1 of 36

A.R. Saperstein

AI4Fusion Summer School – June 4^th, 2026

Disruption Physics

June 4^th, 2026

AI4Fusion Summer School - Summer 2026

2 of 36

Outline

What is a disruption?

The anatomy of a disruption

What causes disruptions?

C-Mod database indicators

June 5^th, 2024

2

AI4Fusion Summer School

3 of 36

Outline

What is a disruption?

The anatomy of a disruption

What causes disruptions?

C-Mod database indicators

June 5^th, 2024

3

AI4Fusion Summer School

4 of 36

First, what is a Tokamak?

June 5^th, 2024

4

AI4Fusion Summer School

5 of 36

Now, what is a Disruption?

Disruption

June 5^th, 2024

5

AI4Fusion Summer School

6 of 36

Disruptions are frequent on experimental devices

[deVries 2011 NF] P.C. de Vries et al 2011 Nucl. Fusion 51 053018

[Gerasimov 2020 NF] S.N. Gerasimov etal 2020 Nucl.Fusion 60 066028

1983🡪2010

2011🡪2016

Disruptivity on JET over lifetime

C walls

Metal walls

[deVries 2011 NF]

[Gerasimov 2020 NF]

High performance campaign

June 5^th, 2024

6

AI4Fusion Summer School

7 of 36

Outline

What is a disruption?

The anatomy of a disruption

What causes disruptions?

C-Mod database indicators

June 5^th, 2024

7

AI4Fusion Summer School

8 of 36

Every disruption has the following events

June 5^th, 2024

8

AI4Fusion Summer School

9 of 36

Thermal Quench (TQ)

Confinement properties are lost and profiles flatten

Temperature and pressure drop uniformly
Current does NOT!
Density doesn’t necessarily

Flattening of current profile actually causes spike in total current

Conservation of magnetic energy

Timescale set by new energy confinement time

Very short relative to other disruption timescales

June 5^th, 2024

9

AI4Fusion Summer School

10 of 36

Current Quench (CQ)

Plasma becomes more resistive, so current wants to drop

Inertia of current (self-inductance) prevents this from happening simultaneously with TQ

Timescale is set by L/R time

Longest timescale of the disruption

Usually*

June 5^th, 2024

10

AI4Fusion Summer School

11 of 36

Loss of position control (VDEs*)

← Big kick

June 5^th, 2024

11

AI4Fusion Summer School

12 of 36

Each of these events lead into each other

June 5^th, 2024

12

AI4Fusion Summer School

13 of 36

Outline

What is a disruption?

The anatomy of a disruption

What causes disruptions?

C-Mod database indicators

June 5^th, 2024

13

AI4Fusion Summer School

14 of 36

How is the energy stored in the first place?

Magnetic field-lines and nested flux (Ψ) surfaces (MFE)!
Most of the motion of particles (and heat) is now directed parallel to the field lines and trapped on nested flux surfaces

Disruption

June 5^th, 2024

14

AI4Fusion Summer School

15 of 36

Disruption

June 5^th, 2024

15

AI4Fusion Summer School

16 of 36

What destroys flux surfaces?

Disruption

Magneto Hydro-Dynamic (MHD) instabilities

Instabilities associated with Fluid and Electro-Magnetic (EM) properties of the plasma
Fluid = Pressure
EM = Current

Overlap of perturbations�to the Ψ-surfaces�destroys them

Called Stochasticity �(or Chaos)

Transport across destroyed surfaces is very fast

A single field line can cross the whole region (short circuit)

MHD instabilities

June 5^th, 2024

16

AI4Fusion Summer School

17 of 36

What are instabilities?

Principle of least Action

Systems always want to transition to the lowest energy state possible

(plasma) state

(potential) energy

June 5^th, 2024

17

AI4Fusion Summer School

18 of 36

What are MHD instabilities?

time

June 5^th, 2024

18

AI4Fusion Summer School

19 of 36

MHD instabilities strongly inhibited by bending field lines

June 5^th, 2024

19

AI4Fusion Summer School

20 of 36

How do MHD instabilities destroy flux surfaces?

MHD instabilities perturb the shapes of individual flux surfaces

And the surfaces around them

Overlap of perturbations to the Ψ-surfaces destroys them

Called Stochasticity �(or Chaos)

Transport across destroyed surfaces is very fast

A single field line can cross the whole region (short circuit)

Disruption

MHD instabilities

June 5^th, 2024

20

AI4Fusion Summer School

21 of 36

So… What causes MHD instabilities?

Disruption

MHD instabilities

Direct causes

Current Pressure

Terms we know to consider

June 5^th, 2024

21

AI4Fusion Summer School

22 of 36

What are some indirect causes of MHD instabilities?

Disruption

MHD instabilities

Direct causes

Indirect causes

Current Pressure

Radiation Position Density

June 5^th, 2024

22

AI4Fusion Summer School

23 of 36

How does radiation depend on the plasma?

Terms we know to consider

June 5^th, 2024

23

AI4Fusion Summer School

24 of 36

How do vertical instabilities depend on the plasma?

Terms we know to consider

June 5^th, 2024

24

AI4Fusion Summer School

25 of 36

How do density limits depend on the plasma?

Plasma

Murari, A. et al. Nat Commun 15, 2424 (2024)

Terms we know to consider

Error

June 5^th, 2024

25

AI4Fusion Summer School

26 of 36

What are the most common triggers of disruptions?

Survey of Disruptions on JET between 2000-2011

NTMs are the most common triggers!

Human error is the second most common!

NTMs!

Density Control

Greenwald limit

MHD

Disruptions due to physics instabilities and limits

Disruptions due to Technical/Human errors

VDEs

P.C. de Vries et al 2011 Nucl. Fusion 51 053018

Human Error!

June 5^th, 2024

26

AI4Fusion Summer School

27 of 36

Most common triggers adjusting for danger

Same survey, but now only considering disruptions with a stored plasma energy above 1 MJ

Disruptions that occur when the plasma has more stored energy are typically more dangerous

NTM’s still one of the most dangerous

But ITB’s (MHD) now start making an appearance

NTMs!

Internal Transport Barriers!

VDEs

P.C. de Vries et al 2011 Nucl. Fusion 51 053018

June 5^th, 2024

27

AI4Fusion Summer School

28 of 36

Tokamaks vs Stellarators

Disruptions are far more common in tokamaks, but can* occur in stellarators

They are inherently more stable because most of their equilibrium is set by static external coils

Perturbations to the plasma have little effect on the equilibrium
Tokamaks have control over themselves that is… non-ideal

Stellarators aren’t really prone to current driven instabilities though

Which are often the instabilities to most likely actually disrupt the plasma

But they are still prone to pressure driven and radiative instabilities

Moiraf, David. (2021). Transport and radiation analysis of radiation-limited plasmas in the stellarator Wendelstein 7-X.

Tokamak

Stellarator

Ψ-Surfaces dependent on plasma

Ψ-Surfaces mostly independent of plasma

June 5^th, 2024

28

AI4Fusion Summer School

29 of 36

Outline

What is a disruption?

The anatomy of a disruption

What causes disruptions?

C-Mod database indicators

June 5^th, 2024

29

AI4Fusion Summer School

30 of 36

C-Mod database features (63* total features)

beta_n	i_efc	n_equal_1_mode	n_equal_1_normalized	n_equal_1_phase	pressure_peaking	q0	q95
qstar	beta_p	li	z_times_v_z	zcur	lower_gap	n_over_ncrit	tribot
tritop	upper_gap	v_z	z_error	z_prog	kappa	kappa_area	a_minor
greenwald_fraction	ip	n_e	ne_peaking	p_icrf	p_input	p_lh	p_oh
te_core_vs_avg_ece	te_edge_vs_avg_ece	te_peaking	te_width_ece	te_width	p_rad	prad_peaking	radiated_fraction
tau_rad	wmhd	ip_error	sxr	v_loop	bt	chisq	ip_prog
rmagx	ssep	v_loop_efit	v_surf

time_until_disrupt	disruptive*

shot

time
dbetap_dt
dip_dt
dip_smoothed
dli_dt
dn_dt
dprad_dt
dipprog_dt
dwmhd_dt

(time-independent)

(time-dependent)

VDE

Density Limit (DL)

Radiation limits

MHD

generic

Input features (52 - 9)

Target features (2*)

Metadata (1)

June 5^th, 2024

30

AI4Fusion Summer School

31 of 36

MHD stability related features

beta_n	n_equal_1_mode	q0	ip	i_efc
beta_p	n_equal_1_normalized	q95	li
pressure_peaking	n_equal_1_phase	qstar

Current drive

Pressure drive

Perturbation size

Magnetic resonance

Error field suppression

June 5^th, 2024

31

AI4Fusion Summer School

32 of 36

Power balance (radiation stability) related features

radiated_fraction	n_e	ip_error
tau_rad	ne_peaking	sxr
p_rad	te_width	v_loop
prad_peaking	te_width_ece	wmhd
p_icrf	te_core_vs_avg_ece
p_input	te_edge_vs_avg_ece
p_lh	te_peaking
p_oh

Power balance terms

Radiation sources

Radiation effects

June 5^th, 2024

32

AI4Fusion Summer School

33 of 36

Vertical stability related features

kappa	li	zcur
kappa_area	beta_p	z_prog
tribot	ip	z_error
tritop		v_z
n_over_ncrit		z_times_v_z
		upper_gap
		lower_gap

Plasma position

Shaping

Eddy current stabilization

June 5^th, 2024

33

AI4Fusion Summer School

34 of 36

Density limit related features

greenwald_fraction	p_icrf	te_edge_vs_avg_ece
ip	p_input	te_core_vs_avg_ece
a_minor	p_lh	te_peaking
kappa	p_oh	te_width_ece
kappa_area		te_width
		n_e
		ne_peaking

Edge collisionality

Greenwald limit

Power scaling

June 5^th, 2024

34

AI4Fusion Summer School

35 of 36

Generic features

ip	ip_prog
ip_error	rmagx
v_loop	ssep
sxr	v_loop_efit
bt	v_surf
chisq

June 5^th, 2024

35

AI4Fusion Summer School

36 of 36

References

[1] F C Schuller 1995 Plasma Phys. Control. Fusion 37 A135

[2] P.C. de Vries et al 2011 Nucl. Fusion 51 053018

[3] H. R. Koslowski (2012), DOI: 10.13182/FST12-A13496

[4] Allen H. Boozer, Phys. Plasmas 19, 058101 (2012)

[5] T.C. Hender et al 2007 Nucl. Fusion 47 S128

[6] T. Eich et al, Journal of Nuclear Materials 337–339 (2005) 669–676

[7] R.S. Granetz et al 1996 Nucl. Fusion 36 545

[8] S.N. Gerasimov et al 2015 Nucl. Fusion 55 113006

[9] V. Riccardo and JET EFDA contributors 2003 Plasma Phys. Control. Fusion 45 A269

June 5^th, 2024

36

AI4Fusion Summer School