major size and cost drivers are the magnetic field strengthand the minor radius. In addition the operation is split into two phases, an initialoperational phase at half field (1.25 T), and a full operational phase at 2.5 T. Themagnetic field strength is chosen to take advantage of Electron Cyclotron Heating (ECH)from commercially available 70 GHz gyrotrons at the second harmonic in the initial phase.In the full operational phase, the 70 GHz gyrotrons can be used at the fundamental o-mode harmonic, or 140 GHz gyrotrons can be used to heat at the second harmonic. Thechoice of gyrotron frequency sets the density cut-off, the maximum operational densityfor ECH plasmas. The cutoff density is (cid:15) m e ω /e , where (cid:15) is the permittivity of freespace, m e is the electron mass, e is the fundamental charge and ω is the angular frequencyof the launched wave. For 70 GHz gyrotrons the cutoff is ∼ × m − . For 140 GHzgyrotrons the cutoff is ∼ × m − . The Sudo density limit for 1 MW absorbedpower is 9.0 × m − (Sudo et al. τ E (in ms) is given by the ISS04 empirical scaling law (Yamada et al. τ E = τ ISS04 E = 134 a . R . P − . n . e B . ι . / . (A 1)Here, R and a are the major and minor radii respectively in meters, P is the totalabsorbed power in MW, n e is the electron density in units of 10 m − , B is the magneticfield on axis in T and ι / is the rotational transform value at r/a = 2/3. While theconvention for the ISS04 scaling is to use density in units of 10 m − , from this pointforward, all calculations will use the convention of density in units of 10 m − . Theaverage temperature, T = ( T e + T i ) / in eV is given by, T = τ e ( P i + P e )3 nV (A 2)where V is the plasma volume, P i is the power absorbed by ions (taken here as power New Optimized Quasihelically Symmetric Stellarator P e is the power transmitted to the electrons. (taken here asECH power). In the table a confinement improvement factor H is included anticipatingpotential advances in understanding how to reduce turbulent transport.The electron-ion energy equilibration time, τ ie is τ ie = m i m e τ e ; τ e = 3 (2 π ) / (cid:15) m / e T / nZ e ln Λ (A 3)Here, (cid:15) is the permittivity of free space, T is the mean temperature in joules, n is thedensity, Z is the particle charge (taken to be 1), e is the fundamental electric charge, m e and m i are the electron and ion masses respectively, and ln Λ is the coulomb logarithm,taken to be 17. For the values calculated in table 2, we assume main species hydrogenwith n = n e = n i .The energy partition between ions and electrons is calculated assuming, P e nV = (cid:20) τ E + 1 τ ie (cid:21) T e − T i τ ie ; 23 P i nV = (cid:20) τ E + 1 τ ie (cid:21) T i − T e τ ie (A 4)Where, ion and electron temperature, T e and T i respectively are in joules.The normalized ion collisionality, ν ∗ i is given by, ν ∗ i = 1 τ ii (cid:114) m i T i R(cid:15) / ( N − ι ) ; τ ii = 12 π / √ m / i T / i (cid:15) nZ e ln Λ (A 5)where, (cid:15) is the inverse aspect ratio evaluated at the mid-radius, ( a/ /R , the iontemperature, T i is evaluated in joules, and the usual tokamak safety factor, q has beenreplaced by the stellarator equivalent for a QHS stellarator, / ( N − ι ) with N the numberof field periods. The 0-D analysis indicates that for relatively modest amounts of externalheating, small ν ∗ i regimes can be realized.The normalized pressure, β is β = nTB / (2 µ ) (A 6)with µ the permeability of free space.The midscale design employs water cooled copper coils, and thus pulse length isexpected to be limited by coil heating.The parameters for the coil quality of fit and relevant engineering parameters for themidscale realization are given in table 3. These data include a preliminary analysis ofthe coils including a finite build made up of multiple filaments and a winding pack sizecommensurate with realizable current densities. The error f B is given by f B ≡ (cid:34) (cid:90) S (cid:18) B · n | B | (cid:19) ds (cid:35) (cid:18)(cid:90) S ds (cid:19) − , (A 7)where S represents the boundary surface, and B · n represents the normal field on theboundary. More information on the procedure for generating these coils can be found in(Singh et al. REFERENCESAnderson, F Simon B, Almagri, Abdulgader F, Anderson, David T, Matthews,Peter G, Talmadge, Joseph N & Shohet, J Leon
Fusion Technology (3T), 273–277. A. Bader et al
Param. Single-filament Multi-filament
Minimum Coil-Plasma Distance 19.5 cm 13.0 cmAverage Coil-Plasma Distance 22.5 cm 16.0 cmIntegrated Normal Field f B . − . − Table 3.
Parameters for the coil set produced for the WISTELL-A configuration using thedimensions in table 2
Bader, A, Boozer, AH, Hegna, CC, Lazerson, SA & Schmitt, JC
Physics of Plasmas (3), 032506. Bader, Aaron, Drevlak, M, Anderson, DT, Faber, BJ, Hegna, CC, Likin, KM,Schmitt, JC & Talmadge, JN
Journal of Plasma Physics (5). Bader, Aaron, Hegna, CC, Cianciosa, M & Hartwell, GJ
Plasma Physicsand Controlled Fusion (5), 054003. Bauer, F., Betancourt, 0. & Garabedian, P.A.
Magnetohydrodynamic Equilibriumand Stability of Stellarators . Springer-Verlag, New York.
Beidler, Craig, Grieger, Günter, Herrnegger, Franz, Harmeyer, Ewald,Kisslinger, Johann, Lotz, Wolf, Maassberg, Henning, Merkel, Peter,Nührenberg, Jürgen, Rau, Fritz & others
Fusion Technology (1), 148–168. Beidler, CD, Harmeyer, E, Herrnegger, F, Igitkhanov, Yu, Kendl, A, Kisslinger,J, Kolesnichenko, Ya I, Lutsenko, VV, Nührenberg, C, Sidorenko, I & others
Nuclear Fusion (12), 1759. Boozer, A H
Physics of Fluids (11), 1999. Boozer, A H & Gardner, H. J.
Physics of FluidsB-Plasma Physics (10), 2408. Boozer, Allen H & Punjabi, Alkesh
Physics ofPlasmas (9), 092505. Brent, Richard P
Algorithms for minimization without derivatives . Courier Corporation.
Canik, J. M., Anderson, D. T., Anderson, F. S. B., Likin, K. M., Talmadge, J. N.& Zhai, K.
Phys. Rev. Lett. , 085002. Carreras, B.A., Dominguez, N., Garcia, L., Lynch, V.E., Lyon, J.F., Cary, J.R.,Hanson, J.D. & Navarro, A.P.
Nuclear Fusion , 1195–1207. Drevlak, M, Beidler, CD, Geiger, J, Helander, P & Turkin, Y
Nuclear Fusion (1), 016010. Drevlak, M, Geiger, J, Helander, P & Turkin, Yu
Nuclear Fusion (7), 073002. Faber, B. J., Pueschel, M. J., Terry, P. W., Hegna, C. C. & Roman, J. E.
Journal of PlasmaPhysics (5), 905840503. Faber, B. J.
Faber, B. J., Pueschel, M. J., Proll, J. H. E., Xanthopoulos, P., Terry, P. W.,Hegna, C. C., Weir, G. M., Likin, K. M. & Talmadge, J. N.
Physics of Plasmas , 072305. Feng, Y, Sardei, F, Kisslinger, J, Grigull, P, McCormick, K & Reiter, D
New Optimized Quasihelically Symmetric Stellarator edge modeling and island divertor physics. Contributions to Plasma Physics (1-3),57–69. Greene, John M
Comments on Plasma Physics andControlled Fusion , 389–402. Grieger, G, Lotz, W, Merkel, P, Nührenberg, J, Sapper, J, Strumberger, E, Wobig,H, Burhenn, R, Erckmann, V, Gasparino, U & others
Physics of Fluids B: Plasma Physics (7), 2081–2091. Hegna, C.C. & Callen, J. D.
Phys. Plasmas (9), 3135. Hegna, C.C. & Hudson, S. R.
Phys. Rev. Lett. (3), 035001. Hegna, C.C., Terry, P.W. & Faber, B.J.
Physics of Plasmas (2), 022511. Henneberg, SA, Drevlak, M & Helander, P a Improving fast-particle confinementin quasi-axisymmetric stellarator optimization.
Plasma Physics and Controlled Fusion (1), 014023. Henneberg, SA, Drevlak, M, Nührenberg, C, Beidler, CD, Turkin, Y, Loizu, J& Helander, P b Properties of a new quasi-axisymmetric configuration.
NuclearFusion (2), 026014. Hirshman, Steven P & Whitson, JC
The Physics of fluids (12), 3553–3568. Jenko, F., Dorland, W., Kotschenreuther, M. & Rogers, B. N.
Physics of Plasmas (5), 1904–1910. König, R, Grigull, P, McCormick, K, Feng, Y, Kisslinger, J, Komori, A, Masuzaki,Suguru, Matsuoka, K, Obiki, T, Ohyabu, N & others
Plasma physics and controlled fusion (11), 2365. Ku, LP, Garabedian, PR, Lyon, J, Turnbull, A, Grossman, A, Mau, TK,Zarnstorff, M & Team, ARIES
Fusion Science andTechnology (3), 673–693. Landreman, Matt
Nuclear Fusion (4), 046003. Landreman, Matt & Paul, Elizabeth
Nuclear Fusion (7), 076023. Landreman, Matt & Sengupta, Wrick
Journal of Plasma Physics (6). Landreman, Matt, Sengupta, Wrick & Plunk, Gabriel G
Journal ofPlasma Physics (1). Landreman, M., Smith, H.M., Mollén, A. & Helander, P.
Physics of Plasmas , 042503. McKinney, IJ, Pueschel, MJ, Faber, BJ, Hegna, CC, Talmadge, JN, Anderson, DT,Mynick, HE & Xanthopoulos, P
Journal of Plasma Physics (5). Miller, R, Team, SPPS & others
University ofCalifornia San Diego Report UCSD-ENG-004 . Mynick, HE, Pomphrey, N & Xanthopoulos, P
Physical review letters (9), 095004.
Nemov, VV, Kasilov, SV, Kernbichler, W & Heyn, MF ν neoclassical transport in stellarators. Physics of plasmas (12), 4622–4632. Nemov, VV, Kasilov, SV, Kernbichler, Winfried & Leitold, GO
Physics of plasmas (5), 052501. Nührenberg, J. & Zille, R.
PhysicsLetters A (2), 113–117.
Pablant, N. & others
Nuclear Fusion (3), 036021. A. Bader et al
Plunk, Gabriel G, Landreman, Matt & Helander, Per arXiv preprintarXiv:1909.08919 . Plunk, G. G., Xanthopoulos, P. & Helander, P.
Phys. Rev. Lett. , 105002.
Pueschel, M. J., Faber, B. J., Citrin, J., Hegna, C. C., Terry, P. W. & Hatch,D. R.
Phys. Rev. Lett. , 085001.
Sanchez, R., Hirshman, S. P. & Ware, A. S.
J. Comp. Phys. (2), 576.
Schmitt, J.C., Talmadge, J.N. & Anderson, D.T.
Nuclear Fusion , 0820001. Schmitt, J.C., Talmadge, J.N., Anderson, D.T. & Hanson, J.D.
Physics of Plasmas , 092518. Shimizu, Akihiro, Liu, Haifeng, Isobe, Mitsutaka, Okamura, Shoichi, Nishimura,Shin, Suzuki, Chihiro, Xu, Yuhong, Zhang, Xin, Liu, Bing, Huang, Jie & others
Plasmaand Fusion Research , 3403123–3403123. Singh, L., Kruger, T.G., Bader, A., Zhu, C., Hudson, S.R. & Anderston, D.T.
Submitted to Journal of Plasma Physics . Strumberger, Erika
Nuclear fusion (5), 737. Sudo, S, Takeiri, Y, Zushi, H, Sano, F, Itoh, K, Kondo, K & Iiyoshi, A
Nuclear Fusion (1), 11. Talmadge, J.N., Sakaguchi, V., Anderson, F. S. B., Anderson, D. T. & Almagri,A. F.
Physics of Plasmas , 5165. Weller, A., Anton, M., Geiger, J., Hirsch, R., Jaenicke, R., Werner, A.,Team, W7-AS, Nührenberg, J., Sallander, E. & Spong, D. A.
Phys.Plasmas (3), 931. Xanthopoulos, P, Mynick, HE, Helander, P, Turkin, Yu, Plunk, GG, Jenko, F,Görler, T, Told, D, Bird, T & Proll, JHE
Physical review letters (15), 155001.
Yamada, H, Harris, JH, Dinklage, Andreas, Ascasibar, E, Sano, F, Okamura, S,Talmadge, J, Stroth, U, Kus, A, Murakami, S & others
Nuclear Fusion (12), 1684. Zarnstorff, MC, Berry, LA, Brooks, A, Fredrickson, E, Fu, GY, Hirshman, S,Hudson, S, Ku, LP, Lazarus, E, Mikkelsen, D & others
Plasma Physics and Controlled Fusion (12A),A237. Zhu, Caoxiang, Hudson, Stuart R, Song, Yuntao & Wan, Yuanxi
Plasma Physics and ControlledFusion60