Published in Physics of Plasmas 23, 032111 (2016)

Visco-Resistive Plasmoid Instability

Luca Comisso1, 2, ∗ and Daniela Grasso2
1
Department of Astrophysical Sciences and Princeton Plasma Physics Laboratory,
Princeton University, Princeton, New Jersey 08544, USA
2
Istituto dei Sistemi Complessi - CNR and Dipartimento Energia,
Politecnico di Torino, Torino 10129, Italy
arXiv:1603.00090v2 [physics.plasm-ph] 6 Apr 2016

The plasmoid instability in visco-resistive current sheets is analyzed in both the linear and non-
linear regimes. The linear growth rate and the wavenumber are found to scale as S 1/4 (1 + Pm )−5/8
and S 3/8 (1 + Pm )−3/16 with respect to the Lundquist number S and the magnetic Prandtl num-
ber Pm . Furthermore, the linear layer width is shown to scale as S −1/8 (1 + Pm )1/16 . The growth
of the plasmoids slows down from an exponential growth to an algebraic growth when they enter
into the nonlinear regime. In particular, the time-scale of the nonlinear growth of the plasmoids
is found to be τNL ∼ S −3/16 (1 + Pm )19/32 τA,L . The nonlinear growth of the plasmoids is radically
different from the linear one and it is shown to be essential to understand the global current sheet
disruption. It is also discussed how the plasmoid instability enables fast magnetic reconnection in
visco-resistive plasmas. In particular, it is shown that the recursive plasmoid formation can trigger
a collisionless reconnection regime if S & Lcs (ǫc lk )−1 (1 + Pm )1/2 , where Lcs is the half-length of the
global current sheet and lk is the relevant kinetic length scale. On the other hand, if the current
sheet

remains
 in the collisional regime, the global (time-averaged) reconnection rate is shown to be
dψ/dt|X ≈ ǫc vA,u Bu (1 + Pm )−1/2 , where ǫc is the critical inverse aspect ratio of the current sheet,
while vA,u and Bu are the Alfvén speed and the magnetic field upstream of the global reconnection
layer.

PACS numbers: 52.35.Vd, 52.30.Cv, 52.35.Py

Keywords: magnetic reconnection, magnetohydrodynamics

I. INTRODUCTION paper, when Bhattacharjee and coworkers [7] showed that

the predictions of the Sweet-Parker model [8] break down
for high-Lundquist-numbers as a consequence of the plas-
During the last few years there have been great ad- moid instability. The Sweet-Parker model gives a recon-
vances in understanding the nature and role of the plas- nection rate ∼ S −1/2 vA,u Bu , where Bu is the upstream
moid instability [1], i.e., the instability of reconnecting magnetic field, but Bhattacharjee et al. [7] showed that
current sheets leading to the formation of secondary mag- this cannot be valid for arbitrarily large S-values. In
netic islands (plasmoids). Numerical simulations provid- particular, they found from numerical simulations that
ing clear indications that thin reconnecting current sheets the reconnection rate becomes nearly independent on
may be unstable to the formation of secondary islands the Lundquist number when the current sheet exceeds
date back at least to the mid-eighties [2–5], but it has a critical Lundquist number Sc . Note that the recon-
been only after a seminal paper by Loureiro and cowork- nection rate calculated by Bhattacharjee and coworkers
ers [6] that the scientific community has begun to show was averaged over time, since in the high-Lundquist-
a distinct interest in the plasmoid instability and its ef- number regime the reconnection process is strongly time-
fects on magnetic reconnection. In their paper, Loureiro dependent due to the continuous formation and ejec-
et al. [6] presented a linear theory of the plasmoid in- tion of plasmoids [7, 9, 10]. An estimation of the time-
stability in Sweet-Parker current sheets, showing that averaged reconnection rate in this regime was proposed
the growth rate and the wavenumber of the instability by Huang and Bhattacharjee [11] as well as by Uzden-
scale as γmax Lcs /vA,u ∼ S 1/4 and kmax Lcs ∼ S 3/8 , re- sky and coworkers [12]. They showed that, as a result of
spectively. Here, S is the Lundquist number based on the plasmoid instability, the time-averaged reconnection
the half-length of the current sheet (Lcs ) and the Alfvén rate in statistical steady-state is ∼ 0.01vA,u Bu , indepen-
speed upstream of the sheet (vA,u ). This implies that dent of the Lundquist number and much higher than the
the linear growth of the plasmoids is surprisingly fast for Sweet-Parker rate for very large S-values. These crucial
large values of the Lundquist number. implications of the plasmoid instability caused a rethink-
The effects of the plasmoid instability on the magnetic ing of the traditional magnetic reconnection theory, with
reconnection rate became clear shortly after Loureiro’s a particular impetus for the investigation of the linear
properties of this instability [13–21] and its effects on the
reconnection rate [22–32] and particle acceleration [33–
37].
∗ Electronic address: [email protected]
 2

In the present paper we have three main objectives. are the electric current density and plasma vorticity, both
First, we intend to generalize the linear analysis pre- in the z direction. Furthermore, E0 is a constant of in-
sented by Loureiro and coworkers [6] to non-negligible tegration representing the equilibrium electric field. The
values of the plasma viscosity. Then, we want to extend mass density is assumed to be uniform and is normalized
the analysis of the plasmoid instability also to the nonlin- such that ρ = 1. Lengths are normalized to a convenient
ear regime. Finally, we aim to estimate the effects of this scale length L, magnetic field to a convenient scale field
instability on the global reconnection rate. We point out strength B0 , and time to the Alfvén time τA = L/vA ,
that for all these issues we will consider arbitrary values where vA is the Alfvén speed based on B0 . In this normal-
of the magnetic Prandtl number. ization, the resistivity η and the viscosity ν are recipro-
The linear analysis will show that the plasmoid in- cals of the Lundquist number and the (kinetic) Reynolds
stability remains fast also in the general case of non- number, respectively. Therefore, the ratio Pm = ν/η in-
negligible plasma viscosity. This means that its rapid- dicates the magnetic Prandtl number.
ity is a robust feature in the linear regime. However, we We consider the equilibrium configuration employed by
must recall that the linear layer width has to be much Loureiro et al. [6] in order to investigate the stability of a
smaller than the current sheet width for the linear anal- Sweet-Parker type current sheet. As pointed out in their
ysis to apply. Thus, one has to consider also the non- landmark paper, this configuration is not intended to re-
linear regime to evaluate the overall growth rate of the tain all the features of such a sheet, but only those which
plasmoid instability. We will see that the growth of the may be regarded as the most important for its stability;
plasmoids slows down when their width becomes compa- namely, a linearly increasing velocity in the ouflow direc-
rable or larger than the linear layer width. In particular, tion and a sheared magnetic field in the inflow direction.
we will show how the time-scale of the plasmoid nonlin- Assuming an equilibrium flow profile v0 = (−Γ0 x, Γ0 y)
ear growth depends on the Lundquist number and the for |x| 6 x0 (inside the current sheet), v0 = (−Γ0 x0 , 0)
magnatic Prandtl number. It will also be shown that for x > x0 , and v0 = (Γ0 x0 , 0) for x 6 −x0 , the equilib-
the recursive plasmoid formation (due to the instability rium stream function turns out to be
of the secondary current sheets) may result in a speed 
up of the global current sheet disruption. The break up +Γ0 xy , |x| 6 x0 ,

of the global current sheet has the effect of accelerating φ0 (x, y) = +Γ0 x0 y , x > x0 , (5)
the magnetic reconnection process with respect to the

−Γ x y , x 6 −x .
0 0 0
case of a current sheet that would remain stable for ar-
bitrary values of the Lundquist number. We will show Here, Γ0 ≡ vd /Lcs is a constant velocity gradient, with
that for high-Lundquist-numbers the (time-averaged) re- vd = v0y (0, Lcs ) and Lcs indicating the downstream ve-
connection rate becomes independent of the Lundquist locity and the half-length of the current sheet, respec-
number but not on the magnetic Prandtl number. In tively. In the presence of non-negligible viscosity effects,
particular, the reconnection rate decreases for increasing the force balance along the current sheet gives that [38]
values of the magnetic Prandtl number. Finally, we will
see that plasma viscosity has also the effect of increas- vd = vA,u (1 + Pm )
−1/2
, (6)
ing the value of the global Lundquist number required to
trigger collisionless reconnection. where vA,u is the Alfvén speed upstream of the sheet.
−1/2
The downstream velocity decreases as vd ≈ vA,u Pm
for Pm ≫ 1, while the standard Sweet-Parker result [8]
II. MODEL EQUATIONS AND EQUILIBRIUM vd ≈ vA,u is recovered in the limit Pm ≪ 1. Also, mass
conservation for an incompressible flow implies that the
We consider the two-dimensional (∂z = 0) incompress- upstream velocity is related to the downstream veloc-
ible MHD equations, which can be conveniently written ity and the aspect ratio of the current sheet as vu =
in terms of the magnetic flux function ψ(x, y, t) and the v0x (δcs , 0) = −vd δcs /Lcs , where [38]
stream function φ(x, y, t) as [8]
1/2
(1 + Pm )1/4

Lcs
∂t ψ + v · ∇ψ = −ηj + E0 , (1) δcs = η = Lcs (7)
vd S 1/2

is the characteristic half-width of the current sheet. Here,

∂t ω + v · ∇ω = B · ∇j + ν∇2 ω , (2)
S ≡ Lcs vA,u /η is the Lundquist number based on the cur-
rent sheet length and the upstream Alfvén speed. Again,
where
the standard Sweet-Parker width [8] δcs ≈ Lcs S −1/2 is
B = ∇ψ × ez , v = ez × ∇φ (3) recovered in the limit Pm ≪ 1, while the opposite limit
1/4
Pm ≫ 1 yields δcs ≈ Lcs S −1/2 Pm .
are the magnetic and velocity fields, while The equilibrium magnetic field, which is assumed to
be in the form B0 = (0, B0y (x)), can be determined by
j = −∇2 ψ , ω = ∇2 φ (4) solving Eq. (1) with ∂t = 0 and the equilibrium flow
 3

profiles already specified. In this case, Eq. (1) reduces φ(x, y, t) = φ0 (x, y) + φ1 (x, t) eik(t)y , (14)
to
where ψ1 and φ1 are small perturbations to the equilib-
2 dB0y E0 rium. As in Ref. [6], the wavenumber along the y direc-
δcs + xB0y = (8)
dx Γ0 tion is supposed to depend only on t, which is a valid as-
sumption if k varies sufficiently slowly across the current
for |x| 6 x0 . Introducing the new variable sheet [14]. Therefore, inside the current sheet (|x| 6 x0 )
x the perturbations ψ1 and φ1 satisfy the equations
ξ≡ (9)
δcs
∂t ψ1 + iy (∂t k + Γ0 k) ψ1 + ikB0y φ1 − Γ0 x∂x ψ1
and using B0y (0) = 0 as boundary condition, the solution = η ∂x2 − k 2 ψ1


(15)
of Eq. (8) is
r   and
−ξ 2 /2 π ξ
B0y ( ξ) = α e erfi √ , (10)
∂t ∂x2 − k 2 φ1 + iy (∂t k + Γ0 k) ∂x2 − k 2 φ1



2 2
where d2 B0y
−2k (∂t k) φ1 − Γ0 x∂x ∂x2 − k 2 φ1 = ik


ψ1
dx
2
x
2
Z
−ikB0y ∂x2 − k 2 ψ1 + ν ∂x4 − 2k 2 ∂x2 + k 4 φ1 .(16)
2


erfi (x) ≡ −i erf (ix) = √ ez dz (11)
π 0
These equations depend on both x and y. However, the
is the imaginary error function and α = E0 /Γ0 δcs . y-dependence can be removed by choosing k(t) to satisfy
Therefore, since for |x| > x0 (outside the current sheet) ∂t k + Γ0 k = 0 inside the current sheet. This gives that
the magnetic field is constant by virtue of Eq. (1) k(t) = k0 e−Γ0 t [6], and the above equations reduce to
and the specified equilibrium flow profiles, we can write
Loureiro’s 1-D equilibrium [6] as ∂t ψ1 + ikB0y φ1 − Γ0 x∂x ψ1 = η ∂x2 − k 2 ψ1 ,


(17)
 Z ξ
2 2
α e−ξ /2

 ez /2
dz , |ξ| 6 ξ0 ,
∂t ∂x2 − k 2 φ1 + 2Γ0 k 2 φ1 − Γ0 x∂x ∂x2 − k 2 φ1



0
B0y (ξ) = (12)
 +1 , ξ > ξ0 ,  2
d B0y

2 2



−1 , ξ 6 −ξ0 . = ik − B0y ∂x − k ψ1
dx2
+ ν ∂x4 − 2k 2 ∂x2 + k 4 φ1 .


Here, ξ0 ≡ x0 /δcs and the normalizing magnetic field (18)
strength is chosen to be the equilibrium magnetic field
at ξ = ξ0 . The value of the constant α can be spec- Actually, in our analysis we can ignore also the time-
ified by matching the magnetic field inside the current dependence of k since we look for fast growing modes
sheet (|ξ| 6 ξ0 ) with the magnetic field outside the cur- [6, 14]. Alternatively, we could have started directly with
rent sheet (|ξ| > ξ0 ). Logical matching points ±ξ0 are a time-independent k, but we have adopted this approach
where the solution of Eq. (8) has its maximum and its to illustrate the domain of validity of the const-k approx-
minimum, i.e., where the current density vanishes. These imation. In particular, we search for solutions of the form
points correspond to ±ξ0 = ±1.307. Therefore, requiring
B0y (±ξ0 ) = ±1 yields α = 1.307. ψ1 (x, t) = Ψ(x)eγt , φ1 (x, t) = −iΦ(x)eγt , (19)

with growth rate

III. LINEAR THEORY vd
γ≫ ≡ Γ0 . (20)
Lcs
We now show that the presented equilibrium is prone
to the visco-resistive plasmoid instability. To this pur- Indeed, in order for a tearing mode to grow it is necessary
pose, differently from Ref. [6], we perform a boundary that its growth rate exceeds the shearing rate [8, 39]. In
layer theory that takes into account both plasma resis- this limit, k(t) ≈ k0 and the terms proportional to Γ0
tivity and viscosity. can be neglected. Therefore, the linear shear flow con-
tributes only via the equilibrium profile B0y (x) [6]. Un-
der these approximations, and with the change of vari-
A. Linearized equations able ξ ≡ x/δcs , the equations that govern the perturbed
quantities Ψ(ξ) and Φ(ξ) are
We start by linearizing Eqs. (1) and (2) according to
d2
 
1 2 2
λΨ + B0y Φ = −κ ǫ Ψ, (21)
ψ(x, y, t) = ψ0 (x) + ψ1 (x, t) eik(t)y , (13) κ(1 + Pm )1/2 dξ 2
 4

d2
 2
d2 B0y
  
2 2 d 2 2 first derivative at ξ = ±ξ0 yields C1± = ±αΨ(0)/κǫ and
λ − κ ǫ Φ = B 0y − κ ǫ Ψ− Ψ
dξ 2 dξ 2 dξ 2 C3± = (αΨ(0)/κǫ) eκǫξ0 . Therefore, the ideal-MHD mag-
Pm
 4
d 2
 netic flux eigenfunction for Loureiro’s 1-D equilibrium is
2 2 d 4 4
+ − 2κ ǫ + κ ǫ Φ,(22) [6]
κ(1 + Pm )
1/2 dξ 4 dξ 2
Zξ
 !
where we have introduced the parameters  α Ψ(0)B0y (ξ) ±1 − κǫ

−2

B0y (z)dz , |ξ| 6 ξ0 ,

Ψ± (ξ) = κǫ
γ δcs ±ξ0
λ≡ , ǫ≡ , κ ≡ k0 Lcs . (23)  α Ψ(0) eκǫ(ξ0 ∓ξ) ,


k0 vA,u Lcs |ξ| > ξ0 .

κǫ
We recall that ǫ ≪ 1, since we are considering highly (29)
elongated current sheets. Furthermore, we assume This solution has a discontinuous first derivative at the
rational surface ξ = 0. In particular, the jump in its first
λ ≪ 1, κ ≫ 1, κǫ ≪ 1 , (24) derivative gives the tearing stability parameter [6, 40]
2α2
   
but also ′ 1 dΨ  dΨ 
∆ ≡ − ≈ . (30)
Ψ(0) dξ 0+ dξ 0− κǫ
λ ≫ κ−1 (1 + Pm )−1/2 , (25)
In the following, as in the standard tearing mode the-
since we are looking for perturbations such that γ ≫ Γ0 . ory, we will perform a matching between the solutions in
All these assumptions are satisfied by the fastest growing the outer and inner regions by means of the parameter
mode, as can be verified a posteriori. Then, as in tearing ∆′ . Note that while the expression (30) is the same as
mode theory, the problem can be made more tractable by in Loureiro et al. [6], here the inverse aspect ratio of
dividing the spatial domain into two regions: an “outer the current sheet ǫ depends not only on the Lundquist
region” (|ξ| & 1), where non-ideal effects and plasma in- number S but also on the magnetic Prandtl number Pm .
ertia are negligible, and a thin “inner region” (|ξ| ≪ 1)
centered on the resonant surface at x = 0, where resis-
C. Inner region
tivity, viscosity, and plasma inertia can be important.

In the region |ξ| ≪ 1 we can assume κ2 ǫ2 ≪ d2 /dξ 2

B. Outer region and B0y ≈ αξ. Therefore, Eqs. (21) and (22) reduce to
1 d2 Ψ
In the region |ξ| ∼ 1, neglecting plasma inertia and λΨ + αξΦ = , (31)
1/2
κ(1 + Pm ) dξ 2
non-ideal effects, Eq. (21) reduces to
λ d2 Φ d2 Ψ Pm d4 Φ
Φ=− Ψ (26) λ = αξ + . (32)
B0y dξ 2 dξ 2 1/2 dξ 4
κ(1 + Pm )
while Eq. (22) becomes Following a standard procedure [43–46], we introduce the
! Fourier transform of the perturbed quantities
d2 B0y dξ 2

d2 Ψ 2 2
= +κ ǫ Ψ. (27) h i Z +∞
dξ 2 B0y Ψ̂(θ), Φ̂ (θ) = [Ψ (ξ) , Φ (ξ)] e−iθξ dξ , (33)
−∞
Eq. (27) can be solved perturbatively by exploiting the
which should be understood in a generalized sense [41],
smallness of κ2 ǫ2 , as it has been done by Loureiro et al.
since Ψ (ξ) and Φ (ξ) are not square integrable. Then, the
[6]. Its solution can be written as
Fourier transformed Eqs. (31) and (32) can be combined
Z ξ to give the layer equation
Ψ± (ξ) = C1± B0y (ξ) + C2± B0y (ξ) −2
B0y (z)dz , (28) !
±ξ0 d θ2 dΦ̂
dθ λ + κ−1 (1 + Pm )−1/2 θ2 dθ
where ± refers to the solution at ξ ≷ 0. The constants (34)
of integration C2± can be found by requiring the solu- λ Pm
= 2 θ2 Φ̂ + θ 4
Φ̂
tion to be continuous across the rational surface ξ = 0 α α2 κ(1 + Pm )1/2
and adopting the approximation B0y (ξ) ≈ αξ for small
The appropriate boundary condition that ensures the
ξ. This gives that C2+ = C2− = −αΨ(0). The integra-
asymptotic matching to the outer solution must behave
tion constants C1± can instead be found by matching the
as
solution (28) with the solution of Eq. (27) in the region
|ξ| > ξ0 , which is Ψ± (ξ) = C3± e∓κǫξ . Matching this solu-
 
1 π
Φ̂(θ) → Φ0 + sgn (θ) for θ → 0 , (35)
tion and its first derivative with the solution (28) and its θ ∆′
 5

where Φ0 is a constant. Furthermore, the layer equation where Φ̂< (θ) indicates the function that approximate
must be solved subject to the condition Φ̂(θ) for relatively small values of θ. The general solution
of this equation is
Φ̂ (θ) → 0 for θ → ∞ . (36)
 
λ
In the limit Pm ≪ 1, the eigenvalue problem (34)-(36) Φ̂< = C6 − + κ−1 Pm −1/2
θ + C7 , (43)
θ
yields the dispersion relation [6, 42]
 3/2
 where C6 and C7 are integration constants. Imposing
5/4 Γ  Λ − 1 /4 8

 = − (κα)−1/3 ∆′ , the boundary conditions (35) and (36) and matching
Λ 3/2
(37)
Γ Λ + 5 /4 π the functions Φ̂> and Φ̂< in the overlapping interval
1/4
(λκ)1/2 Pm < θ < (ακ)1/3 , we can finally obtain
where Λ ≡ α−2/3 κ1/3 λ and Γ indicates the gamma func-
tion. This is the regime considered by Loureiro et al. [6], 1 α1/3
who found the growth rate and the wavenumber of the λ= 1/2
∆′ , (44)
πc∗ κ2/3 Pm
fastest growing mode by balancing two relevant limits of
Eq. (37), namely where
4/5
Γ (5/6)

Γ (1/4) 4/5
λ= α2/5 κ−3/5 ∆′ for Λ ≪ 1 , c∗ ≡ 62/3 . (45)
2πΓ (3/4) Γ (1/6)
(38)
and We now turn our attention to the large-∆′ regime
(∆′ → ∞). In this case, we can assume the “visco-
2π 1/2 α resistive” ordering, for which all terms in Eq. (34) are
λ = α2/3 κ−1/3 − for Λ → 1− . (39)
3∆′ comparable over the entire width of the inner layer, with
the exception of the inertial term, which is taken to be
In this way, they showed that for the fastest growing negligible [45]. With this ordering, estimating θ ∼ 1/δin
mode γmax Lcs /vA,u ∼ S 1/4 and kmax Lcs ∼ S 3/8 . and d/dθ ∼ δin , we obtain
Here, on the other hand, we are interested also in plas-
mas with non-negligible viscosity. In the case Pm > 1 an
δin ∼ (ακ)−1/3 (46)
exact analytic solution of the layer equation (34) has not
been found, but we can rely on analytic approximations and
that are valid in different relevant asymptotic regimes
[45]. λ ∼ α2/3 κ−1/3 Pm
−1/2
. (47)
We first consider the small-∆′ regime, which is char-
acterized by the condition ∆′ δin ≪ 1, with δin indicat- A more precise multiplicative constant can be found
ing the inner layer width. In this case, Eq. (34) can be through the numerical analysis of the layer equation (34),
solved by asymptotic matching the solutions obtained for which leads asymptotically (for large Pm and ∆′ → ∞)
large θ and relatively small θ. For values of θ such that to
1/4
θ > (λκ)1/2 Pm , Eq. (34) may be approximated as
α2/3
λ = c∗∗ , (48)
d2 Φ̂> θ4 1/2
κ1/3 Pm
2
− 2 2 Φ̂> = 0 , (40)
dθ α κ
where c∗∗ ≈ 1.53 [45].
where Φ̂> (θ) indicates the function that approximate
Φ̂(θ) in the considered interval. The general solution of
this equation is [31] D. Fastest growing mode
   
1 3 3 1 3 3 The most unstable mode, which is the one that domi-
Φ̂> = C4 θ1/2 I1/6 θ δin + C5 θ1/2 K1/6 θ δin ,
3 3 nates the plasmoids growth, occurs at the intersection of
(41) the small-∆′ and large-∆′ regimes. Therefore, by equat-
where I1/6 and K1/6 are the modified Bessel functions ing Eqs. (44) and (48), with the help of Eq. (30), we find
of the first and second kind of order 1/6, C4 and C5 are the wavenumber of the fastest growing mode
integration constants, and δin = (ακ)−1/3 . On the other
hand, if we consider values of θ such that θ < (ακ)1/3 ,  3/4 5/4 3/8
2 α S
Eq. (34) may be approximated as κmax = kmax Lcs ≈ 3/16
. (49)
π Pm
!
d θ2 dΦ̂< From this relation we have also the number of plasmoids
−1/2 2 dθ
= 0, (42)
dθ λ + κ−1 Pm θ that initially appear in the unstable current sheet, which
 6

is N ≈ κmax /2π. The λ-value of the fastest growing mode with increasing magnetic Prandtl number may occur as
can be easily evaluated as a dissipative effect of viscosity, it is also possible for a
stable current sheet to be destabilized by the viscosity in
γmax Lcs 3  π 1/4 α1/4 analogy to what happens for the Orr-Sommerfeld equa-
λmax = ≈ 7/16
, (50)
vA,u κmax 2 2 S 1/8 Pm tion in fluid dynamics. The plasma viscosity may indeed
destabilize the stable branches of the dispersion relation
which means that the growth rate of the plasmoid insta- that follows from Eqs. (21) and (22). For this reason, we
bility is believe that a treatment of the role of viscosity on the
effects of marginal stability is warranted.
Lcs 3 α3/2 S 1/4 It is easy to verify that the obtained relations
γmax ≈√ 5/8
. (51)
vA,u 2π Pm (53)-(55) for the fastest growing mode justify the
assumptions adopted to derive these relations, i.e.,
Finally, the inner layer width of the fastest growing mode κ−1 −1/2
max (1 + Pm ) ≪ λmax ≪ 1, 1/κmax ≪ 1, and
is
κmax ǫ ≪ 1 (recall that ǫ = S −1/2 (1 + Pm )1/4 ≪ 1 since
δin,max  π 1/4 Pm
1/16 we are considering highly elongated current sheets).
≈ . (52) Furthermore, if we suppose that the stabilizing effect of
δcs 2 α3/4 S 1/8
the flow becomes ineffective for current sheets exceeding
Note that plasma viscosity has the effect of decreasing the critical aspect ratio Lc /δc ≡ 1/ǫc , it is straightfor-
the growth rate and the wavenumber of the plasmoid ward to see that
instability, while it increases (slightly) the inner layer 1/2
width. We observe that the same dependence of κmax , Sc = ǫ−2
c (1 + Pm ) , (56)
γmax and δin,max on S and Pm was found in a heuris-
tic way by Loureiro and coworkers [17]. Their scalings i.e., the critical Lundquist number for the stability of the
were obtained (non-rigorously) by applying the classical sheet increases with increasing plasma viscosity due to
tearing mode expressions for large Pm [45] to a Harris the reduction of the outflow velocity [31, 62]. In particu-
current sheet with equilibrium scale length L ≡ δcs ≈ lar, in the limit Pm ≪ 1 we get Sc ≈ ǫ−2c [7, 10], whereas
1/2
1/4
Lcs S −1/2 Pm [38], in analogy to what was done by Bhat- in the limit Pm ≫ 1 it follows that Sc ≈ ǫ−2c Pm , which
tacharjee et al. [7] to reproduce the scaling relations de- agrees with the condition proposed by Loureiro et al.
rived by Loureiro et al. [17] for plasmas with Pm ≪ 1. [17].
Here, instead, we have adopted a more rigorous approach
that confirms the S and Pm dependencies and gives us the
remaining multiplicative factors. Moreover, we observe IV. NONLINEAR THEORY
that since the relations (49)-(52) for high-Pm plasmas
have the same α factors and S dependencies of those re- The previous linear analysis breaks down and non-
sulting from Eqs. (38)-(39) for low-Pm plasmas, general linearities must be considered when the plasmoid chain
relations for arbitrary Pm can be represented as width w grows to a size comparable or exceeding the
linear layer width δin,max. Then, in order to determine
α5/4 S 3/8 the nonlinear evolution of the plasmoids, we have to
kmax Lcs ∼ , (53)
(1 + Pm )
3/16 consider the relevant nonlinear regime that characterizes
their growth.

Lcs α3/2 S 1/4

γmax ∼ 5/8
, (54) A. Nonlinear evolution
vA,u (1 + Pm )

To ascertain the proper nonlinear regime, we observe

1/16
δin,max (1 + Pm ) that for the fastest growing mode
∼ . (55)
δcs α3/4 S 1/8
2α2 2α3/4 S 1/8
It is clear that, despite the damping effect of the ∆′max δcs ≈ ∼ 1/16
, (57)
κmax ǫ (1 + Pm )
plasma viscosity, the plasmoid instability turns out to be
super-Alfvénic in plasmas of interest to this work, i.e., while the plasmoid half-width is w ∼ δin,max at the be-
for very large S-values. Note that the growth rate of ginning of the nonlinear evolution. Therefore, when the
the plasmoid instability is super-Alfvénic with respect fastest growing mode enters into the nonlinear regime
to the inverse Alfvén time τA,L = Lcs /vA,u (along the
current sheet), but it is slower than the inverse Alfvén ∆′max w ∼ 1 (58)
time τA,δ = δcs /vA,u (transverse to the current sheet),
as is required for the tearing mode analysis. It should (ignoring a factor 2), which implies that the plasmoids
also be noted that while the decrease of the growth rate evolve nonlinearly according to a Waelbroeck regime [47]
 7

(∆′max w & 1) instead of a Rutherford one [48] (∆′max w ≪ (quadratic) growth in time. Since kmax is given by Eq.
1). This means that the inter-plasmoids X-points col- (53), after restoring the dimensions, one can obtain the
lapse to form thin inter-plasmoids current sheets soon relation
after entering the nonlinear regime. 2
α9/4 S 3/8

The nonlinear growth of the plasmoids can be deter- w t
∼ , (65)
mined by matching a model of magnetic reconnection δcs 16 (1 + Pm )19/16 τA,L
within current sheets with Waelbroeck’s solution for the
magnetic configuration of rapidly reconnecting islands which tells us how the nonlinear evolution of the plas-
[49]. In this case, the temporal rate of change of magnetic moids depends on the Lundquist and magnetic Prandtl
flux at the separatrix of a plasmoid can be evaluated as numbers. In the case of negligible plasma viscosity, this
9/4 2
relation reduces simply to w/δcs ∼ α16 S 3/8 (t/τA,L ) .
dψs By∗ It is important to point out that since γmax τA,L ∝
≈η ∗ , (59)
dt δcs S 1/4 and δin,max /δcs ∝ S −1/8 , in high-Lundquist-number
systems the plasmoids spend an extremely short period
where By∗ is the magnetic field upstream of the inter-
∗ of time in the linear regime. It is therefore the nonlinear
plasmoids current sheet and δcs is the half-width of this
regime that sets out the time scale for the disruption of
sheet. Considering Eq. (7) for the case of the secondary
∗ the current sheet. From Eq. (65) we can see that w ∼ δcs
reconnecting current sheet, we can estimate δcs to obtain
on the time scale
[38]
!1/2 τN L ∼ S −3/16 (1 + Pm )19/32 τA,L . (66)
dψs η 1/2 By∗3
≈ 1/4
. (60) Therefore, the time scale for the current sheet disrup-
dt (1 + Pm ) L∗cs
tion decreases with increasing Lundquist number but in-
The time scale of the magnetic flux change across the creases with increasing magnetic Prandtl number.
inter-plasmoid layer is It should be noted that for extremely large S-values
also the secondary current sheets connecting the plas-
ψs w moids may themselves be subject to the plasmoid insta-
τψ = ≈ , (61)
dψs /dt η∆′max bility if

which is much larger than the transverse Alfvén time  ∗ ∗ −1/2

∗
δcs 1/4
Lcs vA,u
τA,δ = δcs /vA,u . Hence, outside the singular layer, ∗
= (1 + Pm )
Lcs η
the magnetic configuration proceeds through a sequence "
5/16 1/2
# (67)
of MHD equilibria. Furthermore, in this regime τψ is δcs (1 + Pm )
∼ < ǫc .
smaller than the characteristic time for current diffusion, w S 5/8
the plasmoid skin time τw = w2 /η. This implies that
the flux becomes frozen-in and the magnetic helicity is The same reasoning can be applied to the tertiary current
conserved within pairs of flux tubes [49, 50]. We can sheets connecting the secondary plasmoids, and so forth.
therefore consider Waelbroeck’s solution for rapidly re- This fractal-like cascade process [52] towards smaller
connecting islands [49, 51], which tells us that the plas- scales ends when the (n)th current sheets are sufficiently
moids become self-similar as w grows for w ≫ 1/∆′max thick to avoid the plasmoid instability. Therefore, using
(and w < δcs ), with a plasmoid magnetic configuration S (n) ∼ Sc one can find that, after the linear growth, the
given by plasmoids separated by the (n)th current sheets destroy
 1/2 the (n − 1)th current sheets on the nonlinear time-scale
ψs 1
w∼ , By∗ ∼ αw , L∗cs ∼ . (62) (n) (n)
α kmax τN L ∼ ǫ3/8
c (1 + Pm )
1/2
τA,L . (68)

Substituting these relations into Eq. (60), we obtain This sets off a “domino effect” that speeds up the disrup-
the plasmoid width evolution equation tion of the global current sheet, which, in turn, allows for
" #1/2 a very rapid release of magnetic energy.
dw 1 ηα
∼ kmax w , (63)
dt 2 (1 + Pm )1/2
B. Implications for fast magnetic reconnection
which yields the algebraic growth law
Plasmoids have a great impact on the global recon-
α ηkmax nection rate if they grow to a size w & δcs before being
w∼ t2 . (64)
16 (1 + Pm )1/2 expelled from the global current sheet. In this case, the
global current sheet breaks up and is replaced by a chain
Therefore, the growth of the plasmoids slows down from of plasmoids of different sizes separated by smaller cur-
the exponential growth of the linear stage to an algebraic rent sheets [7, 12, 24, 52, 53]. In a plasmoid-dominated
 8

−1/2
reconnection layer the dynamics can be particularly com- [11, 12] and Pm ≫ 1 (h dψ/dt|X i ≈ ǫc vA,u Bu Pm ) [28]
plex, with plasmoids constantly being generated, ejected to arbitrary magnetic Prandtl numbers. Note that the
and merging with each other. This leads to a strongly statistical steady-state reconnection rate in the plasmoid-
time-dependent reconnection process. However, we may dominated regime becomes nearly independent of the
expect this process to reach a statistical steady state microscopic plasma parameters only for low magnetic
with a marginally stable current sheet located at the Prandtl numbers. For large magnetic Prandtl numbers
main X-point. The statistical steady state expectation this behavior changes because the viscous energy dissi-
is supported by many numerical simulations from differ- pation leads to a decrease of the outflow velocity.
ent research groups [7, 10, 11, 24, 25, 28, 31, 33, 54– Finally, it is important to point out that the plasmoid
56]. The supposition of a marginally stable current sheet instability may also lead to an even faster reconnection
at the main X-point is also well-founded. Indeed, the regime by triggering a transition to collisionless reconnec-
fractal-like cascade process towards smaller scales due to tion [9, 12, 22, 23, 63]. Indeed, the fractal-like cascade
the plasmoid instability ends when the length of the fi- process caused by the plasmoid istability produces local
nal local current layer is shorter that the critical length reconnection layers that may be in the collisionless regime
Lc . Also, the local current sheet at the main X-point is even if the initial global current sheet is collisional, i.e.,
continuously stretched by the plasmoids moving in the even if δcs ≫ lk , being lk a relevant kinetic length scale.
outflow direction, thereby being regularly subject to the This transition occurs if δc . lk . Therefore, by rewriting
plasmoid instability every time its length exceeds Lc . Eq. (70) as δc ≈ Lcs (1 + Pm )1/4 (Sc /S 2 )1/2 , we find that
Hence, it is reasonable to expect that the length of the the plasmoid instability leads to the collisionless regime
local current sheet at the main X-point should always be if
quite close to Lc . Lcs
We can evaluate the global reconnection rate in the S& (1 + Pm )1/2 . (72)
plasmoid-dominated regime as the rate of change of the ǫ c lk
magnetic flux reconnected at the main X-point [12]. This This relation exhibits a significant dependence on the
is because only the open-flux parcels matter when cal- magnetic Prandtl number, which has the effect of in-
culating the total reconnection rate, and summing and creasing the value of the global Lundquist number re-
subtracting the contributions of all the reconnection lay- quired to reach the collisionless regime. Note also that
ers one can find that the global reconnection rate is the relevant kinetic length scale lk depends on the value
given only by the reconnection region at the main X- of the out-of-plane magnetic field. In the case of negligi-
point. Therefore, the reconnection rate in statistical ble guide magnetic field lk = di (see, e.g., [64–66]), being
steady state can be calculated as di = c/ωpi the ion skin depth, while in the opposite case
   of strong guide magnetic field lk = ρτ (see, e.g., [67–
dψ  By 69]), where ρτ = cs /ωci represents the ion sound Larmor
≈η , (69)
dt X δc radius based on both the electron and ion temperatures.
where h. . .i denotes time-average and
 1/2 V. CONCLUSIONS
Lc
δc = ǫ c L c = η . (70)
vd In this paper we have presented a theory of the plas-
Using the global reconnecting magnetic field Bu as an moid instability in visco-resistive plasmas, i.e., plasmas
estimation of By upstream of the current sheet at the in which both resistivity and viscosity are important. We
main X-point, it follows that the local downstream ve- have considered both the linear and nonlinear regimes of
−1/2 the plasmoid instability, and we have also evaluated its
locity can be written as vd = vA,u (1 + Pm ) . Furter-
effects on the global reconnection rate.
more, using Eq. (56) for the critical Lundquist number,
The linear analysis presented here generalizes the
the reconnection rate in statistical steady state becomes
theoretical work performed by Loureiro et al. [6]
simply
to account for non-negligible values of plasma viscos-
  
dψ  vA,u Bu ity. The linear growth rate and the wavenumber of
≈ ǫc 1/2
. (71) the visco-resistive plasmoid instability are found to
dt X (1 + Pm )
be γmax τA,L ∼ α3/2 S 1/4 (1 + Pm )−5/8 and kmax Lcs ∼
−3/16
Note that numerical simulations [4, 7, 10, 11, 13, 22, 24, α5/4 S 3/8 (1 + Pm ) , respectively. Therefore, plasma
57–60] indicate 1/ǫc ∼ 102 , implying that the plasmoid- viscosity has the effect of decreasing the linear growth
dominated regime is a fast reconnection regime provided rate and the wavenumber of the instability. However,
that Pm is not too high, as it could be in some astrophys- despite its damping effect, for very high Lundquist num-
ical environments like the warm interstellar medium and bers the plasmoid instability turns out to be very rapid
protogalactic plasmas [61]. (super-Alfvénic) in the linear regime.
Expression (71), first obtained in Ref. [31], extends the Nonlinearities begin to be important when the width
expressions valid for Pm ≪ 1 (h dψ/dt|X i ≈ ǫc vA,u Bu ) of the plasmoids becomes of the order of the lin-
 9

ear layer width, which is found to be δin,max /δcs ∼ state is h dψ/dt|X i ≈ ǫc vA,u Bu (1 + Pm )−1/2 , indepen-
α−3/4 S −1/8 (1 + Pm )1/16 . Since δin,max /δcs decreases dent of the Lundquist number but not on the mag-
with increasing Lundquist number, the nonlinear growth netic Prandtl number. For not excessively high Pm -
of the plasmoids turns out to be fundamental in set- values, the reconnection rate is fast, i.e., a significant
ting out the time-scale for the disruption of high-S cur- fraction of the out-of-plane electric field evaluated up-
rent sheets. For w ≫ 1/∆′max (soon after the be- stream of the global reconnection layer. We have also
ginning of the nonlinear regime) we have found that shown that the plasmoid instability may allow fast recon-
the half-width of the plasmoids grows as w/δcs ∼ nection by leading to a collisionless reconnection regime
α9/4 3/8 −19/16 2 if S & Lcs (ǫc lk )−1 (1 + Pm )1/2 .
16 S (1 + Pm ) (t/τA,L ) . Therefore, the growth
of the plasmoids slows down from the exponential growth
of the linear regime to an algebraic growth characterized
by a time scale τN L ∼ S −3/16 (1 + Pm )19/32 τA,L . For
Acknowledgments
extremely large Lundquist numbers, a fractal-like cur-
rent sheet structure with hierarchical plasmoid chains can
speed up the growth of the larger plasmoids and conse- We benefited from stimulating discussions with Ami-
quently also the destruction of the global current sheet. tava Bhattacharjee, Yi-Min Huang, Manasvi Lingam,
Finally, we have shown that also in visco-resistive plas- Nuno F. Loureiro and François L. Waelbroeck. We ac-
mas the plasmoid instability is pivotal in allowing fast knowledge financial support from the European Commu-
magnetic reconnection. In particular, we have calcu- nity under the contracts of Association between Euratom
lated that the reconnection rate in statistical steady- and ENEA.

[1] N.F. Loureiro and D.A. Uzdensky, Plasma Phys. Control. [19] F. Pucci and M. Velli, Astrophys. J. Lett. 780, L19
Fusion 58, 014021 (2016). (2014).
[2] R.S. Steinolfson and G. Van Hoven, Phys. Fluids 27, 1207 [20] A. Tenerani, A.F. Rappazzo, M. Velli and F. Pucci, As-
(1984). trophys. J. 801, 145 (2015).
[3] W.H. Matthaeus and S.L. Lamkin, Phys. Fluids 28, 303 [21] M.J. Nemati, Z.X. Wang, L. Wei, and B.I. Selim, Phys.
(1985). Plasmas 22, 012106 (2015).
[4] D. Biskamp, Phys. Fluids 29, 1520 (1986). [22] L.S. Shepherd and P. A. Cassak, Phys. Rev. Lett. 105,
[5] L.C. Lee and Z.F. Fu, J. Geophys. Res. 91, 6807 (1986). 015004 (2010).
[6] N.F. Loureiro, A.A. Schekochihin and S.C. Cowley, Phys. [23] Y.-M. Huang, A. Bhattacharjee, and B.P. Sullivan, Phys.
Plasmas 14, 100703 (2007). Plasmas 18, 072109 (2011).
[7] A. Bhattacharjee, Y.-M. Huang, H. Yang, and B. Rogers, [24] N.F. Loureiro, R. Samtaney, A.A. Schekochihin, and
Phys. Plasmas 16, 112102 (2009). D.A. Uzdensky, Phys. Plasmas 19, 042303 (2012).
[8] D. Biskamp, Magnetic Reconnection in Plasmas (Cam- [25] L. Ni, U. Ziegler, Y.-M. Huang, J. Lin, and Z. Mei, Phys.
bridge University Press, Cambridge, 2000). Plasmas 19, 072902 (2012).
[9] W. Daughton, V. Roytershteyn, B.J. Albright, H. [26] Y.-M. Huang and A. Bhattacharjee, Phys. Plasmas 20,
Karimabadi, L. Yin, and K.J. Bowers, Phys. Rev. Lett. 055702 (2013).
103, 065004 (2009). [27] N.A. Murphy, A.K. Young, C. Shen, J. Lin and L. Ni,
[10] P.A. Cassak, M.A. Shay, and J.F. Drake, Phys. Plasmas Phys. Plasmas 20, 061211 (2013).
16, 120702 (2009). [28] L. Comisso, D. Grasso and F.L. Waelbroeck, J. Phys.:
[11] Y.-M. Huang and A. Bhattacharjee, Phys. Plasmas 17, Conf. Ser. 561, 012004 (2014).
062104 (2010). [29] Q. Yu, S. Günter and K. Lackner, Nucl. Fusion 54,
[12] D.A. Uzdensky, N.F. Loureiro, and A.A. Schekochihin, 072005 (2014).
Phys. Rev. Lett. 105, 235002 (2010). [30] L. Ni, B. Kliem, J. Lin and N. Wu, Astrophys. J. 799,
[13] R. Samtaney, N.F. Loureiro, D.A. Uzdensky, A.A. 79 (2015).
Schekochihin, and S.C. Cowley, Phys. Rev. Lett. 103, [31] L. Comisso, D. Grasso and F.L. Waelbroeck, Phys. Plas-
105004 (2009). mas 22, 042109 (2015).
[14] L. Ni, K. Germaschewski, Y.-M. Huang, B.P. Sullivan, H. [32] F. Ebrahimi and R. Raman, Phys. Rev. Lett. 114, 205003
Yang and A. Bhattacharjee, Phys. Plasmas 17, 052109 (2015).
(2010). [33] L. Sironi and A. Spitkovsky, Astrophys. J. Lett. 783, L21
[15] S.D. Baalrud, A. Bhattacharjee, Y.-M. Huang, and K. (2014).
Germaschewski, Phys. Plasmas 18, 092108 (2011). [34] F. Guo, H. Li, W. Daughton, and Y.-H. Liu, Phys. Rev.
[16] S.D. Baalrud, A. Bhattacharjee, and Y.-M. Huang, Phys. Lett. 113, 155005 (2014).
Plasmas 19, 022101 (2012). [35] J.T. Dahlin, J.F. Drake and M. Swisdak, Phys. Plasmas
[17] N.F. Loureiro, A.A. Schekochihin and D.A. Uzdensky, 21, 092304 (2014)
Phys. Rev. E 87, 013102 (2013). [36] F. Guo, Y.-H. Liu, W. Daughton, and H. Li, Astrophys.
[18] Y.-M. Huang, A. Bhattacharjee, and G. Terry, Phys. J. 806, 167 (2015).
Plasmas 20, 082131 (2013). [37] X. Li, F. Guo, H. Li, and G. Li, Astrophys. J. Lett. 811,
 10

L24 (2015). J. Geophys. Res. 116, A09226 (2011)

[38] W. Park, D.A. Monticello and R.B. White, Phys. Fluids [55] Y.-M. Huang and A. Bhattacharjee, Phys. Rev. Lett.
27, 137 (1984). 109, 265002 (2012).
[39] S.V. Bulanov, J. Sakai and S.I. Syrovatskii, Fiz. Plazmy [56] L. Ni, J. Lin and N.A. Murphy, Phys. Plasmas 20, 061206
5, 280 (1979) [Sov. J. Plasma Phys. 5, 157 (1979)]. (2013).
[40] H.P. Furth, J. Killeen and M.N. Rosenbluth, Phys. Fluids [57] N.F. Loureiro, S.C. Cowley, W.D. Dorland, M.G. Haines,
6, 459 (1963). and A.A. Schekochihin, Phys. Rev. Lett. 95, 235003
[41] M.J. Lighthill, Introduction to Fourier Analysis and Gen- (2005).
eralised Functions (Cambridge University Press, London, [58] E. Lazzaro, L. Comisso and L. Valdettaro, Phys. Plasmas
1958). 17, 052509 (2010).
[42] B. Coppi, R. Galvão, M.N. Rosenbluth and P. Ruther- [59] M. Skender and G. Lapenta, Phys. Plasmas 17, 022905
ford, Fiz. Plazmy 2, 961 (1976) [Sov. J. Plasma Phys. 2, (2010).
533 (1976)]. [60] H. Baty, Phys. Plasmas 19, 092110 (2012).
[43] B. Coppi, J.M. Greene and J.L. Johnson, Nucl. Fusion [61] A. Brandenburg and K. Subramanian, Phys. Rep. 417,
6, 101 (1966). 1 (2005).
[44] A. Bondeson and J.R. Sobel, Phys. Fluids 27, 2028 [62] L. Comisso, D. Grasso and F.L. Waelbroeck, J. Plasma
(1984). Phys. 81, 495810510 (2015).
[45] F. Porcelli, Phys. Fluids 30, 1734 (1987). [63] H. Ji and W. Daughton Phys. Plasmas 18, 111207 (2011).
[46] F. Pegoraro and T.J. Schep, Plasma Phys. Controlled [64] Z.W. Ma and A. Bhattacharjee, Geophys. Res. Lett. 23,
Fusion 28, 647 (1986). 1673 (1996).
[47] F.L. Waelbroeck, Phys. Rev. Lett. 70, 3259 (1993). [65] P.A. Cassak, M.A. Shay, and J.F. Drake, Phys. Rev. Lett.
[48] P.H. Rutherford, Phys. Fluids 16, 1903 (1973). 95, 235002 (2005).
[49] F.L. Waelbroeck, Phys. Fluids B 1, 2372 (1989). [66] A. Simakov and L. Chacón, Phys. Rev. Lett. 101, 105003
[50] B.B. Kadomtsev, Fiz. Plazmy 1, 710 (1975) [Sov. J. (2008)
Plasma Phys. 1, 389 (1975)]. [67] R.G. Kleva, J.F. Drake, and F.L. Waelbroeck, Phys.
[51] L. Zakharov, B. Rogers and S. Migliuolo, Phys. Fluids B Plasmas 2, 23 (1995).
5, 2498 (1993). [68] P.A. Cassak, J.F. Drake, and M.A. Shay, Phys. Plasmas
[52] K. Shibata and S. Tanuma, Earth Planets Space 53, 473 14, 054502 (2007).
(2001). [69] L. Comisso, D. Grasso, F.L. Waelbroeck, and D. Bor-
[53] M. Bárta, J. Büchner, M. Karlický and J. Skála, Astro- gogno, Phys. Plasmas 20, 092118 (2013).
phys. J. 737, 24 (2011).
[54] R.L. Fermo, J.F. Drake, M. Swisdak, and K.-J. Hwang,