Isotope effects on L-H threshold and confinement in tokamak plasmas

The dependence of energy and particle transport on the main ion isotope mass (A) is of fundamental interest for the understanding of turbulent transport and, therefore, for accurate predictions of confinement in burning devices. Future experiments, such as ITER, will operate in deuterium–tritium mixtures, while current experiments typically use a single hydrogen isotope, usually deuterium. Knowledge of the dependence of plasma properties on isotope species is important for the initial, low-activation phase of ITER operations (hydrogen or helium).

For the analysis of transport processes in tokamak plasmas one can think of three main regions of the plasma volume, as shown schematically in figure 1. In the very core of the plasma, physics processes are determined primarily by sawteeth (core MHD events) and by collisional (neoclassical [1]) transport. The core confinement region carries most of the plasma volume and therefore plays a key role for the overall confinement. In this region, turbulent transport—much higher than neoclassical—dominates and is thought to be due to small scale, ion Larmor radius size turbulence: ρ ∼ (A T)^1/2/B, where T is the ion temperature and B the magnetic field. In this region, the local heat transport is predicted by quasi-linear models to follow the gyro-Bohm scaling: χ_gB ∼ ρ* (with ρ* = ρ/a the normalized ion Larmor radius, a the plasma minor radius and χ_gB the gyro-Bohm heat diffusivity), which results in an unfavourable dependence of local transport on isotope mass, χ_gB ∼ A^1/2. In H-modes, characterized by a region with steep pressure gradient at the plasma edge and improved energy and particle confinement above L-mode values, the gyro-Bohm scaling has been confirmed in numerous experiments in D plasmas, for the local electron transport see e.g. [2] and references therein. The local ion heat transport is also consistent with gyro-Bohm scaling in conditions when electron and ion heat diffusivities can be resolved, but departs from it depending on the variation of the fractional heat flux in the electron and ion heat channels [2].

Zoom In Zoom Out Reset image size

In L-mode, tokamak experiments have shown in deuterium that the local transport follows the gyro-Bohm scaling for electrons and the Bohm scaling for ions, see e.g. [3, 4]: χ_i ∼ χ_B ∼ T/e B (with χ_B the Bohm heat diffusivity). The Bohm scaling is therefore consistent with the absence of isotope dependence in the ion heat channel: τ_E ∼ 1/χ_B ∼ A⁰.

The transition from low (L-mode) to high mode of confinement (H-mode) is characterized by a threshold in input power, P_L-H, whereby for input powers in excess of P_L-H ion scale turbulence is strongly suppressed at the plasma edge, leading to the formation of an edge transport barrier (ETB) for both density and temperature in a region with large velocity shear, see e.g. [5, 6]. In the edge pedestal region, the transport is reduced to roughly neoclassical levels and periodic MHD events, the edge localized modes (ELMs), limit the maximum pressure achievable at the top of the pedestal between subsequent ELM crashes. The pedestal region is strongly influenced by scrape-off-layer (SOL) physics: neutrals, radiation and impurity effects all have an impact on the pedestal. As the pedestal is influenced by SOL physics, also core and pedestal are strongly coupled in a wide range of H-modes, as observed in many tokamaks [7] and confirmed by modelling [8]: the pedestal pressure increases with the total plasma pressure as the core temperature gradients develop at constant gradient length, L_T = ∇T/T ≅ constant. The pedestal, therefore, sets the boundary condition in H-modes. The ELMy H-mode is chosen as the primary mode of operation for ITER.

Over the years, several tokamak experiments and theoretical work have addressed the plasma transport properties with varying hydrogen isotope mass, primarily in hydrogen and deuterium plasmas, see e.g. [9–21]. The TFTR tokamak [22] and the JET tokamak (JET DTE1) [23] are the only two tokamaks to have conducted experiments also in tritium and in the 50:50 deuterium–tritium mixtures required for burning plasmas. Different dependencies of the energy confinement, τ_E, on isotope mass have been measured in L-mode and in H-mode [2]. The relationship between the local heat diffusivity and the global energy confinement is taken to be τ_E ∼ a²/χ. Although the strength of the isotope dependence often varies from tokamak to tokamak experiment, even in the same mode of confinement, all experiments consistently find a positive isotope dependence of confinement, τ_E ∼ A^α, with α ranging from 0 to 0.85, which projects favourably to deuterium–tritium burning plasmas. On the other hand, the isotope effects observed experimentally are in contradiction of theoretical predictions of the gyro-Bohm scaling, from which one would expect an unfavourable dependence of τ_E on isotope mass with α = −0.5. This discrepancy between theory and experiments still remains largely unexplained to date and presents a main challenge to the understanding of plasma turbulent transport. The main explanation put forward by theory regarding core heat transport has been linked to the effect of different isotopes on the shear flow and the resultant modification of the ion temperature gradient (ITG) driven turbulence [24]. Recently, gyrokinetic simulations with the GENE code have found a weak isotope effect in core heat transport in ITG/trapped electron mode regimes, but of smaller intensity than found in experiments [25]. Nonlinear gyrokinetic simulations have shown that at high beta the core plasma micro-turbulence can depend on the isotope mass via multi-scale nonlinear effects involving zonal flows, electromagnetic effects and E × B shear, which counteract the intrinsic, local, gyro-Bohm scaling of plasma micro-turbulence [26]. The isotope effect on particle transport has been less studied. The recent work of [20] on the FT-2 tokamak reports an increase in particle confinement time by a factor of 1.4 from hydrogen to deuterium in comparable Ohmic plasmas, while no difference is observed in the energy confinement time [20, 21]. Global and local gyrokinetic simulations reproduce the isotope effect in particle transport [20]. However, gyrokinetic predictions for the total energy fluxes also indicate an isotope effect on energy transport, in contradiction of the ASTRA power balance analysis [20]. Nonetheless, in these plasmas geodesic acoustic modes (GAM) are found with larger amplitude in deuterium than in hydrogen both in experiment and in simulations, as well as a high correlation between turbulent transport bursts and GAM [21], suggesting the need to account for zonal flows in the analysis of the energy fluxes [20].

As the JET tokamak prepares for new deuterium–tritium experiments in 2019/20, with improved edge, neutrons and fast particles diagnostics, as well as with the ITER-like Wall (W in the divertor and mainly Be in the main chamber), a thorough experimental investigation of isotope effects in hydrogen, deuterium and tritium plasmas is being carried out, in order to provide stringent tests to plasma energy and particle transport models. Experiments in deuterium, hydrogen and hydrogen–deuterium mixtures were carried out in 2016. Between 2018 and 2019 a series of campaigns will cover studies in deuterium and tritium plasmas, followed by a second campaign in hydrogen and hydrogen–tritium and hydrogen–deuterium mixtures. A final campaign in deuterium is then foreseen, to complete the scenario optimization before the deuterium–tritium phase.

In the next three sections of the paper the isotope effects on L-mode (section 2), L-H threshold (section 3) and H-mode (section 4) are briefly reviewed and recent findings of isotope effects in tokamak plasmas are presented and discussed. A summary and conclusions are drawn in section 5.

Over the past two decades several tokamak experiments have confirmed that in L-mode the local energy transport follows the short-wavelength gyro-Bohm scaling for electrons, χ_e ∼ χ_gB, see e.g. [3], while the ions follow the long-wavelength Bohm scaling, χ_i ∼ χ_B, see e.g the review of [2] and references therein. As the Bohm scaling is independent of isotope mass, τ_E ∼ 1/χ_B ∼ A⁰ would be predicted for the ion energy transport, namely consistent with no isotope dependence. On the other hand, regression of multi-tokamak data for the L-mode thermal energy confinement time has yielded the ITER97-L scaling [27], τ_E,_th (ITER97-L) = 0.023 ${{{I}}_{{\rm{P}}}}^{0.96}\,{{{B}}_{{\rm{T}}}}^{0.03}\,{{{P}}_{{\rm{abs}}}}^{-0.73}\,{{{n}}_{{\rm{e}}}}^{0.40}\,{{A}}^{0.2}$ (with I_P the plasma current [MA], B_T the toroidal magnetic field [T], P_abs the absorbed input power [MW] and n_e the central line averaged density [10¹⁹ m⁻³]), which exhibits a positive isotope mass dependence, in contradiction of both gyro-Bohm and Bohm scaling.

Recent L-mode experiments in ASDEX Upgrade (AUG) with electron cyclotron resonance heating (ECRH) have compared two L-mode discharges in hydrogen and in deuterium, where the normalized ion gyroradius ρ* differed by √2, which is the square root of the mass ratio, while other dimensionless parameters were kept constant [18]. τ_E,th (H) = 83 ms was lower than τ_E,th (D) = 110 ms. Power balance and perturbative transport analyses revealed that the electron heat transport was unaffected by the change in isotope mass. Due to the central electron heating scheme used, T_e > T_i in the plasma core, thus allowing for measurement of the electron–ion collisional exchange term in the energy balance equation with sufficient accuracy: P_e–i ∼ n² (T_e – T_i)/(A ${{{T}}_{{\rm{e}}}}^{3/2}$). Due to the mass dependence of P_e–i, P_e–i(H) = 2 × P_e–i(D) and it is found that the additional ECRH heating required in hydrogen to match the profiles in deuterium is transported by the ion channel and Q_i(H) > Q_i(D), as shown in figure 2 [18]. The uncertainties in the measured ion temperature gradients are too large to distinguish between hydrogen and deuterium in the power balance transport analysis and ${{\chi }_{{\rm{i}}}}^{{\rm{PB}}}$ = 1–3 m² s⁻¹ are reported for both ion species. The reduced τ_E,th in hydrogen in these low β plasmas (where β is the ratio of the plasma kinetic pressure to the magnetic pressure) is explained by the mass dependence in the collisional energy exchange between electrons and ions and not by a gyroradius effect. It is also noted in [18] that an increase in ion heat transport due to isotope mass effects is not expected for such ITG turbulence dominated plasmas. This result was also confirmed with ASTRA simulations, which show that the experimental profiles and global energy confinement can be reproduced when using the same coefficients for hydrogen and deuterium in the transport code [18].

Zoom In Zoom Out Reset image size

This same mechanism is found to explain the weak, but positive isotope effect on thermal energy confinement observed from hydrogen to deuterium Ohmic discharges in JET-ILW, where T_i < T_e is measured in the plasma core, although in this case an isotope effect on the ion heat transport is not excluded by the data [28].

The mass dependence of the collisional ion–electron heat exchange cannot, however, be invoked to explain the lower τ_E,th in hydrogen than in deuterium when T_i ≅ T_e, that is when the ion–electron heat exchange term cannot be quantified within experimental uncertainties. This is the case of L-mode plasmas obtained in recent JET-ILW experiments with neutral beam (NBI) heating at 2.5 MA/3.0 T. The NBI power was scanned from 1 to 9 MW in both hydrogen and deuterium while keeping the plasma density constant to n_e ∼ 3.0 × 10¹⁹ m⁻³ (obtained with feedback control on the injected gas). The accuracy in the NBI power calibration is estimated to be of order 10%, with the largest errors originating from the beam neutralization and beam transmission measurements (obtained by a combination of test bed measurements and simulations). The transmission in H is lower (70%) than that in D (75%) and it is accounted for in the calibration factor. Based on the available data, the relative accuracy of the NBI power calibration is believed to be the same for hydrogen and deuterium beams. We note that at all NB power levels in the scan the ratio of electron to ion NB heating was the same in hydrogen and in deuterium, as shown in figure 3, due to H-NBI operation at lower NB voltage (60–70 kV) than with D-NBI (80–90 kV). The calculations of figure 3 were carried out with the PENCIL neutral beam deposition code [29].

Zoom In Zoom Out Reset image size

In the experiments, a larger gas puff rate and NBI fuelling were needed in hydrogen to achieve the same density as in deuterium, indicating worse particle confinement in hydrogen. Higher NBI power was required in hydrogen to achieve the same total stored energy as in deuterium. After subtraction of the fast ion energy contribution from the stored energy inferred by the MHD equilibrium reconstruction EFIT, the thermal stored energy is still found to be lower in hydrogen than in deuterium and regression of the measured τ_E,th data gives: τ_E,th ∼ A^0.15±0.02 ${{{P}}_{{\rm{abs}}}}^{-\mathrm{0.63\pm 0.02}},$ as shown in figure 4, in broad agreement with the favourable mass dependence of the ITER97-L scaling. The thermal stored energy calculation described above is in agreement, within 10%, with that obtained by the TRANSP code [30] with input the measured kinetic profiles integrated over the plasma volume. The positive isotope dependence of τ_E,th observed in JET-ILW L-modes is in contradiction of both gyro-Bohm and Bohm scaling.

Zoom In Zoom Out Reset image size

The JET-ILW plasmas of this dataset are of relatively high collisionality, leading to strong coupling of ions and electrons. A simple equipartition analysis shows that if T_i were to exceed T_e by 20%, the net power to the ions would be negative over a significant portion of the radial profile in over half of the cases analysed. The core T_i profile data have large systematic uncertainties (of order 20% overall). They were measured either by Ne X, n = 11–10, charge exchange recombination spectroscopy (CXRS), using diagnostic Ne puffs, and/or by bulk ion CXRS. Within the uncertainties of the available data, T_i ≅ T_e, but potential, small differences between core ion and electron temperatures cannot be resolved. Conversely, the edge T_i profiles are measured by CVI, n = 8–7, CXRS and have higher accuracy. At the plasma edge T_i = T_e, consistently through the L-mode dataset. Under these conditions, both in hydrogen and in deuterium plasmas for the transport analysis we have assumed T_i ≅ T_e within experimental uncertainties of the T_i profile data and the transport analysis has been carried out based on the electron kinetic profiles. The sawtooth inversion radius is at ρ_tor = 0.22–0.26 (corresponding to ρ_pol = 0.3–0.35), depending on the individual discharge, where q_ψ ∼ 1. In the transport region, the inverse normalized temperature gradient length is measured for electrons, ${R/L}_{{T}_{{\rm{e}}}}\cong {R/L}_{{T}_{{\rm{i}}}}$, as shown in figure 5. Therefore, under these conditions, the electrons and/or ions are stiff. An error analysis, with the above uncertainties, shows that a meaningful species-resolved transport analysis is unfortunately not possible. Only the one-fluid, effective heat diffusivity, χ_eff, can be extracted and figure 5 shows that at mid-radius (ρ_tor = 0.5) hydrogen and deuterium plasmas have comparable χ_eff at all input power levels of the L-mode power scan, in contradiction of the unfavourable gyro-Bohm dependence predicted by theory, χ_gB ∼ A^1/2.

Zoom In Zoom Out Reset image size

While no significant isotope effect is found in the L-mode core heat transport under these conditions, a strong, favourable isotope effect is found both in particle and heat transport at the plasma edge from detailed EDGE2D/EIRENE [31, 32] interpretative simulations. Two representative discharges of the 2.5 MA/3.0 T power scans were analysed, in the phase marked by the black, dashed rectangle in figure 6, where higher NBI power was required in hydrogen to achieve the same total stored energy as in deuterium and larger gas puff rate and NBI fuelling were required in hydrogen to achieve the same central line averaged density as in deuterium. The corresponding electron kinetic profiles, in the steady time window analysed, are shown in figure 7, where n_e is measured by reflectometry and T_e by the electron cyclotron emission diagnostic (ECE). The n_e and T_e profiles measured by high resolution Thomson scattering (HRTS) are in very good agreement with those by reflectometry and ECE, respectively (not shown in figure 7 for simplicity).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The edge anomalous transport coefficient of electron particle diffusion, D_perp (Γ_e = −D_perp ∇n_e), and of ion and electron heat diffusivity, χ_e,i (q_e,i = −n_e,i χ_e,i ∇T_e,i) were obtained by fitting the EDGE2D/EIRENE solutions to the experimental upstream profiles. The latter were measured by profile reflectometry and HRTS for n_e and by HRTS for T_e (as the ECE diagnostic does not have sufficient spatial resolution near the last closed flux surface (LCFS)). The edge experimental profiles and the EDGE2D/EIRENE profiles fitted to the experimental data (black lines) are shown in figure 8 for the two hydrogen and deuterium discharges of figure 6. The experimental data are averaged in the steady time window (14.0–16.0 s). While the edge T_e profile is more uncertain, in particular for the H shot, the n_e profile data from reflectometry have high accuracy and allow for a clear distinction in H and D edge particle diffusion coefficients. Figure 9 shows that larger transport coefficients D and χ_e,i are required in hydrogen than in deuterium—both inside and outside the LCFS—to fit the modelled n_e and T_e profiles to the experimental data. In particular, the good spatial resolution of the edge n_e profile measured with reflectometry allows for the computation of a radial structure in D_perp.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

An additional aspect of the 2D modelling is that for a given neutral source it is not possible to fit the experimental density profiles by only adjusting D_perp. Values of albedo at the cryopump surfaces (in the divertor corners of the EDGE2D grid) also require adjustment. In particular, the solutions shown in figures 8 and 9 were obtained with albedo @ pump = 0.925 for the hydrogen case and albedo @ pump = 0.91 for the deuterium case. In EDGE2D/EIRENE the pumping albedo is defined to be equal to [1—sticking probability of H₂ (or D₂) gas to the cryosurface]. The sticking probabilities were shown to be higher for deuterium than for hydrogen from neutral beam test bed results [33], which implies lower albedo values for deuterium than for hydrogen, in qualitative agreement with the EDGE2D/EIRENE simulations.

Consistent with the larger edge particle transport in hydrogen than in deuterium, larger edge density fluctuations (relative to their value in the Ohmic phase of the deuterium discharge) were measured with Doppler reflectometry [34] in hydrogen than in deuterium for the same P_NBI. The probed wavenumbers for the pulses in this paper are in the range 3–4 cm⁻¹. Figure 10 illustrates the case of two discharges in hydrogen and deuterium with P_NBI = 3 MW, although this result was obtained at all power levels in the L-mode dataset.

Zoom In Zoom Out Reset image size

In summary, recent experiments in the L-mode regime in hydrogen and deuterium plasmas show that in conditions where T_i and T_e are decoupled in the plasma core, such as ECRH heated plasmas in AUG or Ohmic plasmas in JET-ILW, the lower thermal energy confinement time measured in hydrogen can be explained by the isotope mass dependence of the i–e heat exchange term in the energy balance equation. On the other hand, this mechanism cannot be invoked with certainty to explain the favourable isotope effect on τ_E,th in JET-ILW L-modes with NBI heating, where T_i and T_e are strongly coupled and therefore the e–i heat exchange term cannot be measured outside experimental error bars. In these plasmas a negligible isotope effect is found in the plasma core from transport analysis setting T_i = T_e (within experimental error bars of the T_i data), while a strong, positive isotope effect is found at the plasma edge, both inside and outside the LCFS, on both energy and particle transport. The combination of negligible core and strong edge isotope effects yield the weak, but favourable isotope dependence on τ_E,th ∼ A^0.15 measured in JET-ILW L-modes, in broad agreement with the isotope mass dependence of the multi-tokamak ITER-97L scaling (A^0.2).

Despite much progress in theoretical work and a growing body of detailed experimental measurements on L-H transition dynamics in many tokamaks, a first principles physics model of the L-H transition with predictive capability is still missing. Extrapolations to future experiments therefore still rely on empirical scalings, such as the 2008 ITPA scaling: ${{P}}_{{\rm{L}} \mbox{-} {\rm{H}},08}\sim {{{n}}_{{\rm{e}}}}^{0.717}\,{{{B}}_{{\rm{T}}}}^{0.803}\,{{S}}^{0.941}$ [35], with n_e [10²⁰ m⁻³] the line averaged density, B_T [T] the toroidal magnetic field and S [m²] the plasma surface area. This scaling carries at least a factor of two uncertainty, due to additional dependencies of P_L-H not included in the scaling.

It has been known for several decades that the H-mode power threshold has a favourable dependence on isotope mass: P_L-H ∼ 1/A, as measured in many tokamak experiments in hydrogen and deuterium plasmas, see e.g. [9, 36–38] as well as in hydrogen, deuterium, tritium and deuterium–tritium in JET [39]. These observations typically pertain to the high density branch of P_L-H, namely where P_L-H increases with density. This result projects to H-mode access in a fusion reactor with 50:50 deuterium–tritium mixture with about 20% less power than in a deuterium plasma.

Recent isotope experiments on the L-H threshold in JET-ILW in hydrogen and deuterium plasmas, with ICRH heating, have highlighted a stronger isotope dependence of P_L-H in the low density branch [40], of order a factor of three, as shown in figure 11, which may be indicative of edge L-mode turbulence of different nature in the low and high density branch [34, 41]. The JET-ILW measurements differ from results reported from AUG, where the ∼1/A isotope dependence of P_L-H was observed both in the low and in the high density branch [38]. The difference in H-mode threshold between hydrogen and deuterium plasmas in JET-ILW is much larger for NBI-heated plasmas, reaching up to a factor of 4. It is currently being investigated whether this is due to an effect of NB torque on P_L-H, as observed on DIII-D [37], or to the partition between ion and electron energy fluxes at the L-H transition, as proposed in [42].

Zoom In Zoom Out Reset image size

L-H transitions in 50:50 hydrogen–deuterium mixtures in JET-ILW resulted in P_L-H values intermediate between the pure hydrogen and deuterium cases, which had isotope purity >98%–99%, as measured by Balmer-α spectroscopy at the plasma edge. The largest variation in P_L-H when varying the isotope ratio is observed at high and low H/(H + D) ratios, while little variation is found for 0.2 < H/(H + D) < 0.8 [40]. This was an unexpected result and remains as yet unexplained. The strong variation in P_L-H at small levels of hydrogen in deuterium and of deuterium in hydrogen plasmas may indicate a role of ion–ion collisions, where the product of the hydrogenic isotope densities gives a parabolic dependence for the collision rate [40]. An important implication of this finding is that it is imperative to achieve the highest isotope purity (>98%) in experiments when studying hydrogen versus deuterium plasmas, in order to prevent spurious conclusions from the experimental observations.

A similar, nonlinear dependence on isotope mixtures is observed for H-mode energy confinement: the thermal stored energy, W_th, is lower in hydrogen than in deuterium (see section 4) and the largest variations in W_th between the two species is observed for 0 < H/(H + D) < 0.1 and 0.9 < H/(H + D) < 1.0 [43].

The isotope dependence of P_L-H remains as yet unexplained. Here we put forward the hypothesis, based on the JET-ILW observations, that, at least in the high density branch, the favourable isotope effect on P_L-H is a consequence of the favourable isotope effect on transport at the L-mode edge. As discussed in section 2, a positive isotope effect is found in L-mode at the plasma edge, both for particle and heat transport and the ETB forms in the plasma region where the isotope effect is found in L-mode. In the high density branch, at the L-H transition, the same edge kinetic profiles (n_e, T_e) are measured in hydrogen and deuterium when P_loss(H) ∼ 2 × P_loss(D) [41], as shown in figure 12. As evidenced by a large body of experimental and theoretical work, the L-H transition is triggered when the E_r × B shearing rate is of order of the growth rate of the L-mode edge turbulence, γ_E = ∇E_r/B ∼ γ_turb. In addition, considering that E_r = ∇_pi/(Z e n_i) – v_i × B does not depend on isotope mass (except for mass effects in the coefficients for the neoclassical poloidal rotation), then the larger P_L-H measured in hydrogen than in deuterium may only be a consequence of the larger input power required to suppress the higher levels of L-mode edge turbulence in hydrogen, via the dominant diamagnetic shear, and not of an isotope effect on the L-H transition per se.

Zoom In Zoom Out Reset image size

Isotope experiments in JET with the C wall (JET-C) in hydrogen, deuterium, tritium and deuterium–tritium showed for ELMy H-modes no isotope dependence of thermal energy confinement time τ_E,th ∼ A^0.03±0.1, due to strong, positive isotope dependence at the pedestal, W_ped = 0.45 B^1.2 (A/2)^0.96 and weak, negative isotope dependence in the plasma core, τ_th,core ∼ A^−0.16±0.1 [14] (the latter is in approximate agreement with the gyro-Bohm dependence, τ_gB ∼ A^−0.5). The strong isotope dependence at the pedestal was also supported by ELM losses measurements: in the type I ELMy H-mode regime larger ELMs were observed in tritium than in deuterium in discharges at 2.7 T/2.6 MA at the same input power of ∼10 MW of NBI heating [44], while hydrogen discharges did not have type I ELMs due to the limited H-NBI power (and larger H-mode threshold). These experiments were, however, hampered by diagnostic resolution at the plasma edge.

A rather different picture, on the other hand, has been put forward by hydrogen and deuterium isotope studies in JT-60U, where, as in other tokamaks, τ_E,th is found to increase with isotope mass, but, unlike in JET-C, it is concluded that the favourable isotope effect stems from the plasma core, with no direct isotope effect in the pedestal region [15–17]. Comparative hydrogen and deuterium type I ELMy H-modes were obtained with matching thermal stored energy, W_th, by raising P_NBI in hydrogen to match the electron and ion temperature profiles, while the n_e profile was automatically matched in the two species and the toroidal rotation profile was similar, although co-current in deuterium and ctr-current in hydrogen, as shown in figure 13. It is thus likely that the radial electric field would be somewhat different in the two cases. The effective ion charge Z_eff increased for deuterium up to 2.4 at the maximum input power, while it remained approximately constant with power at Z_eff = 1.5 in hydrogen. The two H-modes in hydrogen and deuterium at same W_th had similar Z_eff ∼ 1.5 [17]. The larger ion heat flux, Q_i, in hydrogen, at same W_th, resulted in larger χ_i in hydrogen than in deuterium over the entire plasma minor radius [17], see figure 13. The same pedestal structure was measured in the edge T_i profile in hydrogen and deuterium (same pedestal height, width and gradient) at matched normalized pressure β_pol in the two ion species [15], but differing ρ*_pol, which led to the conclusion that no direct isotope effect is observed in the pedestal region in JT-60U. The poloidal beta is defined as β_pol = p/(〈B_p〉²/2μ₀), with p the plasma kinetic pressure and 〈B_p〉 the average poloidal magnetic field. It is however always the case that when hydrogen and deuterium ELMy H-modes are compared at the same I_P/B_T and same input power a larger pedestal pressure is observed in deuterium than in hydrogen [17]. This is explained by: (i) the measurement of a positive isotope effect of core ion heat transport [16, 17], whereby profile stiffness is observed in both hydrogen and deuterium H-modes, but with a smaller ${L}_{{T}_{{\rm{i}}}}$ for the heavier isotope mass, see figure 14; (ii) the pedestal pressure increasing with total β_pol in both hydrogen and deuterium; (iii) deuterium plasmas achieving larger β_pol, due to reduction in core ion heat transport compared to hydrogen (see (i)) and larger fast ion energy, as the fast ion slowing down time, τ_S, increases with isotope mass: τ_S ∼ A^1/2 T^3/2; (iv) extension of the pedestal stability boundary with larger β_pol for deuterium than for hydrogen [17].

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

In recent isotope experiments in JET-ILW in hydrogen and deuterium plasmas, Type I ELMy H-modes could be obtained in hydrogen, both at 1 MA/1 T and at 1.4 MA/1.7 T. At 1 MA/1 T, H-mode discharges were obtained with H-NBI and D-NBI heating only, respectively. We note that the maximum power delivered by H-NBI was 10 MW in these experiments. Therefore, for the 1.4 MA/1.7 T dataset the discharges had NBI heating only in the deuterium plasmas and either H-NBI only (up to 10 MW) or a combination of 9–10 MW H-NBI and 2.5–5.5 MW ICRH (51 MHz, H majority, 2nd harmonic) heating in the hydrogen plasmas in order to maximize the available auxiliary heating at this magnetic field and hence obtain enough data for the type I ELMy H-mode dataset at B_T = 1.7 T. An additional deuterium dataset at 1.7 MA/1.7 T, with D-NBI only, was also obtained. H-NBI operation was at lower NB voltage (70 kV) than with D-NBI (90–100 kV). In the hydrogen H-modes at 1.4 MA 1.7 T with NBI + ICRH heating, the fraction of heating (as computed by the ICRH deposition code PION [45] including ICRH + NBI synergy [46]) delivered to ions is slightly reduced and the fraction of power delivered to electrons is slightly increased compared to the case with NBI-only heating (as computed by the neutral beam deposition code PENCIL [29]). However, the relative variation is of order 5%–10%, depending on the discharge, and thus unlikely to compromise the conclusions drawn on the isotope scaling of energy confinement presented below.

As the input power is increased in experiment after the L-H transition, the plasma first enters the type III ELMy H-mode, at low input power above P_L-H and high ELM frequency, f_ELM. Subsequently, as the auxiliary heating is increased further, the type I ELMy H-mode regime is accessed, at higher β and confinement. The power threshold for the transition from type III to type I ELMs, characterized phenomenologically by f_ELM decreasing with power for type III ELMs and f_ELM increasing with power for type I ELMs [47], doubles from deuterium to hydrogen, as shown in figure 15. This is in agreement with findings in JET-C [39] and consistent with the favourable 1/A scaling found for the H-mode threshold in the high density branch. The variation in f_ELM at a given P_loss, shown in figure 15, is a consequence of the varying injected gas rates in the hydrogen and deuterium power and gas scan experiments.

Zoom In Zoom Out Reset image size

The thermal energy confinement is strongly reduced in hydrogen H-modes, in contradiction of the gyro-Bohm scaling, resulting in a stronger, favourable dependence with isotope mass, τ_E,th ∼ A^0.4, than for the IPB98(y, 2) scaling τ_{E,th,IPB98(y,2)} = 5.62 × 10⁻² A^0.19 ${{{I}}_{{\rm{P}}}}^{0.93}\,{{{B}}_{{\rm{T}}}}^{0.15}$ R^1.97 ε^0.58 ${{{k}}_{{\rm{a}}}}^{0.78}\,{{{n}}_{{\rm{e}}}}^{0.41}\,{{{P}}_{{\rm{abs}}}}^{-0.69}$ (with R the major radius [m], ε the inverse aspect ratio and k_a the plasma elongation) [48]. The hydrogen and deuterium JET-ILW dataset was obtained in plasmas with the same shape, therefore regressions of the τ_E,th data were performed with respect to I_P, B_T, P_abs and A, n_e and gas rate (or f_ELM). Although isotope mass, plasma density and gas rate (or ELM frequency) are correlated—in addition n_e and I_p are also correlated—the A^0.4 dependence of τ_E,th on the isotope mass is robust against which combination of this variables is chosen in the regression of the data.

A distinct feature of hydrogen plasmas is the lower plasma density and the fact that the density can be varied with injected gas rate, unlike in the deuterium H-modes, where the density is weakly affected by gas injection [49, 50]. This indicates degraded particle confinement with decreasing isotope mass, as observed in L-mode. In H-modes at the same input power and gas injection rate, a larger pedestal pressure is observed in deuterium than in hydrogen. The difference in pedestal pressure, and therefore total pressure, is primarily due to the larger pedestal density in deuterium than in hydrogen, while the difference in temperature profiles is more modest, as shown in figure 16 for two type I ELMy H-modes at 1.4 MA/1.7 T in H and D with P_NBI = 10 MW.

Zoom In Zoom Out Reset image size

H-modes in hydrogen and deuterium at the same thermal stored energy, W_th, do not have matched density and temperature profiles: in hydrogen the lower density is compensated by higher temperature. This contrasts JT-60U experiments, where the density and temperature profiles were matched for the two species when W_th was matched by raising the H-NBI heating, as discussed above. This difference suggests different particle transport in the two tokamaks and warrants further investigations in order to enable extrapolation of present-day experimental results to ITER and beyond.

Across the isotope dataset obtained in JET-ILW, the electron pedestal pressure, p_e,PED, is larger in deuterium than in hydrogen type I ELMy H-modes, as shown in figure 17 for the 1.4 MA/1.7 T data. Similarly, also the ion pedestal pressure is larger in deuterium than in hydrogen, as T_i,PED ∼ T_e,PED consistently throughout the dataset for both ion species and Z_eff (primarily determined by the Be intrinsic impurity) varies between 1.1 and 1.4 in deuterium and between 1.1 and 1.7 in hydrogen, depending on I_P/B_T and input power. Interestingly, hydrogen and deuterium pedestals are found to react differently to variations in gas rate and input power: whereas hydrogen pedestals evolve along the same isobar, exchanging temperature for density as the gas rate is increased and/or the input power is reduced, in deuterium it is primarily T_e,PED to be reduced with increasing gas rate, while n_e,PED is little affected by varying levels of injected gas. An increase in input power in deuterium—at a given gas rate—leads to an increase in T_e,PED and a decrease in n_e,PED (via the increase in f_ELM for type I ELMs). Thus, especially at the lowest gas rate in the scan (see figure 17) the pedestal pressure of deuterium plasmas increases with power, see e.g. [49].

Zoom In Zoom Out Reset image size

As discussed for the L-mode dataset in section 2, also the H-modes of this study have relatively high collisionality, leading to strong electron–ion coupling, and the core T_i profile data have large uncertainties (∼20%). Therefore, in these conditions, a meaningful species-resolved transport analysis is unfortunately not possible. Across the entire type I ELMy H-mode dataset, both in hydrogen and in deuterium, the transport analysis has been carried out based on the electron kinetic profiles. As for the L-mode discharges, also in the H-mode dataset the sawtooth inversion radius is at ρ_tor = 0.22–0.26, depending on the individual discharge, where q_ψ ∼ 1. In the transport region, the inverse normalized temperature gradient length is measured for electrons, ${R/L}_{{T}_{{\rm{e}}}}\cong {R/L}_{{T}_{{\rm{i}}}}$, for both isotope species. Also in type I ELMy H-modes the ions and/or the electrons are stiff. In this sense, there is no difference between L-mode and H-mode core behaviour. Figure 18 shows the one fluid, effective heat diffusivity, χ_eff, at mid radius (ρ_tor = 0.5) for the hydrogen and deuterium type I ELMy H-mode dataset. There is little overlap in χ_eff in hydrogen and deuterium H-modes, due to the large difference in auxiliary heating required to access type I ELMy H-mode in the two species and due to the plasma density being systematically lower in hydrogen than in deuterium.

Zoom In Zoom Out Reset image size

In contrast to JT-60U, the combination of larger pedestal pressure in deuterium and constant R/L_T for both plasma species suggests for JET-ILW type I ELMy H-modes that the favourable isotope effect on τ_E,th propagates from the pedestal through the plasma core via constant critical temperature gradient.

Possible contributing factors to the isotope effect observed in the pedestal region could be due to the isotope mass dependence in the edge bootstrap current, which plays a key role in the H-mode edge stability, and/or in the micro-instabilities that drive the residual edge transport levels inter-ELMs. Both these aspects will be addressed in future studies.

Recent isotope experiments in hydrogen and deuterium in JET-ILW have yielded a wealth of new results concerning the isotope effects on L-mode and H-mode confinement and transport and on the L-H power threshold.

In L-mode, τ_E,th is found to scale weakly with isotope mass, τ_E,th ∼ A^0.15, in contradiction of the gyro-Bohm (and the Bohm) scaling, and in analogy with findings from most tokamak L-mode experiments, as expressed by the ITER-97L scaling, τ_E,th ∼ A^0.2. A negligible isotope effect on heat transport is found in the plasma core and a strong, positive isotope effect is found at the plasma edge, both inside and outside the LCFS, on energy and particle transport. Larger edge density fluctuations in hydrogen than in deuterium at the same input power corroborate the larger particle transport coefficient obtained by interpretative EDGE2D/EIRENE modelling of upstream edge density profiles. The favourable 1/A scaling of P_L-H is confirmed in the high density branch in ICRH heated plasmas, whereas a stronger isotope dependence is observed in the low density branch, which is suggestive of L-mode edge turbulence of different nature in the two density branches of P_L-H. It is suggested that the isotope effect on P_L-H in the high density branch, P_L-H ∼ 1/A, may be linked to the favourable isotope effect on transport at the L-mode edge. In turn, the isotope effect on P_L-H is likely to propagate to the isotope effect measured on the power threshold for type III to type I ELMs, P_{typeIII-typeI} ∼ 1/A, and to the strong, favourable isotope effect on thermal energy confinement time measured in type I ELMy H-modes, τ_E,th ∼ A^0.4. This is in contradiction of the gyro-Bohm scaling, which would predict τ_E,th ∼ A^−0.5. On the other hand, in deuterium plasmas the dimensionless energy confinement time follows the gyro-Bohm scaling, Ωτ_Ε ∼ ρ*⁻³ [51], in agreement with experimental results from several other tokamaks, including JET-C, see e.g. [2] and references therein (where Ω = eB/A is the cyclotron frequency, τ_E ∼ a²/χ and χ/χ_B ∼ ρ*, with χ_B = T/ZeB the Bohm diffusivity [2]). As previously shown in JET-C, and in contrast to JT-60U findings, in JET-ILW the isotope effect is found at the pedestal and it propagates through the plasma core via constant critical temperature gradient in conditions where ions and electrons are strongly coupled and/or the ion heat transport is stiff. On the other hand, it is possible that in plasmas with weak ion–electron coupling the ion transport could be different and, consequently, an additional isotope effect on core ion transport may arise in such conditions.

Overall, a favourable isotope effect on confinement and P_L-H is observed in tokamak experiments, which projects favourably to a future reactor that will operate in deuterium–tritium mixtures. The isotope dependence observed in experiment is in contradiction of the gyro-Bohm scaling, both in the L-mode and in the H-mode regime, while gyro-Bohm scaling of the thermal energy confinement is consistently confirmed in deuterium in H-mode experiments from many tokamaks. This discrepancy still presents a main challenge to the understanding of plasma turbulent transport and, despite many recent advances in theory, needs to be investigated further with extensive isotope studies across multiple tokamaks.

Indeed, recent numerical studies of an ITER modelling scenario, with nonlinear gyrokinetic simulations, have shown that the core plasma micro-turbulence can depend on the isotope mass via multi-scale nonlinear effects involving zonal flows, electromagnetic effects and E × B shear [26]. The local, gyro-Bohm scaling of plasma micro-turbulence is thus counteracted by the above effects, leading to isotope effects on heat transport on longer scale lengths [26, 52]. On the other hand, these theoretical analyses are highly relevant to the interpretation of isotope effects observed in high beta plasmas, such as the hot ion H-modes of TFTR [53] or future hybrid deuterium–tritium plasmas in JET, where observation of these effects likely requires that ions are thermally decoupled from the electrons, unlike the JET experiments reported in this paper. Further theoretical work is required to understand the strong isotope effect measured at the pedestal of JET-ILW H-modes, or the isotope effect found in Ohmic and low beta JET-ILW L-modes.

Different results and interpretations of the isotope effects obtained across tokamak experiments, e.g. the contrasting results in H-mode from JT-60U and JET-C and JET-ILW discussed in section 4, may be related to a variety of factors, such as different plasma heating methods, the role of T_i/T_e on ion-scale core turbulence, the impact of E × B flow shear stabilization in the plasma core, high beta stabilization of ion-scale core turbulence, the impact of impurities. Such studies will be undertaken in JET in the upcoming isotope campaigns leading to deuterium–tritium experiments with the ITER-like wall, in support of ITER research.

Useful discussions with Dr M Mantsinen are gratefully acknowledged.

This work was carried out within the framework of the EUROfusion Consortium and received funding from the Euratom research and training programme 2014–2018 under grant agreement No. 633053 and from the RCUK Energy Programme grant No. EP/P012450/1. The views and opinions expressed herein do not necessarily reflect those of the European Commission. To obtain further information on the data and models underlying this paper please contact PublicationsManager@ukaea.uk.