Nuclear Materials and Energy 12 (2017) 1259–1264

Contents lists available at ScienceDirect

Nuclear Materials and Energy

journal homepage: www.elsevier.com/locate/nme

Type I ELM filament heat fluxes on the KSTAR main chamber wall

M.-K. Bae

a , R.A. Pitts b , J.G. Bak

c , S.-H. Hong

c , H.S. Kim

c , H.H. Lee

c , I.J. Kang

a , K.-S. Chung

a , ∗

a Hanyang University, 222 Wangsimni-ro, Seongdong-gu, Seoul, Republic of Korea b ITER Organization, Route de Vinon-sur-Verdon, CS 9 046 13067 St. Paul-lez-Durance, France c National Fusion Research Institute(NFRI), 169-148 Gwahak-ro, Yuseong-gu, Daejeon, Republic of Korea

a r t i c l e i n f o

Article history:

Received 16 July 2016

Revised 14 March 2017

Accepted 16 April 2017

Available online 3 May 2017

a b s t r a c t

Heat loads deposited on the first wall by mitigated Type I ELMs are expected to be the dominant con-

tributor to the total thermal plasma wall load of the International Thermonuclear Experimental Reactor

(ITER), particularly in the upper main chamber regions during the baseline H-mode magnetic equilib-

rium, due to the fast radial convective heat propagation of ELM filaments before complete loss to the

divertor. Specific Type I ELMing H-mode discharges have been performed with a lower single null mag-

netic geometry, where the outboard separatrix position is slowly ( ∼7 s) scanned over a radial distance

of 7 cm, reducing the wall probe–separatrix distance to a minimum of ∼9 cm, and allowing the ELM fila-

ment heat loss to the wall to be analyzed as a function of radial propagation distance. A fast reciprocating

probe (FRP) head is separately held at fixed position toroidally close and 4.7 cm radially in front of the

wall probe. This FRP monitors the ELM ion fluxes, allowing an average filament radial propagation speed,

found to be independent of ELM energy, of 80–100 ms −1 to be extracted. Radial dependence of the peak

filament wall parallel heat flux is observed to be exponential, with the decay length of λq, ELM ∼25 ± 4 mm

and with the heat flux of q ‖ , ELM = 0.05 MWm

−2 at the wall, corresponding to q ‖ ∼ 7.5 MWm

−2 at the sec-

ond separatrix. Along with the measured radial propagation speed and the calculated radial profile of the

magnetic connection lengths across the SOL, these data could be utilized to analyze filament energy loss

model for the future machines.

© 2017 The Authors. Published by Elsevier Ltd.

This is an open access article under the CC BY-NC-ND license.

( http://creativecommons.org/licenses/by-nc-nd/4.0/ )

1

e

t

a

E

(

b

a

t

t

m

m

a

(

t

c

a

a

o

t

w

F

h

D

h

S

U

h

s

h

2

. Introduction

Edge localized modes (ELMs) are periodic relaxation phenom-

na due to the steep edge gradients of plasma pressure and lead

o ejection of high energy and particles. In recent years, which

re the intermittent filamentary structure of ELM filaments during

LMs has been observed in the edge plasma and scrape-off layer

SOL) [1–7] . These ELM filaments, with higher density than am-

ient plasma, are mostly extended along the magnetic field lines,

nd partially propagate outward to the wall due to negative curva-

ure and gradient related with E ×B drift [8,9] .

Since Type I ELM heat flux mainly reaches along the field lines

o the divertor plates, heat fluxes on the first walls of present toka-

aks would be negligible. However, if the heat flux of ELM fila-

ent exceed the thermal limitation of divertor ( ∼20 MW/m

2 for W

nd CFC) and first wall ( ∼0.5 MW/m

2 for Be mockups) [10,11] due

∗ Corresponding author.

E-mail addresses: bmin999@hanyang.ac.kr (M.-K. Bae), kschung@hanyang.ac.kr

K.-S. Chung).

h

u

l

t

ttp://dx.doi.org/10.1016/j.nme.2017.04.006

352-1791/© 2017 The Authors. Published by Elsevier Ltd. This is an open access article u

o fast radial convective transport of ELM filament [12] , it would

ause severe damage to the first wall in advanced tokamaks such

s ITER and DEMO reactors. The specific parameters of ELM fil-

ments such as density and temperature would be essential not

nly to understand the basic mechanisms of ELM filaments and

heir radial transport but also to estimate the damage to the first

all for the extrapolation to future devices.

Propagation of ELM filament is analyzed by the fast camera and

RP in MAST [1] . A comparison between ELM filament transport at

igh and low field side SOL is carried out in JT-60U [13] . In DIII-

, far SOL transport and plasma interaction with the main wall

ave been investigated [6] . Characteristics of ELM filament and far

OL transport have been investigated by electric probes in ASDEX

pgrade [3–5] . Recently, particle and heat fluxes in the far SOL

ave been studied using fast reciprocating probe (FRP) and Thom-

on scattering in ASDEX Upgrade [14] . So far, a number of studies

ave been reported ELM filament transport in SOL and far SOL, but

nderstanding of effect of ELM filament on the first wall is still

imited.

In this paper, the results of first measurements of heat fluxes

oward the first wall of KSTAR device by electric probes are

nder the CC BY-NC-ND license. ( http://creativecommons.org/licenses/by-nc-nd/4.0/ )

1260 M.-K. Bae et al. / Nuclear Materials and Energy 12 (2017) 1259–1264

Fig. 1. Horizontal and vertical cross section of KSTAR and position of fast reciprocating probe (FRP), toroidal limiter (TL) and poloidal electric probe (PEP) are presented. The

dark green dashed line indicates the D α view direction (DVD). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of

this article.)

h

s

i

a

w

e

e

T

s

b

r

a

t

u

w

a

γ

3

3

o

A

presented. Poloidal electric probes (PEPs), installed on the outer

main chamber wall of KSTAR, are used to measure the far-SOL

plasma parameters such as plasma density, electron temperature,

ion saturation current, parallel Mach number and floating poten-

tial. From these measurements, ELM filament parameters such as

heat and particle fluxes, radial velocity are calculated, and the

effect of ELM filament on the first wall are analyzed. In addition,

a separate FRP head is held at a fixed position in front of the wall

probe. This FRP monitors the ELM ion fluxes, allowing an average

filament radial propagation speed, found to be independent of

plasma radial position.

2. Experiment

We have installed a poloidal electric probe (PEP) set for the

measurement of far-SOL parameters of KSTAR, which is composed

of 8 electric probes on a rectangular plate with 125 × 280 mm

2 .

The probe tips are made of carbon fiber composite and insulated

by boron nitride covers. The diameter of each probe tip is 4 mm

and the tip is protruded 1 mm from the graphite cover tile surface.

PEPs are fixed on the outboard midplane wall and located 7.5 cm

behind the outboard toroidal limiters, which is also located ∼16 cm

behind from the typical separatrix position of KSTAR diverted plas-

mas as shown in Fig. 1 . Probe tips are arranged poloidally to be

composed of two triple probes and one Mach probe, which al-

lowed direct measurements of electron temperatures, particle and

eat fluxes, and Mach numbers at far SOL region with a fast acqui-

ition rate of 2 MHz. Detail structure of the PEP assembly is shown

n Fig. 1 .

As for the triple probe system, a fixed bias voltage −200 V was

pplied to the two probes ( p 1 and p 3 ), and the other probe ( p 2 )

as for measuring the floating potential ( V f ). Potential differ-

nce between the p 1 and floating probe p 2 is proportional to the

lectron temperature ( T e ), which is given by simple formula as

e = [ e ( V 1 − V f ) ] / ln 2 , where V 1 is positive bias voltage of p 1 . As-

uming T e ≈ T i at the edge plasma, the plasma density ( n e ) is given

y n e = I is ( αe A e f f

√

T e / M i ) , where I is , M i , A eff and α are ion satu-

ation current, ion mass, effective projected area for parallel flow

nd coefficient of I is (for magnetized plasma, α ≈ 0.49), respec-

ively. Parallel particle flux ( �‖ ) and heat flux ( q ‖ ) can be derived

sing following equations: �‖ = I is /e A e f f and q ‖ = γs T e I is /e A e f f ,

here γ s is the sheath transmission coefficient as given in γs = 7 ,

lthough the average value of various tokamaks is to be given as

s = 6 [15] .

. Analysis

.1. Radial propagation of ELM filament

Cross-field transport of heat and particles were experimentally

bserved by an electric probe on outboard midplane wall (PEP).

lthough three limiters exist between A and N bay of KSTAR outer

M.-K. Bae et al. / Nuclear Materials and Energy 12 (2017) 1259–1264 1261

Fig. 2. (a) Typical Single ELM structure observed through change of magnetic signal, D α and probe measurements. From the probe measurements, plasma density, electron

temperature, heat flux and Mach number were calculated. The time difference between the initiation of the filaments (dashed red line: dB/dt) and the arrived (dashed blue

line: V f or I is ) to the poloidal electric probes (PEPs) is about 0.5 ms. D α signal is from the private region of the bottom divertor. (b) Ion saturation current measurements

by poloidal electric probes (PEPs) and fast reciprocating probe (FRP) and position of separatrix during the discharge. (shot #13117). We have not used the data for the time

range of 6 < t < 7 (s) due to sudden current ramp up in (b) FRP measurement. (For interpretation of the references to color in this figure legend, the reader is referred to

the web version of this article.)

m

fi

t

p

c

f

d

c

7

t

P

∼

r

v

a

b

E

s

t

c

M

p

o

I

a

c

a

w

a

F

P

t

t

f

t

a

T

A

c

a

i

3

d

A

4

u

o

A

t

w

F

a

o

1

idplane (refer to Fig. 1 ), we can know considerable number of

laments do not hit the limiter, by comparison of probe signal at

he wall with ELM peaks in D α signal which came out from the

lasma. ELM filament structures were observed clearly by PEP be-

ause the background plasma density and electron temperature at

ar-SOL are much lower than those of ELM filament.

Since the probes are fixed on the wall to investigate the ra-

ial propagation of ELM filament on the main wall, we have slowly

hanged the outboard separatrix position in the radial direction for

s during an ELMy H-mode discharge with the following condi-

ions of shot 13117: I P = 600 kA, B T = 2.7 T, n e = 1.8 −4 × 10 19 m

−3 ,

NB = 4.5 MW, κ ∼1.6, W Plasma = 30 0–40 0 kJ, H-mode flattop phase

10 s, midplane separation between primary and secondary sepa-

atrices ∼1 cm. The distance between the separatrix and PEP was

aried from 9 to 16 cm, which produced the changed of the sep-

ratrix to be 7 cm. The position of the separatrix was verified

y using the equilibrium reconstruction code (EFIT). The error of

FIT on radial position control of LCFS is within 1 cm. Fig. 2 (a)

hows typical parameters of ELM filament structures based on

ime trace of magnetic fluctuation, D α signal and ion saturation

urrent signal. D α and magnetic signal which was measured by

irnov coil were used to match the filament peaks to those of

robe signals. This Mirnov coil is mounted behind PFC tiles in

utboard divertor region which is near the midplane and probes.

n Fig. 2 (b), ion saturation currents were measured with notice-

ble change during movement of the separatrix. A drastic in-

rease of I is was observed as separatrix approached to the wall

round 4.5 s ( r s ∼ 2270 mm). Peak ELM heat and particle fluxes

hich are calculated by measurement of I is ( Fig. 2 (b)) were an-

lyzed as a function of radial propagation distance as shown in

fi

ig. 4 . �r s is gap distance between moving separatrix and fixed

EP. Radial dependence of the peak filament heat flux, parallel

ransport to the wall were reasonably fitted into exponential func-

ions, and the decay lengths of heat and particle fluxes were

ormed to be λq, ELM

= 25 ± 4 mm and λ�, ELM

= 31 ± 1 mm, respec-

ively. Absolute values of heat and particle fluxes were also given

s q ‖ , ELM

= 0.05 MWm

−2 and �‖ , ELM

= 3 × 10 21 m

−2 s −1 at the wall.

he deduced decay length of heat flux is comparable to those of

UG device ( ∼2–30 mm) [3] . One has to keep in mind that the

hange of r s not only can affect the plasma density and temper-

ture at the separatrix, but also in the near and far-SOL, which can

nfluence filament dynamics [16] .

.2. Velocity of ELM filament

Radial velocities of ELM filament were measured from the time

elay between two radially separated electric probes, PEPs and FRP.

n FRP head is held at a fixed position toroidally close to PEP and

.7 cm radially in front of the PEP. In Fig. 3 (a), signals of ion sat-

ration currents from both FRP and PEP show a good correlation

f the ELM structures on the ion saturation current measurement.

pproximately 200 ELM filament peaks, which can be clearly iden-

ified and corresponded with D α , were selected and these data

ere averaged over 4 ms centered on the D α peak as shown in

ig. 3 (b). There is a time delay of ∼ 0.5 ms between first large fil-

ment peaks of both probes. From the time delay and difference

f distance, radial ELM filament velocity can be deduced as 80–

00 ms −1 .

Fig. 5 (a) and (b) show the trend that the radial velocity of

lament increases with the amplitude of ion saturation current

1262 M.-K. Bae et al. / Nuclear Materials and Energy 12 (2017) 1259–1264

Fig. 3. (a) ELM filament peaks in FRP and PLP measurements were selected by matching with D α peaks. (b) Selected peaks were sliced for 4 ms based on the highest of

peak and averaged for velocity calculation.

Fig. 4. Profile of ELM peak particle flux ( �‖ = I is /e A e f f = J is /e ) and heat flux at the far-SOL measured by PEP during the variation of separatrix in 7 cm. The decay lengths

of heat and particle flux are determined by the least square fitting (red), which are given as about 25 mm and 31 mm, respectively. (shot #13117). (For interpretation of the

references to color in this figure legend, the reader is referred to the web version of this article.)

a

a

q

i

o

r

c

density ( J is = I is / A e f f ) and heat flux during the change of separatrix

position. This trend indicates that v r _ ELM

depends on the J is and on

the relative separatrix position ( �r s ), since J is decreases exponen-

tially with �r s as observed in Figs. 2 (b) and 4 . It means v r _ ELM

is strongly correlated with the particle flux of filaments and sep-

aratrix position, while it is weakly correlated with the heat flux.

In addition, radial propagation of ELM filament can be derived

s v r _ ELM

∼ λ/ τ‖ ∼ C s λ/ L ‖ ( C s =

√

T e + T i / M i , L ‖ = πRq ), where τ ‖ nd L ‖ are time scale for transport and parallel connection length,

is the safety factor, C s is ion sound speed [2] , estimated assum-

ng T e = T i = 10 eV which is measured by PEP. The radial velocity

f the ELM filament ( v r _ ELM

) is calculated as 43–73 ms −1 , from the

esults of moving separatrix experiment (#13117) based upon the

alculated decay length of particle flux.

M.-K. Bae et al. / Nuclear Materials and Energy 12 (2017) 1259–1264 1263

Fig. 5. The radial velocity of filaments as a function of (a) ion saturation current density, (b) parallel heat flux, and (c) ELM energy loss ratio ( �W ELM / W tot ). Peak values of

J is and q ‖ are averaged over 10 measurements. In case of (c), 8 discharges with different plasma current were used.

�

c

d

t

t

s

t

m

r

�

e

w

m

c

o

o

m

d

t

r

t

t

o

s

s

v

3

W

a

θ

t

r

T

b

w

E

1

E

E

p

r

e

(

1

r

c

4

o

u

In order to investigate the correlation with v r _ ELM

and

W ELM

/ W tot , the radial velocity has been obtained from 8 dis-

harges, which contains about 200 ELMs in each discharge, with

ifferent plasma current ( I P = 400–750 kA), at the same separa-

rix position. The average ELM energy loss is determined from

he drop of total stored energy ( W tot ) during ELMs. Fig. 5 (c)

hows v r _ ELM

as a function of the fraction of energy loss due

o ELMs ( �W ELM

/ W tot = 0 . 015 − 0 . 03 ). The range of average fila-

ent radial velocity is 80–100 ms −1 , while separatrix is fixed at

s = 2.23 m. There seems to be no clear correlation with v r _ ELM

and

W ELM

/ W tot . The reason could be that the only largest filament in

ach ELM period has been used for analysis of the radial velocity,

here could lead the uncertainty of v r _ ELM

because each ELM fila-

ent has different ener gy and size. Therefore, to obtain the valid

orrelation between v r _ ELM

and �W ELM

/ W tot , we have to include all

f filaments for analysis during ELM.

Radial velocity of the filaments has been also calculated by time

f flight (TOF) method from the time delay between the start of

agnetic fluctuation signal and PEP signal. The mean value of ra-

ial velocity by TOF ( v r _ T OF ) is about 200 ms −1 , which is about 2

imes larger than the v r _ ELM

calculated by the above method. The

eason for this difference seems to be due to the time delay be-

ween the magnetic fluctuation, indicating the time of the forma-

ion of filaments within the pedestal, and the time of departure

f filaments from the LCFS, which would usually be happened in-

ide of core plasma with respect to separatrix. Hence, v r _ ELM

mea-

ured at far-SOL would relatively be expected to be smaller than

r _ T OF .

t

.3. Energy loss to the wall by ELM filament

Total ELM energy deposition to the wall can be derived as

wal l _ ELM

= q ⊥ S out �t ELM

, where S out is approximated by the total

rea of midplane outer wall (13.3 m

2 ) and q ⊥ �t ELM

( q ⊥ = q ‖ sinθ ,

≈ 5 °) is calculated by integrating the heat flux as a function of

ime measured by PEP. The same method as used to determine the

adial velocity is utilized to find the ELM peaks in heat flux data.

he average energy of eight discharges, deposited on the first wall

y the filament ( W wal l _ ELM

) during the flat-top phase, is 1 ± 0.3 kJ,

hich is significantly lower than those of divertor targets. Since,

LM energy ( �W ELM

) ejected from the core plasma is generally 5–

0 kJ in KSTAR, the ratio of energy deposited on the wall to the

LM energy loss ( W wal l _ ELM

/ �W ELM

) is about 15%. Then ratio of the

LM deposited energy on the wall to the total stored energy of

lasma, W wal l _ ELM

/ W tot is about 0.5%. If we simply extrapolate this

atio to the case of ITER, with 100 MJ plasma energy, deposited

nergy by ELM on outer midplane wall can be about 2.5 MW/m

2

assuming outer midplane wall area is ∼200 m

2 , ELM duration is

ms), which does not exceed thermal limitation of Be, which indi-

ectly validates the results of Kocan’s: neither melting nor signifi-

ant evaporation of Be [17] .

. Conclusion

To investigate the ELM filamentary phenomena in the far SOL

f KSTAR during type-I ELMs, poloidal electric probes (PEPs) were

sed and measured the far-SOL plasma parameters such as elec-

ron temperature, plasma density, heat and particle flux, and radial

1264 M.-K. Bae et al. / Nuclear Materials and Energy 12 (2017) 1259–1264

A

t

f

(

p

(

c

R

velocity of ELM filaments to the wall by the configuration of two

triple probes and one Mach probe. During the slow change of sepa-

ratrix position by 7 cm, the radial decay length of heat and particle

fluxes are deduced as λq, ELM

= 25 ± 4 mm and λ�, ELM

= 31 ± 1 mm,

respectively. The mean radial velocity of ELM filaments for eight

discharges are measured as 80–100 ms −1 by the time of flight

method with two radially separated electric probes, PEPs and FRP,

which a fixed in far-SOL region. Radial velocity of the filaments

have been also calculated as ∼200 ms −1 by TOF method, i.e., using

the time delay between start of magnetic fluctuation signal and

PEP signal, which seems to indicate the faster formation of fila-

ments inside the pedestal far from the separatrix than the depar-

ture of the filaments from the separatrix.

These radial velocities of ELM in KSTAR are smaller than the

other tokamaks, e.g. MAST [1] , AUG [5] and JET [18] , because this

result is based on the measurements in the far-SOL region. There-

fore, energy loss of the ELM filament will be larger and it can have

a relatively smaller value than the other tokamaks. Moreover, typ-

ical ELM and plasma energy in KSTAR is smaller than the other

tokamaks.

Radial velocity is strongly dependent upon the particle flux

of filament and separatrix position, while it is weakly dependent

upon the heat flux. Besides, v r _ ELM

seems to be independent of the

ratio of ELM energy loss to the total stored energy ( �W ELM

/ W tot ).

From the heat flux measurement, energy deposition to the wall

due to convective transport of ELM filament is estimated as 0.5%

of the total stored energy, which could be utilized to energy loss

transport model for the application to ITER [17,19] .

cknowledgement

This research was supported by National R&D Program

hrough the National Research Foundation of Korea (NRF)

unded by the Ministry of Science, ICT & Future Planning

2015M1A7A1A01002784 ), and was partly supported by KSTAR

roject and National Research Council of Science and Technology

NST) under the international collaboration & research in Asian

ountries ( PG-1314 ).

eferences

[1] A. Kirk , Phys. Rev. Lett. 96 (2006) 185001 . [2] A. Kirk , Plasma Phys. Control. Fusion 53 (2011) 035003 .

[3] A. Herrmann , et al. , J. Nucl. Mater. 363-365 (2007) 528 . [4] H.W. Müller , Nucl. Fusion 51 (2011) 073023 .

[5] A. Schmid , Plasma Phys. Control. Fusion 50 (2008) 045007 .

[6] D.L. Rudakov , Nucl. Fusion 45 (2005) 1589 . [7] G.S. Yun , Phys. Rev. Lett. 107 (2011) 045004 .

[8] D.A. D’lppolito , Phys. Plasmas 18 (2011) 060501 . [9] S.I. Krasheninnikov , Phys. Lett. A 283 (2001) 368 .

[10] T. Hirai , et al. , Mater. Trans. 46 (2005) 412 . [11] T. Hirai , Phys. Scr. T167 (2016) 014072 .

[12] A. Herrmann , J. Nucl. Mater. 313-316 (2003) 759 .

[13] N. Asakura , Plasma Fusion Res. 4 (2009) 021 . [14] H.W. Müller , J. Nucl. Mater. 463 (2015) 739 .

[15] J. Marki , et al. , J. Nucl. Mater. 363-365 (2007) 382 . [16] A. Kirk , Plasma Phys. Control. Fusion 48 (2006) B433 .

[17] M. Kocan , J. Nucl. Mater. 463 (2015) 39 . [18] C. Silva , Plasma Phys. Control. Fusion 51 (2009) 105001 .

[19] W. Fundamenski , R.A. Pitts , Plasma Phys. Control. Fusion 48 (2006) 109 .