Professional Documents
Culture Documents
Shedding Light On The MRI Driven Dynamo in A Stratified Shearing Box
Shedding Light On The MRI Driven Dynamo in A Stratified Shearing Box
17 August 2023
ABSTRACT
We study the magneto-rotational instability (MRI) dynamo in a geometrically thin disc (H/R ≪ 1) using stratified zero net flux
(ZNF) shearing box simulations. We find that mean fields and EMFs oscillate with a primary frequency fdyn = 0.017 (≈ 9
orbital period), but also have higher harmonics at 3fdyn . Correspondingly, the current helicity, has two frequencies 2fdyn and
4fdyn respectively, which appear to be the beat frequencies of mean fields and EMFs as expected from the magnetic helicity
density evolution equation. Further, we adopt a novel inversion algorithm called the ‘Iterative Removal Of Sources’ (IROS), to
extract the turbulent dynamo coefficients in the mean-field closure using the mean magnetic fields and EMFs obtained from the
shearing box simulation. We show that an α−effect (αyy ) is predominantly responsible for the creation of the poloidal field from
the toroidal field, while shear generates back a toroidal field from the poloidal field; indicating that an α − Ω-type dynamo is
operative in MRI-driven accretion discs. We also find that both strong outflow (v̄z ) and turbulent pumping (γz ) transport mean
fields away from the mid-plane. Instead of turbulent diffusivity, they are the principal sink terms in the mean magnetic energy
evolution equation. We find encouraging evidence that a generative helicity flux is responsible for the effective α-effect. Finally,
we point out potential limitations of horizontal (x − y) averaging in defining the ‘mean’ on the extraction of dynamo coefficients
and their physical interpretations.
Key words: accretion,accretion discs - dynamo - instabilities - magnetic fields - MHD - turbulence - methods: numerical.
1 INTRODUCTION the density scale height and R is the local radius. Therefore, this
approach is valid only for geometrically thin discs with H/R ≪ 1.
The problem of angular momentum transport is a key concept in a ro-
Depending on whether the vertical component of gravity (gz =
tationally supported accretion disc (for a review, see Balbus & Hawley
−Ω2 z) (producing a vertically stratified gas density) is considered
1998). The current consensus is that a weak magnetic field instability,
in the momentum equation or not, shearing box simulations are of
namely magneto-rotational instability (MRI; Velikhov 1959; Chan-
two types: stratified (gz ̸= 0) and unstratified (gz = 0). Further,
drasekhar 1960; Balbus & Hawley 1991; Balbus & Hawley 1992)
depending on whether the computational domain can contain net
is responsible for outward angular momentum transport and drives
magnetic fields, shearing box models can be classified into zero
mass accretion in a sufficiently ionized accretion disc. Although lin-
net flux (ZNF) and net flux (NF) models. Therefore, four possible
ear MRI ensures outward angular momentum transport, it must be
combinations of the shearing-box model are: i) unstratified ZNF, ii)
studied in the non-linear phase to account for different observable
unstratified NF, iii) stratified ZNF and iv) stratified NF. This work
phenomena in accretion discs.
considers a stratified ZNF shearing box model to explore the MRI
MRI in an accretion disc is either studied in a local set-up (shearing
dynamo in saturation.
box; Balbus & Hawley 1992; Brandenburg et al. 1995; Hawley et al.
1995; Davis et al. 2010; Shi et al. 2010; Bodo et al. 2014; Bhat et al. Shearing box simulations provide a wide range of behaviour (e.g.,
2016) or in a global simulation (Stone et al. 1999; Hawley 2001; convergence, turbulence characteristics etc.) depending on the shear-
Hawley et al. 2013; Beckwith et al. 2011; Parkin & Bicknell 2013; ing box model used (for details, we refer to readers to see Table 1 in
Hogg & Reynolds 2016; Dhang & Sharma 2019; Dhang et al. 2023). Ryan et al. (2017)). However, it is to be noted that we will restrict
While a global approach is more desirable, it is computationally our discussion to the isothermal (i.e. sound speed is constant) models
expensive. On the other hand, the shearing box approach offers an where there is no explicit dissipation and the numerical algorithms
alternate path which is computationally less costly and can provide provide the dissipation through truncation error at the grid scale. In
deep insights into the local processes in MRI-driven turbulence. the presence of an NF, unstratified shearing box simulations show
In the shearing-box approach (Goldreich & Lynden-Bell 1965), a convergence (in terms of accretion stresses) and sustained turbu-
we expand fluid equations to the lowest order of H/R, where H is lence (Hawley et al. 1995; Guan et al. 2009; Simon et al. 2009). On
the other hand, stratified NF simulations present different accretion
stresses depending on the net flux strength and sustained turbulence
⋆ E-mail:prasundhang@gmail.com (Guan & Gammie 2011; Bai & Stone 2013). Unstratified ZNF mod-
αRey , αMax
E (z ′ , t ) Ey (z ′ , t1 )
x 2
y(z ′ , t) =
.. .. 0.02
. .
Ex (z ′ , tN ) ′
Ey (z , t1 ) 0.01
0.00
0 50 100 150 200 250 300
B̄x (z ′ , t1 ) B̄x (z ′ , t2 ) . . . B̄x (z ′ , tN )
B̄y (z ′ , t1 ) 0.015 B 02 B̄ 2
B̄y (z ′ , t2 ) . . . B̄y (z ′ , tN )
′ ′ ′
B 02 , B̄ 2
⊺ ′
C x (z , t 1 ) C x (z , t 2 ) . . . C x (z , t N ) 0.010
A (z , t) =
′ ′ ′
C
y (z , t 1 ) C y (z , t 2 ) . . . C y (z , t N )
¯ ′ 0.005
−Jx (z , t1 ) −J¯x (z , t2 ) . . . − J¯x (z , tN )
′ ′
ηxy (z ′ ) ηyy (z ′ ) the EMF associated with the best fitted coefficient, also multiplied by
the ϵ is subtracted from the EMF component. For instance using the
Here the terms Ci (z ′ , tl ) = B̄i (z ′ , tl ) cos (2Ωdyn tl + ϕ) (∀i ∈ same example a factor of ϵ αxy (z ′ ) B̄y (z ′ , t1 . . . tN ) is subtracted
{x, y}). For simplicity, we assume ϕ to be zero. To then determine the from Ex (z ′ , t1 . . . tN ). This residual EMF is then used as an actual
dynamo coefficients (x) we pseudo-invert equation 17. This task is EMF component, and this process is repeated a suitable number
complicated firstly by the fact that both components of mean-field and of times until either all the dynamo coefficients converge to their
current have additive correlated noise and secondly by the fact that the respective constant values or all four chi-squared errors get smaller
y component of the mean-field is typically much stronger compared than a certain predefined threshold. All the aforementioned steps are
to the x component, due to the rotational shear (and by consequence then repeated at every z = z ′ .
the x component of current is much stronger than its y component). We apply this method with ϵ = 0.1 for five hundred refinement
Typical schemes of the least square minimisation in such cases tends loops to the time series of EMFs, mean-fields and currents obtained
to underestimate the dynamo coefficients that are associated with from the DNS data. While constructing these time series (from t =
x and y components of mean-field and current respectively (i.e. 100Ω−1 to 300Ω−1 ) with data dumping interval ∆tdump = 0.2 Ω−1
0 1
αix , αix and ηiy ). To circumvent these issues, we rely upon the we make sure that they correspond to the quasi-stationary phase of
IROS method (Iterative removal of sources) (Hammersley et al. 1992) the magnetic field evolution.
that we have recently adapted for such inversions in the dynamo IROS method does not provide an estimate of errors on the calcu-
context (Bendre et al. 2023). This method is based on Högbom lated coefficients directly. We, therefore, calculate a statistical error of
clean algorithm used in Radio Astronomy to construct an image the dynamo coefficient by considering the five different realizations
by convolving multiple beams, iteratively locating and subtracting of time series. We construct five different time series of mean fields,
out the strongest source to model the rest of the dirty image. It is currents and EMFs by skipping four data points in the time series.
particularly useful when the relative contribution of some of the Specifically, the time series (t1 , t2 , . . . tN ) (of all components of
beams to the final image happens to be negligible. Such a situation mean-field, current and EMF) are split into (t1 , t6 . . .), (t2 , t7 . . .),
is analogous to have only a few of the columns of A (the beams) (t3 , t8 . . .), (t4 , t9 . . .) and (t5 , t10 . . .). We use these time series to
largely contribute to y (an image). A brief outline of the method is calculate five sets of dynamo coefficients and calculate their standard
as follows. deviations to represent the errors on the calculated coefficients.
Firstly, at any particular z = z ′ we set all the dynamo coefficients,
0
αij (z ′ ), αij
1
(z ′ ) and ηij (z ′ ) to zero, i.e we set x(z ′ ) = 0. Then
to derive their zeroth order estimates we fit every ith column of
y(z ′ , t) (say yi (z ′ , t)), against the individual columns of A(z ′ )
(denoted as Ak (z ′ )) separately as lines. The best amongst these four
3 RESULTS: SATURATION OF MRI, MEAN FIELDS AND
linear fits is decided basedPon the least of chi-square errors of the
EMF-S
individual fits (χ2ik (z ′ ) = i (yi − Ak xik )2 ). Then the best fitted
dynamo coefficient is updated by adding to it, its zeroth order estimate We now turn to the results of our shearing box simulation of MRI in
multiplied by a small factor (ϵ < 1), called the loop-gain, while other a geometrically thin disc, investigate its dynamo action in addition to
coefficients are kept constant. For example if the chi-squared error discussing several important properties which illuminates the nature
associated with the line fit Ex (z ′ , t1 . . . tN ) versus B̄y (z ′ , t1 . . . tN ) of the MRI dynamo. Most of our analysis of magnetic field generation
(i.e. χ212 (z ′ )) is the least then x2,1 (z ′ ) (i.e. αxy
0
) is updated by a will focus on the saturated state of MRI, when the disc is in the quasi-
factor of the slope multiplied by ϵ. Subsequently, the contribution to stationary phase.
4 4
B̄x Ēx
0.01 0.01
2 2
0.00 0.00
0 0
z
z
−2 -0.01 −2 -0.01
−4 −4
0 10 20 30 40 0 10 20 30 40
4 4
B̄y /10 10 × Ēy
0.01 0.01
2 2
0.00 0.00
0 0
z
z
−2 -0.01 −2 -0.01
−4 −4
0 10 20 30 40 0 10 20 30 40
t (2π/Ω) t (2π/Ω)
Figure 2. Spatio-temporal variation of mean magnetic fields, B̄x (top left panel), B̄y (bottom left panel) and mean EMFs Ēx (top right panel) and Ēy (bottom
right panel). Mean magnetic field component B̄y and y-component of EMF Ēx show a coherent change in space and time (with a time period ≈ 9 orbital period
(2π/Ω)), while the spatio-temporal patterns in B̄x and Ēy are less coherent.
4 mean EMFs Ēx , Ēy (middle panels) and helicities Khel , Chel (bottom
αkin panels). It is noticeable that power spectra for mean fields and spectra
0.02 peaks at the primary frequency fdyn = 0.017, which was also visible
2
in the butterfly diagrams. In addition to the primary frequency, the
power spectra also show the presence of higher harmonics (at 3fdyn ),
0.00
0 which got unnoticed in the earlier works of MRI dynamo. Similarly,
z
0.00
0
z
−2 -0.10
101
B̄x 103
0−H B̄y 0−H
100 H − 2H H − 2H
2H − 3.5H 102 2H − 3.5H
10−1 101
P
10−2 100
10−3 10−1
10−2
10−4 fdyn 3fdyn fdyn 3fdyn
10−3
10−2 10−1 100 10−2 10−1 100
−1
10
Ēx 0−H Ēy 0−H
100 H − 2H −2 H − 2H
2H − 3.5H
10 2H − 3.5H
10−1
10−3
10−2
P
10−4
10−3
10−4 10−5
fdyn 3fdyn fdyn 3fdyn
10−5 10−6
10−2 10−1 100 10−2 10−1 100
f (Ω) f (Ω)
4
10
0−H 0−H
104
103 H − 2H H − 2H
2H − 3.5H 103 2H − 3.5H
102
102
P
1
10
101
0
10 100
10−1 10−1
Khel 2fdyn 4fdyn Chel 2fdyn 4fdyn
10−2 10−2
10−2 10−1 100 10−2 10−1 100
f (Ω) f (Ω)
Figure 4. Power spectra of mean fields B̄x , B̄y (top panels), mean EMFs Ēx , Ēy (middle panels) and helicities Khel , Chel (bottom panels). Spatial averages
are done over different heights: z = 0 − H (black lines), z = H − 2H (green lines), and z = 2H − 3.5H (red lines). The zeroth frequency values are denoted
by ‘asterisks’. Vertical dashed lines denote the dynamo frequency fdyn = 0.017 and its multiples.
4.1 Time independent dynamo coefficients local (Brandenburg 2008; Gressel 2010; Gressel & Pessah 2015) and
global (Dhang et al. 2020) frameworks also found a similar trend in
Fig. 5 shows the vertical profiles of time-independent dynamo coef- 0
αyy . However, it is to be noted that our study suggests a stronger
0
ficients αij and ηij for different values of fc . Four panels at the top negative αyy0
in the upper-half plane compared to that in the earlier
0 0
illustrate the vertical profiles of coefficients (αxx , αxy , ηxx , ηxy ) studies. The negative sign in the upper half plane is attributed to
associated with the x-component of EMF Ēx , while four panels at the buoyant rise of magnetic flux tubes under the combined action
0 0
the bottom show profiles of those (αyx , αyy , ηyx , ηyy ) associated of magnetic buoyancy and shear (Brandenburg & Schmitt 1998;
with the y-component of EMF Ēy . Brandenburg & Subramanian 2005; see also Tharakkal et al. (2023)).
0
The ‘coefficient of most interest’ out of the calculated ones is αyy , Brandenburg & Schmitt (1998) also suggested that negative αyy is
which plays a vital role in producing the poloidal field (here B̄x ) responsible for the upward propagation direction of dynamo waves
out of the toroidal field (B̄y ) (also see section 4.4). The coefficient seen in the butterfly diagrams of MRI-driven dynamo simulations
0
αyy shows an anti-symmetric behaviour about the z = 0 plane, (e.g. see Fig. 2). Another different way of looking at the origin of the
with a negative (positive) sign in the upper (lower)-half plane (for effective α is to link it to the helicity flux as envisaged by Vishniac
0
|z| < 2). For |z| > 2, the sign of αyy tends to be positive (negative)
in the upper (lower)-half plane. Earlier studies of MRI dynamo in
/10−2
10 10
0 0
αxx
αxy
−10 −10
0
0
−20 −20
−3 −2 −1 0 1 2 3 2 −3 −2 −1 0 1 2 3
3
ηxx /10−2
ηxy /10−2
1
2 0
1 −1
−2
0
6 −3 −2 −1 0 1 2 3 −3 −2 −1 0 1 2 3
1.0
/10−2
/10−2
3 0.5
0 0.0
αyx
αyy
−0.5
0
0
−3
−1.0
−6
0.4 −3 −2 −1 0 1 2 3 1.0 −3 −2 −1 0 1 2 3
z z
ηyx /10−2
ηyy /10−2
0.3 0.5
0.2 0.0
0.1 −0.5
0.0 −1.0
−3 −2 −1 0 1 2 3 −3 −2 −1 0 1 2 3
z z
Figure 5. Vertical profiles of time-independent turbulent dynamo coefficients (α0ij , ηij
0 ) in MRI simulation calculated using IROS method. A low-pass Fourier
filter with a cut-off frequency fc removes the contribution from the small-scale dynamo. We used two values of fc : fc = 0.05 and fc = 0.12. The results are
compared to the case when IROS is applied to the unfiltered data obtained from DNS.
2015 and Gopalakrishnan & Subramanian 2023. We discuss this noted that MRI turbulence in a stratified medium is neither isotropic
possibility in section 5.3. nor homogenous. Minimal τ -approximation (MTA) suggests that in
The off-diagonal terms of the α-coefficients are related to turbu- a stratification and rotation-induced anisotropic turbulent medium,
lent pumping. This effect is responsible for transporting large-scale which includes the quasi-linear back reaction due to Lorentz forces,
magnetic fields from the turbulent region to the laminar region. We
0 0 0 0 1 1
found αxy and αyx to be antisymmetric and αxy > αyx unlike γzMTA = − τ ∇z (v ′2 − B ′2 ) − τ 2 Ωẑ × ∇z (v′2 + B′2 ), (24)
the previous studies (Brandenburg 2008; Gressel & Pessah 2015) 6 6
0 0
which found αyx ≈ αxy . This resulted in a strong turbulent pump- where τ is the correlation time and it is assumed that ρ = 1 (see
0 0
ing γz = (αyx − αxy )/2, transporting large-scale magnetic fields equation (10.59) in Brandenburg & Subramanian 2005). The last
from the disc to the corona as shown in the top panel of Fig. 6. term in equation 24 vanishes because all the variables are functions
We also compare the relative importance of turbulent pumping (γz ) of z alone. Therefore, equation 24 and the bottom panel of Fig. 6
′2
and wind (v̄z ) in advecting the magnetic field upward (in the upper illustrating the vertical profiles of v ′2 and vA imply that sign of
half-plane) at different heights. Vertical profiles of γz and v̄z in the turbulent pumping obtained from MTA supports that obtained from
top panel of Fig. 6 shows that at low heights (|z| < 2.5), turbulent extracted dynamo coefficients.
pumping is the dominant effect over the wind, while the effects of We found turbulent diffusion tensor ηij to be anisotropic with
wind become comparable or larger over the pumping term at large ηxx > ηyy and having a significant contribution from the off-
scale-heights (see also Fig. 11). diagonal components ηxy and ηyx . Different values of diagonal
The theory of isotropic kinematically forced turbulence predicts components of ηij imply that mean field components B̄x and B̄y
that γz is supposed to be in the direction of negative gradient of are affected differently by the vertical diffusion (also see section
turbulent intensity (v ′2 ) (Krause & Raedler 1980), that is, in the 4.4). It is worth mentioning that ηyy ≈ 0 for the fc = 0.05 case,
negative z-direction (in the upper-half plane) in our simulation. This while it is slightly negative for the other two cases. This is somewhat
is opposite to what has been found in Fig. 6. However, it is to be different from the earlier studies (Gressel 2010; Gressel & Pessah
0.0
20 5
/10−2
/10−2
10
0 0
αxx
αxy
1
1
−10
−20 −5
6 −3 −2 −1 0 1 2 3 0.5 −3 −2 −1 0 1 2 3
3
/10−2
/10−2
0 0.0
αyx
αyy
1
1
−3
−6 −0.5
−3 −2 −1 0 1 2 3 −3 −2 −1 0 1 2 3
z z
Figure 7. Vertical profiles of time-independent turbulent dynamo coefficients (α1ij ) in MRI simulation calculated using IROS method. A low-pass Fourier filter
with a cut-off frequency fc is used to remove the contribution from the small-scale dynamo. We used two values of fc : fc = 0.05 and fc = 0.12. The results
are compared to the case when IROS is applied to the unfiltered data obtained from DNS.
Ēx 10 × Ēy
-0.01 0.00 0.01 -0.01 0.00 0.01
3 Ēx,f 3 Ēy,f
0 0
z
−3 −3
20 30 40 20 30 40
3 Ēx,r 3 Ēy,r
0 0
z
−3 −3
20 30 40 20 30 40
t (2π/Ω) t (2π/Ω)
Figure 8. Left panels: Comparison between x-component of EMF Ēx,f used to determine the turbulent dynamo coefficients,and EMF Ēx,r reconstructed using
the turbulent dynamo coefficients. Right panels: Same as figures in left panels, but for the y-component of EMF.
Figure 9. Histograms of the residual EMFs, δ Ēi = Ēi,f − Ēi,r calculated within region of different heights for fc = 0.05 case. We normalise δ Ēi with the
absolute values of the corresponding EMFs at the respective points. The Red dashed line shows a normal distribution fitting the histogram.
where
−1 1 ∂
-0.01 TBx ,vz = − B̄x v̄z B̄x ,
−2 2 ∂z
1 ∂
−3 Tαyx = − B̄x αyx B̄x ,
2 ∂z
20 30 40 50 60
1 ∂
t (2π/Ω) Tαyy = − B̄x αyy B̄y ,
2 ∂z
1 ∂ ∂
Figure 10. Butterfly diagrams of the mean magnetic fields B̄x and B̄y Tηyx = − B̄x ηyx B̄y ,
obtained by running 1D dynamo model. Both B̄x and B̄y flip sign regularly 2 ∂z ∂z
with a cycle of ≈ 9 orbital period, similar to that found in shearing box 1 ∂ ∂
Tηyy = B̄x ηyy B̄x ,
simulations (see Fig. 2). 2 ∂z ∂z
1 ∂
TBy ,vz = − B̄y v̄z B̄y , (29)
2 ∂z
1 ∂
Tαxy = B̄y αxy B̄y ,
2 ∂z
1 ∂
Tαxx = B̄y αxy B̄y ,
2 ∂z
1 ∂ ∂
′
diffusion as follows η0 = vrms ∆x, where we consider the smallest Tηxy = − B̄y ηxy B̄x ,
one among the relevant velocities (vrms ′
, cs , vA ) in the problem. 2 ∂z ∂z
Therefore, we add a correction term η0 ≈ 10−3 (with ∆x = 1/32 1 ∂ ∂
′
Tηxx = B̄y ηxx B̄y ,
and vrms = 0.1) to the diagonal components of diffusivity tensor 2 ∂z ∂z
ηij to consider the contribution from the mesh to the magnetic field 1
TS = qΩ B̄x B̄y .
diffusion. This also helps us to stabilize the 1D dynamo solution. 2
With this setup, we solve the system of equations with a finite dif- Fig. 11 shows the space-time plots of different terms involving
ference method over a staggered grid of resolution ∆z = 1/32, same mean flow (v̄z ) and turbulent dynamo coefficients (αij , ηij ) in the
as the z resolution of DNS. Outcome of this analysis are presented in mean magnetic energy evolution equations. The top six panels in
Fig. 10, where top and bottom panels show the butterfly diagrams of Fig. 11 describe the terms in the x-component of the magnetic energy
B̄x and B̄y obtained using the 1D dynamo model respectively. We equation (equation 27), while the bottom seven panels illustrate terms
find both x and y-components of mean fields flip sign regularly with a in the y-component of the magnetic energy equation ( 28) at different
cycle of ≈ 9 orbital period, similar to what is found in DNS (see Fig. heights and times.
2). Thus, applying calculated coefficients to the 1D dynamo model Fig. 11 provides a fairly complicated picture to account for the
successfully reproduces broad features of spatiotemporal variations generation-diffusion scenario of the mean magnetic fields. Broadly
mean magnetic fields. speaking, the poloidal field (B̄x ) is predominantly generated by an
1 2
2 B̄x TBx ,vz Tαyx
2 2 2
50.0 2.0 2.0
0 0 0
z
0 0 0
z
−2 −2
-50.0 -4.0 −2 -4.0
20 30 40 20 30 40 20 30 40
Tαxx t (2π/Ω) Tηxy t (2π/Ω) Tηxx
2 4.0 2 4.0 2 4.0
20 30 40 20 30 40 20 30 40
TS
t (2π/Ω) t (2π/Ω) t (2π/Ω)
2 4.0
0 0.0
z
−2 -4.0
20 30 40
t (2π/Ω)
Figure 11. Contributions of different terms involving mean flow (v̄z ) and turbulent dynamo coefficients (αij , ηij ) to the x-(top six panels) and y-(bottom seven
panels) components of mean magnetic energy evolution equation (equations 27 and 28). Poloidal field (B̄x ) generation is primarily attributed to an α-effect (the
term Tαyy ), while shear (the term TS ) dominates the toroidal field generation; thus implying and α − Ω type of dynamo. Winds carry mean fields out of the
computational box and contribute largely as the sink term in the mean magnetic energy evolution equation.
α-effect (the term Tαyy in Fig. 11). However, there is a significant of the dynamo in an MRI-driven geometrically thick accretion disc
contribution from αyx (the term Tαyx in Fig. 11) in generating B̄x (Dhang et al. 2020), implying universal action of α − Ω-dynamo in
in larger scale-heights. Toroidal field generation is mainly due to the MRI-driven accretion flows.
presence of shear, here differential rotation, (TS in Fig. 11) which
converts poloidal fields to the toroidal fields. However, it is worth
noting that there is a minute contribution from the αxx , generating a
toroidal field out of the poloidal field by an α-effect, (as in an α2 -Ω
dynamo). The dominance of α-effect in generating a poloidal field Generally, it is expected that diagonal components of the diffusion
and that of Ω-effect (shear) in generating a toroidal field imply the tensor, ηyy and ηxx are primarily responsible for the diffusion of B̄x
presence of an α − Ω type dynamo in MRI-driven geometrically thin and B̄y respectively. However, our simulation finds that winds carry
accretion disc. This is similar to what has been found in the study mean fields out of the computational box and act as a sink in the
mean magnetic energy evolution equation, not the η-s.
2 (34)
−0.02 η̃xx + κ̃xyz = ηxx ,
η̃yy − κ̃yxz = ηyy ,
−0.04
1 1
δ̃z = (ηxy − ηyx ) + (κ̃xxz + κ̃yyz ) .
−0.06 2 2
−3 −2 −1 0 1 2 3
z Therefore it is evident from equation 34 that it is impossible to
decouple a few coefficients (coefficients in the last three identities)
Figure 12. Vertical profiles of α0yy (for fc = 0.05 case) obtained from IROS as there are more unknown coefficients than independent variables
inversion and (α0yy )hc , expected from helicity flux. (B̄, Ē) and the actual diffusion coefficients (η̃ij ) might be different
from the calculated ones (ηij ).
/10−2
0.5
0 0.0
αyx
αyy
0
0
−3 −0.5
−1.0
−6
−3 −2 −1 0 1 2 3 1.5 −3 −2 −1 0 1 2 3
0.2
ηyx /10−2
ηyy /10−2
0.1 1.0
0.0
−0.1 0.5
−0.2
0.0
−3 −2 −1 0 1 2 3 −3 −2 −1 0 1 2 3
z z
Figure A1. Vertical profiles of the time-independent αij and ηij related to the Ēy calculated imposing the constraint ηyy = ηxx for fc = 0.05 case. As
expected, αij are not affected with the change in fη , but ηij s are. A clear trend has been found; more positive the ηyy is more negative is ηyx .
Figure A2. Histograms of the residual EMFs δ Ēx (top panels) and δ Ēy (bottom panels) for feta = 0 and fη = 1. We consider the fc = 0.05 case. Imposition
of constraints on ηyy compromises the quality of fits, but not significantly because αij are the main contributors in the fitting of EMFs, not the ηij .
Shi J.-M., Stone J. M., Huang C. X., 2016, MNRAS, 456, 2273 e-prints, p. arXiv:2305.03318
Shukurov A., Subramanian K., 2021, Astrophysical Magnetic Fields: From Velikhov E., 1959, Sov. Phys. JETP, 36, 995
Galaxies to the Early Universe. Cambridge Astrophysics, Cambridge Vishniac E. T., 2015, in American Astronomical Society Meeting Abstracts
University Press, doi:10.1017/9781139046657 #225. p. 229.08
Simard C., Charbonneau P., Dubé C., 2016, Advances in Space Research, 58, Wissing R., Shen S., Wadsley J., Quinn T., 2022, A&A, 659, A91
1522 Zier O., Springel V., 2022, MNRAS, 517, 2639
Simon J. B., Hawley J. F., Beckwith K., 2009, ApJ, 690, 974 von Rekowski B., Brandenburg A., Dobler W., Dobler W., Shukurov A., 2003,
Simon J. B., Beckwith K., Armitage P. J., 2012, MNRAS, 422, 2685 A&A, 398, 825
Squire J., Bhattacharjee A., 2016, Journal of Plasma Physics, 82, 535820201
Stepanovs D., Fendt C., Sheikhnezami S., 2014, ApJ, 796, 29
Stone J. M., Pringle J. E., Begelman M. C., 1999, MNRAS, 310, 1002
Subramanian K., 2002, Bulletin of the Astronomical Society of India, 30, 715
Subramanian K., Brandenburg A., 2006, ApJL, 648, L71
Tharakkal D., Shukurov A., Gent F. A., Sarson G. R., Snodin A., 2023, arXiv