The Astrophysical Journal, 941:79 (15pp), 2022 December 10 https://doi.org/10.

3847/1538-4357/ac9e5a
© 2022. The Author(s). Published by the American Astronomical Society.

STITCH: A Subgrid-scale Model for Energy Buildup in the Solar Corona

J. T. Dahlin1,2 , C. R. DeVore1 , and S. K. Antiochos3
1
Heliophysics Science Division, NASA Goddard Space Flight Center, Greenbelt, MD 20771, USA; joel.t.dahlin@nasa.gov
2
Astronomy Department, University of Maryland, College Park, MD 20742, USA
3
CLaSP, University of Michigan, Ann Arbor, MI, 48109, USA
Received 2021 December 1; revised 2022 October 24; accepted 2022 October 26; published 2022 December 13

Abstract
The solar corona routinely exhibits explosive activity, in particular coronal mass ejections and their accompanying
eruptive flares, which have global-scale consequences. These events and their smaller counterparts, coronal jets,
originate in narrow, sinuous filament channels. The key processes that form and evolve the channels operate on
still smaller spatial scales and much longer timescales, culminating in a vast separation of characteristic lengths and
times that govern these explosive phenomena. In this article, we describe implementation and tests of an efficient
subgrid-scale model for generating eruptive structures in magnetohydrodynamics (MHD) coronal simulations.
STITCH—STatistical InjecTion of Condensed Helicity—is a physics-based, reduced representation of helicity
condensation: a process wherein small-scale vortical surface convection forms ubiquitous current sheets and
pervasive reconnection across the sheets mediates an inverse cascade of magnetic helicity and free energy, thereby
forming the filament channels. We have developed a formalism, STITCH, that abstracts these complex processes
into a single term in Ohm’s law and the induction equation that directly injects tangential magnetic flux into the
low corona. We show that our approach is in very good agreement with a full helicity condensation calculation that
treats all of the dynamics explicitly, while enabling substantial reductions in temporal duration and spatial
resolution. In addition, we illustrate the flexibility of STITCH at forming localized filament channels and at
energizing complex surface flux distributions that have sinuous boundaries. STITCH is simple to implement and
computationally efficient, making it a powerful technique for physics-based modeling of solar eruptive events.
Unified Astronomy Thesaurus concepts: Solar magnetic reconnection (1504); Solar corona (1483); Solar filament
eruptions (1981); Solar coronal mass ejections (310); Solar flares (1496); Magnetohydrodynamical simula-
tions (1966)
Supporting material: animations

1. Introduction Despite such diversity in eruptive manifestations, the energy

source is generally agreed (at least for the largest events) to be
Solar eruptions are the most energetic explosive events in the
filament channels, highly stressed magnetic structures that form
solar system. The largest of these events, known as coronal
above reversals in the photospheric line-of-sight field known as
mass ejections (CMEs), release vast quantities of solar plasma polarity inversion lines (PILs; Gaizauskas 1998; Martin 1998).
and magnetic fields into the solar wind, generating shocks that Filament channels are force-balanced structures consisting of a
disturb the entire heliosphere, driving hazardous energetic highly sheared magnetic field (that is, approximately parallel to
particles and magnetospheric storms at Earth. A major goal of the PIL) restrained by an overlying unsheared field that links
solar modeling efforts is operational prediction of the events the opposite-polarity fluxes to either side of the PIL. Many of
that present the greatest threats to human assets and these channels are observed to exhibit S-shaped sigmoids in
technology. A critical obstacle to achieving this goal, however, their hot coronal loops and/or cool filament threads (if
is the paucity of quantitative measurements of the coronal present), with associated strong electric currents signaling the
magnetic field. Further complicating matters are the remarkable storage of magnetic free energy (Komm et al. 2015). An
diversity of eruptions in size, from CMEs that fill the eruption occurs when this force balance is disrupted by some
heliosphere down to coronal jets that span only a few trigger, either an ideal instability (Fan 2001; Linker et al. 2003;
megameters (Raouafi et al. 2016); the morphology of the Kliem & Torok 2006) or magnetic reconnection (Antio-
erupting material in coronagraphs, including both so-called chos 1998; Antiochos et al. 1999; Moore et al. 2001), which
stealth CMEs (Ma et al. 2010) and halo CMEs (Howard et al. initiates ejection of the built-up shear field. Understanding how
1982); and the accompanying signatures in the low atmosphere, the eruption will occur and the hazard it presents reduces to
in particular the chromospheric flare ribbons, which can be as understanding how the generic filament channel formation and
simple as the classic, linear two-ribbon structure or as complex eruption processes occur in the complex and diverse magnetic
as circular (Masson et al. 2009) or multiribbon (Wang et al. field configurations that house the PILs.
2014) configurations, and the overlying arcades of flare loops Several models have been proposed for filament channel
in the corona. formation. The simplest example, perhaps, is large-scale shear
flows or sunspot/active region rotation (e.g., Cheung &
Original content from this work may be used under the terms
DeRosa 2012; Sun et al. 2012). However, such large-scale
of the Creative Commons Attribution 4.0 licence. Any further coherent motions are not always observed and are unlikely to
distribution of this work must maintain attribution to the author(s) and the title explain all examples of filament channel formation (Gaizaus-
of the work, journal citation and DOI. kas 1998; Martin 1998). For the quiet Sun especially, it has

1
The Astrophysical Journal, 941:79 (15pp), 2022 December 10 Dahlin, DeVore, & Antiochos

often been suggested that a combination of differential rotation be used more generally, even in cases where turbulent
and flux cancellation generates coronal flux ropes that convective motions are not expected to be the dominant source
constitute filaments and their channels (van Ballegooijen & of energy and helicity.
Martens 1989; van Ballegooijen 2004; Mackay & Yeates 2012; In Mackay et al. (2014, 2018), a statistically averaged
Yeates & Mackay 2012). However, the chirality of these approximation to the full helicity condensation (FHC) process
structures, if arising from these processes alone, has been was derived and employed to investigate the formation of
shown (Mackay et al. 2014, 2018, and references therein) to be filament channels across the full Sun and over multiple solar
inconsistent with the global helicity patterns observed at all rotations. Those studies used a magnetofrictional model for the
scales within eruptive coronal structures. For example, Ouyang corona that accurately reproduces the large-scale and long-term
et al. (2017) found that 92% of their sample of 571 erupting formation of filament channels, but which cannot self-
filaments followed the empirical rule that those in the northern/ consistently capture explosive eruptive dynamics because the
southern hemisphere exhibit negative/positive magnetic model output is a sequence of force-free coronal magnetic field
helicity. equilibria. In this article, we extend and test this statistical
The recently developed helicity condensation model (Anti- approach for general use in full MHD simulations. The
ochos 2013) posits injection of helicity/energy into the corona resulting model, which we call STatistical InjecTion of
via the small-scale flows associated with photospheric Condensed Helicity (STITCH), encapsulates the injection of
convective motions. In a sufficiently “turbulent” system helicity into the corona by vortical convection cells, the
supporting widespread reconnection across ubiquitous small- formation of current sheets at boundaries between adjacent
scale current sheets, this structure and associated energy will be cells, and the reconnection of the induced horizontal magnetic
transferred to increasingly larger scales via the well-known fields across the sheets. STITCH has three critical advantages
inverse cascade of helicity (Berger 1984; Berger & Field 1984; over the MHD treatment of the FHC process with all its
Finn & Antonsen 1985; Biskamp 1993). In the context of the attendant small-scale dynamics and structures. First, it is
Sun, this cascade causes small-scale twist injected throughout numerically efficient in both memory and computations,
an area enclosed by a PIL to eventually “condense” at that PIL requiring substantially lower spatial resolution and temporal
in the form of a large-scale shear. This theory has been cadence. Second, it is simple and ease to use. Third, it is highly
rigorously explored through MHD simulations that demon- flexible, and it is readily applicable to the complex magnetic
strate the formation of filament channel structures consistent field distributions observed on the Sun.
with those observed on the Sun (Knizhnik et al. We discuss the STITCH model and its numerical imple-
2015, 2017a, 2017b, 2018; Zhao et al. 2015) and evolution mentation in Section 2. Our magnetohydrodynamics code,
ARMS, and initial configuration are described in Section 3. We
of the structure through the generation of a fast CME with an
compare STITCH to the comprehensive helicity condensation
eruptive flare (Dahlin et al. 2019). Importantly, the filament
model in Section 4. Thereafter, we show examples of the
channels generated by helicity condensation, when it is added
flexibility of STITCH in localizing the filament channel
to the global-scale processes of differential rotation, meridional
(Section 5) and in being used to energize more complex and
flow, flux cancellation, and newly emerged active region flux,
realistic magnetic configurations (Section 6). STITCH is then
are consistent with the observed chirality patterns on the Sun placed in the larger context of methods used to model solar
(Mackay et al. 2014, 2018). eruptive events in Section 7. To conclude, we summarize our
All of the previous full MHD studies of helicity condensa- results and discuss the implications of using STITCH in
tion, cited above, have used close-packed (i.e., tessellated), simulations of coronal energy buildup in Section 8.
boundary-driven vortical flows to represent helicity injection
via convective granule/supergranule motions. Such calcula-
tions are computationally expensive, as they must resolve all of 2. Description of the STITCH Method
the cellular flows and the small-scale reconnection dynamics In his exposition of the helicity condensation model,
that drive the helicity cascade. Additionally, the flows cannot Antiochos (2013) described how an ensemble of vortical
overlap with the PILs lest they trigger excessive small-scale motions at the photosphere builds up magnetic shear in the
diffusion across structures generated at the grid scale, a corona. Current sheets form at the interfaces between pairs of
challenging constraint for driving the highly complex magnetic neighboring vortical cells that have the same sense (clockwise
configurations typical of eruptive events. An important or counterclockwise) of rotation. Magnetic reconnection across
implication of the helicity condensation model, however, is the sheets depletes the twist component of magnetic flux
its relative insensitivity to the particular mechanism or spatial between the cells, leaving behind the twist flux that envelopes
scale of the helicity injection (Knizhnik et al. 2017a, 2019). So both cells jointly. This process repeats in an inverse cascade
long as there is a net helicity injection with sufficiently that transports the twist flux to ever-larger spatial scales. The
complex turbulent dynamics, the shear will be transferred to the cascade halts at the boundary of the flux system—defined by a
PILs. Hence, while resolving the small-scale reconnection is reversal of either the sign of the vertical magnetic field or the
critically important for understanding nanoflare-type coronal sense of rotation of the vortical cells—because the twist flux
heating, ultimately the large-scale energy buildup is insensitive has the same direction on both sides of such a boundary and,
to the underlying details of the helicity injection and cascade hence, cannot reconnect away. The twist flux encircling the
processes. This motivates the search for a method that captures flux system is said to “condense” at that boundary, and the
the essential large-scale features of the energy buildup without linkage between the twist flux and the enclosed vertical flux
explicitly calculating all of the small-scale structure and imparts magnetic helicity to the structure. The term “helicity
dynamics. Such a method, if sufficiently flexible in its condensation” had been coined previously to describe this
implementation, should be agnostic with respect to the process when it occurs in laboratory experiments and numerical
particular mechanism for helicity injection. Hence, it could simulations (Biskamp 1993).

2
The Astrophysical Journal, 941:79 (15pp), 2022 December 10 Dahlin, DeVore, & Antiochos

Several subsequent investigations of helicity condensation as The change in the horizontal field component Bs is
applied to the corona have been carried out (Knizhnik et al. ¶Bs
2015, 2017a, 2017b, 2018; Zhao et al. 2015; Dahlin et al. 2019) = - n ´ s (zBn) = + s ´ n (zBn) ; (5 )
in which numerous small-scale vortical cells—as many as 100 ¶t
+—were resolved on the computational grid. The needed we note that the exchange of curl and gradient operators in
helicity-preserving simulations were performed with the Equation (5) is valid in both spherical and Cartesian
Adaptively Refined Magnetohydrodynamics Solver (ARMS; coordinates. Third, the total rate of injection of relative helicity,
DeVore & Antiochos 2008). The processes of twist-flux Hr, into the atmosphere is given by the surface integral of the
accumulation, current-sheet formation and reconnection, and
local rate due to the individual vortical cells (for details, see
helicity condensation that were predicted theoretically were
confirmed quantitatively and in detail. However, the required Mackay et al. 2014),
well-resolved, long-duration calculations of this type are very dHr
resource intensive. dt
= -2 ò ò zBn2 ds. (6 )
In a separate line of investigation that applied the helicity
condensation concept to the full-Sun magnetic field over many This is required in order for the STITCH term to be consistent
weeks of evolution, we developed and employed a subgrid- with helicity conservation throughout the processes of
scale representation of the model (Mackay et al. 2014, 2018). reconnection and transport of twist flux to the flux system
The essential assumptions that were made are as follows: (1) boundaries.
the vortical cells are small, numerous, and nominally identical, STITCH is implemented numerically by computing the rate
although their properties may exhibit large-scale variations; and at which horizontal magnetic flux is injected into the domain
(2) the reconnection across current sheets between the cells is according to Equation (5). This flux is calculated on the cell
so efficient that the oppositely directed twist fields can be faces oriented perpendicular to the two horizontal coordinates
treated as simply canceling algebraically. As derived in the s1 and s2, where sˆ1 ´ sˆ2 = nˆ . Integrating Equation (5) over
Appendix of Mackay et al. (2014, Equation (A1)), the resulting differential cell areas dAs1 and dAs2 , respectively, we obtain the
subgrid-scale model is expressed as an additional term in the flux change rates
induction equation, ¶Fs1 ¶Bs
= ò ò dA s1 sˆ1 ·
¶B ¶As ¶t ¶t
= ´ = -  ´ s (zBn). (1 )
¶t ¶t ¶ ¶
= + ò ò ds2 dn (zBn) , (7 )
¶s2 ¶n
Here B is the magnetic field; A is the vector potential;
subscripts n and s denote components normal and parallel to ¶Fs 2 ¶Bs
= ò ò dA s 2 sˆ2 ·
the surface, respectively; and ζ is a parameter characterizing the ¶t ¶t
vortical flows. Specifically, ¶ ¶
= - ò ò ds1 dn (zBn). (8 )
¶s1 ¶n
z= á ℓ 2w lñ 2, (2 )
These expressions are coordinate system independent. We
where ℓ is the radius and ωl is the angular rotation rate of the localize the angular rotation of the convection cells near the
cells, and the angle brackets denote a local average in space and coronal base, which is positioned at height n = n0, by assuming
time. The parameter ζ has dimensions of diffusivity, and, as that ζ falls to zero at the top of the first layer of grid cells at
will be seen below, the local rate of injection of helicity (per height n = n1 = n0 + h. Integrating Equations (7) and (8) over n
unit area and time) into the corona is determined directly by the from n0 to n1, they simplify to
product zBn2 (Mackay et al. 2014). From the form of
¶Fs1 ¶
Equation (1), it is clear that the changes in B are strongest
where the product ζBn varies most rapidly across the surface. In ¶t ò
= - ds2
¶s2
(zBn∣n 0 ) , (9 )

particular, this occurs at the flux system boundaries described ¶Fs 2 ¶

previously. We refer to the subgrid-scale model represented by ¶t ò
= + ds1
¶s1
(zBn∣n 0 ) , (10)
Equation (1) as STITCH.
The STITCH term has several notable properties. First and where zBn∣n 0 denotes the value of ζBn evaluated at the surface,
most fundamentally, by construction it preserves the diver- n = n0. These expressions integrate immediately to yield the
gence-free character of the magnetic field, flux changes
¶Fs1
¶ = -D s 2 (zBn∣n 0 ) , (11)
 · B = -  · [ ´ s (zBn)] = 0, (3 ) ¶t
¶t
¶Fs 2
subject only to the requirement that ζBn be sufficiently smooth ( = +D s1(zBn∣n 0 ) , (12)
¶t
i.e., differentiable) everywhere. Second, it prescribes an
injection of twist flux into the corona that leaves the normal where Ds1 (Ds2 ) denotes a finite difference whose argument is
magnetic field component Bn unchanged, evaluated at cell vertices along coordinate s1 (s2).
The flux changes on the four faces of any grid cell involve
¶Bn pairwise differences of the argument (zBn∣n 0 ) evaluated at the
= - nˆ ·  ´ s (zBn) = - nˆ · s ´ s (zBn) º 0. (4) four vertices of the (s1, s2) grid. Summing these changes
¶t
algebraically, to calculate the change in the net flux leaving the

3
The Astrophysical Journal, 941:79 (15pp), 2022 December 10 Dahlin, DeVore, & Antiochos

cell through its four faces, yields zero; each vertex contribution We can associate the rate of change of δBs with an effective
appears twice in the sum with opposite signs (+, −). This is shear speed Vs applied to the footpoints of the magnetic field at
simply a discrete expression of Gauss’s law, Equation (3), the line-tied surface, with the result
integrated over the cell volume.
The final step in the calculation is to update the horizontal ¶dBs V dB
» s n. (16)
magnetic field components using the flux changes in ¶t h
Equations (11) and (12),
Equating the above two rates, we find the relationship between
¶Bs1 1 ¶Fs1 1 Vs and ζ,
= =- D s 2 (zBn∣n 0 ) , (13)
¶t A s1 ¶t A s1 z
Vs » . (17)
Ds
¶Bs 2 1 ¶Fs 2 1
= =+ D s1(zBn∣n 0 ) , (14) With this identification of the grid-dependent (due to Δs)
¶t A s 2 ¶t As2
effective shear speed Vs, we now define a flux injection
where As1 ( As2 ) is the area of the cell face perpendicular to the Courant number òfi according to
coordinate s1 (s2). Evaluating the (zBn∣n 0 ) vertex values and Vs z
multiplying them by the finite-difference time increment, Δt,  fi º º , (18)
VA VA Ds
yields the required changes to the field components, DBs1
and DBs2 . where VA is the characteristic Alfvén speed in the active region.
Using STITCH in a numerical simulation requires specifying As noted below, we found empirically that òfi ≈ 1/6 was a
the magnitude of the parameter ζ to be used. By appeal to satisfactory upper limit on ζ for the STITCH simulations
observations of supergranules (Duvall & Gizon 2000; Gizon & reported in this paper. This result is entirely in line with the
Duvall 2003; Komm et al. 2007), together with modeling numerical time increments Δt (relative to the minimum signal
results from flux transport simulations, Mackay et al. (2014) crossing time on the grid) that are adopted in MHD
estimated the solar value to fall in the range (6 ± 4) × 102
simulations, where fractional values on the order of a few
km2s−1. As expected, this is also the magnitude of the
corresponding diffusion coefficient with which supergranules tenths are acceptable and are used routinely to obtain
disperse the surface magnetic flux. The value is set by the numerically stable solutions.
supergranules’ relatively large size (ℓ ≈ 2 × 104 km) but very The net result of the above considerations is that we can set
low vortical speed (ℓωℓ ≈ 6 × 10−2 km s−1). In MHD an upper limit on the allowed STITCH parameter ζ simply by
calculations that are performed over much shorter time adopting a suitable value for the flux injection Courant number
intervals, of necessity, this value is impractically small to have òfi,
any significant effect. Fortunately, adopting substantially larger z max =  fi VA Ds. (19)
values in MHD simulations is no barrier, so long as the
resulting shear flux injection occurs more slowly than the This numerically constrained upper bound depends on the
governing fast MHD processes in the system, such as Alfvén- physical Alfvén speed, VA, as should be expected, but also on
wave propagation. Accelerating the helicity condensation the assumed grid spacing, Δs. The physical spatial scales in the
process by enhancing the STITCH coefficient is akin to simulation, such as the width w of the filament channel or the
accelerating the surface flows imposed in many other diameter L of the active region, are larger than the grid spacing.
simulations to form and energize filament channels (e.g., Consequently, the Courant numbers for processes at these
Karpen et al. 2012). So long as the injection speed is sub-
scales are progressively smaller than òfi; for example, that for
Alfvénic—preferably, if practical, also subsonic—the system
evolves quasi-statically as the magnetic free energy builds up, the shear injection of horizontal flux at the channel-width scale
until it transitions to fast, dynamical behavior that may include w is (Δs/w)òfi, and that for injecting an amount of horizontal
a violent eruption. The only caveat is that the duration of the flux equal to the vertical flux in the active region at its diameter
energy buildup phase is artificially shortened compared to that L is (Δs/L)òfi. The physically intuitive requirements that these
required on the Sun; the critical point, however, is that both the two subsidiary parameters be small are well satisfied by our
real and model buildup phases are much more slowly evolving constraint on ζ expressed in Equation (19).
than the eruption phase.
We calculate a constraint on the allowed value of the 3. Numerical Simulations
STITCH parameter ζ by limiting the speed of propagation
Our numerical calculations were performed with ARMS
across the grid of any induced horizontal field component δBs
(DeVore & Antiochos 2008), which has been used extensively
to the local Alfvén speed, as follows. We assume that the grid
to model both CMEs/eruptive flares (see also Lynch et al.
spacings are Δs horizontally and h vertically (in practice, these
2008, 2009, 2016, 2021; Karpen et al. 2012; Masson et al.
three are similar if not identical), the magnetic field
2013) and the formation of filament channels (Knizhnik et al.
components are δBs horizontally and δBn vertically (these two
2015, 2017a, 2017b, 2018; Zhao et al. 2015). Our fiducial
may differ substantially in magnitude), and δBn varies at the
calculations adopt the magnetic field configuration shown in
smallest accessible scale (i.e., the horizontal grid spacing Δs).
Figure 1. It consists of an elliptical, bipolar active region
The STITCH induction Equation (5) then yields directly
centered at 22°. 5N latitude with peak radial field |Br| ≈ 50 G,
¶dBs zdBn embedded in a background 10 G solar dipole field (for full
» . (15) description see Dahlin et al. 2019). The resulting topology is
¶t hDs
the well-known embedded bipole with a separatrix dome and a

4
The Astrophysical Journal, 941:79 (15pp), 2022 December 10 Dahlin, DeVore, & Antiochos

Figure 1. Initial embedded-bipole configuration. Surface gray shading indicates the sign and strength of the radial magnetic field component, Br; its PILs are colored
cyan. Magnetic field lines are colored according to whether they close entirely within the active region (red), between the active region and the background magnetic
field (blue and magenta), or entirely within the background field (green), or are separatrix lines bounding these flux systems (gray/silver).

pair of spine lines emanating from a 3D null point necessary to drive eruption differ by a substantial factor of 15.
(Antiochos 1990; Lau & Finn 1990; Priest & Titov 1996). A One reason for this is that the width of the injection region is
region of maximally refined grid enclosed the entire separatrix much larger, by a factor of about 4, for STITCH versus FHC;
dome (gray/silver field lines arching over the arcade of red the rates of shear flux accumulation differ by the same factor.
loops; Figure 1) and a substantial volume above it, in order to Whereas the STITCH profiles are smooth and space filling, the
resolve the eventual small-scale reconnection dynamics. The FHC profiles are patchy owing to the cellular structure; hence,
initial atmosphere was a spherically symmetric hydrostatic FHC has a filling factor of about 0.5, compared to unity for
equilibrium with an inverse-r temperature profile at a base STITCH. Lastly, the changing direction of the vortical flows in
temperature Ts = 2 × 106 K and pressure Ps = 4 × 10−1 dyn the FHC case lowers the average effective injection rate parallel
cm−2. We solved the ideal MHD equations with an adiabatic to the PIL; this provides yet another reduction factor of about 2
temperature equation. The domain extents were r ä [1Rs, 30Rs], in efficacy of FHC compared to STITCH. These cumulative
θ ä [π/16, 15π/16], and f ä [ − π, + π]. effects account for the major difference in the required drive
For the FHC calculation, the system was driven by 107 times to eruption in the two cases.
tessellated, vortical, Br-conserving cellular flows within the The stated rate of field injection and time interval in the
black shaded minority-polarity region. The flows were STITCH case suggest that a tangential field strength on the
subsonic and sub-Alfvénic, attaining a maximum speed order of 6 kG should build up; however, this does not occur.
|V⊥| ≈ 50 km s−1 after an initial sinusoidal ramp-up interval The tangential field is injected just above the surface, but then it
1000 s in duration. Elsewhere on the surface, the magnetic field is redistributed all along the length of the filament channel field
was line-tied at rest (V⊥ = 0) for the entire duration of the lines by Alfvén waves. This reduces the accumulated
simulation. The resultant rates of change of the horizontal maximum tangential field strength to approximately 100 G,
magnetic field components are shown in Figures 2(a) and (d). roughly twice the peak radial field strength inside the minority
An extended energy buildup phase occurred over about polarity. Expressed another way, we find that the tangential
90,000 s, which required approximately two continuous flux injected by STITCH over the simulation’s duration is
months of computational time on the NCCS discover roughly equal to the normal flux within the minority-polarity
supercomputer at NASA GSFC. region.
For our first STITCH calculation described below, we
assumed that ζ was uniform at ζ0 = 1.4 × 106 km2 s−1 within
the minority-polarity region and zero everywhere outside of 4. Comparing STITCH to Full Helicity Condensation
that region. This corresponds to a flux injection Courant Figures 3 and 4 show snapshots of the filament channel
number òfi ≈ 1/6 (Equation (18)), given the characteristic formation and eruption phases for the FHC and STITCH cases,
Alfvén speed VA = 4 × 103 km s−1 and minimum grid spacing respectively. Both generate low-lying, highly sheared orange
Δs = 2 × 103 km. field lines (the longest of which stretch more than halfway
The resulting profiles of injected horizontal magnetic field, along the PIL) that are restrained by overlying, weakly sheared
whose maximum rate was about 1 G s−1, are shown in red field lines (Figures 3(a) and 4(a)). These snapshots are
Figures 2(b) and (e) (Figures 2(c) and (f) show a STITCH case taken about two-thirds of the way through the filament channel
with localized ζ that is discussed in detail in Section 5). In this formation phase for the two cases. Eventually, the entire
and other STITCH calculations, the magnetic field was line-tied filament channel erupts as a quasi-circular filament with
at rest (V⊥ = 0) over the whole surface for the entire duration maximum speed V > 1000 km s−1, indicated by dark red
of the simulation. To achieve a smooth start-up, we ramped up shading in the plane of the sky (Figures 3(b) and 4(b)). The
the tangential field injection sinusoidally over the first 1000 s of kinetic energy is near its maximum value at the times shown
elapsed time in the simulation. The energy buildup phase was for both cases.
much shorter with STITCH, in this case being only about One significant point of contrast between the two cases is the
6000 s in total duration and requiring approximately 4 days of shear imparted to the red loops overlying the PIL of the active
computational time. region under FHC; under STITCH, the same loops are
Although the maximum rates of horizontal field injection are essentially shear-free. The helicity-injecting flows in the former
similar in the FHC and STITCH cases, as shown by Figure 2, fill the minority-polarity portion of the active region, inducing
the timescales for accumulating the helicity and free energy twist everywhere that subsequently condenses at the PIL, but

5
The Astrophysical Journal, 941:79 (15pp), 2022 December 10 Dahlin, DeVore, & Antiochos

Figure 2. Comparison of helicity injection by FHC vs. STITCH for the idealized embedded-bipole configuration. (a, d) Horizontal components of magnetic field, Bs,
generated by vortical flows imposed on the boundary to drive the FHC case. Horizontal components of magnetic field, Bs, injected just above the boundary to drive
STITCH case with uniform ζ (panels (b) and (e)) and localized ζ (panels (c) and (f)). Black contours indicate radial magnetic field values Br = [−50, −40, −30, −20,
−10, 0] G.

with a finite time delay and with a spatial distribution that is other hand, there are both close similarities and significant
smoothed across the region of flow. In contrast, the STITCH differences, as we now discuss in detail.
model with uniform ζ instantaneously injects tangential field The accumulation of magnetic energy from its initial value,
adjacent to the PIL, across a zone whose characteristic width EM … EM0, and its later reduction in conjunction with eruption,
w = |Br|/|∇sBr| is determined solely by the magnetic field are shown in Figure 5(a). We find that the accumulated
profile. The constant stirring of the field within the filament magnetic free energy is some 20% higher in the STITCH case
channel by the helicity condensation flows means that the than in the FHC case, EM/EM0 ≈ 1.24 and 1.20, respectively.
effective minimum width of the FHC channel will be the This STITCH excess is expected, based on its concentrating all
characteristic diameter of the vortical cells, d = 2ℓ (recall of its shear adjacent to the PIL, which we discussed above. The
Equation (2)). Due to the limited resolution available for use in more broadly distributed shear in the FHC case distorts the
our simulations, these two measures of the width are overlying coronal null point into a current patch, facilitating the
comparable, d ≈ w, as can be seen directly in Figure 2. onset of breakout reconnection and removal of the high-lying
Therefore, the resulting filament channels exhibit similar magnetic tethers, earlier in the energy injection process than is
widths of their strongly sheared core fields, as can be seen in true for the STITCH case. We emphasize that the elapsed time
Figures 3(a) and 4(a). is so much longer for FHC only because its imposed vortical
Figure 5 shows the evolution of the volume-integrated flows are so slow. Were we to inject magnetic free energy in
magnetic and kinetic energies, along with the maximum value the STITCH case at a comparably slow rate, we would expect
of |Bf|, a proxy for the strength of the shear magnetic field. the STITCH energy buildup phase to require more elapsed time
Throughout the entire evolution, except for a brief period than FHC, in order to accumulate the greater energy needed to
lasting about 100 s at the beginning of the driving phase, this initiate the STITCH eruption.
maximum occurs at the center of the southern half of the PIL The strong concentration of the STITCH shear adjacent to
near the inner boundary. the PIL can account for two more contrasting features of its
In order to compare the macroscopic evolution most energy profile compared to the FHC case. First, the STITCH
conveniently, we have scaled the respective time axes to align magnetic energy release is only about 40% of that for FHC,
the approximate eruption times in the two cases, t ≈ 90,000 s ΔEM/EM0 ≈ 0.04 and 0.10, respectively, culminating in post-
(FHC) and t ≈ 6000 s (STITCH). The reduction in the nominal eruption energies EM/EM0 ≈ 1.20 and 1.10. The much
driving time is matched by a corresponding decrease by a factor narrower shear distribution achieved by STITCH reduces the
of 15 in the computational cost of the STITCH run. volume of free energy containing magnetic flux, relative to that
Qualitatively, all of the trends in the displayed global variables of FHC, and throttles back the amount of flux that is processed
are shared between FHC and STITCH. Quantitatively, on the and energy that is released by the eruption. Second, the

6
The Astrophysical Journal, 941:79 (15pp), 2022 December 10 Dahlin, DeVore, & Antiochos

Figure 3. Filament channel (a) formation and (b) eruption for the FHC calculation. Magnetic field lines are colored according to the scheme used in Figure 1, with
orange lines added to show the filament channel field. Color shading in the plane of the sky (right) shows radial velocity Vr. An animation of this figure is available,
showing first the driving phase as in panel (a) from t = 0 s to t = 74,000 s at 1000 s cadence. This occurs in the first 5 s of the animation. The eruptive phase in panel
(b) is in the next 15 s of the animation and shows the t = 86,600 s to t = 97,950 s at 50 s cadence.
(An animation of this figure is available.)

STITCH energy release timescale is only Δt ≈ 1000 s, a factor CME flux rope, a lower-lying portion is left behind in the
of four less than the Δt ≈ 4000 s timescale for the FHC energy reconnected flare loops. These loops relax downward, and their
release. (Notice that the STITCH energy release phase only shear field strength increases as the occupied volume contracts.
appears to be much more gradual than that for the FHC phase, This generates a second, post-eruption peak in the shear proxy.
due to the difference in the absolute time axes for the two The FHC and STITCH cases exhibit entirely similar behaviors,
cases.) We would expect the smaller volume processed in the but due to the greater concentration of the STITCH filament
STITCH case to correspond to a shorter timescale, and this is channel, both of its peaks—during the early expansion and
exactly what we find. The greater concentration of the STITCH after the eruption (110/80 G, respectively)—are noticeably
shear flux leads to a rather more explosive energy release, higher than those of the FHC filament channel (70/70 G). The
however, by a factor of 0.4 × 4.0 = 1.6, or a 60% increase local-minimum shear field strength at eruption onset, on the
relative to the FHC energy release rate. other hand, is very nearly the same in the two cases (about
For clarity, the energy storage and release rates dEM/dt, 40 G).
normalized to the initial magnetic energy, are displayed These results demonstrate that driving the filament channel
separately in Figure 5(b). The storage rate for STITCH is just formation process via the STITCH model produces an
over 15× faster than that for FHC, whereas, as mentioned, its evolution leading to eruption that is very similar to our
release rate is only about 60% faster. The energy release previous findings using FHC (Dahlin et al. 2019). The major
appears to subside about four times slower for STITCH versus difference between the two is the highly compressed
FHC on this plot, due to the disparate time axes used, but in characteristic timescale for the STITCH energization: whereas
fact it subsides about four times faster, as we also noted above. FHC relies on slow, subsonic vortical flows that gradually build
The kinetic energy in the simulation, normalized to the initial up the sheared filament channel, STITCH relies on the much
magnetic energy, is shown in Figure 5(c). This ratio is faster, but sub-Alfvénic, direct injection of horizontal magnetic
Ek EM0 = áV 2ñ áVA2ñ = áMA2ñ, where VA and MA are the flux into the corona. A more minor, but physically very
Alfvén speed and Alfvén Mach number, respectively. In both significant, difference between them is in the width of the
cases, the ratio during the long phase of slow driving is on the resulting filament channel: the minimum FHC channel width is
order of 10−4, or 〈MA〉 ≈ 1%. The value jumps steeply during determined by the characteristic diameter d of the vortical
the explosive eruption phase to a few times 10−2, or flows, whereas the STITCH channel width is determined
〈MA〉 ≈ 15%–20%. This rather high range highlights the entirely by the characteristic scale w of the vertical magnetic
pervasive Alfvénic outflows of the fast eruption, whose plasma field profile. The limited numerical resolution available to us
for our FHC calculation produced results for a case in which d
occupies only a small fraction of the total volume. The FHC
is comparable to, but slightly larger than, w. In the true subgrid-
case has a larger volume of ejected flux and a higher peak
scale regime where d = w, in contrast, we would expect the
kinetic energy content, some 50% more, than the STITCH case.
resulting FHC channel width to be determined by the magnetic
The shear field proxy, ∣Bf∣max , is shown in Figure 5(d). This
field length scale w. The FHC and STITCH results then should
evolution, discussed previously in our work on the FHC
come into even better alignment than for the cases presented
simulation (Dahlin et al. 2019), exhibits a strong initial increase
here, and they should show an even greater disparity in their
as the shear flux begins to accumulate above the PIL in the low
respective timescales, further enhancing the advantages of
corona. Eventually, although the total amount of shear flux STITCH for studies of filament channel formation.
continues to rise, the peak shear strength begins to decline, as
the accumulating shear magnetic pressure inflates the overlying
5. Localized Helicity Injection
arcade loops and expands the volume occupied by the flux.
This steady decrease continues until eruption onset. Although The STITCH case presented above assumed uniform ζ
much of the shear flux is ejected upward as core field of the throughout the minority-polarity region, best matching the

7
The Astrophysical Journal, 941:79 (15pp), 2022 December 10 Dahlin, DeVore, & Antiochos

Figure 4. Filament channel (a) formation and (b) eruption for the STITCH calculation with uniform ζ. Magnetic field lines are colored according to the scheme used in
Figure 1, with orange lines added to show the filament channel field. Color shading in the plane of the sky (right) shows radial velocity Vr. An animation of this figure
is available, showing the simultaneous evolution of both panels (a) and (b) throughout the driving and eruptive phases from t = 0 s to t = 7200 s at 50 s cadence. (The
animation duration is 10 s.)
(An animation of this figure is available.)

uniform vortical flows in the FHC simulation. The resulting minority-polarity region, due to contributions to the injection
eruptions displayed in Figures 3 and 4 (best seen in the online rates made by the ζ gradients. Based on the results of our
animations) encompass the entire elliptical filament channel, simulation shown previously, we ramped down the tangential
proceeding first on the southern segment of the PIL and field injection between elapsed times t = 6000 s and t = 7000 s,
subsequently on the northern segment. This scenario is not reducing the injection rate during the eruption. We also reran
universal, but it is observed routinely on the Sun (e.g., Wang
the uniform-ζ case with this change in the time profile, for
et al. 2012). The sequence implies that the energy buildup is
relatively uniform all along the PIL; hence, the entire sheared consistency.
Snapshots of the evolving configurations at time t = 6200 s
structure approaches its critical point and transitions to eruption
nearly instantaneously. For the configuration discussed in are shown in Figure 6 for the uniform-ζ (left; panels (a)–(c))
Section 4, for example, the gradient in Br across the PIL varies and localized-ζ (right; panels (d)–(f)) cases. Several significant
relatively little (by less than a factor of two) along the length. differences are clearly evident from this comparison. In the
Therefore, the rate of STITCH flux injection likewise is quite uniform-ζ case, strongly sheared field lines populate the
uniform (Figures 2(b) and (e)), and the resulting eruption is a filament channel all along the PIL (Figure 6(a)); in the
two-sided ejection of the entire channel. localized-ζ case, on the other hand, such field lines are
More typically, however, solar eruptions occur along observed only along the southern segment of the PIL, with
restricted segments of the filament channels and present one- weak shear present along the other segments by design
sided ejections, even along quasi-circular PILs. For example, (Figure 6(b)). The red arcade field lines overlying the southern
such an event is described by Mason et al. (2021). The implied filament channel are inflated to a greater extent in the localized-
local concentration of energy and shear buildup could arise as a ζ case, due in part to less competition from the much more
result of several mechanisms, e.g., variable vortical granular/ weakly sheared, northern segment of the PIL, and in part to the
supergranular and/or large-scale shear flows, disordered flux reversed shear that has been imparted to these field lines. This
cancellation, flux emergence, etc. The STITCH model allows reversed shear arises from the relatively small, but finite,
us to abstract any and all such mechanisms into our simulations injection of oppositely signed Bf flux near the center of the
by simply adopting a spatially varying ζ profile, thereby minority-polarity region due to the ζ gradients in the STITCH
localizing the helicity and energy injection to a selected region. term (noted in the description of Figure 2(f)).
We modified our uniform-ζ simulation in this way by centering Side views of the azimuthal magnetic field Bf show how
a simple cosine variation on the southern PIL, localizing the STITCH injection (Figure 6(d)) greatly reduces
the shear along the northern PIL segment and introduces the
f - f0 ⎞ reversed shear in the center of the minority polarity (both
z = z 0 cos ⎛⎜p ⎛ y - y 0 ⎞.
⎟ cos p (20)
⎜ ⎟
shaded light red), relative to the uniform STITCH injection
2 f ⎝ 2y w ⎠
⎝ w ⎠ (Figure 6(c)). Obvious features of the uniform-ζ case
Here f is longitude, ψ is latitude, f0 = 0°, ψ0 = 22°. 5, (Figure 6(c)) are the roughly equal strengths of its northern
fw = 22°. 5, and ψw = 11°. 25. The amplitude ζ0 and the and southern filament channels and the extents to which they
bulge upward and outward as they inflate toward nearly
restriction to the minority-polarity region are the same as in
simultaneous eruptions. In contrast, only the southern filament
the previous simulation. Figures 2(c) and (f) show the resulting channel is clearly primed for eruption in the localized-ζ case
injection rates for Bf and Bθ. The principal feature is that the (Figure 6(d)). These views also highlight a slight northward tilt
injection is weakened substantially along the eastern, northern, of the southern filament channel field and its overlying arcade
and western segments of the PIL relative to the uniform-ζ case. (shaded blue) in this case.
A secondary feature is that the distribution of flux injection Side views of magnetic field lines in the four flux systems of
spreads from those segments of the PIL into the interior of the the configuration further illustrate these features in the uniform-

8
The Astrophysical Journal, 941:79 (15pp), 2022 December 10 Dahlin, DeVore, & Antiochos

Figure 5. Global diagnostics for STITCH (black curves and time-axis labels) with uniform ζ and FHC (red curves and time-axis labels): (a) EM/EM0; (b)
d (EM EM0 ) dt; (c) Ek/EM0; and (d) ∣Bf∣max .

ζ (Figure 6(e)) and localized-ζ (Figure 6(f)) cases. The arcade does not drop quite as low prior to eruption, nor does it increase
of blue field lines in the north, along with its enclosed orange quite as high after eruption, compared to the uniform-ζ case.
field lines, resides much lower in height in the latter versus the These two features indicate that the filament channel and its
former. The color shading of radial velocity Vr in these panels overlying arcade neither expand quite as freely nor contract
captures the outward motion of the central arcade of red field quite as much, when the shear is localized to the southern PIL.
lines in both cases, and it also illustrates breakout reconnection
downflows along the flanks of this arcade in the localized-ζ
case (Figure 6(f)). 6. Energizing Complex Flux Distributions
Global diagnostics for these two simulations, uniform-ζ The idealized flux distribution used in the preceding
(black curves) and localized-ζ (green curves), are compared in examples has sufficed to show some of the capabilities of the
Figure 7. (In these plots, unlike the previous similar Figure 5, STITCH model. Actual solar flux distributions can be much
the time axes for the two cases [both STITCH] are identical.) more complex, even fragmented, posing far greater challenges
Figures 7(a), (b), and (c) show the energies for the uniform-ζ to eruptive event modeling of such regions and requiring far
results at 50% amplitude, drawn with dashed curves. greater computational resources. To explore the potential
Coincidentally, these reduced curves track very well with usefulness of STITCH for investigations of this kind, we
those for the localized-ζ case, in the buildup to eruption for the created a synthetic active region with a corrugated PIL by
magnetic energy (Figures 7(a) and (b)) and beyond into the superposing 30 subsurface magnetic dipoles, all centered
eruption phase for the kinetic energy (Figure 7(c)). This nominally at latitude ψ0 = 22°. 5 and longitude f0 = 0°. We
agreement indicates that, during the buildup, the northern and randomized their displacements away from this center,
southern segments of the PIL store roughly equal amounts of |ψ − ψ0| „ 7°. 5 and |f| „ 22°. 5, and their horizontal orienta-
energy for uniform ζ. The magnetic energy release rate during tions away from due north, within ±45°. Each dipole was
the eruption is essentially the same in the two cases, although placed at a depth of 80 Mm below the surface and was given a
the release subsides somewhat faster for localized ζ tangential field strength of 6.7 G at the surface. Figures 8(a)
(Figure 7(b)). The kinetic energy for that case peaks slightly and (b) show views of the resulting initial magnetic field with
above 50% of the uniform-ζ value (Figure 7(c)). Peak strengths its corrugated PIL.
of the shear field attained early in the energization phase are The active region was energized using the STITCH profiles
essentially identical for the two cases (Figure 7(d)). Early in the of tangential magnetic field injection shown in Figure 9. We
evolution for localized ζ, a transient expansion of the field adopted ζ as expressed in Equation (20), but here we used
overlying the channel briefly reduces the shear strength, fw = 45° and ψw = 27°. As can be seen in the figure, STITCH
producing a double peak. Late in the evolution, the shear field flux injection was applied in both polarities of the active region,

9
The Astrophysical Journal, 941:79 (15pp), 2022 December 10 Dahlin, DeVore, & Antiochos

Figure 6. Selected filament channel magnetic field lines plus gray shaded radial magnetic field component (panels (a) and (d)), azimuthal magnetic field Bf saturated at
±15 G (panels (b) and (e)), and selected magnetic field lines plus color shaded radial velocity component (panels (c) and (f)) for uniform ζ (panels (a)–(c)) and
localized ζ (panels (d)–(f)) at t = 6200 s. An animation of this figure is provided online, showing the driving and eruptive phases of both calculations (the figure above
represents a snapshot from the animation) from t = 0 s to t = 7200 s at 50 s cadence. The animation duration is 10 s.
(An animation of this figure is available.)

as well as in the surrounding background magnetic field. In this STITCH driving procedure from the simple case was required
case, we used a smaller value of ζ0 = 5 × 105 km2 s−1, and the for handling this more “realistic” PIL, whereas driving by
gradients in the radial magnetic field Br were gentler owing to small-scale twists, shear flows, or flux rope insertion would
our constructing the active region from a set of well-submerged have required the tedious construction of drivers that followed
dipoles. Consequently, the resulting values of dBs/dt were the detailed geometry of the meandering PIL.
about an order of magnitude smaller than in our idealized cases
(see Figure 2). We allowed the usual ramp-up interval of 1000 s
duration in this simulation and then held the tangential field 7. Utility for Event Modeling
injection rates fixed thereafter. In this section, we briefly describe the utility of STITCH to
The elapsed time required to build a strongly sheared event modeling of solar eruptions (for reviews see, e.g.,
filament channel and induce it to erupt was about 2.5× larger Toriumi & Wang 2019; Jiang et al. 2022). A full discussion of
for this case, i.e., some 15,000 s. Snapshots of the formation the topic is beyond the scope of our paper and ambition. To
and eruption stages in the evolution are shown in Figures 8(c) frame the presentation, we note that event modeling attempts to
and (d). The development is very similar qualitatively to that replicate an observed solar eruption by taking one of three
shown in our previous examples, with the filament channel general approaches (for details see, e.g., Chintzoglou et al.
highly localized to the PIL and generating a fast CME when it 2019; Toriumi & Wang 2019): (1), (2) Energize some
erupts. The key point here is that essentially no change in the (arbitrarily) chosen initial configuration with the model,

10
The Astrophysical Journal, 941:79 (15pp), 2022 December 10 Dahlin, DeVore, & Antiochos

Figure 7. Global diagnostics for uniform ζ (black curves) and localized ζ (green curves): (a) EM/EM0; (b) d (EM EM0 ) dt; (c) Ek/EM0; and (d) ∣Bf∣max . Dashed black
curves in panels (a)–(c) show halved quantities for uniform ζ.

generating a single-snapshot end state that resembles the match the observed filament channel (e.g., Torok et al. 2018).
observed pre-eruptive solar structure; this modeled state may Such intricate iterative procedures would be challenging to use
correspond only qualitatively to the observed state (“data- in real time for space-weather predictive modeling. A third
inspired”), or it may be based quantitatively on measurements problem is that after an equilibrium has been calculated the
of the observed state (“data-constrained”). (3) Evolve the eruption must be triggered without disturbing the original
observed solar structure within the model over an extended normal flux distribution. By construction, STITCH avoids all
series of snapshots, constraining the model evolution so that it of these difficulties. Unlike flux rope insertion, the shear
replicates the multistep evolution observed in the data at each injection by STITCH is continuous; hence, it can be halted
step (“data-driven”). readily at any point to examine the pre-eruptive structure and
The usefulness of STITCH for energizing filament channels possible trigger mechanisms near marginal stability. Note that
is demonstrated by the results presented in the preceding STITCH could be run automatically as part of any 3D MHD
sections. It has a number of key advantages in comparison to heliospheric model. Furthermore, STITCH is so flexible and
some other methods commonly used in data-inspired or data- easily implemented that it can be used in combination with the
constrained models to generate the coronal structure preceding other filament channel formation and energization mechanisms.
a solar eruptive event (for a review, see Patsourakos et al. For example, one could insert one or more equilibrium flux
2020). These methods include flux emergence from below the ropes where desired and then use STITCH to drive the system
surface (e.g., Manchester et al. 2004; Leake et al. 2013), to eruption while preserving the matched surface boundary
surface shear flows (e.g., Lynch et al. 2008; Wyper et al. 2017), conditions. Beyond its foundation in a compelling physical
surface flux cancellation (e.g., Amari et al. 2010; Aulanier et al. model for filament channel formation and energization, helicity
2010; Hassanin et al. 2022), and procedures for direct insertion condensation, STITCH clearly has substantial practical utility
of coronal flux ropes (e.g., Titov & Démoulin 1999; van for use in single-snapshot, data-inspired or data-constrained
Ballegooijen 2004; Fan 2005; Borovikov et al. 2017; Titov eruptive event modeling.
et al. 2021). One problem for these methods is matching the Multiple-snapshot, data-driven modeling of solar events is a
normal flux at the boundary of the numerical system to much more challenging undertaking, requiring the model to be
photospheric observations. This is particularly difficult for constrained at each step over a time sequence of observational
emergence, shear flows, and cancellation, because these measurements. Numerous approaches have been developed and
processes generally change the flux distribution. Another investigations pursued with this aim in recent years (e.g.,
problem is dealing with complex-geometry PILs, which can Mackay et al. 2011; Cheung & DeRosa 2012; Fisher et al.
be especially challenging for insertion methods that require the 2012; Kazachenko et al. 2014; Cheung et al. 2015; Lumme
specification of multiple different flux ropes carefully fitted to et al. 2017; Hayashi et al. 2018, 2022; Pomoell et al. 2019;

11
The Astrophysical Journal, 941:79 (15pp), 2022 December 10 Dahlin, DeVore, & Antiochos

Figure 8. Initial configuration (panels (a) and (c)) of a more complex flux distribution with a corrugated PIL and later snapshots during (b) formation and (d) eruption
of its filament channel. Magnetic field lines are color-coded and radial velocity is color shaded as in Figure 3.

Hoeksema et al. 2020; Toriumi et al. 2020). Unlike any single-

snapshot method, including STITCH, data-driven models must
incorporate a horizontal electric field Es that replicates the
evolution of the normal magnetic field component Bn at the
Sun’s surface,
¶Bn
= - s ´ Es. (21)
¶t
In order to fully constrain the evolution of the surface magnetic
field, observational data must be available and reliable for the
rate of change of both Bn and Bs. Due to observational
limitations, these boundary values are frequently underspeci-
fied, and a variety of methods have been developed to
complement the measurements (see discussion in Jiang et al.
2022). We note that it is precisely the horizontal components of
the magnetic field that are critical to the formation and
energization of the filament channels that culminate in solar
eruptions.
Among those models that do not attempt to constrain the
evolution of the horizontal field Bs, one class (Cheung &
DeRosa 2012; Cheung et al. 2015; Lumme et al. 2017; Pomoell
et al. 2019) imposes a mathematical closure on the electric field
Es by supplementing the constraint on its curl (Equation (21))
with a specification for its divergence, via one of a hierarchy of
prescribed forms:

⎧C0;
s · Es = C1n · B ; (22)
⎨
⎩C2 n · (s ´ B).
The constant Ci is determined empirically by comparing the
evolution of the observed and modeled systems. The STITCH
Figure 9. Tangential magnetic field injection for a more complex surface flux term in our induction equation (Equation (5)) corresponds to
distribution with a corrugated PIL. Black contours show values of Br = [−30, the potential (curl-free) electric field
−20, −10, 0, +10, +20] G.
EsS = s (zBn) , (23)

12
The Astrophysical Journal, 941:79 (15pp), 2022 December 10 Dahlin, DeVore, & Antiochos

whose divergence is reconnection-driven transport of the shear field across the

surface.
s · EsS = 2s (zBn). (24)

In regions where ζ is uniform, this expression may be recast 8. Discussion

into the simple form
STITCH is our acronym for STatiscal InjecTion of
s · EsS = C3 n · (2s B). (25) Condensed Helicity, a subgrid-scale physical model for the
formation and evolution of filament channels on the Sun. In
Mathematically, STITCH is similar to the next element in the this article, we have used STITCH to inject magnetic helicity
and free energy into simulated coronae that evolve self-
Cheung hierarchy (Equation (22)) at one higher level in
consistently according to the otherwise-familiar equations of
derivatives of B (Equation (25)). Mackay et al. (2014)
magnetohydrodynamics. The only modification to standard
estimated the solar value for ζ to be (6 ± 4) × 102 km2 s−1. MHD that is required to implement STITCH is the addition of a
It would be highly informative to see the results of an event new, mathematically simple term to the induction equation
modeling experiment that assumed Equation (25) with a value (Faraday’s law). Including the STITCH term amounts to
for C3 in this ζ range. Notice that the STITCH mathematical adjusting the electric field, i.e., modifying Ohm’s law in the
form is motivated by the physics of helicity condensation, and coronal plasma, at the base of the atmosphere.
the value of its parameter is constrained a priori by The mathematical simplicity of the STITCH term obscures,
observations. Hence, a successful experiment could lead to a but simultaneously encapsulates, a great deal of physical
substantial advance in our broad physical understanding of complexity. As derived originally (Mackay et al. 2014, 2018),
filament channel formation and evolution on the Sun. it represents the injection of helicity by vortical cells
A second class of models in which the horizontal field Bs is characterized by rotation rate ωℓ and radius ℓ, the formation
unconstrained is surface flux transport simulation (e.g., Mackay of current sheets at the boundaries between adjacent cells, and
et al. 2011, 2014, 2018; Hoeksema et al. 2020), in which efficient reconnection of the induced horizontal magnetic fields
prescribed physical processes of differential rotation, meridio- across the current sheets. The STITCH model distills these
nal flow, flux cancellation due to supergranular diffusion, and detailed processes of helicity condensation described by
emergence of fresh magnetic flux from below are all included. Antiochos (2013) into a single term with one parameter,
The evolution of the normal field Bn in such models is more ζ ≡ 〈ℓ2ωℓ〉/2, plus an assumption about the height h over which
loosely constrained, on a rather long (typically one Carrington the helicity is injected into the magnetic field. In practice, the
rotation) timescale, by periodically assimilating up-to-date simplest approach is to perform the injection into the bottom-
maps of the normal surface field into the simulation to correct most grid cell adjacent to the lower boundary, as has been done
any accumulating errors. STITCH was introduced initially in here and by others (Mackay et al. 2014, 2018; Lynch et al.
collaborative studies with Mackay (2014, 2018; referred to 2021).
therein as Statistically Averaged Helicity Condensation) that Implementing and using the STITCH term in MHD models
showed how adding our subgrid-scale model to his simulations are straightforward:
improved the agreement between the observed and simulated 1. Assign a value to the helicity injection parameter ζ.
magnetic maps. In particular, sufficiently fast helicity con- 2. Evaluate the STITCH product ζBn at the bottom
densation was able to counter the shear induced by differential boundary, n = n0, where Bn is the normal component of
rotation along the high-latitude PILs of the polar crown. the magnetic field and n = x (Cartesian) or n = r
Following on these studies, the STITCH formalism was used (spherical).
by Mackay and coworkers to predict the state of the corona 3. Difference this product along the two horizontal coordi-
during the 2017 August 21 solar eclipse (Mikic et al. 2018; nates, (y, z) or (θ, f), setting the sign according to the
Yeates et al. 2018). Those simulations included an electric field rules for the curl (see Equations (11) and (12)), to
of the STITCH form (Equation (23)), in which ζ was replaced compute the tangential flux change rates.
by a masking coefficient M employed to localize the induced Bs 4. Divide the flux change rates by the appropriate vertical
to the neighborhood of the PIL (Mikic et al. 2018, p. 919), cell-face areas to calculate the corresponding tangential
similar to the procedure that we have described in detail here in field change rates (see Equations (13) and (14)).
Section 5 on localized helicity injection. The eclipse simula- 5. Multiply these rates by the numerical time increment, Δt,
tions (and others) described the masked electric field as finishing the calculation of the tangential field changes.
corresponding to “emerging” transverse (horizontal) magnetic
field locally within the filament channels (Yeates et al. 2018, p. To our knowledge, the procedure above would be trivial to
99). All of the simulations shown in this paper assume a implement in any of the MHD codes currently used by the
reflecting condition (i.e., a zero value) on the normal space science community. Indeed, it already has been
component of the plasma velocity V at the inner radial implemented in one such model (Mikic et al. 2018; Yeates
boundary; consequently, flux emergence plays no role in the et al. 2018). Moreover, the helicity injection via STITCH may
formation of our filament channels. Furthermore, we are not be tailored spatially to occur only within one polarity of the
convinced that the highly complex photosphere−corona magnetic field, or only within one or more restricted region(s)
interaction accompanying flux emergence can be captured by of the bottom-boundary surface, or any combination of both.
a simple STITCH-like formalism. Rather, the STITCH term This is accomplished by introducing suitable spatial variations
that we originally derived in Mackay et al. (2014) and employ into the ζ parameter adopted in step 1, as we have done in this
here seeks to capture effectively the cumulative effects of paper. The results for the tangential field changes will be
surface vortical flows, induced horizontal shear field, and mathematically smooth so long as the product ζBn goes to zero

13
The Astrophysical Journal, 941:79 (15pp), 2022 December 10 Dahlin, DeVore, & Antiochos

smoothly at the boundary of the tailored region, i.e., either ζ by a NASA Postdoctoral Program fellowship administered by
vanishes (by construction) or Bn vanishes (at a PIL). Universities Space Research Association at NASA Goddard
Although the STITCH model is derived from the intricate Space Flight Center. J.T.D. acknowledges support from Grant
physical processes inherent to the FHC model, we emphasize No. 80NSSC21K1313. S.K.A. acknowledges support from
that its essence is to accumulate helicity at the large-scale Grant No. 80NSSC22K0892 to the University of Michigan.
boundaries of magnetic flux systems, in particular at PILs.
Therefore, in general it represents the outcome of an inverse ORCID iDs
cascade of magnetic helicity, from small scales to large, due to
any mechanism that injects helicity into the corona: vortical J. T. Dahlin https://orcid.org/0000-0002-9493-4730
motions, shearing motions, flux emergence, flux cancellation, C. R. DeVore https://orcid.org/0000-0002-4668-591X
and so forth. Our results in this article demonstrate that S. K. Antiochos https://orcid.org/0000-0003-0176-4312
STITCH can be used as a generic tool for coronal MHD
modeling that forms filament channels and energizes magnetic References
field configurations. Here we have used it to evolve both highly
Amari, T., Aly, J. J., Mikic, Z., & Linker, J. 2010, ApJL, 717, L26
idealized and more elaborately corrugated active region Antiochos, S. K. 1990, MmSAI, 61, 369
structures, and we have compared it successfully, both Antiochos, S. K. 1998, ApJL, 502, L181
qualitatively and quantitatively, with a fully detailed helicity Antiochos, S. K. 2013, ApJ, 772, 72
condensation calculation. In other work, we have used STITCH Antiochos, S. K., DeVore, C. R., & Klimchuk, J. A. 1999, ApJ, 510, 485
Aulanier, G., Torok, T., Demoulin, P., & DeLuca, E. E. 2010, ApJ, 708, 314
to perform full-Sun, sunspot-cycle scale modeling of filaments Berger, M. A. 1984, GApFD, 30, 79
in a flux transport model (Mackay et al. 2014, 2018) and to Berger, M. A., & Field, G. B. 1984, JFM, 147, 133
energize a high-latitude filament channel leading to the Biskamp, D. 1993, Nonlinear Magnetohydrodynamics (Cambridge: Cambridge
eruption of a “stealthy” CME (Lynch et al. 2021). Others Univ. Press)
Borovikov, D., Sokolov, I. V., Manchester, W. B., Jin, M., & Gombosi, T. I.
have employed the method to form filament channels in a 2017, JGRA, 122, 7979
global corona model used to predict solar eclipse observations Cheung, M. C. M., & DeRosa, M. L. 2012, ApJ, 757, 147
(Mikic et al. 2018; Yeates et al. 2018). Cheung, M. C. M., Pontieu, B. D., Tarbell, T. D., et al. 2015, ApJ, 801, 83
The STITCH model possesses the following efficiencies and Chintzoglou, G., Zhang, J., Cheung, M. C. M., & Kazachenko, M. 2019, ApJ,
871, 67
advantages compared to fully detailed MHD descriptions of Dahlin, J. T., Antiochos, S. K., & DeVore, C. R. 2019, ApJ, 879, 96
filament channel formation and eruption: DeVore, C. R., & Antiochos, S. K. 2008, ApJ, 680, 740
Duvall, T. L. J., & Gizon, L. 2000, SoPh, 192, 177
1. Straightforward implementation within existing MHD Fan, Y. 2001, ApJL, 554, L111
models. Fan, Y. 2005, ApJ, 630, 543
2. Easy application to realistic, complex PILs, as well as to Finn, J. M., & Antonsen, T. M. 1985, CoPPC, 9, 111
localized regions along extended PILs, as is likely to be Fisher, G. H., Welsch, B. T., & Abbett, W. P. 2012, SoPh, 277, 153
Gaizauskas, V. 1998, in New Perspectives on Solar Prominences, ed.
required for modeling observed events. D. F. Webb, B. Schmieder, & D. M. Rust (San Francisco, CA: ASP), 257
3. Energizes the system while leaving the normal comp- Gizon, L., & Duvall, T. L. 2003, in ESA Spec. Publ. 517, GONG+ 2002.
onent of the magnetic field untouched, thereby greatly Local and Global Helioseismology: the Present and Future, ed.
facilitating event modeling. H. Sawaya-Lacoste (Noordwijk: ESA), 43
Hassanin, A., Kliem, B., Seehafer, N., & Torok, T. 2022, ApJL, 929, L23
4. Reduced computational expense (15× in our detailed Hayashi, K., Feng, X., Xiong, M., & Jiang, C. 2018, ApJ, 855, 11
comparison) relative to FHC because STITCH is Hayashi, K., Wu, C.-C., & Liou, K. 2022, ApJ, 930, 60
amenable to sub-Alfvénic (flux injection) driving rather Hoeksema, J. T., Abbett, W. P., Bercik, D. J., et al. 2020, ApJS, 250, 28
than requiring subsonic (vortical-flow) driving, allowing Howard, R. A., Michels, D. J., Sheeley, N. R., Jr., & Koomen, M. J. 1982,
ApJL, 263, L101
much faster formation and eruption of filament channels. Jiang, C., Feng, X., Guo, Y., & Hu, Q. 2022, Innov, 3, 100236
5. Reduced computational expense (not exploited in our Karpen, J. T., Antiochos, S. K., & DeVore, C. R. 2012, ApJ, 760, 81
detailed comparison) relative to FHC because the small- Kazachenko, M. D., Fisher, G. H., & Welsch, B. T. 2014, ApJ, 795, 17
scale vortical flows and current structures need not be Kliem, B., & Torok, T. 2006, PhRvL, 96, 255002
Knizhnik, K. J., Antiochos, S. K., & DeVore, C. R. 2015, ApJ, 809, 137
resolved, allowing the use of coarser grids. Knizhnik, K. J., Antiochos, S. K., & DeVore, C. R. 2017a, ApJ, 835, 85
6. Elimination of complex small-scale structure and Knizhnik, K. J., Antiochos, S. K., DeVore, C. R., & Wyper, P. F. 2017b, ApJL,
dynamics, so that the large-scale topology and dynamics 851, L17
of the filament channel evolution can be ascertained more Knizhnik, K. J., Antiochos, S. K., Klimchuk, J. A., & DeVore, C. R. 2019,
ApJ, 883, 26
readily. Knizhnik, K. J., Uritsky, V. M., Klimchuk, J. A., & DeVore, C. R. 2018, ApJ,
In summary, STITCH demonstrates substantial utility for 853, 82
Komm, R., De Moortel, I., Fan, Y., Ilonidis, S., & Steiner, O. 2015, SSRv,
physics-based modeling of filament channel formation and 196, 167
evolution culminating in CMEs and, potentially, great promise Komm, R., Howe, R., Hill, F., et al. 2007, ApJ, 667, 520765
as an operational tool for future space-weather prediction Lau, Y.-T., & Finn, J. M. 1990, ApJ, 350, 672
capabilities. The work required to demonstrate and validate this Leake, J. E., Linton, M. G., & Torok, T. 2013, ApJ, 778, 99
Linker, J. A., Mikic, Z., Lionello, R., et al. 2003, PhPl, 10, 1971
promise is already underway. Lumme, E., Pomoell, J., & Kilpua, E. K. J. 2017, SoPh, 292, 191
Lynch, B. J., Antiochos, S. K., DeVore, C. R., Luhmann, J. G., &
Our work was sponsored by NASA’s H-LWS, H-SR, and Zurbuchen, T. H. 2008, ApJ, 683, 1192
H-ISFM research programs. Resources supporting this work Lynch, B. J., Antiochos, S. K., DeVore, C. R., Luhmann, J. G., &
Zurbuchen, T. H. 2009, ApJ, 697, 1918
were provided by the NASA High-End Computing (HEC) Lynch, B. J., Masson, S., Li, Y., et al. 2016, JGRA, 121, 10
Program through the NASA Center for Climate Simulation Lynch, B. J., Palmerio, E., DeVore, C. R., et al. 2021, ApJ, 914, 39
(NCCS) at Goddard Space Flight Center. J.T.D. was supported Ma, S., Attrill, G. D. R., Golub, L., & Lin, J. 2010, ApJ, 722, 289

14
The Astrophysical Journal, 941:79 (15pp), 2022 December 10 Dahlin, DeVore, & Antiochos

Mackay, D., & Yeates, A. 2012, LRSP, 9, 6 Priest, E. R., & Titov, V. S. 1996, RSPSA, 354, 2951
Mackay, D. H., DeVore, C. R., & Antiochos, S. K. 2014, ApJ, 784, 164 Raouafi, N. E., Patsourakos, S., Pariat, E., et al. 2016, SSRv, 201, 1
Mackay, D. H., DeVore, C. R., Antiochos, S. K., & Yeates, A. R. 2018, ApJ, Sun, X., Hoeksema, J. T., Liu, Y., et al. 2012, ApJ, 748, 77
869, 62 Titov, V. S., & Démoulin, P. 1999, A&A, 351, 707
Mackay, D. H., Green, L. M., & van Ballegooijen, A. 2011, ApJ, 729, 97 Titov, V. S., Downs, C., Torok, T., et al. 2021, ApJS, 255, 9
Manchester, W., IV, Gombosi, T., DeZeeuw, D., & Fan, Y. 2004, ApJ, Toriumi, S., Takasao, S., Cheung, M. C. M., et al. 2020, ApJ, 890, 103
610, 588 Toriumi, S., & Wang, H. 2019, LRSP, 16, 3
Martin, S. F. 1998, SoPh, 182, 107 Torok, T., Downs, C., Linker, J. A., et al. 2018, ApJ, 856, 75
Mason, E. I., Antiochos, S. K., & Vourlidas, A. 2021, ApJL, 914, L8 van Ballegooijen, A. A. 2004, ApJ, 612, 519
Masson, S., Antiochos, S. K., & DeVore, C. R. 2013, ApJ, 771, 82 van Ballegooijen, A. A., & Martens, P. C. H. 1989, ApJ, 343, 971
Masson, S., Pariat, E., Aulanier, G., & Schrijver, C. J. 2009, ApJ, 700, 559 Wang, H., Liu, C., Deng, N., et al. 2014, ApJL, 781, L23
Mikic, Z., Downs, C., Linker, J. A., et al. 2018, NatAs, 2, 913 Wang, S., Liu, C., Liu, R., et al. 2012, ApJL, 745, L17
Moore, R. L., Sterling, A. C., Hudson, H. S., & Lemen, J. R. 2001, ApJ, Wyper, P. F., Antiochos, S. K., & DeVore, C. R. 2017, Natur, 544, 452
552, 833 Yeates, A. R., Amari, T., Contopoulos, I., et al. 2018, SSRv, 214, 99
Ouyang, Y., Zhou, Y. H., Chen, P. F., & Fang, C. 2017, ApJ, 835, 94 Yeates, A. R., & Mackay, D. H. 2012, ApJL, 753, L34
Patsourakos, S., Vourlidas, A., Torok, T., et al. 2020, SSRv, 216, 131 Zhao, L., DeVore, C. R., Antiochos, S. K., & Zurbuchen, T. H. 2015, ApJ,
Pomoell, J., Lumme, E., & Kilpua, E. 2019, SoPh, 294, 41 805, 61

15