communica ions physics A icle
A Na u e Po olio jou nal
h ps://doi.o g/10.1038/s42005-025-02121-1
Di e o shaping wi h neu al ba fling as a
solu ion o he okamak powe exhaus
challenge
Check o upda es
Ke in Ve haegh 1,2 , James Ha ison1, Da id Moul on1, B uce Lipschul z 3,NicolaLonig o 1,3,
Nick Osbo ne1,4, Pe e Ryan1, Ch is ian Theile 5,TijsWijkamp 2,6, Dominik B ida7,CydCowley
1,3,
Gijs De ks 2,6,RhysDoyle 8, Fabio Fede ici 9, Bob Kool 2,6, Oli ie Fé ie 5, An i Hakola10,
S ua Hende son 1, Holge Reime des5, And ew Tho n on1, Nicola Vianello 11,Ma coWischmeie
7,
Lingyan Xiang1, he EURO usion Tokamak Exploi a ion Team*& he MAST Upg ade Team*
Exhaus ing powe om he ho usion co e o he plasma- acing componen s is one usion ene gy’s
bigges challenges. The MAST Upg ade okamak uniquely in eg a es s ong con ainmen o neu als
wi hin he exhaus a ea (di e o ) wi h ex eme di e o shaping capabili y. By sys ema ically al e ing
he di e o shape, his s udy shows he s onges e idence o da e o ou knowledge ha long-legged
di e o s wi h a high magne ic field g adien ( o al flux expansion) deli e key powe exhaus benefi s
wi hou ad e sely impac ing he ho usion co e. These benefi s a e al eady achie ed wi h ela i ely
modes geome y adjus men s ha a e mo e easible o in eg a e in eac o designs. Benefi s include
educed a ge hea loads and imp o ed access o, and s abili y o , a neu al gas bu e ha ‘shields’
he a ge and enhances powe exhaus (de achmen ). Analysis and model compa isons shows hese
benefi s a e ob ained by combining mul iple shaping aspec s: long-legged di e o s ha e expanded
plasma-neu al in e ac ion olume ha d i e educ ions in pa icle and powe loads, while o al flux
expansionenhancesde achmen access ands abili y.Con aining heneu als in heexhaus a eawi h
physical s uc u es u he augmen s hese shaping benefi s. These esul s demons a e s a egic
a ia ion in he di e o geome y and magne ic opology is a po en ial solu ion o one o usion’s
powe exhaus challenge.
Sus ainable nuclea usion is one o he mos p omising solu ions o he
wo ld’s ene gy challenges, o e ing an essen ially limi less and clean ene gy
sou ce. Howe e , one o he c i ical hu dles in de eloping iable usion
eac o s is e ficien ly managing i s powe exhaus : emo ing hea and pa -
icles om he ho using co e while educing su ace hea fluxes o su fi-
cien ly low le els o p e en damaging he eac o ’s componen s1,2.
Combining expe imen s, analysis and model esul s om he MAST
Upg ade okamak, his s udy no only demons a es ha inno a i e shaping
o he powe exhaus egion can sol e his c i ical challenge, bu also explains
he physics and syne gy be ween combining di e en powe exhaus
shaping s a egies.
In magne ic confinemen usion, such as okamaks and s ella a o s3,
he ho usion co e plasma is confinedwi hinnes edmagne icfield lines
(‘closed flux ubes’). Hea and pa icles a e expelled om he co e in o he
edge egion, whe e hey ollow he ‘open flux ubes’ o ming he Sc ape-O
Laye (Fig. 1a). Coils enable al e ing he magne ic opology o hese open
flux ubes o c ea e a magne ic null poin (‘X-poin ’), which di e s hea and
pa icle fluxes o a dedica ed egion called he ‘di e o ’(Fig. 1b). Since he
powe exhaus is ca ied by cha ged pa icles ollowing he flux su aces,
he na ow wid h o he SOL esul s in ex eme hea fluxes (150 MW m−2 o
he DEMO eac o design1,2) due o he na ow plasma we ed a ea, a
exceeding enginee ing limi s (5–10 MW m−21,2) i unmi iga ed.
1Uni ed Kingdom A omic Ene gy Au ho i y, Culham, UK. 2Eindho en Uni e si y o Technology, Eindho en, The Ne he lands. 3Yo k Plasma Ins i u e, Uni e si y o
Yo k, Yo k, UK. 4Uni e si y o Li e pool, Li e pool, UK. 5Swiss Plasma Cen e, École Poly echnique Fédé ale de Lausanne, Lausanne, Swi ze land. 6Du ch Ins i u e
o Fundamen al Ene gy Resea ch DIFFER, Eindho en, The Ne he lands. 7Max Planck Ins i u e o Plasma Physics, Ga ching, Ge many. 8Dublin Ci y Uni e si y,
Dublin, I eland. 9Oak Ridge Na ional Labo a o y, Oak Ridge, TN, USA. 10VTT Technical Resea ch Cen e o Finland, Espoo, Finland. 11Conso zio RFX, Pado a, I aly.
*Lis s o au ho s and hei a filia ions appea s a he end o he pape . e-mail: [email p o ec ed]
Communica ions Physics | (2025) 8:215 1
1234567890():,;
1234567890():,;
To educe a ge hea loads, he powe mus be sp ead o e a la ge
a ea.This isfi s achie ed by injec ing adia ing impu i y gasses o hyd ogen
uel o cool he di e o plasma and con e he hea ca ied by cha ged
pa icles in o hea ca ied by pho ons ( adia ion) ha do no ollow he
magne ic field lines and hus dissipa e he powe olume ically4. Howe e ,
cooling he di e o plasma inc eases he ion a ge fluxes and associa ed
powe loading om su ace ecombina ion, limi ing he o al possible powe
educ ion o a ac o ~ 45.‘Di e o de achmen ’is a p ocess ha educes he
ion a ge flux, enabling u he powe dissipa ion and o de -o -magni ude
a ge hea flux educ ions4–7. A elec on empe a u es o ≤~3–5eV, he
ionising plasma ‘de aches’ om he a ge , o ming a neu al bu e below
he ionising plasma o ‘de achmen ’/ionisa ion on (Fig. 1c, d). Plasma-
a om/molecule in e ac ions, wi hin ha bu e , cause simul aneous powe ,
pa icle (e.g. ion), and momen um losses, ha collec i ely d i e de achmen
(see Me hods sec ion). Recombining he ions in o neu al a oms, h ough
ion sinks like Molecula Ac i a ed Recombina ion (MAR) and Elec on-Ion
Recombina ion (EIR) (Fig. 1c, d), plays a c i ical ole in de achmen 5–8.One
d awback o de achmen is ha i can be highly sensi i e o changes in co e
powe , impu i y seeding and uelling. A high sensi i i y could mo e easily
esul in a loss o de achmen , damaging he eac o walls, o in he de ached
egion eaching he ho using co e, esul ing in a adia i e collapse o he
plasma9 ha can ca as ophically damage a eac o 10. The e o e, we e e o
educing he de achmen sensi i i y as an inc eased de achmen s abili y o
simplici y.
Despi e ad ances in unde s anding and maximising he mi iga ion o
hea /pa icle fluxes h ough plasma de achmen , main aining co e pe o -
mance while e ec i ely exhaus ing powe emainsamajo challengeand
key unce ain y o u u e eac o s2. Compac eac o designs like STEP11,12,
SPARC13,ST-F1/E1
14 and ARC15,16, aiming o accele a e he pa hway o
usion ene gy14,15, ace e en la ge powe exhaus challenges. Inno a i e
powe exhaus solu ions a e hus equi ed o compac usion eac o s11,13,16
and, as isk mi iga ion, o DEMO and beyond17–19. This includes (combi-
na ions o ) liquid me al a ge s20,21, high impu i y injec ion o induce
X-poin adia o s22–24, and Al e na i e Di e o Configu a ions (ADCs)25–27.
ADCs use coils o op imise he di e o magne ic opology o educe hea
loads whils main aining a ho usion co e25; inc ease he ange o co e
condi ions o which de achmen can be achie ed25 and imp o e he s a-
bili y o de achmen 28 (see ‘Me hods’sec ion).
One p omising ADC app oach combines long-legged di e o s27,29,
achie ed by inc easing he dis ance be ween he X-poin and he a ge o
inc ease he powe dissipa ion olume, wi h using he di e o magne ic
opology o sp ead he powe o e a la ge a ge a ea: poloidal and o al flux
expansion. Poloidal flux expansion inc eases he dis ance be ween poloidal
magne ic flux ubes (poloidal flux expansion Fx¼Bu
θB
ϕ
B
θBu
ϕsee Me hods),
whe eas o al flux expansion inc eases he magne ic field g adien by
inc easing he a ge adius ( o al flux expansion FR¼B
Bxp ). ADCs can be
u he op imised by con aining he neu al pa icles wi hin he di e o
chambe using ba flepla es(Fig.1a), boos ing plasma-neu al in e ac ions
and p e en ing neu als escaping o he co e whe e hey cool he usion co e
plasma, hus enhancing co e-edge compa ibili y29,30.
MAST Upg ade is he UK’s na ional usion expe imen , newly buil o
ackle usion’s powe exhaus challenge. MAST-U’s design uniquely in e-
g a es s ong neu al ba fling, long-legged di e o s ((poloidal) di e o leg
leng h/majo adius >1), and high o al flux expansion (up o 2.5). In con-
as , con en ional di e o solu ions (on JET, Asdex-Upg ade31)ha e
sho -legged di e o s (di e o leg leng h/majo adius <0.1), negligible
o al flux expansion (F
R
~ 1) and no neu al ba fles. The sphe ical (‘apple-
shaped’)na u eo MAST-UenablesF
R
a ia ions o e a much la ge ange
(1–2.5) han possible in con en ional (‘doughnu -shaped’) okamaks wi h
flexible shaping, such as TCV (1–1.6)25. P elimina y MAST-U esul s unde
low powe (Ohmic) condi ions (P
SOL
= 0.4 MW) demons a e he benefi s
o he ‘Supe -X Di e o ’(SXD)32,33, which has he highes F
R
achie able,
o e he con en ional di e o (F
R
=1.2)
8,34,35; consis en wi h simula ions35.
This wo k shows he key expe imen al esul s o explo ing al e na i e
di e o solu ions on MAST-Upg ade. Since he e is a con inuum o ADC
solu ions, his wo k s udies he impac o a ying o al flux expansion and
di e o leg leng h ins ead o ocusing only on he Supe -X opology, using
plasmas wi h highe powe (1.5–1.7 MW Neu al Beam Injec ion (NBI)
hea ing, P
SOL
= 1.2 MW). This no only p o ides he s onges expe imen al
e idence o da e o he benefi s o ADCs o ou knowledge by combining
o al flux expansion, di e o leg leng h and neu al ba fling; bu also shows
hese benefi s a e main ained a mo e mode a e di e o shaping (lowe F
R
and sho e leg leng hs han he maximum alues). These powe exhaus
benefi s a e ob ained wi hou any ad e se co e impac and include a ge
hea flux educ ions; imp o ed access o, and s abili y o , plasma de ach-
men ; as well as imp o ed co e-edge compa ibili y. Since enginee ing
complexi y is a c i ical hu dle o in eg a ing ADCs in eac o s, he finding
ha hei benefi scanbemain aineda mo emode a eshapingiso key
impo ance o usion eac o enginee ing and design18,19, ad ancing he
pa h o using ADCs o achie e sus ainable usion ene gy.
Low empe a u e plasma physics and plasma chemis y a e cen al o
plasma de achmen . In his wo k we un a el hese p ocesses as a unc ion o
di e o shape h ough ad anced analysis echniques8and model com-
pa isons. This shows insigh s in o how di e en shaping aspec s wo k
oge he o achie e he obse ed benefi s. The di e o poloidal leg leng h/
olume esul s in addi ional powe /pa icle losses wi hou impac ing he
plasma ups eam (‘Volume long-legged di e o s d i es powe and pa icle
a)
R (m)
Z (m)
02
-2
2
D2
DMS
0
6
0.6 0.8 1.0 1.2
Pol. dis ance X-poin (m)
Ion sou ce/sink
(1021 pa /m2/s)
Ion sou ce MAR ion sink
ion sink
De achmen on EIR
b)
c)
d)
X-poin
Lowe
di e o
#46866 CD =0.45 s
#46860 SXD =0.45 s
#47079 ED =0.45 s
Neu al ba le
Fig. 1 | O e iew o MAST-U plasma shapes and di e o p ocesses. a O e iew o
he magne ic geome y o he Supe -X Di e o (SXD34, blue), Elonga ed Di e o
(ED, g een) and Con en ional Di e o (CD, g een), om he indica ed discha ges/
imes oge he wi h he uelling (D
2
) loca ion and fi s wall geome y. bLowe
di e o wi h diagnos ic co e age o he Di e o Moni o ing Spec ome e (DMS,
g ey)8,40, X-poin posi ion and neu al ba fle loca ion (shaded cyan). cSchema ic
illus a ion o he cha ac e is ic p ocesses in a de ached MAST-U Supe -X di e o
ea u ing he ion sou ce (magen a), Molecula Ac i a ed Recombina ion (MAR) ion
sink (o ange) and Elec on Ion Recombina ion (EIR) ion sink (cyan). dMeasu ed
1D p ofile o he schema ic di e o p ocesses in (c), ob ained om line-in eg a ed
spec oscopic in e ences (# 46860 a 45% G eenwald ac ion34) (ions m−2s−1)as
unc ion o poloidal dis ance om he X-poin o he a ge , indica ed wi h a g ey
a ow in bo h cand d. The de achmen (o ionisa ion) on posi ion is indica ed
wi h a do ed magen a line in bo h cand d. The expe imen al esul s in (d) a e
de i ed om a p obabilis ic sample ob ained om a Bayesian spec oscopic analysis,
showing he median (solid lines) and he 68% equal- ailed confidence in e al
(shaded egion). See ‘Me hods’sec ion o mo e in o ma ion abou he analysis and
unce ain y p opaga ion.
h ps://doi.o g/10.1038/s42005-025-02121-1 A icle
Communica ions Physics | (2025) 8:215 2
losses’sec ion). To al flux expansion imp o es access o, and s abili y o ,
de achmen (‘To al flux expansion imp o es de achmen access and s abi-
li y’sec ion). S ong neu al ba fling enables o al flux expansion benefi s
and augmen s plasma-neu al in e ac ions, maximising he benefi s o
di e o poloidal leg leng h (‘Neu al apping enables powe exhaus
shaping benefi s’sec ion). This imp o es ou unde s anding o how di e o
shaping can imp o e powe exhaus in ag eemen wi h educed model
p edic ions (‘To alflux expansion imp o es de achmen access and s abili y’
sec ion) and simula ions (‘Exhaus simula ions’sec ion), u he ad ancing
he pa h o using ADCs o achie e sus ainable usion ene gy.
Resul s
By sys ema ically compa ing h ee di e o geome ies: he Con en ional
Di e o (CD); Elonga ed Di e o (ED) and Supe -X Di e o (SXD)
(Fig. 1, di e o shape pa ame e s shown in Table 1), we ob ain fi e benefi s
o combined o al flux expansion, poloidal leg leng h and di e o neu al
ba fling.
(i) Imp o ed access o de achmen : de achmen occu s a lowe co e
densi y.
(ii) Inc eased ope a ional egime o de ached di e o ope a ion: he
ange o co e densi y and powe s a which he di e o de ached is
la ge .
(iii) Imp o ed de achmen s abili y: he sensi i i y o de achmen o
changes in co e densi y is educed.
(i ) Reduced a ge hea fluxes and powe loads.
( ) Imp o ed powe exhaus wi hou ad e se co e impac : co e pe o -
mance in de ached condi ions is imp o ed.
Fo eachdi e o configu a ion, he e olu ion o hei powe exhaus
and de achmen p ope ies a e diagnosed as he co e elec on densi y is
g adually inc eased and he di e o condi ions g ow colde , whils o he
pa ame e s a e held as cons an as possible. Fo ease o e e ence and o
compa ison agains li e a u e, he line-a e aged co e elec on densi y
(ob ained om in e e ome y) is exp essed as a ac ion o he maximum
co e densi y, G eenwald, limi 36 which depends on he plasma cu en and
okamak size.
Powe exhaus benefi s
The longe legged, o ally flux expanded, di e o s ha e imp o ed access o
de achmen a lowe co e plasma densi ies. In he CD, he in eg a ed a ge
pa icle fluxes (Fig. 2a) inc ease as unc ion o co e densi y, indica i e o an
a ached di e o plasma, up o a co e G eenwald ac ion o
GW
≈40%. A
his poin , bo h he pa icle flux a he a ge dec eases and he ionisa ion
on de aches om he a ge (Fig. 2b), indica i e o he onse o de ach-
men . In con as , he ED and SXD a e de ached h oughou he scanned
co e densi y ange: he pa icle flux does no inc ease wi h inc easing densi y
whils he ionisa ion on emains de ached. Since he e is no di e ence in
he densi y limi achie able be ween he di e en geome ies, he ope a-
ional window (in e ms o he co e densi y ange o which he discha ge is
de ached) o de ached ope a ion is inc eased o he long-legged di e o s
compa ed o he CD.
The longe -legged, o ally flux expanded, di e o s ha e a highe
de achmen s abili y o quasi-s eady-s a e changes in co e pa ame e s,
quali a i ely consis en wi h educed (s eady-s a e) models (‘De achmen
Loca ion Sensi i i y (DLS) model’sec ion28). The sensi i i y o he de ach-
men on o changes in co e densi y, i.e, he slope o he de achmen on
posi ion (Fig. 2b) is a ac o 5 s eepe o he CD (
GW
≈40%), compa ed o
he ED and SXD a his co e densi y: he de achmen on is much mo e
sensi i e o changes in co e densi y o he CD. These s eady-s a e esul s
indica e an inhe en s abilisa ion, akin o a shock abso be , o he de ach-
men on o he ED and SXD, in con as o he CD whe e he ionisa ion
egion mo es wi h minimal co e changes ou o he di e o chambe a e
de achmen , inc easing co e and X-poin adia ion (Fig. 2h). Al hough he
educed de achmen sensi i i y and inc eased ope a ional window o
de achmen a e ela ed, hey a e no iden ical. Using he analogy o a shock
abso be , a educed de achmen sensi i i y (enabled by o al flux expansion)
co esponds o a s onge damping, whe eas hewide de achedope a ional
egime (enabled by poloidal leg leng h combined wi h o al flux expansion)
inc eases he displacemen he sp ing can unde go be o e he elas ic limi is
exceeded: bo h wo k in unison enhancing de achmen s abili y o long-
legged, o ally flux expanded di e o s.
These benefi s ex end o dynamic a ia ions in uelling37 and hea ing
pe u ba ions34: indica ing an inhe en lesse esponse o he de achmen
loca ion o uelling/hea ing ansien sandimp o edde achmen con ol o
he ED and SXD37. In con as o ou quasi-s eady-s a e expe imen s, hese
dynamic a ia ions ea u e mo e han se en imes as e uelling changes o
which he di e o esponds dynamically37.
The longe -legged, o ally flux expanded, di e o s esul in la ge hea
flux educ ions han expec ed. Based on he magne ic geome y25,a
educ ion in pe pendicula hea flux by ~5.8× and ~2.1× o he SXD and
ED is expec ed, compa ed o he CD, mainly due o inc eased poloidal
(F
x
~3.0 × (SXD) and ~1.6× (ED)) and o al (F
R
~2.0 × (SXD) and ~1.4×
(ED)) flux expansion. Howe e , a much la ge educ ion in a ge hea flux is
obse ed: ~18.5× and ~7 × o he SXD and ED, compa ed o he CD
(Fig. 2c). The longe -legged di e o s esul in addi ional hea flux dis-
sipa ion h ough olume ic and/o adial/c oss-field anspo by a ac o
~3.2× , quali a i ely consis en wi h bo h SOLPS-ITER simula ions
(‘Exhaus simula ions’sec ion)35 and olume ic powe loss es ima es
(‘Volume long-legged di e o s d i es powe and pa icle losses’sec ion).
The longe -legged, o ally flux expanded, di e o s enable di e o
de achmen wi hou ad e se co e impac , in con as o he CD - which
needs high densi ies o de ach (
GW
>40%). The co e densi ies, empe a u es
and P
SOL
a e simila o he CD, ED and SXD (Fig. 2d–h). These esul s
indica e a s ong decoupling be ween he di e o shape and he ob ained
co e condi ions, e en when he ou e a ge is de ached.
A e ha ing shown he benefi s o s ongly ba fled, long-legged,
o ally flux-expanded, di e o s, we will explo e why hese di e o
configu a ions ha e a supe io exhaus pe o mance using spec oscopic
analysis38, educed models and simula ion compa isons. This shows
poloidal leg leng h, o al flux expansion and s ong neu al ba fling
s eng hen each o he ’s impac and all wo k oge he o achie e he
obse ed benefi s. The addi ional leg olume in he SXD and ED, com-
pa ed o he CD, esul s in hei supe io powe dissipa ion and d i es he
educ ion o he ion a ge flux du ing de achmen h ough ion sinks,
whe eas o al flux expansion d i es educ ions in de achmen onse and
imp o es de achmen on s abili y. Neu al ba fling augmen s plasma-
neu al in e ac ions, s eng hening he benefi s o poloidal leg leng h, and
Table 1 | Summa y o di e o magne ic shape pa ame e s
Discha ge R
(m) F
x
F
R
L(m) L
pol
(m) Desc ip ion
46860 1.45 9 2.3 19 1.3 Supe -X Di e o (SXD)
47079 1.11 6 1.7 17 1.1 Elonga ed Di e o (ED)
46762 0.79 3.3 1.2 13 0.64 Con en ional Di e o (CD)
46895 0.81–1.39 4–6 1.2–2.2 13–19 0.65–1.3 CD ->ED ->SXD scan
Ta ge adius (R
), poloidal flux expansion (F
x
), o al flux expansion (F
R
), connec ion leng h om he ups eam midplane o he a ge (L) and poloidal leg leng h om he X-poin o he a ge (L
pol
) o he
discha ges s udied. Discha ge # 46895 keeps he co e densi y and powe cons an as he di e o opology is changed o e he ange o he fi s h ee discha ges.
h ps://doi.o g/10.1038/s42005-025-02121-1 A icle
Communica ions Physics | (2025) 8:215 3
enables o al flux expansion benefi s by p e en ing neu al leakage o
he co e.
Volume long-legged di e o s d i es powe and pa icle losses
Pa icle balance analysis shows inc eased ion sinks educe he a ge fluxes in
heSXDandEDcompa ed o heCD(Fig.3a–c), whe eas he o al ion
sou ce is he same wi hin unce ain ies o all h ee di e en geome ies. Ion
sinks a e significan in bo h he SXD and he ED om he s a o hose
discha ges, bo h h ough MAR as well as EIR (in he SXD). Ou spec o-
scopic analysis e eals plasma condi ions o n
e
=2–4×10
19 m−3and
T
e
≈0.2 eV in he egion whe e EIR becomes obse able (
GW
> 33% in he
SXD and
GW
> 40% in he ED)34,39. MAR only appea s in he CD a he
highes co e densi ies a e i s ionisa ion on de aches om he a ge
(
GW
> 40%), bu i s magni ude emains limi ed downs eam he ba fle.
The o al ion sou ce, in e ed h ough pa icle balance, is ob ained by
adding he ion a ge flux and he ion sinks obse ed in he di e o chambe
(see Me hods sec ion). This consis s ou o he di e o chambe ion sou ce
and any ne inflow o ions in o he di e o chambe . The ion sou ce is
gene a ed be ween he X-poin , p edominan ly downs eam he mos
ups eam-end o he ba fle(∣Z∣=1.3m, Fig. 1): he neu al ba fling is
e ec i e in limi ing he ion sou ce ups eam o he X-poin o all h ee
configu a ions, consis en wi h simula ion esul s (‘Exhaus simula ions’
sec ion). Howe e , only up o 40% o he o al ion sou ce is gene a ed
downs eam he mos downs eam-end o he ba fle(∣Z∣=1.55m,Fig.1)in
he di e o chambe 34.
Analogously o he pa icle flux educ ion, i is he addi ional olu-
me ic powe dissipa ion in hei di e o olume (Fig. 3d– ) ha d i es he
educ ion in a ge powe loads o he SXD and ED. The in e ed powe
flowing in o he di e o chambe is simila o all h ee geome ies
(Fig. 3d– o
GW
<40%). As he in e ed hyd ogenic adia ion is simila o
he o al measu ed adia ion om an imaging bolome e (no shown)40, he
di e o chambe powe losses mos ly a ise om hyd ogenic p ocesses.
These hyd ogenic powe losses educe P
a ge
by a ac o ~×4 (SXD) and
~×2 (ED) compa ed o he CD; consis en wi h he a ge hea load
educ ion being la ge han expec ed based on geome y (Fig. 2c). A sig-
nifican pa o he ED and SXD hyd ogenic powe losses o igina e in he
de ached egime om Molecula Ac i a ed Dissocia ion (MAD). Elas ic
collisions be ween he plasma and he neu al cloud, which is neglec ed
abo e, can u he augmen he powe losses in he ED and SXD h ough
adial anspo o neu als by up o 15% o P
SOL
41 acco ding o he SOLPS-
ITER simula ions shown in ‘Exhaus simula ions’sec ion. This would occu
in he de ached egion and depend on he plasma-neu al in e ac ion
olume downs eam he ionisa ion on , highligh ing he benefi s o
inc eased di e o leg leng h41.
The ED and SXD main ain s ong adia ion in he di e o chambe .
Tha is in con as o he CD configu a ion whe e he adia ion ups eam o
he di e o ba flelowe sPdi a e he de achmen onse (
GW
> 40%),
consis en wi h he much highe de achmen on loca ion sensi i i y o
changes in he co e densi y (Fig. 2b). These esul s illus a e longe -legged
di e o s can 1) inc ease maximum (di e o ) powe dissipa ion; 2) main-
ain powe losses away om he X-poin owa ds he di e o a ge ; and 3)
u he enhance powe losses a e he onse o de achmen .
The impac o di e o shaping on powe and pa icle exhaus is
e ealed when s udying he 1D p ofiles o ion sou ces and sinks (a–c),aswell
as powe flows (g), along he di e o leg as unc ion o poloidal dis ance o
he X-poin a a fixed co e densi y (
GW
= 35%) (Fig. 4). Bo h p ofiles a e
simila be ween he di e en geome ies (up un il he CD de achmen
onse ) a he same poloidal dis ance o he X-poin : he plasma is hus
p edominan ly al e ed in he ex ended egion. The deepe de achmen and
lowe powe loadsin heSXDandEDa eb ough onbyin e ac ionsin he
addi ional olume a ailable downs eam o he ionisa ion egion when he
di e o leg is ex ended. Plasma-chemis y occu ing in his egion, esul ing
in MAR and MAD, plays a key ole explaining he di e ences be ween he
di e en di e o geome ies.
To al flux expansion imp o es de achmen access and s abili y
To gain u he insigh s in o he impac o di e o shaping on de achmen ,
he expe imen al esul s a e compa ed agains he DLS analy ical
Fig. 2 | Imp o ed di e o pe o mance wi hou
ad e se impac co e o long-legged di e o s.
Compa ison o di e o (a,b,c) and co e (d,e, ,g,h)
pe o mance as unc ion o co e G eenwald ac ion
(
GW
in %) o he CD ( ed), ED (g een) and SXD
(blue). Di e o pa ame e s: aIn eg a ed ion a ge
flux (symbols) (wi h polynomial fi s (solid, shaded
line)), bde achmen (ionisa ion) on posi ion as
poloidal dis ance o he a ge , ces ima ed pe pen-
dicula a ge hea load on a loga i hmic scale,
combining Langmui p obe and spec oscopy
measu emen s (see Me hods)8,40. The esul s in
(band c) a e de i ed om a p obabilis ic sample
ob ained om a Bayesian spec oscopic analysis,
showing he median and he 68% equal- ailed con-
fidence in e al (shaded egion). See ‘Me hods’
sec ion o mo e in o ma ion abou he analysis and
unce ain y p opaga ion. Co e pa ame e s: d–gco e
elec on empe a u es and densi ies a wo di e en
co e G eenwald ac ions (co esponding o e ical
do ed lines in (a,b,c,h)) indica ed by blue c osses
(SXD), g een do s (ED) and ed plusses (CD), hP
SOL
(solid lines) deduced om he ollowing con-
ibu o s: NBI abso p ion (TRANSP, dashed lines);
Ohmic hea ing (EFIT, no shown); changes o s o ed
ene gy (EFIT, no shown) and co e adia i e losses
(bolome y, do ed lines).
h ps://doi.o g/10.1038/s42005-025-02121-1 A icle
Communica ions Physics | (2025) 8:215 4
model28,42,43. The DLS model p edic s de achmen occu s i he pa ame e s
d i ing de achmen (in ou case co e densi y (
GW
)andpowe (P
SOL
),
lumped oge he as C/ne;uffiffiffi
I
p
q5=7
k/ GW
P5=7
SOL
—see ‘Me hods’sec ion) eaches he
de achmen h eshold, C
.C
is a unc ion o he magne ic geome y:
C /1
FRðBxp
<B>Þ2=71
L2=7
k
, depending mos ly on o al flux expansion (F
R
), con-
nec ion leng h (L
∥
, pa allel o he field line) and he a e aged magne ic field
s eng h <B>.To alflux expansion and connec ion leng h bo h lowe C
,
educing he densi y (and impu i y con en - see Me hods sec ion) equi ed
o de achmen a a fixed P
SOL
.
The impac o o al flux expansion on he de achmen onse p edic ed
by he DLS model is consis en wi h ou obse a ions (Table 2). The DLS
p edic ed benefi s o he SXD and ED o e he CD a ise mos ly om an
inc ease in o al flux expansion. Gi en he densi y a which he CD de aches,
he DLS p edic s ha he SXD and ED a e al eady de ached a he lowes
co e densi y achie ed, consis en wi h he expe imen . To compa e he SXD
and ED agains each o he , he onse o EIR is used as a colde e e ence
poin o DLS compa isons, showing ag eemen wi hin 10% o he expe i-
men . Fu he mo e, he DLS model p edic s ha he de achmen on
posi ion only depends on he magne ic field opology ups eam o he
de achmen on (see Me hods sec ion), consis en wi h he obse ed
in a iance o he ups eam pa ame e s o he downs eam di e o magne ic
opology.
Al hough only he magne ic geome y is used o ob aining DLS p e-
dic ions on he di e ences be ween he SXD, ED and CD, i should be no ed
ha he DLS model de i a ion assumes ha : 1) he de achmen on is
infini ely hin; 2) all powe dissipa ion is d i en by impu i y adia ion. Bo h
hese assump ions a e in alid o he MAST-U condi ions shown, al hough
he alidi y o hese assump ions would inc ease in mo e eac o - ele an
condi ions. The ag eemen be ween he DLS model and he MAST-U
esul s sugges s, howe e , ha he impac o di e o opology on he
de achmen onse may be mo e gene ally applicable ou side o impu i y
adia ion dominan condi ions. Fu he mo e, he DLS model neglec s adial
hea anspo , which may be enhanced due o o al flux expansion44.Fu -
he wo k is equi ed gene alising he DLS model o MAST-U like con-
di ions and in es iga ing he impac o adial anspo .
Ou powe and pa icle balance analysis showed he addi ional olume
(o poloidal leg leng h) in he SXD and ED is c i ical o explain he educ ion
o powe and pa icle loads du ing de achmen . To al flux expansion,
ins ead, d i es 80% o he imp o ed access o de achmen acco ding o DLS
model compa isons, wi h he longe di e o leg leng h d i ing he
emaining 20%. The DLS model also p edic s ha o al flux expansion
educes de achmen sensi i i y, quali a i ely consis en wi h ED and SXD
obse a ions (benefi iii): he combina ion o inc easing di e o leg leng h/
olume and o al flux expansion esul in s ong, syne gis ic, powe exhaus
benefi s.
Neu al apping enables powe exhaus shaping benefi s
Al hough he poloidal di e o leg leng h downs eam he ba fle en ance is
e y di e en be ween he he di e en opologies, ou expe imen s sugges
ha he ba fling has a simila ly s ong impac on he CD (up un il i s
de achmen onse ), ED and SXD. The obse ed and simula ed (‘Exhaus
simula ions’sec ion) di e o neu al p essu es a e simila be ween he h ee
geome ies. Likewise, he neu al apping, defined as he a io be ween he
ion sou ce downs eam he X-poin o he o al ion sou ce33, is simila
be ween he h ee di e o opologiesinsimula ions(78% o heSXDand
ED and 75% o he CD, espec i ely), despi e he SXD and ED config-
u a ions being deeply de ached (‘Exhaus simula ions’sec ion). In con as ,
SOLPS-ITER simula ions o a co e densi y amp o he (open, un-ba fled)
con en ional TCV di e o indica ed a neu al apping o 32–45% in
de achmen onse condi ions, dec easing du ing deepe de achmen o
Fig. 3 | Powe and pa icle balance shows addi-
ional olume long-legged di e o d i es powe
and pa icle losses. Pa icle (a–c)andpowe (d– )
balance compa isons be ween di e en di e o
shapes as unc ion o co e G eenwald ac ion.
a–cPa icle balance showing he ion a ge flux
(lowe ou e di e o ) - black, o al ionisa ion
sou ce - magen a, Molecula Ac i a ed Recom-
bina ion (MAR - o ange) and Elec on-Ion
Recombina ion (EIR - cyan) ion sinks (bo h ion
sinks a e in eg a ed o e he di e o chambe )
o he Supe -X Di e o (SXD) (a), Elonga ed
Di e o (ED) (b) and Con en ional Di e o
(CD) (c). d– Powe balance showing hyd ogenic
powe losses Phyd o
loss (o ange, in eg a ed o e he
di e o chambe ), a ge powe deposi ion P
a ge
(black, ob ained om spec ocopically in e ed
empe a u es and Langmui p obe pa icle fluxes)
and es ima ed powe flow in o he di e o
chambe (magen a, Pdi Phyd o
loss þP a ge )
assuming ha he di e o chambe powe losses
a e dominan ly hyd ogenic, in ag eemen wi h
imaging bolome y measu emen s34.Unde he
assump ion ha he lowe and uppe di e o s
a e simila (consis en wi h Langmui p obe
esul s34), Pdi ,Phyd o
loss and P
a ge
ha e been mul-
iplied by wo o ob ain in eg a ed alues o he
uppe and lowe ou e di e o s. The esul s a e
de i ed om a p obabilis ic sample ob ained
om a Bayesian spec oscopic analysis, showing
he median and he 68% equal- ailed confidence
in e al (shaded egion). See Me hods sec ion o
mo e in o ma ion abou he analysis and unce -
ain y p opaga ion.
h ps://doi.o g/10.1038/s42005-025-02121-1 A icle
Communica ions Physics | (2025) 8:215 5
11%33,45. Howe e , he inc eased de achmen on sensi i i y o he CD
likely esul s in a educ ion o neu al apping a e i s de achmen onse ,
diminishing he impac o neu al ba fling on he CD a e i s de ach-
men onse .
The benefi s o long-legged di e o s, o al flux expansion, di e o
shaping and neu al ba fling ha e been indi idually s udied on TCV25,27;
showing benefi s o neu al ba fling30,46–48 and long di e o leg leng hs27.The
benefi s o o al flux expansion (bo h in e ms o de achmen onse and on
sensi i i y/s abili y) we e, howe e , much smalle han p edic ed by he DLS
model49,50. Escape o neu als om he di e o o he SOL ups eam o he
X-poin can lead o s ong plasma flows om he midplane o he a ge 51,52,
which diminishes he impac o o al flux expansion on he de achmen
onse 49. SOLEDGE2D-EIRENE simula ions sugges ha he neu al ba fling
on TCV may be insu ficien o eco e he ull benefi o o alflux
expansion53, consis en wi h p e ious SOLPS-ITER simula ions33,54 which
showed ha he neu al apping o he (open, unba fled) TCV Supe -X
di e o was wo se han ha o he con en ional di e o . This no only
nega es pa o he benefi s o o al flux expansion due o an ion flow om
ups eam he X-poin owa ds he a ge , bu also educes he benefi o he
neu als in powe /pa icle dissipa ion be ween he Supe -X di e o com-
pa ed o he con en ional di e o .
This di e ence be ween MAST-U and TCV, as well as he absence o
s ong flows om he midplane o he X-poin in MAST-U simula ions,
sugges s ha s ong neu al apping, ob ained by ba fling on MAST-U,
enables he shaping benefi s o o al flux expansion. Addi ionally, neu al
ba fling augmen s powe /momen um/pa icle losses om plasma-neu al
in e ac ions such as MAR and MAD by con aining he neu als in he
di e o chambe , ampli ying he benefi s o long-legged di e o s (‘Volume
long-legged di e o s d i es powe and pa icle losses’sec ion). Neu al
ba fling is no equi ed o ob ain s ong neu al apping on high powe
de ices (i.e., JET, AUG) nea a ached condi ions as he neu al mean- ee-
pa hs a e dec eased, leading o a sho e ex en o he ionisa ion egion.
Howe e , ba fling would s ill be equi ed o main ain high neu al apping
in cases whe e he ionisa ion sou ce is significan ly ups eam o he a ge ,
mo i a ing he STEP55,SPARC
13 and ARC16 di e o designs. The MAST-U
esul s sugges ha neu al ba fling can be placed downs eam he X-poin -
he esul ssugges heba fle s uc u e should be: 1) ups eam o he in ended
loca ion o he ionisa ion on ; 2) su ficien ly p ohibi ing he escape o
neu als o he SOL.
Exhaus simula ions
MAST-U56 uniquely in eg a es s ong ba fling and ex eme le els o o al
flux expansion. The DLS de achmen onse p edic ions (C
) a yby110%
be ween he MAST-U CD and SXD. Compa a i ely, C
a ies by 70% o a
ange o TCV di e o geome ies ( a ge adius scan, poloidal flux
Table 2 | DLS educed model p edic ions a e in ag eemen wi h
obse a ions
De achmen onse EIR onse
Model Expe imen Model Expe imen
SXD −55% <−37% 0% Re e ence
ED −40% <−37% +36% +27%
CD 0% Re e ence +106% >+52%
Measu ed and De achmen Loca ion Sensi i i y (DLS) model p edic ed ela i e di e ences (in co e
densi y) be ween he di e en di e o opologies (Supe -X (SXD), Elonga ed (ED) and Con en ional
(CD) Di e o ) o he de achmen onse (T
e
=~3–5 eV) and he onse o Elec on Ion Recombina ion
(EIR), se ing as a colde e e ence poin (T
e
≪1 eV). The expe imen ally obse ed densi y a which
he CD de aches ( e ;CD;de ach
GW ) and a which he EIR occu s in he SXD ( e ;SXD;EIR
GW ) a e used as
e e ence densi ies. The pe cen ages shown a e he obse ed and DLS modelled ela i e densi y
di e ences o he e e ence o de achmen onse ( ED;SXD;de ach
GW = e ;CD;de ach
GW ) and EIR onse
( CD;ED;EIR
GW = e ;SXD;EIR
GW ). These DLS modelled di e ences only depend on magne ic opology, see
Me hods sec ion. The densi y ange ob ainable in he expe imen limi s he di e ences in
de achmen andEIRonse ha can be explo edbe ween hedi e en opologies.The e o e,when<
(o > ) is indica ed, he de achmen o EIR onse is no obse ed in he expe imen and he ela i e
di e ence is la ge han indica ed.
Fig. 4 | Powe flow and 1D ion sou ces/sinks show
simila di e o condi ions a same poloidal dis-
ance o X-poin . Spec oscopically in e ed line-
in eg a ed ion sou ces (magen a) and sinks (Mole-
cula Ac i a ed Recombina ion (MAR) - o ange;
Elec on-Ion Recombina ion - cyan) (pa . m−2s−1)
o he Supe -X (SXD) (a), Elonga ed (ED) (b) and
Con en ional (CD) (c) Di e o s a
GW
= 35% as
unc ion o poloidal dis ance o he X-poin . The ed
(CD), g een (ED) and blue (SXD) e ical colou ed
do ed lines indica e hei espec i e s ike poin
posi ions, indica ed by hei magne ic geome y
(d– ). The 1D ion sou ce/sink p ofiles (a,b,c) a e
ex ended downs eam o hei espec i e s ike-
poin s due o con olu ion o he adial-ex en o he
SOL/ a -SOL wi h he spec oscopic lines-o -sigh ,
whe e he plasma is colde han a he sepa a ix.
gPowe flow (W) owa ds he di e o a ge s as
unc ion o poloidal dis ance o he X-poin om he
di e o en ance o he a ge o he CD ( ed), ED
(g een) and SXD (blue) a
GW
= 35%, wi h e ical
do ed lines indica ing hei espec i e s ike poin s.
The pa whe e he di e o leg is de ached is shaded
in g ey in g ey. The powe flow is in e ed by sub-
ac ing om Pdi he cumula i e sum o he
hyd ogenic powe losses om ups eam o he a -
ge . The esul s a e de i ed om a p obabilis ic
sample ob ained om a Bayesian spec oscopic
analysis, showing he median and he 68% equal-
ailed confidence in e al (shaded egion). See
Me hods sec ion o mo e in o ma ion abou he
analysis and unce ain y p opaga ion.
h ps://doi.o g/10.1038/s42005-025-02121-1 A icle
Communica ions Physics | (2025) 8:215 6
expansionscan,as wellas o he X-poin a ge di e o )25,27.Byin eg a ing
s ong ba fling, long-legged di e o s and o al flux expansion, MAST-U
e ie es he ull benefi o i s shaping capabili y on he de achmen onse
and s abili y ( ac o 5 × educed sensi i i y o ED and SXD s CD, Fig. 2b),
as well as powe and pa icle exhaus capabili y. Wewill now in es iga e he
esul s o a single discha ge unde cons an cons an co e densi y (30%
G eenwald ac ion, nsep
e0:8×1018m3) and powe (P
SOL
≈1.0 MW),
whe e C
was al e ed by 110% by slowly sweeping he ou e s ike poin
om a CD o an ED o a SXD geome y. No significan di e ences a e
obse ed, a he same s ike poin posi ion and co e densi y, du ing his
s ike poin scan a cons an co e densi y wi h he densi y amps p esen ed
p e iously.
The esul s in Fig. 2b sugges ed ha he ionisa ion on posi ion,
once de ached om he a ge , is in a ian o he magne ic opology
downs eam o i : i s loca ion depends only on he ups eam magne ic
opology. This is confi med by he s ike poin sweep discha ge, whe e
he co e condi ions emain mos ly unchanged, wi h a 5-10% inc ease in
ups eam and co e T
e
when ansi ioning om CD o SXD (po en ially
due o he longe connec ion leng h4). A e he D
2
Fulche band emis-
sion on , which is a p oxy o he ionisa ion on 8,57,58, de aches om
he a ge ( a ge adius ≈0.95 m), i emains close o his adial posi ion
as he s ike poin is swep u he and u he ou wa ds and bo h o al
flux expansion and poloidal leg leng h is inc eased (Fig. 5d– ). This
implies ha he ionisa ion on posi ion ( o his P
SOL
and
GW
) emains
a a cons an poloidal dis ance o he X-poin as he di e o leg leng h is
u he inc eased.
This beha iou ag ees wi h SOLPS-ITER p edic ions35 o he CD, ED
andSXDconfigu a ions (Fig. 5a–c). The CD simula ion is a ached, whe eas
he SXD and ED simula ions a e de ached. The adius o bo h he D
2
Fulche
emission on as well as he 5 eV con ou , o he ED and SXD, emains nea
=0.95m. The D
2
Fulche emission nea he X-poin egion is also in
ag eemen be ween expe imen s (Fig. 6), sugges ing ha he s ong neu al
apping ob ained in he simula ions is consis en wi h he expe imen .
Howe e , he simula ions ea u e an a ached inne a ge (Fig. 6a–c) wi h
s ong D
2
Fulche emission nea he inne s ike poin (Fig. 6d– ), which is
no obse ed expe imen ally (Fig. 6g–i). This sugges s he inne a ge powe
loading is negligible in he expe imen s and o e es ima ed in SOLPS-ITER
simula ions. This equi es u he s udy including using mul i-diagnos ic,
Bayesian, o in e plasma pa ame e s ou side he di e o chambe analysis
echniques59.
SOLPS-ITER simula ions sugges adial plasma anspo is d i en
p ima ily in he ionising egime and is hus simila be ween he h ee di -
e en di e o shapes, whils adial hea anspo om neu als occu s
downs eam o i 35. This seems consis en wi h 2D emissi i y and elec on
densi y measu emen s60, which is o be s udied in mo e de ail in u u e wo k
using mo e a ached condi ions and mul i-diagnos ic echniques o in e 2D
plasma pa ame e s59.
A mo e de ailed compa ison be ween expe imen s and simula ions is
ob ained by compa ing hei ion sou ces and sinks (Fig. 5g-i) in he ou e
di e o chambe , indica ing a quan i a i e ag eemen be ween expe imen s
and simula ions o he ion sou ce and EIR. The MAR ion sinks a e
unde es ima ed in he simula ion in he de ached egion. This disc epancy is
esol ed when a co ec ed a e o molecula cha ge exchange34 is used in
SOLPS-ITER (Fig. 5g).
Discussion
Using he unique capabili ies o MAST-U as a es bed o in es iga ing
al e na i e di e o opologies educes unce ain y in ex apola ing cu en
knowledge o eac o class de ices by alida ing bo h educed models and
exhaus simula ions: a c ucial miles one o explo ing ADCs as a eac o
solu ion. The e a e, howe e , key di e ences be ween MAST-U and a
eac o ha mus be add essed.
Fig. 5 | Inc easing he di e o leg leng h does no
al e he ionisa ion egion a e de achmen , in
ag eemen wi h simula ions. Syn he ic D
2
Fulche
emission om SOLPS-ITER simula ions o he
Supe -X (SXD) (a, blue), Elonga ed (ED) (b, g een)
and Con en ional Di e o s (CD (c, ed), o e laid
wi h 5 eV con ou s (dashed lines) and he sepa a ix
(solid line). d– Expe imen ally measu ed D
2
Ful-
che band emission (595-605 nm) o a s ike poin
scan wi h magne ic equilib ium shown, mo ing
om CD o SXD a cons an densi y and powe ,
ob ained h ough in e ing Mul i-Wa eleng h-
Imaging (MWI) imaging da a o # 4689557.g–i1D
ion sou ces and sinks (ionisa ion - magen a, Mole-
cula Ac i a ed Recombina ion (MAR) ion sink -
magen a, Elec on-Ion Recombina ion (EIR) ion
sink - cyan), ob ained om spec oscopic analysis
in eg a ed along he spec oscopic lines o sigh
(Fig. 1b) (pa . m−2s−1), compa ed agains syn he ic
diagnos ic esul s om SOLPS-ITER simula ions
(do ed lines). Fo he SXD (g) wo SOLPS-ITER
simula ion esul s a e shown: one wi h de aul a es
and one wi h co ec ed molecula cha ge exchange
(D2þDþ!D2þþD) a es (‘Sim. Co . Ra e’),
ob ained om34, which inc eases MAR. To guide he
eye, a shaded magen a e ical line has been added a
a adius o 0.95 m and a black a ow has been added
a he s ike poin loca ion (a–i). The expe imen al
esul s in (g,h,i) a e de i ed om a p obabilis ic
sample ob ained om a Bayesian spec oscopic
analysis, showing he median (solid lines) and he
68% equal- ailed confidence in e al (shaded
egion). See Me hods sec ion o mo e in o ma ion
abou he analysis and unce ain y p opaga ion.
h ps://doi.o g/10.1038/s42005-025-02121-1 A icle
Communica ions Physics | (2025) 8:215 7
Fi s , eac o s will ope a e a highe powe inpu and smalle hea flux
wid hs han MAST-U. As a esul , hea loads will become so la ge ha
impu i y seeding will be equi ed o lowe a ge empe a u es su ficien ly o
enable de achmen . Unde s anding how combining o al flux expansion,
di e o neu al ba fling, and poloidal leg leng h a ec s impu i y-d i en
powe dissipa ion equi es u he in es iga ion. Reac o -scale simula ions
ha e, howe e , demons a ed ha ADCs enhance powe dissipa ion,
enabling ope a ion wi h a ge hea fluxes below enginee ing limi s wi h
educed impu i y concen a ions11,16,17.
Al hough highe powe condi ions sho en neu al mean- ee-pa hs -
impac ing neu al anspo , plasma-molecula in e ac ions a e expec ed o
play a key ole in ob aining significan powe , momen um and pa icle
dissipa ion du ing (deep) de achmen acco ding o exhaus simula ions ha
inco po a e long-legged, igh ly ba fled, di e o s11,34,41. The c i ical ole
plasma-molecula chemis yplaysin heEDandSXDillumina eddis-
c epancies wi h simula ions (Fig. 5h) ha ha e been educed wi h imp o ed
a es o D2þDþ!D2þþD. Ex apola ing hese imp o ed a es o
eac o s wi h long-legged, igh ly ba fled, di e o s shows hey can make a
c i ical impac on he eac o scale34.
Secondly, eac o -g ade ope a ion ypically in ol es H(igh confine-
men )-mode ope a ion. Cu en H-mode ope a ion exhibi s iolen ELMs
ha esul in ex emely high hea loads. Al hough eac o s will aim o
minimise o supp ess ELMs, i emains unclea whe he ADCs can e ec-
i ely mi iga e he hea loads o esidual ( as , ELM) ansien s o inc ease
di e o li e imes.
Thi dly, ou MAST-U esul s show a balanced double-null config-
u a ion, wi h simila pa icle fluxes eaching he lowe and uppe ou e
di e o s and negligible powe eaching he inne a ge (Fig. 6). Achie ing
such up/down balance in eac o s poses challenges due o a la ge dis ance
be ween he X-poin and poloidal field coils and smalle sc ape-o -laye
wid hs12 (10–12 mm o he cu en MAST-U expe imen s; 1–2mm o
STEP55 and 0.2–0.4 mm o SPARC13). Al hough he exhaus benefi s o
double null may be limi ed acco ding o eac o exhaus simula ions61 and
TCV expe imen s50, double-null imbalances exace ba es inne a ge powe
loading and may necessi a e solu ions o sphe ical okamak eac o s, such
as a p oposed inne a ge X-Di e o geome y11,55 equi ing u he
expe imen al alida ion.
One conce n o ou e a ge op imisa ion s a egies is ha he inc eased
ou e a ge connec ion leng h will exace ba e he inne a ge hea load in
single null condi ions in a ached condi ions acco ding o educed models62.
This was in con as o de ached DEMO Supe -X di e o simula ions,
which indica ed inne a ge hea loads we e educed wi h ou e di e o
op imisa ion17,19, equi ing u he s udy.
This educed model ( o a ached condi ions) p edic s an inc ease o
inne a ge hea loadsby31%(ED)and 57% (SXD) compa ed o he CD.
Howe e , hese inc eased inne a ge hea loads a e educed when
accoun ing o ou e a ge de achmen . Using he ED and SXD ou e
di e o de achmen on as a i ual a ge , he p edic ed addi ional inne
a ge hea loads a e educed below 15% compa ed o an ou e a ge
a ached CD.
Fu u e MAST-U expe imen s aim o add ess hese key di e ences and
ad ance owa ds mo e eac o - ele an scena ios. Planned upg ades
include inc eased ex e nal hea ing om 4.4 MW o o e 10 MW (>2026),
enabling ho e , mo e a ached di e o condi ions. C yopumping has been
ins alled o educe di e o neu al p essu es and ob ain ho e di e o
condi ions (2025). Ad anced scena io de elopmen (2025) may enable
enable single- o-double-null compa isons. P elimina y esul s sugges ha
he benefi s o ADCs obse ed in his s udy pe sis in H-mode and may e en
bu e small ELMs, equi ing u he in es iga ion40.
O e all, ou esul s demons a e ha ADCs no only imp o e exhaus
pe o mance, enabling educed ups eam densi y and likely hus impu i y
concen a ion/co e adia ion in eac o s17, bu also imp o e co e-edge
compa ibili y when pai ed wi h s ong ba fling. This allows o de ached
di e o ope a ion wi hou comp omising co e condi ions: a majo
Fig. 6 | O e iew o simula ed and expe imen al
ion sou ces and D
2
Fulche emission o di e en
di e o shapes. a–c2D ionisa ion sou ce om
SOLPS-ITER simula ions (shown in Fig. 5) wi h
ho izon al lines a z=−1.6 m (pink) and z= 1.07 m
(magen a), dema king he edge o he di e o
spec oscopy iew and X-poin , espec i ely. The
ac ion o he ion sou ce downs eam hese limi s
compa ed o he o al ion sou ce (ou e leg only) a e
no ed. d– Syn he ic diagnos ic o he D
2
Fulche
emissi i y (a bi a y uni s) ob ained om SOLPS-
ITER simula ions. g–iMeasu ed D
2
Fulche emis-
si i y (595-605 nm) ob ained om combined
di e o imaging57 and X-poin imaging in e sions.
The indica ed ime and discha ges used a e shown
and a e ob ained om epea discha ges o he same
co e densi y as used in Fig. 5. A ho izon al line a he
heigh o he X-poin loca ion is added (magen a).
Only emissi i ies ob ained a he same ,zco e-
sponding o he simula ion g ids a e shown. An
in e sion a e ac is p esen nea = 0.85 m,
z=−1.6 m, whe e he e is a gap in co e age be ween
he X-poin and di e o imaging sys ems. Da a a e
shown o he Supe -X (SXD, blue, a,d,g), Elonga ed
(ED, g een, b,e,h) and Con en ional (CD, ed, c, ,i)
Di e o s.
h ps://doi.o g/10.1038/s42005-025-02121-1 A icle
Communica ions Physics | (2025) 8:215 8
miles one owa ds p o ing he applicabili y o ADCs in eac o s. Howe e ,
any eac o design needs a comp omise be ween enginee ing complexi y
and a ac i e ope a ing egimes18,19. The inc eased enginee ing complexi y
o some ADCs, such as he Supe -X di e o , emains a significan con-
side a ion o eac o designs due o associa ed cos , space cons ain s and
magne ic con ol ole ances18,19.Ou findings unde sco e ha smalle
modifica ions o di e o opology, such as ansi ioning om CD o ED,
can achie e significan pe o mance gains ha a e consis en wi h educed
model p edic ions and exhaus simula ions. ADC design is he e o e a
con inuum—an insigh ha has implica ions o eac o designs o DEMO,
as well as mo e compac machines (SPARC, ARC, STEP)13,16,55,andpa es
he way o designs ha op imise powe exhaus and co e-edge compa -
ibili y wi h educed enginee ing and in eg a ion demands18,19.
Me hods
MAST upg ade, al e na i e di e o configu a ions and he
supe -X di e o
MAST Upg ade is he UK’s na ional usion expe imen ackling one
o usion ene gy’s bigges challenges: plasma exhaus . I is a medium
sized, small aspec a io (i.e., sphe ical) okamak (majo adius:
0.9 m, mino adius: 0.6 m) ope a ed by he Uni ed Kingdom A omic
Ene gy Au ho i y56,63. I has a o oidal field o 0.8 T, i s plasma cu -
en can each up o 1 MA, and ea u es one o -axis and one on-axis
neu al beam injec o o ex e nal hea ing, o up o 2.2 MW each.
TRANSP simula ions a e used o model he neu al beam abso p ion,
equi ed o es ima e he powe en e ing he sc ape-o -laye (SOL).
MAST-U ea u es co e Thomson sca e ing o ob ain co e elec on
densi y and empe a u e p ofiles, uses a -in a ed- eflec ome y
(FIR) o ob ain he line-a e aged elec on densi y and u ilises he
magne ic equilib ium econs uc ion code EFIT++ o econs uc
he magne ic equilib ia based on magne ic p obe measu emen s. I s
di e o is well diagnosed, ea u ing line-o -sigh spec oscopy8,
imaging diagnos ics57, Langmui p obes64,aswellasanimaging
bolome y sys em a he X-poin 65.
In his wo k, uelling injec ion om he low field side o he co e plasma
is used o main ain L-mode condi ions whils con olling he co e densi y in
eal ime by adap ing he low-field side main chambe uelling a e66.The
ad an age o using highe powe L-mode condi ions in his s udy is ha he
ups eam densi y can be educed compa ed o ha in H-mode. This makes
he plasma less de ached and enables a wide ange o ups eam densi y
scans o in es iga e he e olu ion du ing de achmen 34.
MAST-U ea u es uppe and lowe di e o chambe s, enabling
double null di e ed scena ios. The di e o chambe s p e en neu-
al anspo om he di e o o he co e, p o iding neu al ba fling
and con ibu ing o co e-edge compa ibili y. The la ge di e o
chambe , combined wi h a ious di e o coils56, acili a es he
in eg a ion o complex di e o shapes wi h s ong neu al ba fling.
This enables s udying he impac o di e o shaping on powe
exhaus while main aining s ong neu al ba fling.
Wi h his shaping flexibli y, MAST-U can al e he poloidal flux
expansion, connec ion leng h and o al flux expansion. Poloidal flux
expansion, Fx¼Bu
θB
ϕ
B
θBu
ϕ
25), is he a io o he pe pendicula flux su ace spacing
a he a ge and ups eam, whe e Bu;
θ;ϕa e he poloidal (θ) and o oidal (ϕ)
componen s o he magne ic field a ups eam (u) and a he a ge ( ),
espec i ely. Inc easing F
x
educes he a ge hea loads (W m−2)by
sp eading i o e a la ge su ace. Inc easing he connec ion leng h be ween
he midplane and he di e o a ge (L
∥
), p o ides a la ge adia ing olume
and is expec ed o imp o e powe exhaus 25.To alflux expansion
(FR¼Bxp
B ) inc eases he c oss-sec ional a ea o a flux ube, sp eading he
hea o e a la ge adius and lowe ing he a ge empe a u e25,28.The
sphe ical na u e o MAST-U enables a ying o al flux expansion o e an
unp eceden ed ange o he bes o ou knowledge, making i an ideal es bed
o s udying he impac o o al flux expansion in a s ongly ba fled di e o .
Di e o de achmen , ion sou ce/sink in e ences and powe
balance
Powe exhaus can be acili a ed by plasma de achmen , which is a s a e
whe e simul aneous powe , pa icle and momen um losses esul in a
simul aneous educ ion o a ge pa icle fluxes and plasma a ge
empe a u e.
Using hyd ogen a omic Balme line spec oscopic analysis5,8, he
elec on empe a u e, ion sou ces (I
i
) and sinks (I
) om plasma-a om and
molecula in e ac ions, as well as he hyd ogenic adia i e powe losses and
Molecula Ac i a ed Dissocia ion, can be in e ed om he hyd ogen Bal-
me line emission.
Since he line-o -sigh spec oscopy sys em has a se an o iews
h oughou he di e o leg (Fig. 1), spa ial p ofiles o cho dally in eg a ed
ion sou ces and sinks (pa . m−2s−1) along he di e o leg can be ob ained
(Fig. 3d– ). Du ing de achmen , fi s he ionisa ionsou ce de aches om he
a ge (T
e
<3–5 eV, in e ed spec oscopically8) and ul ima ely Elec on-Ion
Recombina ion (EIR) s a s o occu nea he a ge (T
e
≈0.2 eV,
n
e
≈2–4×10
19m−3, acco ding o spec oscopic in e ences o he high-n
(n> 9) Balme line spec a40). By acking he loca ion o he downs eam-
end o he ionisa ion sou ce ((1.5 ± 0.25) × 1021 pa . m−2s−1)and he
ups eam-end o he EIR sink ((3 ± 0.5) × 1020 pa . m−2s−1), he dis ance
be ween he a ge and he ionisa ion on (defined as he de achmen
on ) and EIR on (colde e e ence poin o deepe de achmen ) can be
ob ained. These numbe s a e ob ained as onse poin s based on he spa ial
p ofiles o ion sou ces and sinks p esen ed in Fig. 1d.
Combining spec oscopic in e ences on ion sou ces and sinks wi h
Langmui p obe measu emen s, in o ma iononbo hpa icleandpowe
balance can be ob ained. The o al ion a ge flux (I
in pa . s−1) is ob ained
by in eg a ing he ion a ge flux Γ
(pa . m−2s−1) measu ed by Langmui
p obes. F om conse a ion o pa icles, he o al ion a ge flux should equal
he ion sou cesminus he ion sinks, in addi ion o any ne ion inflow in o he
moni o ed sys em I
u
,Eq.(1). Using pa icle balance, he o al ion sou ce
(I
i
+I
u
) can be in e ed (Eq. (1)).
I ¼IiI þIuð1Þ
The a ge powe loading can be in e ed using a combina ion o
spec oscopy and Langmui p obe measu emen s. To o e come limi a-
ions o es ima ing a ge empe a u es using Langmui p obes in low
empe a u e condi ions64, spec oscopy om lines-o -sigh closes o he
a ge is used o in e a cha ac e is ic a ge elec on empe a u e T
.
Using his empe a u e, he pe pendicula plasma a ge powe deposi-
ion can be es ima ed as P
⊥, a ge
=I
(γT
+ϵ) (in W), whe eas he peak
pe pendicula hea flux can be es ima ed as q
⊥,peak
=Γ
,peak
(γT
+ϵ) (in
Wm
−2). A shea h ansmission ac o o γ= 7 is assumed ( alid o equal
elec on and ion empe a u es) and bo h su ace ecombina ion and
molecula e-associa ion is accoun ed o in he po en ial ene gy
ϵ= 13.6 +2.2 eV.
Assuming ha all olume ic powe losses a e pu ely due o hyd o-
genic adia ion as well as dissocia ion, which is mo i a ed by he obse -
a ion ha hyd ogenic adia ion es ima es om spec oscopic analysis
align wi h he measu ed o al adia ion in hese condi ions34, hepowe in o
he di e o chambe can be es ima ed by summing P
⊥, a ge
and he
in e ed hyd ogenic di e o adia i e powe loss. This igno es powe
ans e om he plasma o he neu al cloud h ough elas ic collisions,
which can become subs an ial in he SOLPS-ITER simula ions shown (up
o 15% o P
SOL
41), in quali a i e ag eemen wi h he obse ed o a ional D
2
o a ional empe a u es41.
Al hough his does include su ace ecombina ion, i does no include
a ge hea loadsdue o pho ons and neu al a oms. Including dissocia ion as
a o al loss channel implies assuming ha he neu al a oms, a e dis-
socia ion, a e mos ly los o he side walls, a he han eaching he a ge
(and hence do no con ibu e as a ge hea ing), which is consis en wi h
findings in SOLPS-ITER simula ions.
h ps://doi.o g/10.1038/s42005-025-02121-1 A icle
Communica ions Physics | (2025) 8:215 9
