PFC/JA-94-28 Thermal Bifurcation of SOL Plasma and Divertor Detachment S. I. Krasheninnikovl, D. J. Sigmar, T. K. Soboleva 1 ,2 , P. J. Catto MIT Plasma Fusion Center Cambridge, Massachusetts 02139 USA September 1994 1Kurchatov Institute of 2 Instituto de Ciencias Atomic Energy, Moscow, Russia Nucleares, UNAM, Mexico DF, Mexico This work was supported by the US Department of Energy under contract DE-FG02-91ER-54109. Reproduction, translation, publication, use, and disposal, in whole or in part, by or for the US Government is permitted. Submitted for publication in: Physics of Plasmas Thermal Bifurcation of SOL Plasma and Divertor Detachment S. I. Krasheninnikov a), D.J. Sigmar, T.K. Soboleva a,b), and P. J. Catto MIT Plasma Fusion Center, Cambridge, MA 02139, USA a) Kurchatov Institute of Atomic Energy, Moscow, Russia b) Instituto de Ciencias Nucleares, UNAM, Mexico D.F., Mexico Abstract Models to investigate the main features of plasma neutral interaction in the recycling region are developed for the two opposite extremes of fluid and Knudsen neutrals. Both neutral models show that a reduction of the heat flux into the hydrogen recycling region below some critical value leads to bifurcation of the plasma parameters near the target. This bifurcation causes behavior in the SOL which is in agreement with all of the main features of detached divertor regimes in current tokamak experiments: i) decrease of the plasma temperature near the target to about 1 eV, ii) plasma pressure drop in the recycling region, and iii) strong decrease of the target heat load and plasma flux onto the target. It is also shown that in the Knudsen limit, the neutral density in the divertor region can not exceed some maximum density, which is of the order of 1-2 x 1013 for current experiments. 1 I. Introduction Recent experiments on most diverted tokamaks have demonstrated so called detached divertor regimes (See Ref. 1 and the references therein). These regimes are characterized by 1) high energy radiation losses from the scrape-off layer (SOL) region; 2) low plasma temperature near the divertor plates; 3) strong decrease of the plasma energy and particle fluxes onto the plates; and 4) strong plasma pressure drop along magnetic field lines in the divertor volume. One of the main attractions of these regimes from the ITER divertor design point of view is the very low heat loads on the divertor plates. However the physics of these regimes is not yet understood and, therefore, it is not clear that detached divertor regimes can be considered as a basis for ITER divertor design. In this work we improve detachment modelling by removing the constant plasma pressure assumption and retaining the neutral energy flux into the target. To get a rough estimate of the dependence of the plasma temperature in front of the target, Td, on the heat flux entering the recycling region, q ,, for a high recycling divertor one can use the global energy balance equation of the recycling region which for sufficiently high temperature can be written in the form (see for example Ref. 2, 3) grc - ndCd sin V (yTd + E) , (1) Here y is the heat transmission factor, nd is the target plasma density, Cd =(Td /M)1/ 2 is the sound speed at the target plasma temperature, M is the ion mass, V is the angle between the target and magnetic field line, and El is the ionization cost of the incident charged particles due to 2 ionization of the neutral outflux and hydrogen radiation energy losses (E 1 > I, with I = 13.6 eV is the hydrogen ionization potential). The factor id = ndCd sin V on the right hand side of Eq. (1) describes plasma specific flux on the target, which for the high recycling regime equals the neutral outflux from the target. The first term in the parenthesis describes the heat flux transferred to the target due to kinetic energy of the charged particles, while the second one describes the energy loss due to ionization of and radiation by the neutral outflux. The total heat flux reaching the target, qd, (qd s qrc) must account for the energy I released by recombinating ion and electron and the portion 6EI of the radiated energy reaching the target. It can be written as qd - ndCd sin V(yTd + I + 6E 1 ) (2) , where 6EI the part of the radiated energy coming to the target. Re-writing the Eqs. (1), ( 2) in terms of target plasma pressure, Pd=ndTd, one gets CrC- Pd sin V (yTd + El) qd - Pd sin V (yTd MTd1/2 U PdGrc (Td), + I + bE)/ (MTd1/2 ) PdGd (Td) - (3) (4) It is easy to see that for these approximate forms the functions G(...)( Td) have a minima, G(...), min , at Td=Tm-5 eV increasing at lower and higher temperatures, but it is known that at small Td < 5 eV these forms must be replaced with more accurate expressions. Before deriving 3 these expressions we discuss why they are needed. If one simply assumes that the plasma pressure near the target Pd can be balanced by upstream plasma pressure, Pu , then for Pu=Pd Eq. (3) can only be satisfied if qrc z qmin - PuGrc,min (5) . If Eq. (3) is satisfied then according to relation (4) qd 2 PuGm, d , (6) and the asymptotic form of jd for Td -+ 0 is jd - Pu/ M Td/2 cx Td-1/2 (7) Note, that supersonic plasma flow at the target does not change the sense of inequality (5). In this case one has to replace temperature Td in Eq. (3) by MV 2 , where V is plasma flow velocity along magnetic field line at the target, and use the momentum balance Pu - nd w 2. Sometimes, the inequalities (5), (6) are interpreted as restrictions on the values of the heat fluxes grc and qd at the prescribed upstream plasma pressure. But these interpretations are not correct. Indeed, let us consider the global energy balance equations (1), (3) of the recycling region. For the stability of this balance the energy sink (right hand sides of these equations) as a function of temperature has to have positive slope so it will decrease with the decreasing target plasma temperature. That means that 4 the solutions (3), (4) are not stable or accessible for Td<Tm 2, 3. To get stable solutions in low temperature region one has to consider more sophisticated models of the SOL plasma- neutral interaction 2, 3 to be discussed in the next chapter (section). In this book we will develop physical model of tokamak SOL and show that the reduction of the heat flux into the hydrogen recycling region below the critical value, qcrit - qmin, automatically leads to a decrease in the plasma particle flux onto the target and plasma pressure near the target, increase of the neutral gas density in the divertor volume, and expansion of the recycling region. The physical mechanisms responsible for the decrease of the plasma flux onto the target and plasma pressure drop along magnetic field lines in the divertor volume (depending on the SOL plasma parameters and divertor geometry) are neutral gas viscosity and pressure influence on plasma flow 2, 3, and friction between the plasma flowing onto the target and the neutral gas scattered by the sidewalls 4. Both the increase of the neutral gas density in the divertor volume and expansion of the recycling region can be the reasons for the increase of the impurity radiation losses from the SOL region and the decrease of the heat flux into hydrogen recycling region. Taken together all these can lead to thermal bifurcation of the SOL plasma parameters and detachment of divertor plasma. The simplified physical model and geometry of the problem are considered in Section II. Plasma/neutral recycling for the different limits of neutral gas flow in the recycling region are considered in Section III (fluid approximation) and in Section IV (Knudsen approximation). The results of the analysis of these limits are discussed in Section V. The main conclusions are summarized in Section VI. 5 II. Model of tokamak SOL plasma We will consider simple slab divertor geometry (see Fig. 1) where the axis x, y and z are the "radial", "poloidal" and "toroidal" coordinates respectively. We will assume that magnetic field is in the y, z plane and b=By/B<<1, where By and B are the poloidal and total magnetic fields strengths respectively. We consider regimes with high recycling of the plasma in the divertor e. g. Fd >> Is , where rd is the particle fluxes on the divertor plates and Is is the particle flux from the bulk plasma to the (SOL) and will assume that the particle flux on the first wall is negligible. In this case it is possible to distinguish two main regions in tokamak edge plasma (see Fig. 2). The first one, labelled the H-region, is the region of neutral hydrogen recycling. It characterized by relatively low plasma temperature, T, and high neutral gas density, NH, which can even become comparable to the plasma density, nH. The second region is the remainder of the edge plasma and denoted as the E-region. The plasma temperature there is higher than in the H-region and the neutral density NE is much lower than the plasma density nE . The main physical processes in the E-region are the energy transport across and along the magnetic field lines, and the energy radiation losses mainly due to the impurity radiation. Thus the processes in the E-region determine the energy flux coming to the hydrogen recycling region, Qrc, and the width of the SOL Ap (where Qrc = Sqrc , S = 4nRAp, and R is the tokamak major radius). To simplify the problem we will assume that plasma/neutral parameters only depend on the poloidal coordinate and the radial scale length, Ap , of the plasma/neutral parameters is fixed or can be found from scaling laws. 6 Assuming that plasma convection does not effect the plasma momentum and heat flux (a good approximation for the high recycling divertor plasma), in the one dimensional (D) case one can write down the momentum and energy balance equations in the E-region in the form (8) P(y) - Pu - const. , b2 dy e d -T dy Wrad, (9) where P is the total plasma pressure, ie is the electron heat conduction coefficient along the magnetic field lines, T is the plasma temperature (we assume that electron and ion temperatures are equal), and Wrad is the specific energy radiation losses due to impurities. Note that in spite of a relatively low neutral density NE < nE, neutrals can strongly affect energy balance in the E-region due to a) an increase of the impurity radiation from hydrogen-impurity charge-exchange; b) an increase in wall sputtering by charge-exchange neutrals; and c) an increase in hydrogen charge-exchange energy losses. The most significant impact of the neutrals on the energy balance is due to a) and c) which one expects to occur between the H-region and X-point and in the vicinity of the X-point where the neutral density is still high enough and plasma temperature is too low. The main physical processes in the H-region are plasma/neutral recycling and energy losses due to hydrogen radiation, charge exchange, and elastic collisions. Since we consider regimes with Fd >> Fs , the plasma flux on the target Fd as well as plasma/neutral parameters near the target are determined by plasma - neutral interaction in H-region. We 7 will see below that the E-region affects the flux rd only by the plasma pressure Pu and heat flux Qrc. We will assume that the plasma flow can be treated by a fluid approximation (e.g. mean free path of the charged particles, Xe, i, smaller than the longitudinal scale length, 1). Then, there are two limits in which the description of neutral gas flow can be drastically simplified: a) the short mean free path (with respect to neutral-ion collision, we assume that neutral density is low enough that neutral-neutral collisions can be ignored) neutral fluid approximation 2, 3, 7, and b) the Knudsen approximation of long mean free path neutral gas flow. The fluid limit can be applied for relatively high plasma density in the H-region, when neutral mean free path, XN, is smaller than AP, while the Knudsen approximation describes the opposite extreme XN >> ApIn the following sections/chapters we consider plasma/neutral recycling for these two limits of neutral particle dynamics assuming 100% particle recycling at the target and Fs=0. We will also assume that plasma parameters in the E-region and the heat flux Qrc into recycling region are prescribed, and that the plasma-neutral mixture consist of only one kind of isotope. III. Neutral recycling (fluid approximation) When both the plasma and neutrals can be treated in a fluid approximation it is convenient to consider the equations for the averaged plasma-neutral velocity, i;, temperature, T (we will assume here that the electron, ion and neutral temperatures equal), and relative velocities of plasma and neutral species. For simplicity, we assume here that Ap = A. Then (see for example Ref. 5 and the references therein) one has 8 N)VyVii)+ b dy - (M(n y dqy_dy dy ... where i V- + (V. -EIKI(T)nN, = NVy (10) (11) (12) - K, (T)n N, Vi )-o, (d9/dy) V(NT) MNnKiN (T) + (13) n + N), (14) nVi, y , (nVP + N9N n (Vp) and N (VN) are the plasma and neutral densities (velocities), P is the total pressure, P - (2n + N)T; H is the viscosity tensor, qy is the total heat flux along y coordinate, (. ..) I is the parallel component of (...) vector, and M is the ion/neutral mass. The ionization, hydrogen radiation/ionization energy loss, and ion-neutral collision (charge exchange and elastic) rate constants are denoted by KR(T), KI(T), and Kiv(T), respectively, and EI(T), ERKR/KI+I is the neutral ionization cost with ER the characteristic energy loss. We omit here the influence of the thermal force on the relative velocities of plasma and neutral species and the heat transport. Since we consider complete particle recycling at the target and F s =0, 9 the net mass flow along y coordinate equals to zero everywhere in the SOL (15) Vy= 0 , and therefore Eqs. (10), (13) can be written as dP b - + (v - fi) I 0, (16a) -v(NT) MNnKiN (16b) dy VN - Vi Assuming that plasma velocity Vp is directed along the magnetic field, from Eq. (14) one has (n + N)Vy - b nVi, 1 + N(bVN,11 + b'VN, y') m 0, (17) where the y' coordinate lies in the y, z plane and is directed perpendicular to the magnetic field, and b' - (1 - b2) . From Eqs. (15), (16b) and (17) one finds j bnVi, I 'M(N d(NT. + n), N dy (18) Since there is no mass flow in the y direction, the heat flux qy is only determined by plasma and neutral conduction 10 NT qy=-(Ke(T)b2+ dT dT _ MnNyN where Ke (T) is the electron heat conduction coefficient along the magnetic field, c, is the formfactor depending on ion-neutral collision cross section features and K is the global heat conduction coefficient. Substituting expressions (18) and (19) into Eqs. (11) and (12) one has d dy 1 d(N -K ()nN() K(T)nN M(N + n)KiN dy(NT) d NT + cX MN dy IMnVjN) d{(Ke(T)b2 dTl I dy - EK(T)n N, (20) (21) which, together with Eq. (16a), describe the plasma-neutral interaction in the fluid approximation. To close system of the equations (16a), (20) and (21) we impose the following boundary conditions qy (y -- 0) - -Crc (22) , qy (y -- 0) - -Td -Ypbnd(Vi, II)d N(y - o) YNNdCd} (23) (24) - 0 , Vi, I (Y - 0) - -C + (25) , 11 where (...)d is the (...) value near the target (y=O), Cd - (Td/M) 1 / 2 , a s 1, Yp and YN are the plasma and neutral gas heat transmission coefficients, and qrc is the specific heat flux coming to the H-region. Notice that the heat flux on target, Eq. (23) accounts for the neutral energy flux to the target. It is possible to show that the influence of the viscosity (mainly neutral viscosity) on the total pressure variation is relatively weak (oc a2 s 1) and does not make s significant difference. Therefore from Eq. (16a) we have (2n + N) T = (26) Pu. To simplify analytic solution of Eqs. (20), (21), (26) we replace the exact temperature dependence of the neutral ionization cost E, and ionization K, and ion-neutral collision KiN rate constants by the following dependencies KI(T) - i (T/Tk) 2 0(T - TI) , KiN (T) = KiN (T/Tk) (27) , El(T) . El - const. where e(x) - 1 for x>0 and e(x) - 0 for x<O; K, and KiN are normalization constants, and Tk and T, are characteristic temperatures (Tk -10 eV>> T - few eV). 12 Let us consider first of all the case Td > TI. Then, from Eq. (20), (26) and (27) one gets .d_ 2 dpN. dy 1 + pN dy p (1 - pN) (28) /AF where pN - NT/PU, and AF - MKiN, d KI, d 2Td I/Pu2 (29) KiN, d = KiN (Td), KI, d a KI (Td)- Multiplying Eq. (20) by El, subtracting it from Eq. (21) and introducing (19) one finds dT dY 2E, Td dpN M(1 + PN )KIN,d dy qrc , (30) where boundary conditions (22) and (24) are taken into account. The physical sense of the Eq. (30) becomes more pronounced when (18) and (19) are employed to rewrite it in terms of heat and particle fluxes -(qy + E, ji) - qr (31) . One can see that Eq. (30) or (31) describe energy balance where "internal" energy transport owing to radiation/ionization is taken into account. Returning to Eq. (28), multiplying by (2/(1 + pN ))(dpN /dy), and 13 integrating once using boundary condition (24) one finds for the root of interest that dpN = - AF dy (32) F(PN) where F(pN = {2[pN - ln(1 + pN - PN 2/2 1/2 (33) As a result, (18) becomes (34) ji - - Pu Cd d(Td)F(PN )/Td, here Od (Td ) = (2K (Td )/KiN (Td ))1/2 1 = jd/Tk) 2KI iN 11/2 Using boundary condition (25) one can rewrite expression (18) for the plasma flux onto the target in terms of Td and pN to find ji, d - ab (1 - PN, d)Pu Cd C/2Td) . (35) Then, from Eqs. (23), (31), (34) and (35) one gets the algebraic system of the equations determining Td and PN, d (1 - PN,d)/F(pN,d) - 2cd(Td)/(ab), (36) qrc - Pu Cd ab(yp (37) + EI/Td)( 1 - PN,d)/2 + YN PN,dt - 14 Notice that (37) reduces to (1) only if PN,d = 0 (and a = 1). The left hand side of Eq. (36) is the monotonicaly decreasing function of PN, d . It varies from oo to 0, when PN, d changes from 0 to 1, so that the Eq. (36) always has the solution. Therefore, Eq. (36) can be solved for pN, d as a function of Td and the result inserted into Eqs. (35) and(37) one can rewrite the right hand sides of the resultant equations as the functions of Td , that is . ji,d - -Pu J(Td) qrc - Pu G(Td), (38) One can get the approximate solution of Eq. (36) analytically, by replacing the function F(pNd) by the first term of its Taylor's expansion to find PN, d (39) {1 + I0d(Td )ab). Then G(Td) - Cd J(Td) - 7 Yp + j Cd Gd(Td) 1 1j+ 1 + + YN Od(Td)P . ' Gd T (40) (41) Let us consider the behavior of Eqs. (39), (40) and (41). The ratio of ionization to ion-neutral collision rate constants is of the order of unity for 15 the high temperatures and very small for low plasma temperatures. Then, for high temperatures Td , such that Gd (Td )/(a b) >> 1 one has G(Td) - Cd YP + J(Td) - Cd ' , (42) PN,d - (ab)/[od(Td)] << 1. For low temperatures Td , bit with Td > E, Od (Td ) G(Td) -EICd Gd(Td EIJ(Td) - const. a b) << 1 one has PN, d - 1 (43) 2Td where (27) is employed to find ad x Td 1 / 2 . From the expressions (42) and(43) one can see that the relations (3) and (7) can only be applied when plasma temperature near the target is relatively high and neutral pressure does not significantly affect plasma flow. Only in the high temperature constant plasma pressure limit will the function G have a parabolic dependence on plasma temperature Td with a plasma flux on the target that increases with decreasing Td . When temperature Td becomes low (Ti < Td < Tm) the relations (3), (7) and (42) can no longer be applied. From Eq. (43) one then sees that for low Td both functions G(Td) and J(Td) are essentially constants and the neutral pressure near the target approximately equals upstream plasma pressure. Next, we consider now the qualitative behavior of the functions G(Td), J(Td), and PN, d(Td) for very low plasma temperatures near the 16 target, such that Td < TI. For these very low temperatures (T < T1 ) the value of plasma/neutral flux on the target i,d is not determined by neutral ionization, but rather is determined by neutral diffusion into the region with high plasma temperature (T > T1 ) where neutral ionization can take place. To estimate the size of jid as PN,d - 1 (recall Eqs. (35) and (43)) we return to (18) and write Ji, d - -Tk 2 yI (44) MKiN), by taking yj = - [d ln(NT)/dy]-l as the width of this very low temperature layer (0 < y < yI) as determined by the relation T(y) - T1 . The magnitude of yi has to be found from the energy balance equation (26), which in the region (0 < y < y 1 ), where there are no radiation losses gives qy (y) = - dT - dy (45) = qy, d To analyze the global heat conduction coefficient dependence on the y coordinate we employ = b2 e (T)+ c, 2 PN Tk MKiN ( 1 - PN) (46) The first term on the right hand side of Eq. (46) increases with increasing temperature (and therefore for increasing y), 17 while the second one increases as T decreases becausepN -1. Therefore, the function K (y) has a minimum, cmin , in the region 0 < y < yI. Then, from Eqs. (23), (25) and (45) one can get the estimate Kmin yi (47) " YN Pu Cd PNd - (48) i d Cd -2M abPU Finally, from Eqs. (31), (38), (44), (47) and (48) one obtains 1 - PN,d '- J(Td) - Cd G(Td) m YN d2 2 YN ab min KiN T( YN ocTd ,(49) Td1 / 2 , N Tk (50) M Kmin fiN T, El Tk id ~ + = EIJ(Td) + YN Cd 1/2 Td11 . (51) M G)min yiN (I Thus, for very low temperatures T < TI, both G(Td) and J(Td) decrease with decreasing Td , while PN, d (Td) approaches unity. Comparing Eqs. (36) and (49) one can piece together an expression for an arbitrary temperature dependence of the rate constants. Consider the following equation for pN, d: 18 1 - PN,d F(pN, d) 2 (Xd(Td )+ Cd T (52) F (1)Kmin KiN T ab where Id (Td) - (2KI (Td )/KiN (Td ))1/2 contains the ratio of exact ionization and ion-neutral collision rate constants. It is easy to see that for high plasma temperatures, Id (Td )/(a b) >> 1, Eq. (52) fits relations (42), since G(Td) and J(Td) do not depend on the quantities inside the parentheses of (52). For low temperatures, Id (Td )/(cc b) << 1, (52) fits relations (49)-(51). A reasonable estimate of the value of Kmin from Eq. (46) as Kmin . (2cKTk )/(MKiN) one can simplify Eq. (52) to read 1 - PN,d 2 F(pN,d) ab d(Td) + Td YN Td 2 F(1)q, T) (53) The form of G(Td) and J(Td), and pN,d(Td) obtained from numerical solutions of Eqs. (37), (53) are shown in Fig. 3. Summarizing the results of analysis of the functions G(Td), J(Td) and PN, d(Td) one can see the following: PN, d (Td) monotonically increases (approaching unity) with decreasing temperature, J(Td) increases with the decreasing Td for the high temperatures and then decreases for lower Td, and G(Td) can have an N-like dependence, with a positive slope for high and low temperatures and a negative slope for intermediate ones. These results are in agreement with the results of Ref. 2, 3 where simple zero dimensional and 1D numerical models of the recycling region were employed. 19 IV. Neutral recycling (Knudsen flow approximation) In this section/chapter we consider the plasma-neutral interaction in the recycling region for the case X N >> A p (Knudsen flow approximation). We will assume, however, that plasma still can be described by fluid equations consisting of the plasma continuity equation (12) and d 2 , ( MnVi, 1 dq + P - -MnNKiN (Vi, (54 1 - VN(54) - -EnNKiN - EjnNKj . (55) d111 Here, P=2nT ~' qp, (11' 2 + 5T nVj, j+qeJ1 , lii is the coordinate along the magnetic field with parallel vector component, (56) ... )i indicating the d(...)/dll, - bd(...)/dy, F is the characteristic energy loss due to ion-neutral collisions, and &p and &e are the total plasma heat flux and the heat flux due to electron heat conduction. We assume random scattering of the neutrals by the sidewalls, N = N(y) only, and describe neutral transport by a diffusion equation that neglects the effects of neutrals traveling nearly parallel to the y coordinate. Balancing the divergence of the channel - DNVdN/dy with the ionization in the SOL (bKinN) gives 20 d 6 -- dNDN - = where 6 - A/Ap Z 1, (57) KnN, DN - CD A V0 ,O is characteristic velocity of the neutrals, and CD is the numerical coefficient of the order of unity. We impose boundary conditions (24), (25) and qp, iI(Y ~ 00) = - qrc /b , qP, II y 0) = -~ -y pdd (58) (Vi, II)d. (59) where the neutrals transfer their energy to the sidewalls and so do not contribute to (59). Since the net mass flow is zero (i. e. ANVN, y + ApbnVi, 11 = 0) and there is no neutral toroidal velocity one has Vi, - VN,I1 - Vi, 11 1 + N. (60) Considering subsonic plasma flow and assuming b2/6 < N/n , Eqs. (54) and (60) gives the y component of the flux to be b2 dP . ji - - MNKiN --2 dy (61) Therefore, using NVN,y =-DNaN/ay, the zero mass flow condition can be written as 21 b2 dN dP - + DN -MNK. (62) - ~ = 0. N Y iMN Equations (55), (57), (62) and the boundary conditions describe plasma - neutral interactions for the large neutral mean free path XN >> Ap . Let us analyze them assuming the following dependencies of the rate constants on plasma temperature K1(T) - K, (T/Tk)e(T - TI) , KiN (T) - KiN r const. , (63) E1(T) m E1 - const. . E(T) - T Te(T - TO), where E is the numerical coefficient of the order of unity, and To is the effective temperature of the neutrals, (To < TI). Then from Eqs. (24) and (62) one obtains P - =Pu ) (64) , where i - N/Nmax and N 2 - 2b 2 P N (65) 22 Multiplying plasma continuity equation on E1, adding it to the energy balance equation (55), and integrating over all y, and employing the boundary conditions gives (b Pu) I Y ~d+ Cic + ji, d - ji,d ctbPuCd (1 _ d2 ddn 1(6 E 0 PU J(Td), /Jid (67) where relation (64), 'Id = Nd/Nmax, and dy =- DNbdN/i are employed. The main difference between the global energy balance equation (66) and its short mean free path counterpart (37) is the presence of a volumetric energy sink due to ion neutral collisions (second term on the right hand side). For the case Td > T the 'j dependence of ji can be evaluated. Then from Eqs. (57) and (64) one obtains 1 (68) 2 AK dy 2)1/ 2 where 2 2 AKI Using ji,d Tk DN (69) u = - 8DNNmaxdq/dyly=O along with Eqs. (65), (67), (68) and (69) gives 23 . Td Tk C2 2 2 2 'q iN i (70) Tld 2 1 - 'qd2/' Inserting expression (68) into Eq. (66) to eliminate ji,d = - 6DNNDdi/dy and using the definitions for Nnd Ak and ji,d gives G(Td) - b2 ( YP + E 1 - '9d ) 1/2 1/2j d +Mf 1 -,2 _ 2/2), (71) To simplify expressions (67) and (71) wer first consider high plasma temperatures, Td >> Tk (a/2 )2 (KiN 2 (C2 Wl KiN K, G(Td) - abCd I ). From Eq. (70) one finds Tk (72) Td<1 (Yp + E Tk KiNKI + EI + Td ) (73) so that (67) and (71) reduce to J(Td) J~dJ 0'2b Cd Td (74) 24 2 For lower plasma temperatures, T 1 s Td << Tk (a/2 )2 (KiN /Kl), id -1 in Eq. (70) giving ld d G(Td) 1 1 -b (75) Td KI 1/2 Tk KiN/ cc Tk M /2 El K, 2 Tk KiN + 0. 35T J(Td) J~dj 2b Tk / iN - const. , . const. . (LZ M KiN) 1/2 (76) (77) The exact solutions of Eqs. (67), (70) and (71) are shown in Fig. 4. In the plasma temperature range Td a TI, the functions G(Td) and J(Td) saturate at low temperatures and lid approaches unity, Note that depending on the values of F, KiN/KI, etc. the function G(Td) may increase at low temperatures. Analysis of Eqs. (57) and (66) shows that function G(Td) does not decrease in the temperature range T1 2 Td 2 To . We omit the details here because of the complicated algebra. However, qualitatively one can see directly from Eq. (66) that G(Td) does not decrease. Notice from Eq. (66) that G(Td) has the following structure 25 d + f dy (..)j, G G(Td) (78) '' 0 di,d where Iji/ji,di < 1. Therefore, G(Td) can decrease with decreasing Td only when both i, d and id decrease, but this would contradict relation (67) and so it is not possible. This contradiction can only be removed when Td decreases below the value To. Indeed in this case the structure of equation (66) becomes Crc = b (79) 0 Ji, d d + fcdi (-- (79) . where T10 corresponds to the coordinate yo such that T (y) = To for y s yo. Since ii, d is determined by the value of r1d, it is now possible to decrease the right hand side of Eq. (66) by decreasing both qo and ji, d. General analysis of this case is very cumbersome, therefore we consider here the limit of very small rc (qrc << Pu G(T d a To)). First of all let us determine the energy losses within the region yo 5 y ! yl, where yi corresponds to the temperature T(yj) - T1 . The energy balance equation in this region, where there is no neutral ionization, can be written in the form dJ 5 T ji d + b dy ' e(T) - Pu Nmax KiN 2 dy Eq. (80) can be simplified for small (80) rc . Indeed, in this case the ion flux ji, d is small and the energy heat flux is mainly transported by electron 26 heat conduction. Moreover, we will see that the neutral density variation in the region yo z y s Yi is strong and YiI Then, imposing the boundary condition (dT/dy) = (I << 1) =n (Y I) << i = 1. 0 and using the linear dependence of the neutral density on the coordinate y (see Eq. (57)), from Eq. (80) one obtains the length 110 - y - Yo (110 << yo), and the expression for the heat flux qO dissipated in the region yo r y z yj 1/2 -1/2 CI 11/2 - 24 b2 Ce(TI)TI 7 r PU Nmax KiN (3/14)b2 E Pu Nmax KiN ICe(TI )TI 1/ We assume here that e(TI)TI >> 2 (82) e(To )To. The expressions for the ion flux ji, d and the heat flux qj dissipated in the region y a yj can be found from Eq. (55) and the plasma continuity equation (rid - 1): ji, d q, b bP - Tk djd M Ei + E Tk 1/2 - DN Nmax 6110 rio - DN Nmax , (83) 6 YI (84) - The values of qi, qi, and yi can be found from global energy balance in the recycling region 27 (85) qrc - q 0 +q. From Eqs. (82)-(84) one sees that for low grc values (rj1 qrc < < «< 1) {(3/14) T b2 Pu Nmax KiN Ice (TI )TI 1/2 3 4 cx(b2 PU) / 1/2 A)1/2 (86) so that practically all the energy flux is dissipated in the region yo r Y s Yi due to ion-neutral collisions (and the part related to ionization losses in the region y a yi is small). The ion flux on the target is 3 fr rP i d o yDNa iN 3b2 ice(T ) Ti c qrc Pu 62 Ap b 4 (87) The estimates for 1, 0 , yl, and ql -1 -3 110 , qrc << Y1cqrC, 0 2 3 Il Crc (88) <<110 q rc << 1, (89) obtained from Eqs. (81) and (83), show that (as assumed above) for low qrc , neutral density variation in the region yo r Y r Yi is strong for TII << 710. Thus, for the case XN >> Ap (Knudsen flow approximation) G(Td) 28 (J(Td)) decrease (increases) with the decreasing of plasma temperature and saturates at low temperatures Td > To, while 'qd approaches unity (as we noted above function G(Td) may even increase at low temperatures). When plasma temperature Td decreases below To, the plasma flux on the target, length of the recycling region, etc. are directly determined by the heat flux qrc (see for example Eq. (87)). Therefore, in Fig. 4 functions G(T d ) and J(Td ) have vertical lines at Td =To. V. Bifurcations in the SOL plasmas and divertor detachment We have shown in the previous Chapters that generally speaking functions G(Td) and J(Td) have more or less similar dependencies in the opposite extremes of the fluid and Knudsen limits. Therefore, it seems reasonable to assume that in the general case the functions G(Td) and J(Td) have the same type of behavior: G(Td) increases with increasing temperature for low (G(O)=0) and high Td values and decreases for intermediate Td ; and J (Td ) increases for low temperatures (J(O)=0) and decreases at high ones, reaching the maximum at intermediate Td . Consider now the global behavior of the SOL plasma assuming G(Td) has an N-like dependence with the local minimum Gmin corresponding to the temperature Td - Tm (one can see from Figs. 3 and 4 that for both the fluid and Knudsen approximations Tm - 5 eV). Then, the global energy balance of the SOL plasma can be written in the following way Pu G(Td )S - Qrc R Qheating - Qrad, bulk - Qrad, SOL , . (90) where S = 4nRAP , R is the tokamak major radius, Qheating is the total tokamak heating power, Qrad, bulk is the radiation losses in the bulk 29 plasma, and Qrad, SOL is the energy radiation losses in the SOL plasma (Eregion). Note that the value of Pu determined by the upstream SOL plasma density and temperature depends the edge plasma transport and on the heat flux QSOL Qheating - Qrad, bulk entering the SOL. Q Assume that QSOL , Pu and S are constants. If radiation losses in the SOL are low and QSOL is high enough, then Eq. (90) gives QSOL > Qbif, where Qbif = Pu Gmin S - (91) As a result, one finds regimes with high plasma temperature near the target Td > Tm, balanced plasma pressure Pd - Pu, and high plasma ( d - Pu J(Td ) S) and heat flux on the divertor target (Qid) , which can be esmitted from Eqs. (1), (2) to be +SOL Qd, d " QSOL Fd (92) Q YPTd + El (93) QSOL YPTd + El With the increase of radiation losses in the SOL, the heat flux Qrc will decrease causing the decrease of Td (Td > Tm) and the increase of plasma density near the target, nd - Pu/Td, and plasma flux rd. When radiation losses from the SOL plasma becomes high enough that QSOL - Qrad, SOL < Qbif , bifurcation of solution of Eq. (90) occurs and the temperature Td drops well below Tm (according to the Figs. 3 and 4, 30 the temperature Td after bifurcation to the low temperature branch of G(Td) is smaller than 1 eV). Simultaneously, there is a strong decrease of plasma pressure near the target. As far as the heat load on the divertor target is concerned, there is a big difference between fluid and Knudsen neutrals. Indeed, in the fluid approximation the heat flux qrc going to the divertor plate can only be dispersed to the side walls by radiation if the length of the recycling region, 4c - Yi + AF, becomes less than the plasma column width AP . For the case 4c target < Ap the heat flux on the Qx, d cannot be smaller than Qrc/2 (which corresponds to half of the radiated photons hitting the plate). Therefore, for fluid neutrals a strong decrease in the heat flux on the target is only possible for the case when Lrc > Ap . From Eqs. (29) (47), (51) one can see that AF x Pu 1 and yi ( Qrc~1. Thus, for fluid neutrals one can only expect to have a strong decrease of Q7, d for low Qrc Qrc < Qcrit, F (Ap ) , (94) In the Knudsen approximation energy can easily be dispersed to the whole target surface and to the sidewalls by the neutrals if the length of recycling region is high enough. This is exactly corresponds to the situation after the bifurcation to the low temperature branch of the curve G(Td), when Td =To (see Fig. 4). Moreover, for the case Td =To the plasma flux on the target Id is proportional to Qrc3 (see Eq. (87)) one has the following scaling for Qx, d O I Fd : 31 QId (95) Qrc Q 3. Thus, the specific heat flux on the target and sidewalls decreases much more rapidly then the heat flux into recycling region. Let us now show qualitatively one effect of the influence of the radiation losses from the bulk plasma on the divertor plasma parameters assuming Qrad, SOL= 0 . As one can see from Eq. (91) Qbif is proportional to the upstream plasma pressure. But for a high recycling divertor, the upstream plasma temperature Tu is determined by the heat conduction equation (see Eq. (9)) and hence by QSOL. (Note that the effect of radiation loses from the SOL plasma on the upstream plasma temperature Tu is much less pronounced, especially if the radiation region is localized in the divertor.) An increase in Qrad, bulk causes a decrease in both QSOL and upstream plasma temperature Tu (assuming fixed upstream plasma density). In this case Pu . Pu (QSOL) MPu (Qheating - Qrad, bulk) and Qbif will decrease with increasing Qrad, bulk. Therefore, the transition to the low temperature branch, which is accompanied by a decrease in Td , plasma pressure drop, etc., can only occur for higher radiation losses then for the case of SOL radiation. So far we have assumed that the radiation losses in the SOL (E region) are not affected by plasma parameter variation in the recycling region. However, as was noticed in Chapter II, even a small fraction of neutrals can significantly influence the impurity radiation. Then, taking into account that the neutral density is a function of the heat flux Qrc (N a N(Qrc )), Eq. (90) can be written in the form 32 Pu G(Td)S = Qheating - Qrad,SOL(G(Td)) , (96) where we put Qrad, bulk =0. Assuming that the radiation loses increase with increasing of neutral density one can see the following: 1) bifurcation of divertor plasma parameters may occur for Qrc higher than Qbif - Pu Gmin S and 2) after bifurcation to the low temperature branch, where the neutral density is much higher than on the high temperature branch, the magnitude of Qrc , plasma temperature, and particle and heat fluxes on the target (even for the fluid neutrals) may drop more strongly than they would have had we not accounted for this effect. A detailed analysis of the influence of the neutrals on impurity radiation and SOL plasma stability will be considered in a separate paper. One can see from the preceding that the plasma/neutral parameters in the divertor and the target heat load corresponding to the low temperature branch of the curve G(Td) (especially for Knudsen neutrals) fit all of the following main features of detached divertor regimes :1) high energy radiation losses from the SOL region, 2) low plasma temperature near divertor plates, 3) strong decrease of the plasma energy and particle fluxes onto the plates, 4) strong plasma pressure drop along magnetic field lines in the divertor volume. We are now ina position to make some estimates. First we estimate Qbif for current tokamaks and ITER. For a typical value of b - Bp/B - 0.05 , and upstream plasma pressure Pu= Po - 10 14 cm-3 MW/m 2 x 100eV one finds from Figs. 3 and 4 that qbif - 6 . The typical S values for current tokamaks (DIII-D, ASDEX-U, JET, JT-60-U) having a single 33 null divertor is S - 4a x 0. 01im x 2 m - 0. 25m 2 with an upstream pressure of about 0.1+0.3 PO . Consequently, for current tokamaks Qbif - 0.15 + 0.5 MW. For C-MOD where R=0.67 m but Pu -0.5 assuming Pu -5 PO, one gets Qbif - 0.5 MW. For ITER, PO, R - 8 m and A P - 0. 02 m one has Qbif - 60 MW. All these numbers look quite reasonable. Next we make some estimates for the neutrals. The tokamaks DIII-D, ASDEX-U, JET, and JT-60-U are characterized by quite moderate plasma density near the target, nd si 014 cm-3 when XN a Ap. Therefore, neutral transport in these machines is closer to the Knudsen regime. For C-MOD where plasma density is higher, the neutrals are closer to the fluid regime (for the attached case). However, in detached case in which the plasma density near the target drops, neutral transport near the target may be in the Knudsen regime, while at some distance from the target it still can be in the fluid regime. We can estimate the magnitude of neutral density Nma reached at low temperature branch of the curve G(Td ). For b-0.05, cD -1, A -5 cm, 6-5, KiN-10- 8 cm- 3 , VO -5105 cm/s , and Pu-0.3+0.5 PO one has Nma - 1+2 1013 cm73 . Again, these values are close to experimental observations. For the ITER parameters (A -10 cm, Pu -5 Nmax " 4 10 13 PO ) one finds cm-3 . V. Conclusions We have developed models to investigate the main features of plasma neutral interaction in the recycling region for the two opposite extremes of fluid and Knudsen neutrals. We have shown that for both neutral models the reduction of the heat flux into the hydrogen recycling region below the critical value Qbif leads to bifurcation of the plasma parameters near the 34 target. This bifurcation causes a decrease of the plasma temperature near the target to about 1 eV, plasma pressure drop in the recycling region, increase of the neutral gas density in the divertor volume, and expansion of the recycling region. For the Knudsen neutrals bifurcation also causes a strong decrease of the target heat load. This behavior in the SOL is in agreement with all the main features of detached divertor regimes in current tokamak experiments. We also find that in the Knudsen limit, the neutral density in the divertor region can not exceed some maximum density, which is of the order of 1-2 x 1013 for current experiments. The physical mechanisms responsible for the decrease in the plasma flux onto the target and plasma pressure drop along magnetic field lines in the recycling region depend on neutral gas regime of interest. For fluid neutrals, the neutral pressure influence on the plasma flow is responsible, while for Knudsen neutrals, friction between the plasma flow onto the target and the neutral gas scattered by the sidewalls causes these reductions. Acknowledgments This work was performed under DOE grant DE-FG02-91-ER-54109. 35 References 1 I.H. Hutchinson, R. Boivin, F. Bombarda, P. Bonoli, S. Fairfax, C. Fiore, J. Goetz, S. Golovato, R. Granetz, M. Greenwald, S. Home, A. Hubbard, J. Irby, B. LaBombard, B. Lipschultz, E. Marmar, G. McCracken, M. Porkolab, J. Rice, J. Snipes, Y. Takase, J. Terry, S. Wolfe, C. Christensen, D. Gamier, M. Graf, T. Hsu, T. Luke, M. May, A. Nemczewski, G. Tinios, J. Schachter, and J. Urban, Physics of Plasmas 1, 1511 (1994). 2 S. I. Krasheninnikov, A.S. Kukushkin, V.I. Pistunovich, and V.A. Pozharov, Pis'ma Zh. Techn. Fiz. 11, 1061 (1985). 12th Europ. Conf. on Plasma Phys., Budapest, 1985. 3 S. I. Krasheninnikov, A.S. Kukushkin, V.I. Pistunovich, and V.A. Pozharov, Nuclear Fusion 27, 1805 (1987). 4 P.C. Stangeby, Nuclear Fusion 33, 1695 (1993). 5 P. Helander, S. I. Krasheninnikov, P.J. Catto, will appear in Physics of Plasmas 1, October (1994). 36 Figure Caption Fig.1. Geometry of the problem. Fig. 2. Different regions in the tokamak SOL plasmas. Fig. 3. Function G(Td) [kW/cm 2 ], J(Td) [1020 cm- 2 s- 1 ], and relative plasma pressure near the target p(Td) - 2ndTd/Pu for the fluid neutrals (T 1 = J (Td) 3 eV, a=0.5, CN= 4 , Yp=6, Y N=0.2, and b=0.05). Functions G(Td) and are multiplied on the pressure Po = 104 cm-3 x 100 eV. Fig. 4. Function G(Td) [kW/cm 2 ], J(Td) [1020 cm- 2 s- 1 ], and relative plasma pressure near the target p(Td) - 2ndTd/Pu for the Knudsen neutrals (a=0.5, y p=6, E=1, and b=0.05). Functions G(Td ) and J(Td ) are multiplied on the pressure P0 = 10 37 14 cm 3 x 100 eV. * 14 - X-point SOL core side wall side Lx wall 'z x A -- p. ' 0 I 1 . . . divertor Fig. 1. 38 . plate * X-point energy transport side side wall wall impurity radiation 4 hydrogen recycling Lrc A I__ I Divertor plate Fig. 2. 39 Fluid Neutrals 1-.0 - - -G T /p (Td) .8 | 4 -I .2 - ....... 0 . 5 10 Td, eV Fig. 3. 40 J(Td ) . 15 20 Knudsen Neutrals 1.00 p(Td) .75 - .7I - ~ - . - - -- ~G(Td--- .50 .25 -- J(Td) . 0 0 5 10 Td, eV Fig. 4. 41 15 20