Você está na página 1de 13

Home Search Collections Journals About Contact us My IOPscience

Impact of resonant magnetic perturbations on zonal modes, drift-wave turbulence and the L–H

transition threshold

This content has been downloaded from IOPscience. Please scroll down to see the full text.

2014 Nucl. Fusion 54 013004

(http://iopscience.iop.org/0029-5515/54/1/013004)

View the table of contents for this issue, or go to the journal homepage for more

Download details:
This content was downloaded by: mleconte0
IP Address: 203.230.125.100
This content was downloaded on 29/11/2013 at 02:24

Please note that terms and conditions apply.


| International Atomic Energy Agency Nuclear Fusion
Nucl. Fusion 54 (2014) 013004 (12pp) doi:10.1088/0029-5515/54/1/013004

Impact of resonant magnetic perturbations


on zonal modes, drift-wave turbulence and
the L–H transition threshold
M. Leconte1 , P.H. Diamond1,2 and Y. Xu3,a
1
WCI Center for Fusion Theory, NFRI, Daejeon, Korea
2
CMTFO and CASS, UCSD, CA 92093, USA
3
Association Euratom-Belgian State, ERM-KMS, Brussels, Belgium
E-mail: mleconte@nfri.re.kr

Received 5 February 2013, accepted for publication 8 November 2013


Published 28 November 2013

Abstract
We study the effects of resonant magnetic perturbations (RMPs) on turbulence, flows and confinement in the framework of
resistive drift-wave turbulence. This work was motivated, in parts, by experiments reported at the IAEA 2010 conference (Xu
et al 2011 Nucl. Fusion 51 062030) which showed a decrease of long-range correlations during the application of RMPs. We
derive and apply a zero-dimensional predator–prey model coupling the drift-wave–zonal-mode system (Leconte and Diamond
2012 Phys. Plasmas 19 055903) to the evolution of mean quantities. This model has both density-gradient drive and RMP
amplitude as control parameters and predicts a novel type of transport bifurcation in the presence of RMPs. This model allows
a description of the full L–H transition evolution with RMPs, including the mean sheared flow evolution. The key results are
the following: (i) the L–I and I–H power thresholds both increase with RMP amplitude |b̃x |, the relative increase of the L–I
threshold scales as PLI ∝ |b̃x |2 ν∗−2 ρs−2 , where ν∗ is edge collisionality and ρs is the sound gyroradius. (ii) RMPs are predicted
to decrease the hysteresis between the forward and back-transition. (iii) Taking into account the mean density evolution, the
density profile—sustained by the particle source—has an increased turbulent diffusion compared with the reference case without
RMPs which provides one possible explanation for the density pump-out effect.

Keywords: drift-wave turbulence, zonal flows, resonant magnetic perturbations


(Some figures may appear in colour only in the online journal)

1. Introduction that drift-wave (DW) turbulence levels and associated ZFs are
modified by RMPs. We used a basic representative model,
The H-mode—which usually exhibits edge localized mode i.e. an extension of the Hasegawa–Wakatani system to include
(ELM) bursts—is the target regime of operation for next RMP effects, neglecting the plasma response. We applied
step fusion devices e.g. ITER [1]. The ELMs pose a a modulational stability analysis to the latter system and
major threat to the plasma-facing components and should coupled it to the evolution equation for turbulence energy. The
be controlled. A means to control ELMs relies on resonant resulting coupled drift-wave–zonal mode (DW–ZM) predator–
magnetic perturbations (RMPs) [2–5]. Since there is evidence prey model exhibits modified ZFs. The underlying physical
of residual turbulence in H-mode, and the turbulence can affect mechanisms of this modification of ZFs are the following.
the profiles and thus stability, there is no reason—a priori—to (i) RMPs tilt the magnetic field lines, thereby inducing a
neglect turbulence effects in the ELM control experiments. In quasilinear δ j × δ b torque. (ii) The projection of the
other words, turbulence persists in H-mode. Moreover, RMPs RMP-induced quasilinear torque yields a quasilinear stress
should be activated before the 1st ELM, as the latter can cause which can compete against the Reynolds stress drive of
large damage to the divertor and wall. Hence, understanding ZFs and hence damp them. We identified two asymptotic
RMP effect on LH power threshold is crucial for ITER [6–8]. regimes, analogous to the—‘adiabatic electron’ and ‘fluid
The RMP effect on turbulence is, to our knowledge, not broadly electron’—regimes of the linear Hasegawa–Wakatani model,
considered. We emphasize that experiments clearly show that depending on the strength of the RMP coupling parameter,
RMPs have an effect on turbulence [9]. Reference [9] showed proportional to the square of RMP amplitude (table 1). In
experimentally that RMPs damp geodesic acoustic mode the regime where electron force balance is satisfied, the
(GAM) zonal flows (ZFs), as revealed by the decrease of long- extended Hasegawa–Wakatani system reduces to a single
range-correlations during RMP. We have shown in [10, 11], equation analogous to the linear Hasegawa–Mima equation.
a On assignment at Plasmaphysik (IEF-4) Forschungszentrum Julich, The DW–ZM model has two possible states, thus a bifurcation
Germany. is possible for a critical value of the linear drive, i.e.

0029-5515/14/013004+12$33.00 1 © 2014 IAEA, Vienna Printed in the UK


Nucl. Fusion 54 (2014) 013004 M. Leconte et al

Table 1. Qualitative impact of RMPs on ZFs, in two asymptotic Table 2. Group of different 0D predator–prey models. Here, ‘KD’
regimes. stands for ‘Kim and Diamond’, LD stands for ‘Leconte and
Diamond’ and ‘MSF’ stands for ‘mean sheared flow’.
RMP coupling
parameter Effect Model Components RMP Ref.
 2
 B̃  D Diamond et al DWs, ZFs No [15]
 x  KD 2003 DWs, ZFs, No [14]
Weak-RMP   2 1 Weakly perturbed ZFs
 B  ρs µ MSF, pi gradient
 2 LD 2012 DWs, ZFs, Yes [11]
 B̃  D
 x  zonal density
Strong-RMP   2 1 Electron force balance
 B  ρs µ This article DWs, ZFs, Yes
Er ZM = −nZM zonal density, MSF,
n and Ti gradient

input power. Below threshold, the system is in a low 2. Model


confinement regime (L-mode like), characterized by a high
turbulence energy and no ZMs. Above threshold, the system We consider the extended Hasegawa–Wakatani model of
is characterized by an enhanced confinement, i.e. lower [10, 11], including RMPs, for density n and electrostatic
turbulence energy, due to the presence of ZMs. This is potential φ:
similar to the reference case without RMPs. However, in ∂
n + {φ, n} = ∇ j (1)
presence of RMPs, the threshold for the bifurcation between ∂t
these two states depends on RMP amplitude. In the weak- ∂
ρs2 ∇⊥2 φ + ρs2 {φ, ∇⊥2 φ} = ∇ j , (2)
RMP regime, the threshold increases linearly with the RMP ∂t
coupling parameter. For sufficiently strong RMPs, the system where ∇ denotes the parallel gradient, given—neglecting the
undergoes a back-transition to L-mode. Another result is plasma response—by: ∇ ∼ ∇0 +(δBx /B)(∂/∂x) with δBx /B
that the DW frequency is radially modulated by the zonal the radial magnetic perturbation [11].
density perturbations. As the LH transition is thought to The model (1), (2) is closed by the generalized Ohm law,
be triggered by ZFs [12, 13], our model predicts that RMPs which can be written as: j = −D ∇ (φ − n) with D the
increase the LH power threshold, a key result consistent electron parallel diffusivity D ∝ 1/νei . RMP effects are best
with experiments near-transition [3, 6–8]. Physically, RMPs described using a two-fluid picture. The basic mechanism of
enhance (collisional) radial transport of electrons by coupling RMP effect that we propose is improving over time. In our
parallel transport to the tilting of magnetic flux-surfaces. previous modelling [10, 11], we first thought that RMPs were
generating a mesoscale electron current leading to an RMP-
The resulting cross-field radial flux can compete with the
induced stress which competes against the Reynolds stress thus
cross-field transport of polarization charge which is the agent
increasing ZF damping. In the present work, we realized that
of guiding-centre ambipolarity breaking responsible for ZF the latter RMP-induced stress corresponds to the projection—
formation. This simple model shows that in presence of along the unperturbed magnetic field B0 —of quasilinear
RMPs, a pure drift-wave–zonal flow (DW–ZF) paradigm is δ j × δ B  torques localized around rational surfaces. Note
no longer sufficient. With RMPs, the ZFs, in addition to being that RMPs do also generate a mesoscale electron current—
nonlinearly coupled to primary DWs, are also linearly coupled through the quasilinear δ v × δ B  electric field—referred to
to secondary zonal density perturbations [11], resulting in as ‘y-independent eddy currents’ in the nonlinear tearing mode
increased damping of (secondary) ZFs. Taking into account theory [16]. However, this mesoscale current is not the main
the mean density evolution, the density profile—sustained by actor in our theory and we neglect it in the present work (it is
the particle source—has an increased turbulent diffusion due a third-order nonlinear effect), the main actors are the RMP-
to the increased turbulence saturation level, compared to the induced quasilinear torques. Without RMPs, the amplitude
reference case without RMPs. The later result provides one of E × B ZFs is set by the competition between mesoscale
possible explanation for the puzzling density pump-out effect. Reynolds stress drive and neoclassical damping due to ion–ion
This coupling mechanism—acting here on mesoscales—is friction. In presence of RMPs, mesocale quasilinear torques
quite universal and also acts on profile scales. More precisely, δ j × δ B  arise—in the poloidal direction—which couple the
with RMPs, the density profile—sustained by the particle E × B ZFs to the (electron) zonal diamagnetic flows, thus
driving zonal density perturbations at the expense of E × B
source—is linearly coupled to the mean flow by the RMPs.
ZFs. Physically, the δ j × δ B  torques are equivalent to an
The key, bottom line result is that, through this direct coupling,
RMP-induced quasilinear stress which competes against the
the mean flow shear is decreased relative to its reference value Reynolds stress. As only E × B ZFs are directly driven by
without RMPs. Based on the RMP effect on ZMs, we include the (ion) Reynolds stress, the ZF damping is enhanced—over
mean flow shear effects, thereby extending the 0D model its neoclassical value—by RMPs. This mechanism can also
of [14] to include RMPs. Table 2 summarizes the group of be understood as the uniformization of the stream function of
different 0D predator–prey models. This list is not exhaustive, the electron fluid φ − n on mesoscale, i.e. mesoscale electron
we focus here on 0D models directly related to the present force balance.
study. This allows to describe RMP effects on the full LH The model considered here is an extension of the DW–
transition scenario, including mean E × B flow. ZM model [10], to include equilibrium density (mean density)

2
Nucl. Fusion 54 (2014) 013004 M. Leconte et al

dynamics, as well as equilibrium E × B flow shear. In the RMP-induced electron diffusivity [11]. We see from equa-
this work, we are concerned with the evolution of turbulence tion (4) that RMPs directly couple the mean density evolution
energy, ZF energy and mean density gradient. We assume a to the mean electric potential φ.
characteristic scale LV ∼ Ln for the profile components and
qr−1 ∼  for the zonal components (i.e. ZMs). The RMP effect 2.1. ZF and turbulence energy dynamics with RMPs
is clear from the flux-surface average ∇ j  of the parallel
gradient of the parallel current, which enters both the vorticity In order to obtain the evolution of the E × B ZF energy, we
equation and the continuity equation: need to express the zonal modulations of density nZM in
d  ˜ terms of that of the poloidal flow vy ZM . In other words,
∇ j  = b̃x j = we use a slaving approximation, justified by the fact that DW
dx   
  turbulence only directly drives ZFs. The ZFs VZF = vy ZM
d d dφ dn
− iD k b̃x (φ̃ − ñ) − b̃x b̃x D − . and turbulence energy  thus evolve according to [10]:
dx MC dx dx dx  2
(3) d DRMP
= γ0 N − γNL  − 1 − C
2
α|VZF |2
˜ dt Dturb + DRMP
Here the helical currents j are evaluated in the vicinity of their
respective resonance surfaces. (5)
 2
The first term on the rhs of equation (3) is the contribution d|VZF | 2
DRMP
ρs2 = 1−C α|VZF |2 − ρs2 µ|VZF |2
from RMP-induced magnetostatic cells (MCs). The latter are dt Dturb + DRMP
due to the modes with helicity = RMP-helicity. At first DRMP Dturb
glance, it seems that this term is negligible—in the vicinity of − |VZF |2 . (6)
DRMP + Dturb
the resonance surface x = 0—compared to the second term
on the rhs, because k ∝ x. In the turbulence energy evolution equation (5), RMP effects
For resistive DWs, the components of the RMP-induced appear only in the last term on the rhs, which represents the
MCs can be shown to satisfy −∇0 δn + δbx ∂x ∂
[φZM − ZM shear.
nZM ] = 0, δφ = 0, which leads to ñ ∝ 1/x and suggests that
the RMP effects cancel out, as in ideal MHD. However, taking 2.2. Zero-dimensional model
into account a uniform mean E × B flow ωE shows that the
Thus, including the mean flow shear effect on turbulence, the
RMP-induced MCs have a density component ∼x/(iωE ) close
model takes the form:
to the resonance surface, and ∼1/x far from the resonance
surface, thus cancelling most of the RMP effects, except near d
= a0 N − a1eff ()|VZF |2 − a2 V 2 − a3  2 (7)
the resonance surface, which shows that RMP effects are dt
radially localized. As we consider here a 0D model, we use a
‘local’ approximation: we model the RMP effects as uniform, d|VZF |2
= b1eff ()|VZF |2 − b2eff ()|VZF |2 (8)
and we take the value at the resonance surface x = 0. This is dt
equivalent to neglecting the effect of MCs, as in [11, 19]. dN
At this point, the reader may question the validity of = −c0 N − c1 N − c2 (V + N) + (9)
dt
neglecting MCs. The study of the interaction between MCs
dT
and ZFs needs a dedicated work to be addressed, one which = −d0 T − d1 T + Q. (10)
is by definition 1D, as opposed to the zero-dimensional model dt
treated here. Nevertheless, we have given some thought on Here, the quantities a1eff (), b1eff () and b2eff () are effective
the physics validity of our approximation. The neglect of MC (turbulence-energy dependent) coefficients, respectively
is only valid if the damping due to parallel electron diffusion given by:
is much smaller compared to the Doppler frequency, i.e. the
2
‘local’ approximation is valid if D k2 (x)  |ωE |. We will DRMP
a1eff () = 1 − C a1 (11)
extend this model to 1D somewhere else, and properly take DQL  + DRMP
into account the effects of MCs then. An extension of the 1D

2
predator–prey model of [21] to include RMPs is planned. DRMP
The term on the rhs of equation (3)—proportional to the b1eff () = 1 − C b1 (12)
DQL  + DRMP
divergence of a flux—is similar to the RMP term coupling
DRMP DQL 
ZMs [11]. This is not surprising, since both ZFs and the b2eff () = µ + ρs−2 (13)
mean flow are modified by the RMPs via quasilinear δ j × δ b DQL  + DRMP
torques. and the parameter c2 is an RMP-induced friction given by
After some algebra, we obtain using equation (3), in c2 = DLRMP 2 .
presence of RMPs, the mean density evolution as: n
The model described by equations (7)–(10) is an extension
dn n DRMP  of the model of [14], to include RMPs, with one additional
+ (Dturb + Dresid ) 2 − Spart = 2
φ − n (4) parameter c2 and three modified parameters a1eff (), b1eff () and
dt L⊥ L⊥
b2eff () in this model. The parameters c2 , a1eff (), b1eff () and
with Dturb ∝  the turbulent diffusivity— the turbu- b2eff () represent the effect of RMPs. They involve the RMP
lence energy—and Dresid represents the surviving—short particle diffusivity DRMP [11]. The model is based upon a
wavelength—turbulence in the pedestal and where DRMP is density-gradient (∇n)-driven turbulence, so we tacitly assume

3
Nucl. Fusion 54 (2014) 013004 M. Leconte et al

Figure 1. (a) Predicted mean electric field = −φ versus mean density gradient N = −n given by expression (14), for different
values of the RMP coupling parameter (µRMP /µ) = (DRMP ρs−2 /µ) and (b) stability diagram showing relative variation of LH power
threshold: ZF energy |VZF |2 versus—normalized—effective growth rate γ eff /γc 0 , for different values of the—normalized—RMP coupling
parameter (µRMP /µ) = (DRMP /ρs2 µ).

that increasing the particle flux corresponds to a power ramp- third term on the rhs of equation (9) and (ii) indirectly via the
up. A more complete model with both ∇n-driven and ∇Ti - damping of ZFs which increases turbulence energy and hence
driven turbulence is desirable, but this is beyond the scope of enhances turbulent particle diffusion, the first term on the rhs
this article. of equation (9).
We are now faced with the task of closing the system of We see from expression (14) that RMPs decrease all
equations (7)–(10). For this purpose, we seek—as in [14]—an driving components of the perpendicular electric field by a
expression for the mean flow shear V ∝ φ . Additionally, factor 1 + (DRMP /ρs2 µ). In a weak-turbulence, H-Mode like
we also need an expression for the mean flow V = φ . regime, this seems to be in accord with experiments which
The closure is obtained by considering radial and toroidal ion show a clear decrease of the perpendicular electric field in
force balance combined with mean charge balance, including presence of RMPs (the so-called shallowing of the ‘Er well’).
the RMP effect on the mean electric potential. The detailed Note however, that RMPs also act to decrease the mean density
calculation of the mean electric potential φ is given in gradient through particle balance, and thus—even for weak
appendix A. turbulence—the system is clearly nonlinear and may undergo
The resulting expression is: bifurcations between different states. The mean electric field
is plotted versus mean density gradient (figure 1).
ρs2 µ − DRMP ρs2 µ ρs2 αM In the weak-RMP regime (DRMP /ρs2 µ)  1, the mean
=− N − T − 
ρs2 µ + DRMP ρs2 µ + DRMP ρs2 µ + DRMP electric field simplifies to:



ρs2 µ mom DRMP DRMP
+ gmom () , (14) ∼− 1−2 2 N − 1− 2 T
ρs2 µ + DRMP ν1 ρs µ ρs µ



where we use the following notations: DRMP αM DRMP mom
− 1− 2 + 1− 2 gmom () ,
ρs µ µ ρs µ ν1
= −φ , N = −n > 0, T = −Ti  > 0 (15) (17)
By 1 whereas in the strong-RMP regime (DRMP /ρs2 µ)  1, it
αM = − kx ky > 0, gmom () = × . (16) simplifies to:
B 1 + λmom 
k

ρ2µ ρ2µ
Here, gmom () represents the effect of turbulent ∼ 1−2 s N− s T
DRMP DRMP
momentum diffusion, where λmom = ν0 /ν1 is the ratio of
turbulent to residual momentum diffusivities. Note that direct ρs2 αM ρs2 µ mom
− + gmom () . (18)
RMP effects on the toroidal flow are beyond the scope of this DRMP DRMP ν1
model, and thus neglected here. In the absence of momentum Note however, that in the strong-RMP regime, direct RMP
drive ( mom = 0), expression (14)—obtained neglecting effects on ion dynamics—beyond the scope of this article—
compressional effects (direct RMP effects on ions)—predicts are no longer negligible, and could play an important role.
that RMPs slave the mean electric field to the sum of three The mean E × B flow is simply the opposite of the mean
components: the mean density gradient N, the mean ion electric field, so expression (14) directly yields:
temperature gradient T , and the turbulence energy . Due ρs2 µ − DRMP ρ2µ
to the strong neoclassical damping of the mean poloidal flow, V = 2
N+ 2 s T
ρs µ + DRMP ρs µ + DRMP
(αM /µ)  1, and hence for mom = 0, the mean electric
field is mainly proportional to the density gradient N and ion ρs2 αM ρ2µ mom
+ − 2 s gmom () . (19)
temperature gradient T , consistent with experiments without ρs2 µ
+ DRMP ρs µ + DRMP ν1
external momentum input. In fact, the RMPs act to modify the To evaluate the associated mean flow shear, we note that
density gradient through two mechanisms: (i) directly via the in the scrape-off layer (SOL)—a region of the plasma not

4
Nucl. Fusion 54 (2014) 013004 M. Leconte et al

accounted for in this simple model—the mean electric potential Table 3. Possible states of the model.
is always positive, due to sheath effects. Since we showed Turbulence
that the mean potential gradient is negative (in accord with State energy  ZF amplitude |VZF |
experiments) in the region where our model is valid (just inside L-mode eff
γ /a3 0
from the SOL), there is clearly a strong shear in the mean I-phase, no RMPs µ/b1 ∼ b1 γ eff − a3 µ
electric field. Hence, the mean electric field shear is clearly an I-phase, weak-RMPs (µ + µRMP )/b1 ∼ b1 γ eff − a3 (µ + µRMP )
increasing function of the magnitude of the mean electric field. quiescent H-mode 0 0
Therefore, we approximate this behaviour by considering a
proportionality relation between the amplitude of the mean
electric field shear and the amplitude of the mean electric field: ZF-dominated state where the ZF evolution near transition ( is
slaved to VZF ) is given—in the weak-RMP regime—by:
d
eff
 
|E⊥ | ∝ |E⊥ |, i.e. V ∝ V (20) 1 d|VZF |2 γ ρ 2 b1 µRMP µ
dx = − 1 + 1 + 2C s
b1 dt a3 DQL µ b1

2
In the weak-RMP regime, expression (19) yields: ρ µRMP a3 a1


  ×|VZF |2 − 1 − 2C s |VZF |4 . (23)
DRMP DRMP mom 2 DQL γ eff a3
V 2 ∼ 1−2 2 N + 1− 2 f T , , , Expression (23) shows that the associated bifurcation is a
ρs µ ρs µ ν1
DRMP transcritical bifurcation. The RMPs do not modify the nature
for  1. (21) of the bifurcation, but they do increase the threshold of
ρs2 µ
the bifurcation. A plot of ZF amplitude |VZF | versus order
Keeping only the density drive, i.e. for f (T , , mom ) = 0 parameter γ eff for different values of the ratio (µRMP /µ) is
in expression (21), we recover in the limit DRMP → 0 the shown (figure 1(b)).
V 2 ∝ N 2 expression obtained in [14]. In presence of RMPs, Figures 2(a)–(c) show the numerical solution of
expression (21) predicts that for increasing RMP amplitudes, equations (7)–(10). Figure 2(a) shows the reference case
the sensitivity of turbulence shearing to the driving gradients without RMPs DRMP = 0. The fields plotted are: ZF
(and Reynolds stress and momentum source) decreases in energy |VZF |2 , turbulence energy , mean density gradient
the weak-RMP regime. In the strong-RMP regime, the N, mean temperature gradient T , and mean electric field
dependence on the mean density gradient is given by = −V . The onset of the L–I bifurcation (from L-mode


  2 like state to ZF-dominated state) determines the onset of the
2 ρs2 µ ρs2 µ mom I–H bifurcation (from ZF-dominated state to H-mode like
V ∼ − 1− N+ f T , ,
DRMP DRMP ν1 state), and hence is potentially more important. The power

2 threshold for the L–I bifurcation is given in table 3. In the
ρ µ DRMP
∼ 1−2 s N 2, for  1. (22) case without RMPs, figure 2(a) shows the three regimes:
DRMP ρs2 µ
L-mode like regime between = 0 and ∼ 0.5, followed
Interestingly, in the strong-RMP regime, the density by the ZF-dominated regime—which exhibits predator–prey
dependence of the—squared—flow shear tends to the one oscillations—for ∼ 0.5–1.5, and the quiescent H-mode like
without RMPs, since the mean electric field switches from regime for > 1.5. In the H-mode like regime, the mean
being negative and close to −φ ∼ n + Cst for weak- electric field is negative, corresponding to an ‘Er well’, and
RMPs, to being positive and close to −φ ∼ −n + Cst its absolute value is the sum of the density and temperature
for strong-RMPs. Note that at the boundary between the two gradients. Figure 2(b) shows the case with RMPs with an
regimes, i.e. at (DRMP /ρs2 µ) = 1, the turbulence shearing has RMP parameter DRMP = 0.1. The power threshold of the L–I
no dependence on the mean density gradient N, due to the exact bifurcation is clearly increased compared to the case without
cancellation of the ion diamagnetic contribution by the electron RMPs. This is due to the RMP-induced friction which modifies
diamagnetic contribution, as clear from the first term on the the power threshold as shown in (figure 1(b)). Moreover,
rhs of equation (19). Hence, (DRMP /ρs2 µ) = 1 corresponds the difference between the thresholds of the L–I and I–H
to a minimum of the slope in the graph of V 2 versus N, or bifurcations is increased by RMPs—i.e. the domain of the
equivalently of φ 2 versus −n . intermediate phase increases. As the I–H bifurcation is due
to the mean flow shear stabilization of turbulence [17], the
increased threshold of the I–H bifurcation is due to RMPs
3. Numerical results decreasing both the mean density gradient through RMP-
enhanced turbulent particle diffusion and the sensitivity of
Before presenting the numerical results, we summarize the mean flow shear to the mean density gradient. Finally, during
possible states of the model (7)–(10) in a table (table 3). the L-mode like regime and the ZF dominated regime (I-phase
Near the transition from the L-mode like state to the like), the mean density gradient is strongly decreased by the
ZF-dominated state (I-phase like), the mean density and RMPs due to the RMP-enhanced turbulent diffusion. The latter
temperature are approximately constant—they only weakly decrease of density gradient is a possible explanation of the
depend on turbulence energy—and the nonlinear system so-called ‘density pump-out’ effect observed in experiments.
reduces to equations (7), (8) with an effective growth rate However, in this simple model, the temperature gradient is
γ eff ( , V 2 ). This system has two possible states: (i) a no-ZF also decreased in the L-mode and I-phase like regimes, due
state corresponding to  = 0, |VZF |2 = 0 and (ii) a to enhanced turbulent heat diffusion, a feature not observed

5
Nucl. Fusion 54 (2014) 013004 M. Leconte et al

Figure 2. Evolution of , |VZF |2 , N and T as a function of ‘input power’ = 0.01t. Parameter values are a0 = c0 = d0 = 1,
a1 = b1 = a2 = 1, a3 = 0.1, c1 = d1 = 0.5, µ = 1 and mom = 0. The mean electric field amplitude E⊥ is also plotted. (a) Reference case
without RMPs, (b) case with RMPs, DRMP = 0.1 and (c) case with RMPs, DRMP = 0.5 and (d) RMP effect on LH–HL hysteresis: evolution
of  during a ‘power’ ramp up and ramp down with (dashed-line) and without (full-line) RMPs, for a different set of parameters.

in experiments. This is likely due to the fact that our model decrease of hysteresis—have unfavourable implications for
does not take into account the heat channel for electrons. As a ITER, it suggests that ELM control experiments should seek
result of RMPs, the mean electric field is also decreased in the the minimum RMP amplitude (within a safety margin) required
H-mode like regime, corresponding to a shallowing of the ‘Er to suppress ELMs.
well’ consistent with experiments on DIII-D [18]. For higher
RMP amplitude (figure 2(c)), the trend is similar, i.e. the power 3.1. Steady-states of the model
threshold for the L–I transition increases further, and the power
To get some insight into the possible bifurcations, we apply a
threshold for the I–H transition also increases. In (figure 2(c)),
steady-state analysis. The steady-states are given by:
the I–H transition is not shown, because it occurs at very high
power. Bearing in mind that in experiments the amount of a0 N − a1eff ()|VZF |2 − a2 V 2 − a3  2 = 0 (24)
power (or fuelling) is limited, this implies that for a high b1eff ()|VZF |2
− b2eff ()|VZF |2 =0 (25)
enough RMP amplitude, the LH transition will not occur, and
the system will instead remain in the I-phase like regime, where (c0  + c1 + c2 )N + c2 V = (26)
the pedestal is limited by turbulent diffusion. Figure 2(d) (d0  + d1 )T = Q (27)
shows the RMP effect on hysteresis. During a ramp up and together with expression (19) for the mean flow V in terms of
ramp down of ‘power’ (or fuelling) (dashed–dotted line), N, T ,  and νmom
1
, and with a1eff (), b1eff () and b2eff () given by
the turbulence energy without RMPs (full-line) shows that the equations (11)–(13).
L–H transition and H–L back-transition occur at a different Here, we used the ansatz (20) V ∝ V and we chose—for
‘power’ threshold , i.e. the system exhibits hysteresis. Note simplicity—the coefficient of proportionality equal to 1.
that for this hysteresis study, we choose a different set of The steady-state density and temperature gradients—near-
parameters than in figures 2(a)–(c), because in this 0D model, transition—are approximately given—for αµM → 0—by:
hysteresis only occurs for a restricted domain in parameter N() ∼ N̂
space. The dashed line in figure 2(d) shows that RMPs

 
decrease the hysteresis. The relative decrease of hysteresis, µRMP −1 c2 Q mom
= − 1+ − = Cst (28)
measured by (h − h0 )/ h0 —where h = down / up —ranges c1 µ c1 d1 ν1
from ∼15% for DRMP = 0.01 to ∼50% for DRMP = 0.05. Q
T () ∼ T̂ = = Cst (29)
Our findings—namely the increase of power threshold and d1

6
Nucl. Fusion 54 (2014) 013004 M. Leconte et al

The detailed calculation is presented in appendix B. Combining equations (37) and (38), we obtain a logistic
Hence, the turbulence energy near-transition is approxi- equation for ZF energy |VZF |2 which—in the weak-RMP
mately given—for αµM → 0—by: regime—reduces to:

eff
 
a3 2 γ a eff () γ eff ρs2 b1 µRMP µ
 − − 1 |VZF |2  = 0, (30) − 1 + 1 + 2C
a1 a1 a1 a1 DQL µ b1


a3 DRMP a1
where γ eff is an effective linear growth rate of the turbulence— − 1 − 2C eff |VZF |2 = 0. (39)
γ DQL a3
modified due to the mean sheared flow stabilization—given by:
  2
1 − µRMP
µ 1 mom For the interested reader, detailed analysis is given in
γ = a0 N̂ − a2
eff
N̂ + T̂ − . appendix B. The nonlinear system of equations (37) and
1 + µRMP
µ
1 + µRMP
µ
ν1
(39) has two possible states: (i) a no-ZF state (L-mode)
(31) corresponding to  = γ eff /a3 , VZF = 0 and (ii) a ZF-
We see that the mean sheared flow clearly stabilizes dominated state (I-phase) with  = 0, |VZF | = 0. The effective
the turbulence, through the second-term on the rhs of power threshold γceff for the L–I bifurcation is given by:
equation (31). This effective growth rate is also modified by
RMP effects through the (µRMP /µ) factors in equation (31).
 
γceff ρs2 b1 µRMP µ
Moreover, RMPs also act by modifying the nonlinear coupling = 1 + 1 + 2C . (40)
a3 DQL µ b1
term a1eff (), i.e. the second term on the lhs of equation (30).
This latter modification can have important effects, as it
depends on the turbulence energy . We now restrict our
steady-state analysis to the weak-RMP regime, in order to keep 4. Discussion
analytical calculations tractable.
The weak-RMP regime is defined by: In this work we coupled the DW–ZM predator–prey model
to the evolution of mean quantities, i.e. mean density, mean
DRMP DRMP
,  1. (32) ion temperature and mean electric field, including direct RMP
DQL  ρs2 µ effects on mean density and mean electric field. The resulting
predator–prey model—an extension of [14]—exhibits a higher
In the weak-RMP regime, the coefficients a1eff (), b1eff ()
power threshold than the reference case without RMPs, as well
and b2eff () reduce to:
as a decrease of the mean density gradient, reminiscent of the


DRMP puzzling ‘density pump-out effect’. Our model also shows a
a1eff () ∼ 1 − 2C a1 (33) shallowing of the Er well in the H-mode like regime, consistent
DQL 
with experiments, e.g. figure 6(b) of [18]. A density pump-out


DRMP mechanism in L-mode was presented in [19]. However that
b1eff ()∼ 1 − 2C b1 (34)
DQL  work neglected turbulence effects which are shown here to play

an important role. Our model suggests that the density pump-
µRMP
b2eff () ∼ 1 + µ = Cst, (35) out and the shallowing of the ‘Er ’ well are two consequences
µ of the same effect: the radial diffusion of electrons due to
where µRMP = DRMP ρs−2 . Replacing a1eff () in equation (30) RMP-induced tilt of the magnetic field lines combined with
by its expression (33), we obtain a quadratic equation for the collisions (D ∝ νei−1 ), else there is no irreversibility. Our
turbulence energy at saturation: model shows that, in the strong-RMP regime, the ‘Er ’ well

eff can even become positive. We note however that this result
a3 2 γ a1 ρ 2 a1 DRMP should be taken with care as, in the strong-RMP regime, RMP
 − − |VZF |2  = 2C s |VZF |2 . (36)
µ a1 µ DQL ρs2 µ effects on ion dynamics—beyond the scope of this article—
are no longer negligible, and could play an important role. We
The physical—positive—solution is given in the weak-RMP
also note that although we showed that RMPs damp ZFs and
regime by:

decrease the mean E × B flow shear, the mean flow itself can
γ eff a3 DRMP a1 screen the RMPs [20]. This effect—beyond the scope of this
= − 1 − 2C eff |VZF |2 ,
a3 γ DQL a3 work—could moderate the flow damping effect. There are
DRMP a1 limitations of our model. First, it is zero-dimensional. This
for  1 and |VZF |2  1. (37) 0D approximation is only valid if the characteristic damping
DQL  γ eff
rate associated with electron parallel diffusion is negligible
Apart from the solution VZF = 0, the ZF equation (25) at
compared to the Doppler-shifted frequency of the mode. An
saturation reduces—in the weak-RMP regime—to:

  extension to 1D, based on the model of [21] is under way.
ρ 2 b1 µRMP µ µRMP Second, it neglects the plasma response, which was shown
 − 1 + 1 + 2C s = 0, for 1
DQL µ b1 µ to be important, e.g. in [22]. Third, our model neglects direct
(38) RMP effects on toroidal flows. Moreover, the numerical results

7
Nucl. Fusion 54 (2014) 013004 M. Leconte et al

Table 4. Scalings of the relative change in the L–I power threshold discussions which contributed to this work. This work was
as a function of experimental quantities. supported by the World Class Institute (WCI) Programme of
RMP amplitude edge collisionality sound gyroradius the National Research Foundation of Korea (NRF) funded by
 2 the Ministry of Education, Science and Technology of Korea
 B̃ 
 r 1 1 (MEST) (NRF Grant No WCI 2009-001), and by the DOE
PLI ∝   PLI ∝ PLI ∝
B ν∗2 ρs2 Grant DE-FG02-04ER54738.

Table 5. RMPs decrease the LH–HL hysteresis. Appendix A. RMP effects on mean electric field

D |b̃x |2 We start from the force balance for ions:


RMP coupling parameter 0 0.02 0.04 0.06
ρs2 µ 1 1 B F
ρs cs−1 ṽi · ∇ ṽi = − ∇pi − ∇φ + vi × +
Decrease in hysteresis strength (%) 0 15 40 55 en0 B B B en0 B
(41)

presented here focus—for the sake of simplicity—on the zero and for electrons:
torque regime. Note however, that—in our model—external 1 1 B en0 η
torque simply decreases the effective growth rate γ eff as a result 0=− ∇pe + ∇φ − ve × − ve , (42)
en0 B B B B
of increased shearing. Finally, we note that our model focuses
on zero-frequency ZFs, rather than GAMs. A heuristic model where we neglected electron inertia, and we included a (NBI-
for GAMs—including the curvature coupling can be found in induced) volume force of the form:
the appendix of [11].
F = Fz ez . (43)

5. Conclusion Here, we depart from the usual toroidal-field approxima-


tion by considering an axisymmetric magnetic field B not
We investigated, in this work, RMP effects on the L–H purely toroidal:
transition. Here are the main results: (i) the density profile— B By
sustained by the particle source—has an increased turbulent = ez + ey (44)
B B
diffusion compared to the reference case without RMPs, one B
possible explanation for density pump-out. (ii) RMPs decrease with | By |  1.
the sensitivity of mean flow shear to the driving density Electron force balance equation (42) yields:


gradient, resulting in a shallowing of the Er well. (iii) As 1 1
ve = − −∇ φ + ∇ pe (45)
a result, RMPs are shown to increase the power threshold en0 η en0
for both the L–I and the I–H transitions, although the latter B ∇⊥ φ B ∇ ⊥ pe
effect seems stronger, and will be investigated deeper in the ve⊥ = × − × . (46)
B B B en0 B
future. For reference, we summarize the scalings of the relative
We now solve ion force balance equation (41) by
change in power threshold for the—analytically tractable—
exploiting the fact that the lhs of equation (41) is small
LI transition (table 4). Bearing in mind that in experiments
(reflected by the FLR effect ρs ).
the amount of power (or fuelling) is limited, this implies that
At order 0, we obtain:
for a high enough RMP amplitude, the LH transition will
not occur, and the system will instead remain in an I-phase (0) B ∇⊥ φ B ∇⊥ p i
like regime, where the pedestal is limited. (iv) RMPs can vi⊥ = × + × , (47)
B B B en0 B
amplify the hysteresis between L–H and H–L transitions. Main
results on hysteresis are summarized (table 5). The latter where the subscript ⊥ states that the flow is perpendicular to the
findings, namely the increase of power threshold and decrease total axisymmetric magnetic field B given by expression (44).
of hysteresis have unfavourable implications for ITER, and Note that the two components of (46) and (47) are the usual
set a constraint for ELM control experiments. As RMPs E × B drift and electron/ion diamagnetic drift, respectively.
should be turned on before the LH transition to avoid even However, here they have components due to both the toroidal
the first large ELM, it suggests that ELM control experiments magnetic field and poloidal magnetic field, e.g. for the E × B
should not only seek an RMP amplitude compatible with drift.:
ELM suppression, but the minimum RMP amplitude (within By
a safety margin) required to suppress ELMs. Otherwise, an viE = veE = ez × ∇⊥ φ + ey × ∇⊥ φ. (48)
B
unnecessary expenditure of power for the L–H transition will
result. Up to order 1 we obtain:

(0) ρs B 
Acknowledgments vi⊥ ∼ vi⊥ + × ṽiE · ∇ ṽiE (49)
cs B
We would like to thank G. McKee, T.E. Evans, S. Nishimura, where ṽiE is given by:
Jayhuyn Kim, G.Y. Park and S. Mordijck for usefull
discussions. We thank the participants of the 2012 ‘Asia- B ∇⊥ φ̃
ṽiE = × . (50)
Pacific Transport Working Group’ Workshop for stimulating B B

8
Nucl. Fusion 54 (2014) 013004 M. Leconte et al

We now apply mean charge balance: mean polarization charge (proportional to the E × B Reynolds
  stress).
∇⊥ · [en0 (vi⊥ − ve⊥ )] + ∇ j  = 0, (51)
(i) First without RMPs, mean charge balance implies that
the mean poloidal flow is set by the competition between
i.e.:

 the mean Reynolds stress drive and the neoclassical flow
(0) ρs B 
∇⊥ · −en0 ve⊥ + en0 vi⊥ + en0 × ṽiE · ∇ ṽiE damping (µ).
cs B (ii) In presence of RMPs, the RMP-induced quasilinear flux
d  ˜ b̃x j˜ —due to the poloidal component of the quasilinear
+ b̃x j = 0 (52)
dx δ j × δ B  torque—can compete against the mean E × B
 
Reynolds stress, thus decreasing the mean poloidal flow.
where the RMP-induced flux b̃x j˜ is given—in a quasilinear
approximation—by: Equation (61) clearly shows that, in presence of RMPs, the

neoclassical damping of the flow plays a fundamental role.
  dφ dpe  In fact, RMPs (mainly acting on electrons), mediated by the
b̃x j˜ = −D |b̃x |2 − . (53)
dx dx neoclassical damping (acting on ions), can modify the (ion)
mean poloidal flow, and thereby modify the mean E × B flow
Replacing the velocities by their expression, we obtain: through (ion) radial force balance.


B B ρs B  Now, from the evolution equation for ions equation (41),
∇⊥ · × ∇ ⊥ pi + × ∇⊥ pe + en0 × ṽiE · ∇ ṽiE
B B cs B we obtain the (ion) radial force balance and (ion) toroidal force
d  ˜ balance:
+ b̃x j = 0. (54)
dx ∂ ∂ By
Now, we note that—since we neglect curvature effects— 0=− pi  − φ + viy  − viz  (62)
∂x ∂x B
the divergence of both the ion and electron diamagnetic drifts
vanishes, and equation (54) reduces to: ∂  E  ∂2
ṽix ṽi = νresid 2 viz  + Fz . (63)

 d 
B   ∂x ∂x
ρs
en0 ∇⊥ · × ṽiE · ∇ ṽiE + b̃x j˜ = 0. (55) Here we used the trivial identity: vi  = viz , and we added
cs B dx a diffusion term due to residual short-wavelength turbulence
We note the following identity valid for any 2D vector (νresid ).
field u and the (divergence-free) magnetic field B : We neglected the direct effects of RMPs on the ion
dynamics, here, supposed to be small compared to RMPs
∇⊥ · [B × u] = −B · (∇⊥ × u). (56) effects on electrons, due to fast electron streaming along the
field lines.
We also use the following approximation: Combining the (ion) radial and toroidal force balance
  equations (62) and (63) with the mean charge balance (61),
∇⊥ × ṽiE · ∇ ṽiE ∼ ṽiE · ∇⊥ ∇⊥ × ṽiE . (57)
assuming constant electron temperature Te = Teref = Cst, we
obtain:
Using identity (56) and approximation (57), the mean
∂ ∂ ∂ By
charge balance (55) can be written: −Ti  n − n Ti  − φ + viy  − viz  = 0
∂x ∂x ∂x B
  d  
(64)
˜ +
− ṽiE · ∇⊥  b̃x j˜ = 0 (58)

dx ∂ ∂
− (νturb + νresid ) viz  = Fz  (65)
where we defined the (normalized) vorticity ∂x ∂x
∂  E E  ˜ 
eB 2 B  ρs2 µviy  = −ρs2 ṽ ṽ + b̃x j , (66)
˜ = ρs · ∇⊥ × ṽiE = −ρs2 ∇⊥2 φ̃. (59) ∂x ix iy
kB T B
where Fz  denotes the toroidal component of the volume force
Equation (58) can be interpreted as a mean vorticity equation F , and the RMP-induced flux b̃x j˜  is given by:
in stationary state.

We now use the Taylor identity: ˜ ∂ Teref ∂
b̃x j  = −DRMP φ − ref n . (67)
∂x en ∂x
  ∂2  
˜ =
ṽiE · ∇⊥  ṽiEx ṽiEy . (60)
∂x 2 Here, DRMP = D |b̃x |2 is the RMP-induced electron
diffusivity
 [11],
 and we approximate the angular momentum
Using the Taylor identity (60), we obtain the mean charge
E
balance as: flux ṽix ṽi by a turbulent momentum diffusion. For
∂  E E  ˜ 
simplicity, we neglect non-diffusive terms in the ion channel,
− ρs2 ṽ ṽ + b̃x j − ρs2 µviy  = 0, (61) since we focus on RMP effects on electrons.
∂x ix iy To obtain equation (64), we used the following
where we integrated radially and added a neoclassical flow approximation for the mean ion pressure: pi  ∼ nTi .
damping term. Our model consists of the three equations (64), (65),
Note that the mean charge balance (61) has two important (66), together with equation (67). This model determines
consequences, which are better understood in terms of the the mean perpendicular electric field φ in terms of mean

9
Nucl. Fusion 54 (2014) 013004 M. Leconte et al

ion temperature gradient Ti  , mean density gradient n , The steady-state turbulence energy is given by:
 
turbulence energy  and torque Fz . 2 a2 1 − µ
µRMP

In order to couple the present model to the DW–ZM model,  − 1+


2
N
1 + µRMP
µ
a1 1 + µRMP
µ
we consider profiles with a characteristic gradient scalelength
Lx :   −1
1 mom αM
   −1 + T − gmom ()
 1 ∂ −1  1 ∂  1 + µRMP ν1 µ
   ∼ φ  = L⊥  Ln , (68) µ
  ∂x   φ ∂x  

a2 1 − µ
µRMP
a0 1
where Ln is the usual density-gradient scalelength defined as × N− N+
a1 a1 1 + µRMP
µ
1 + µRMP
µ
L−1
n = −(1/n)∂n/∂x.   2 
eff
Using this ansatz, the model (64), (65), (66) together with mom a1 ()
× T − gmom () − |VZF |  = 0.
2
equation (67) reduces to: ν1 a1

By (76)
− N − T − + Vy − Vz = 0 (69) In equation (76) remain two unknowns N and T .
B
The steady-state temperature gradient T is obtained using
L⊥ mom equation (27):
Vz = gmom () (70)
νresid Q
T = T () = gheat () , (77)
d1
DRMP  α
− N +
M
Vy = − 2 , (71) where gheat () = (1 + λT )−1 , with λT = d0 /d1 the ratio of
ρs µ µ
turbulent to residual heat diffusivities.

where  = k |φk |2 , T = Ti  , N = n and = φ , The particle balance equation (26) can then be written—
and the function g() represents the turbulent diffusion effect: using (19) and (77)—as:
(c0  + c1 + c2 )N
1

gmom () = , (72) 1 − µRMP


µ 1 Q
1 + λ +c2 µRMP N + µRMP gheat ()
1+ µ 1+ µ d1
with the parameter λ = ν0 /ν1 . We use a quasilinear
1 αM 1 mom
approximation for the perpendicular Reynolds stress: +  − g mom () = . (78)
1 + µRMP
µ
µ 1 + µRMP
µ
ν1
∂  E E 1 After some algebra, we obtain the steady-state density gradient
− ṽix ṽiy ∼ − kx ky  = αM   0. (73) N as:
∂x L⊥ k


µRMP −1
N = N() = gpart () − 1+
Combining equations (69)–(71), the mean electric field c1 µ

 
− is given by: c2 Q mom αM
× gheat () − gmom () +  , (79)
DRMP  α c1 d1 ν1 µ
−N − T − − 2 − N −
M
 1−
µRMP
ρs µ µ where gpart () = [1 + (1 + µ
µRMP
1+ µ
) cc21 ]−1 (1 + λN )−1 , with
By L⊥ mom µ
1− RMP
− = 0. (74) λN = [1 + (1 + µ
µRMP ) cc21 ]−1 cc01 . Now, using expressions (77)
B νresid 1+ µ
and (79), we can write the turbulence energy equation (76) as:
Solving equation (74) for − yields expression (14) given in

a2 1 − µ
µRMP
a0
the main text. gturb () N() − µRMP N()
a1 a1 1 + µ
  2
1 mom a1eff ()
Appendix B. RMP effects on the L–I bifurcation + T () − gmom () − |VZF | 
2
1 + µRMP
µ
ν1 a1
Near transition, the turbulence energy is slowly varying a3
−  2 = 0, (80)
compared to ZMs, and hence, its dynamics is effectively slaved a1
to that of ZMs, i.e. d/dt ∼ 0. Replacing the mean flow V by where gturb () is given by:

a2 1 − µ
its expression (19), equation (24) can be written: µRMP
2
  gturb () = 1 + N()
 2 1 + µRMP a1 1 + µRMP
1 − µRMP 1 mom µ µ
a0 N − a 2
µ
N+ T − gmom ()  
1 + µRMP 1 + µRMP ν1 1 mom αM −1
µ
 
µ
 + T () − gmom () . (81)
1 + µRMP ν1 µ
a2 1 − µ
µRMP µ
2
−a1eff ()|VZF |2  − 1 + µRMP µRMP N Now we note that since all four quantities gpart (), gheat (),
1 + µ a1 1 + µ
g () and gturb () are slowly varying functions of the
  mom

1 mom αM turbulence energy  near-transition, we have:
+ T − gmom () a1  2 = 0. (75)
1 + µRMP
µ
ν 1 µ gpart () ∼ gheat () ∼ gmom () ∼ gturb () ∼ 1. (82)

10
Nucl. Fusion 54 (2014) 013004 M. Leconte et al

Hence, the steady-state density and temperature gradients— After some algebra, the ZF evolution (8) reduces—in the weak-
near-transition—are approximately given—for αµM → 0—by: RMP regime—to:


µRMP −1 1 d|VZF |2
N() ∼ N̂ = − 1+ = |VZF |2
c1 µ b1 dt

 
 
c2 Q mom ρ 2 b1 µRMP µ
× − = Cst (83) − 1+ 1+C s |VZF |2 = 0. (93)
c1 d1 ν1 DQL µ b1
Q Combining equations (92) and (93), we obtain a logistic
T () ∼ T̂ = = Cst. (84)
d1 differential equation for zonal flow energy |VZF |2 :
Hence, the turbulence energy near-transition is approxi-
eff
1 d|VZF |2 γ − γceff
mately given—for αµM → 0—by: = |VZF |2
b1 dt a3

eff

a3 2 γ a1eff () a3 DRMP a1
 − − |VZF |  = 0,
2
(85) − 1 − 2C eff |VZF |4
a1 a1 a1 γ DQL a3


where γ eff is an effective linear growth rate of the turbulence— DRMP a1 2
modified due to the mean sheared flow stabilization—given by: +2C |VZF |6 = 0, (94)
DQL γ eff
  2
1 − µRMP
µ 1 mom where γceff denotes the—RMP dependent—power threshold
γ = a0 N̂ − a2
eff
N̂ + T̂ − .
1 + µRMP
µ
1 + µRMP
µ
ν1 for the bifurcation, given by:
 
 
(86) µRMP ρ 2 b1 µRMP a3 µ
γceff = 1 + 1 + 2C s ,
We see that the mean sheared flow clearly stabilizes µ DQL µ b1
the turbulence, through the second-term on the rhs of µRMP
for  1. (95)
equation (86). This effective growth rate is also modified µ
by RMP effects through the µRMP factors in equation (86).
µ Note that, in addition to the standard quadratic
Moreover, RMPs also act by modifying the nonlinear coupling
nonlinearity |VZF |4 ∝ E 2 —with E the ZF energy—of the
term a1eff (), i.e. the second term on the lhs of equation (85).
logistic differential equation, equation (94) has an additional
This latter modification can have important effects, as it
(RMP-induced) cubic nonlinearity ∝ E 3 . However, since the
depends on the turbulence energy . We now restrict
our analysis to the weak-RMP regime, to keep analytical quadratic nonlinearity is already stabilizing (negative sign),
calculations tractable. The weak-RMP regime is defined by: the cubic term does not affect the nature of the bifurcation
near threshold. Hence, RMPs do not modify the nature
DRMP DRMP of the bifurcation, it remains a transcritical bifurcation, but
,  1. (87)
DQL  ρs2 µ they increase the power threshold. The nonlinear system of
Note that the condition (87) implies finite turbulence  = 0. equations (92), (94) has two possible states: (i) a no-ZF state
In the weak-RMP regime, the coefficients a1eff (), b1eff () and (L-mode) corresponding to  = γ eff /a3 , VZF = 0 and (ii)
b2eff () reduce to: a ZF-dominated state (I-phase) where—neglecting the cubic

term—the ZF energy |VZF |2 is given by:
DRMP

a1eff () ∼ 1 − 2C a1 (88) a1 ρs2 µRMP γ eff − γceff
DQL  |VZF | = 1 + 2C eff
2
,

γ DQL a1
DRMP
b1 () ∼ 1 − 2C
eff
b1 (89) µRMP γ eff
DQL  for 1 and |VZF |2  (96)

µ a1
µRMP
b2eff () ∼ 1 + µ = Cst. (90) with the power threshold γceff given by expression (95).
µ
Replacing a1eff () in equation (85) by its expression (88),
we obtain a quadratic equation for the turbulence energy at References
saturation:

eff [1] Wagner F. 2007 Plasma Phys. Control. Fusion 49 B1
a3 2 γ DRMP [2] Evans T.E. et al 2004 Phys. Rev. Lett. 92 235003
 − − |VZF |2  − 2C |VZF |2 = 0,
a1 a1 DQL [3] Kirk A. et al 2011 Plasma Phys. Control. Fusion
DRMP 53 065011
for  1. (91) [4] Suttrop W. et al 2011 Phys. Rev. Lett. 106 225004
DQL  [5] Jeon Y.M. et al 2012 Phys. Rev. Lett. 109 035004
The physical—positive—solution is approximately given by: [6] Gohil P. et al 2011 Nucl. Fusion 51 103020

[7] Kaye S.M. et al 2011 Nucl. Fusion 51 113019
γ eff a3 DRMP a1
= − 1 − 2C eff |VZF |2 [8] Ryter F. et al 2012 Nucl. Fusion 52 114014
a3 γ DQL a3 [9] Xu Y. et al 2011 Nucl. Fusion 51 063020

[10] Leconte M. and Diamond P.H. 2011 Phys. Plasmas
DRMP a1 2
+2C |VZF |4 , 18 082309
DQL γ eff [11] Leconte M. and Diamond P.H. 2012 Phys. Plasmas
19 055903
DRMP γ eff
for 1 and |VZF |2  (92) [12] Xu G.S. et al 2011 Phys. Rev. Lett. 107 125001
DQL  a1 [13] Manz P. et al 2012 Phys. Plasmas 19 072311

11
Nucl. Fusion 54 (2014) 013004 M. Leconte et al

[14] Kim E. and Diamond P.H. 2003 Phys. Rev. Lett. [18] Burrell K.H. et al 2005 Plasma Phys. Control. Fusion
90 185006 47 B37–52
[15] Diamond P.H. et al 2005 Plasma Phys. Control. Fusion [19] Yu Q. and Gunter S. 2009 Nucl. Fusion 49 062001
47 R35 [20] Fitzpatrick R. and Hender T.C. 1991 Phys. Fluids B
[16] Rutherford P.H. 1973 Phys. Fluids 16 1903 3 644
[17] Malkov M.A. and Diamond P.H. 2009 Phys. Plasmas [21] Miki K. et al 2012 Phys. Plasmas 19 092306
16 012504 [22] Waelbroeck F.L. et al 2012 Nucl. Fusion 52 074004

12

Você também pode gostar