RESCEU-9/22, Prepared for submission to Physics Letters B

New description of the scaling evolution of the

cosmological magneto-hydrodynamic system
Fumio Uchida,∗ Kohei Kamada, and Jun’ichi Yokoyama†
Research Center for the Early Universe (RESCEU) and Department of Physics,
Graduate School of Science, The University of Tokyo, Tokyo 113-0033, Japan

Motoko Fujiwara
Physik-Department, Technische Universität München, James-Franck-Str. 1 85748 Garching, Germany
(Dated: January 2, 2023)
We present a new description of cosmological evolution of the primordial magnetic field under the
condition that it is non-helical and its energy density is larger than the kinetic energy density. We
arXiv:2212.14355v1 [astro-ph.CO] 29 Dec 2022

argue that the evolution can be described by four different regimes, according to whether the decay
dynamics is linear or not, and whether the dominant dissipation term is the shear viscosity or the
drag force. Using this classification and conservation of the Hosking integral, we present analytic
models to adequately interpret the results of various numerical simulations of field evolution with
variety of initial conditions. It is found that, contrary to the conventional wisdom, the decay of the
field is generally slow, exhibiting the inverse transfer, because of the conservation of the Hosking
integral. Using the description proposed here, we may trace the intermediate evolution history of
the magnetic field and clarify whether each process governing its evolution is frozen or not, which
is essential to follow the evolution of relatively weak magnetic fields.

I. INTRODUCTION enhanced in contrast to the decaying short-wavelength

modes. In this letter, we present analytic description of
The cosmological magnetic field can serve as a probe the evolution of the cosmological magnetic field with the
into the early universe. Observations tell that the inter- initial conditions stated above, consistently with the nu-
galactic magnetic field exists even in the void [1–13], and merical results.
its origin may be attributed to the early universe. Pos- The system during its evolution can be classified into
sible formation scenarios include magnetogenesis during four different regimes, according to whether the dynam-
or at the end of inflation [14–39] and at the phase tran- ics is linear or nonlinear with respect to the magnetic
sitions [40–53]. To understand the role of the primordial field and whether the kinetic dissipation is mainly due to
magnetic field and to connect the models of the primor- shear viscosity or drag force. The evolution of cosmolog-
dial universe and observation, describing its magneto- ical magnetic field should inevitably experience these dif-
hydrodynamic (MHD) evolution within the framework ferent regimes since the dominant source of dissipation in
of the standard cosmology is the first and the most im- the early universe vary with time among the collision be-
portant step. tween the particles of the fluid, the friction from the back-
In this context, Banerjee & Jedamzik [54] tried to de- ground free-streaming particles such as photons and neu-
scribe the evolution of the cosmological magnetic field trinos, and the Hubble friction in the matter dominated
comprehensively, which has been commonly accepted. era. As for the linear regimes with either shear viscosity
However, some of their results appear to be inconsistent or drag force, our analysis is partly based on Ref. [54],
with those of recent numerical MHD simulations [55– and for the nonlinear regime with shear viscosity, mostly
60]. As a representative scenario, let us assume that the on Ref. [61, 62], which resolves the inconsistency between
coupled system of the primordial magnetic field and the theory and numerical studies by introducing the Hosking
plasma filling the universe is magnetically dominant and integral (or the Saffman helicity invariant) [61] and the
non-helical initially, e.g., at the electroweak symmetry reconnection time scale. The turbulent regime with drag
breaking temperature. While the conventional analysis force has never been discussed in literature, on which an
assumes that the magnetic energy of the long-wavelength analysis is provided in this letter consistently with the
modes is intact as the short-wavelength modes decay, nu- other three regimes. For the first time, we integrate the
merical simulations exhibit the so-called inverse trans- analyses in these four regimes, which complete the de-
fer [55–60], in which the long-wavelength modes are scription of the evolution of the magnetically-dominant
non-helical cosmological magnetic field throughout the
cosmic history from its generation to the present. While
∗ fuchida@resceu.s.u-tokyo.ac.jp
†
the analysis in Ref. [61, 62] may be sufficient to connect
Also at Kavli Institute for the Physics and Mathematics of the
the properties of magnetic field in the early universe and
Universe (Kavli IPMU),
WPI, UTIAS, The University of Tokyo, Kashiwa, Chiba 277- at present for relatively strong one, our analysis is es-
8568, Japan, sential for its complete history. To see the difference
Trans-scale Quantum Science Institute, clearly, we focus on the case of weak magnetic field where
The University of Tokyo, Tokyo 113-0033, Japan the reconnection is driven by the Sweet–Parker reconnec-
 2

tion [63, 64]. that the probability distribution is almost Gaussian. On

top of these assumptions, we parametrize the magnetic
field by its typical strength, B, and coherence length, ξM .
II. EQUATIONS OF One may define these quantities in terms of the power
MAGNETO-HYDRODYNAMICS spectrum, PB (k, τ ), of the magnetic field,

We consider a coupled system of the magnetic field and hB(k) · B(k0 )i = PB (k, τ )(2π)3 δ 3 (k + k0 ), (6)
the velocity field of the plasma fluid in the early universe.
In the following expressions, τ is the conformal time, x is where B(k) denotes a Fourier mode.
the comoving coordinate, and ρ and p are the comoving
d3 k
Z
energy density and the comoving pressure of the fluid, B 2 := hB 2 i = PB (k, τ ), (7)
respectively. The comoving quantities are related to the (2π)3
d3 k 2π
Z
physical fields denoted by a subscript p as 1
ξM := 2 PB (k, τ ), (8)
2 4 4 B (2π)3 k
B = a Bp , v = vp , ρ = a ρp , p = a pp ,
σ = aσp , η=a −1
ηp , α = aαp , (1) In the same way, we can define the typical velocity, v, and
the coherence length of the velocity field, ξK , in tems of
where a is the scale factor of the universe. The equations the velocity power spectrum, Pv (k, τ ), as
of motions are
d3 k
Z
1 2 2 2
v := hv i = Pv (k, τ ), (9)
∂τ B − ∇ × (v × B) = ∇ B (2) (2π)3
σ
d3 k 2π
Z
1
for the magnetic field B(τ, x) and ξM := 2 Pv (k, τ ). (10)
v (2π)3 k
1 1
∂τ v + (v · ∇)v − (∇ × B) × B + ∇p In the rest of the letter, we are going to determine the
ρ+p ρ+p
  time dependence of these quantities, which shows scaling
1 behavior.
= η ∇2 v + ∇(∇ · v) − αv (3)
3
Conserved quantities play important roles to deter-
for the velocity field v(τ, x). The right hand sides of these
mine the scaling evolution of the magnetic and velocity
equations (2) and (3) represent the dissipation of the en-
fields. In magnetically dominant regimes, where the en-
ergy, in terms of the electric conductivity, σ, the shear
ergy density of the non-helical magnetic field is dominant
viscosity, η, and the drag force coefficient, α. The former
over that of the velocity field, an approximate conserved
two quantities originate from the collision between con-
quantity called Hosking integral,
stituent particles of the plasma, while the last one from
the background of the system, i.e., the Hubble friction
Z
and/or free-streaming particles. As the temperature of IHM := d3 rhhM (x)hM (x + r)i, (11)
V
the universe decreases, the dominant term for the dissi- where hM := A · B, (12)
pation changes. These equations are closed together with
the continuity equation has been recently proposed [61].
Here integral is taken over a volume V much larger
∂τ ρ + ∇ · [(ρ + p)v] = E · (∇ × B) (+aHρ), (4) than the correlation volume, ξM 3
, of the magnetic field.
 
1 Since the correlation length of hM is also expected to be
where E = ∇ − v × B. (5)
σ of order of ξM , on dimentional grounds we find
The last term in the right-hand-side of Eq. (4) is present IHM ' B 4 ξM
5
(13)
in the matter dominated era.
We assume the homogeneity and isotropy on average up to a spectrum-dependent numerical factor. In par-
and express the magnetic and velocity field configurations ticular, we obtain the following constraint between the
on each time slice by a few parameters, by treating them parameters at the temperature of the universe, T .
as stochastic fields. If we look inside a sufficiently small B(T )4 ξM (T )5 = Bini
4 5
ξM,ini , (14)
region, the magnetic field will be almost coherent and it
appears anisotropic. However, the present Hubble patch where the subscript ini denotes the initial condition. In-
is composed of many coherent subpatches that have ran- deed, the recent numerical study confirms that the Hosk-
dom magnetic fields assigned independently by a single ing integral is well conserved [65]. This conservation law
probability distribution. The universality of the proba- restricts the evolution of the system as long as the sys-
bility distribution throughout the space guarantees the tem is magnetically dominant. We will determine the
homogeneity, and the isotropy is imposed on the proba- scaling exponents using Eq. (14) for each regime in the
bility distribution. For later purpose, we further assume succeeding section.
 3

III. REGIME-DEPENDENT ANALYSES formed at each boundary between two regions of coher-
ent magnetic field lines, on which incoming magnetic field
We now turn to the regime-dependent analysis, which lines are dissipated, reconnect, and feed energy into the
is the main result of this letter. As a preparation, we velocity field [63, 64]. See Fig. 1. Let vin and vout de-
define a quantity that determines the boundary between note incoming and outgoing velocities of material carry-
non-linear and linear regimes by comparing the contribu- ing the magnetic field lines, respectively, as illustrated
tions from non-linear and linear terms in the induction in Fig. 1. The mass conservation on each sheet implies
equation, (2). From rough estimates, |∇ × (v × B)| ∼ vin ξM = vout ξK . Comparing the inside and outside of the
vB/ min{ξM , ξK } and |∇2 B/σ| ∼ B/σξM 2
, we define the current sheet, the stationarity condition in the induction
magnetic Reynolds number, equation, Eq. (2), is approximately Bv B
ξK ' σξ 2 , where we
in
K
−1
2
σvξM have approximated that ∇ ∼ ξK at the current sheet.
ReM := . (15) Therefore, we can express vin and vout in terms of B, ξM
min{ξM , ξK } and ξK as
We identify ReM > (<)1 with the non-linear (linear)
1 ξM
regimes. vin = , vout = 2 . (18)
We also define another quantity that determines the σξK σξK
boundary between viscous and dragged regimes by com- The outgoing flow spreads over the whole volume keeping
paring the contributions in the right-hand-side of the the coherence length ξK while decreasing the amplitude.
Navier–Stokes equation, Eq. (3), at the scale of the ki- Taking into account the dilution factor, we obtain
netic coherence length, ξK . By
 estimating that α|v| ∼
αv and η ∇2 v + 13 ∇(∇ · v) ∼ ηv/ξK
 2
, we define the ξK 2
v2 ' v . (19)
quantity that characterises the ratio of dissipation terms ξM out
as
2 Here we assume the high aspect ratio, ξM /ξK  1, of the
αξK current sheet. Since it takes the time,
rdiss := . (16)
η
ξM
and then rdiss < (>)1 corresponds to the viscous (drag) τSP := , (20)
vin
regimes. In the following, we study the evolution of the
system by classifying the regimes according to the criteria to process all the magnetic field within the volume ξM 3
,
determined by these quantities. One can find the results the decay of the magnetic field energy in this regime pro-
of our analysis for each regime in Table I. ceeds, keeping the condition that the time scale equals to
the conformal time at cosmic temperature T ,
A. Nonlinear regime with shear viscosity τ (T ) (= τSP ) = σξM ξK . (21)

First, we consider the case where the magnetic Here we have used Eq. (18). This condition is the origin
Reynolds number is larger than unity and the shear vis- of the explicit time dependence of the evolution of the
cosity is dominant over the drag force, characteristic properties of the magnetic and velocity
field.
ReM  1, rdiss  1, (17)
Now let us take into account the effect of shear viscos-
In this regime, the evolution of the system is determined ity [67] to determine the relation between the energy den-
by the quasi-stationary condition of the magnetic recon- sity of the magnetic and velocity fields. Considering the
nection with dissipation due to the shear viscosity, known energy budget along the fluid motion inside the current
as the Sweet–Paker reconnection [63, 64], which drives sheet, the energy of the magnetic field should be trans-
the transfer between the magnetic and kinetic energy, if ferred to the energy of the outflow. Taking into account
the magnetic field is not so strong (or more precisely if the dissipation due to the shear viscosity, ρ+p ηvout
ξM ,
2
the Lundquist number is not so large)1 . The application 2 ξK

of this physical mechanism to this regime is originally dis- which is evaluated by supposing the balance between the
cussed in Ref. [61]. Based on their discussion, we derived injection term and dissipation term in the Navier–Stokes
the scaling laws for MHD system. equation, together with Eq. (18), the energy conservation
The Sweet–Parker reconnection mechanism is de- leads to another condition,
2
scribed as follows. A current sheet of size ξM × ξK is 1 2 ρ+p 2
B = (1 + PrM ) vout (22)
2 2
is imposed. Here we have defined the magnetic Prandtl
1 For stronger magnetic field, we expect that the reconnection is number,
driven by the fast reconnection [66]. In this case, the analysis in
Ref. [61, 62] may be sufficient. PrM := ση. (23)
 M K
(2) ξM > ξK (Large aspect ratio btw magnetic & velocity sector)

• Mechanism: Energy is transferred from (magnetic field) → (velocity field) through reconnection of magnetic
on the sheet w/ σ <∞
4 flux

vin ξK
ξM B

vout σ < ∞, j = ∇ × B ≠ 0

ξM

ξM
2 3
FIG. 1: An illustration of the current sheet (size of ξM × ξK ) embedded in each patch (size of ξM ). The magnetic
reconnection occurs due to the finite electric conductivity on the current sheet. The red solid and the black dashed
arrows show the magnetic field and velocity flows, respectively.

From Eqs. (18), (21) and (22), we obtain a relation be- B. Nonlinear regime with drag force
tween B and ξM .
For the case where the magnetic Reynolds number is
B 2 τ 4 ' (ρ + p)σ 3 ηξM
6
, (24) larger than unity, while the drag force is dominant over
the shear viscosity,
where we have used the approximation, PrM  1, which
is the case in the early universe [68]. ReM  1, rdiss  1, (31)
Combining Eqs. (14) and (24), we may determine the
the system is driven by the magnetic reconnection with
scaling behaviors of the magnetic field in this regime.
dissipation due to the drag force. To the best of our
12 15 5 10
knowledge, this regime has never been discussed consis-
(ρ + p)σ 3 η 34 τ − 17 ,

B = Bini
17
ξM,ini
17
(25) tently with the other three regimes, so we work it out
4 5 − 2 8 here.
(ρ + p)σ 3 η 17 τ 17 .

ξM = Bini
17
ξM,ini
17
(26) Exactly the same discussion as the previous section,
Sec. III A, holds (Eqs. (18) and (21)) up to the consid-
Note that if the initial coherence length is large, eration about the energy balance. On the other hand,
− 1 32 1 the relation between the magnetic and kinetic energy
(ρ + p)σ 3 η 4 ,

τ < Bini2 ξM,ini (27) (Eq. (22)) changes as follows. In this regime, the dissi-
pation of the kinetic energy along the fluid motion inside
the magnetic field is frozen and eventually starts the scal- the current sheet is replaced by the one due to the drag
ing evolution when the equality for Eq. (27) is satisfied. force, ρ+p
2 αvout ξM . Then, the energy balance leads to a
As for the evolution of the velocity field, we find condition,
1 3
−1 1 2 ρ+p
v = σ 2 τ − 2 ξM
2
, ξK = σ −1 τ ξM , (28) B ' αvout ξM . (32)
2 2
as the formulae describing the evolution in this regime. Here we have used the relation αξM /vout = rdiss PrM 
1 so that the dissipation dominates over the remaining
The conditions for the system to be in this regime of kinetic energy. Using Eqs. (18) and (21), Eq. (32) can be
the scaling evolution, Eqs. (17), can be rewritten as rewritten as

5 5 B 2 τ 2 = (ρ + p)σαξM
4
. (33)
ReM (τ ) = σ 2 τ − 2 ξM
5
(:= ReSP
M )  1, (29)
−2 Combining Eq. (14) and Eq. (33), we obtain
rdiss (τ ) = ασ −2 η −1 τ 2 ξM SP
(:= rdiss )  1, (30)
8 10 5 5
B = Bini
13
ξM,ini
13
[(ρ + p)σα] 26 τ − 13 , (34)
which may be eventually violated so that the system en- 4 5 2
− 13 4
ters another regime. Note that from these conditions, ξM = Bini ξM,ini [(ρ + p)σα]
13 13
τ 13 . (35)
we can confirm the consistency of our solutions with the
assumptions that the system is magnetically dominant, Note that, once more, if the initial coherence length is
SP 2 large with
B 2 /[(ρ + p)v 2 ] = PrM ReM 5  1, and that the aspect
SP 25 −1 2 1
ratio is large, ξM /ξK = ReM  1. τ < Bini ξM,ini [(ρ + p)σα] 2 , (36)
 5

the magnetic field is frozen and starts the scaling evolu- The evolution of the velocity field are derived from
tion when the inequality is saturated. The velocity field Eqs. (39) and (41). Note that if the initial magnetic
evolves in the same manner as described by Eqs. (28) in field is too weak,
this regime.
(ρ + p)η
τ< 2 , (46)
The Reynolds number and the ratio of dissipation Bini
terms are evaluated in the same way as Eqs. (29) and
(30). The conditions for the system to be in this regime, the magnetic field is frozen and starts the scaling
Eqs. (31), can then be rewritten as evolution when the inequality is saturated.

ReSP
M  1,
SP
rdiss  1. (37) The conditions for the system to be in this regime,
In this regime, we can also confirm the domi- Eqs. (38), can be rewritten as
nance of the magnetic energy as B 2 /[(ρ + p)v 2 ] =
SP 2 1 σB 2 ξM
2
SP
PrM ReM 5 rdiss  1. The large aspect ratio is confirmed ReM (τ ) = (:= ReηM )  1, (47)
ρ+p η
in the same way as Sec. III A.
αξ 2 η
rdiss (τ ) = M (:= rdiss )  1, (48)
η
C. Linear regime with shear viscosity
from which the consistency of our solutions with the
assumption that the system is magnetically dominant,
Next, we consider the case where the magnetic −1
Reynolds number is smaller than unity and the shear B 2 /[(ρ + p)v 2 ] = PrM (ReηM )  1, is confirmed.
viscosity is dominant over the drag force,
ReM  1, rdiss  1. (38) D. Linear regime with drag force
In this case, the magnetic reconnection does not occur
as the dominant contribution for the energy transfer be- Finally, we consider the case where the magnetic
tween the magnetic and kinetic energy. Instead, the ve- Reynolds number is smaller than unity and the drag force
locity field is excited by the Lorentz force at the scale of is dominant over the shear viscosity,
the magnetic coherence length
ReM  1, rdiss  1. (49)
ξK ' ξM . (39)
In this case, the velocity field is excited by the Lorentz
The energy of the velocity field is brought to smaller force and dissipated at smaller scales, similarly to the
scales by the Kolmogorov turbulence and finally dissi- case in the previous section, Sec. III C, and Eq. (39) also
pated by the shear viscosity. The quasi-stationarity of holds. The quasi-stationarity of the system implies
the system implies the balance of the injection and the
dissipation of the kinetic energy in the Navier–Stokes 1 B2
' αv, (50)
equation, Eq. (3), ρ + p ξM
1 B2 v from which one can express the typical velocity, v, in
'η 2, (40)
ρ + p ξM ξK terms of the typical strength, B, and the coherence
length, ξM , of the magnetic field as
from which one can express the typical velocity as
1 B 2 ξM 1 B2
v= . (41) v= . (51)
ρ+p η ρ + p αξM
For the Kolmogorov turbulence, it takes It takes the time
2
η ξK (ρ + p)η α ξK (ρ + p)αξM
τeddy := = (42) τeddy := = (52)
v B2 v B2
to break the eddy of the coherence scale. Therefore, the for the Kolmogorov turbulence to break the eddy of the
decay of the kinetic energy in this regime proceeds, keep- coherence scale. The decay of the kinetic energy in this
ing the condition regime proceeds, keeping the condition [54],
η
τ (T ) = τeddy . (43) α
τ (T ) = τeddy . (53)
Combining Eqs. (14) and (41), we obtain the scaling
Combining Eqs. (14) and (51), we obtain
behaviors of the magnetic field.
4 5 5 5
[(ρ + p)α] 18 τ − 18 ,
1 1
B = [(ρ + p)η] 2 τ − 2 , B = Bini ξM,ini (54)
9 9
(44)
4 4 5
− 25 2 − 29 2
ξM = Bini ξM,ini [(ρ + p)η]
5
τ .
5 (45) ξM = Bini ξM,ini [(ρ + p)α]
9 9
τ .
9 (55)
 6

The evolution of the velocity field is derived from Also, we confirm that the time scales of the evolution
Eqs. (39) and (51). Note that if the initial coherence of the system are connected smoothly at the boundary
length is large, between the linear and non-linear regimes,
2  η
(ρ + p)αξM,ini τ , (ReM = 1, viscous regimes)
τ< 2 , (56) τSP = eddyα (62)
BM τeddy , (ReM = 1, dragged regimes)

the magnetic field is frozen and starts the scaling which remains identical to the conformal time of the
evolution when the equality is satisfied. universe, τ (T ), in the scaling regime. This also suggests
that the condition for the system to be frozen is also
The conditions for the system to be in this regime, connected smoothly. Note that these expressions,
Eqs. (49), can be rewritten as Eqs. (60)-(62), are clearly self-consistent at the bound-
ary of the viscous and dragged regimes (rdiss = 1).
1 σB 2
ReM (τ ) = (:= Reα M )  1, (57) We conclude that the decay laws in linear and non-
ρ+p α linear regimes as well as viscous and dragged regimes
αξ 2 α are consistent at their boundary (ReM = 1 and rdiss = 1).
rdiss (τ ) = M (:= rdiss )  1, (58)
η
Now let us summarize the decay laws in Table I. De-
from which the consistency of our solutions with the tailed discussions are found in the sections shown in the
assumption that the system is magnetically dominant, first column. The state of the system is classified into
B 2 /[(ρ + p)v 2 ] = PrM (Reα
M)
−1 α
rdiss  1, is confirmed. four regimes, according to the criteria written in the
second-to-fourth columns. In each regime, the formu-
Note that the linear regimes with both shear viscos- lae describing the decay laws are the equations specified
ity (Sec. III C) and drag force (Sec III D) were studied in the fifth and sixth columns. The decay time scales are
in Ref. [54], and it was claimed that the system in the shown in the last column. When the conformal time of
linear regimes with shear viscosity is frozen in the real- the universe τ (T ) is smaller than these time scales, the
istic situation. On the contrary, here we have performed decay processes are too slow to operate, and the system
a general study, not restricting ourselves to the situation is frozen until when τ (T ) catches up with the decay time
which was discussed in the literature. In Sec. V, we will scales. We have confirmed that each regime including the
argue that the system can evolve according to the scaling condition to be frozen is connected smoothly. Therefore,
law we derived also in realistic situation. we may continuously evolve the system, even involving
multiple regimes, by following the decay laws in Table I.
IV. INTEGRATION OF THE ANALYSES

V. DISCUSSION
In the previous section, we have conducted regime-
dependent analyses. Let us confirm the consistency, by
showing that the solutions coincide at the boundary of In this letter, we have provided a comprehensive analy-
linear and non-linear regimes. sis that describes the evolution of the magnetically domi-
First, we focus on the properties of the magnetic and nant and non-helical magneto-hydrodynamic system. We
velocity field in the scaling evolution regimes. On the limit ourselves to the cases when the magnetic field is
coherence length, we have not very strong so that the Lundquist number is not
very large and that the Sweet–Parker reconnection (but
ξM = ξK , when ReM = 1 (59) not the fast reconnection) is relevant. The results are
summarized in Table I. Importantly, our analysis pre-
in all regimes. We identify v as the representative ve-
dicts quite different evolution history of the primordial
locity in each regime. For the non-linear regime, we use
magnetic field, compared with the one by Banerjee and
characteristic velocity in the Sweet–Parker reconnection.
Jedamzik [54]. The difference mainly comes from the fact
We may confirm that these expressions match velocities
that the approximately conserved Hosking integral was
for linear regimes on the boundary between the linear
not taken into account there, which was first pointed out
and non-linear regimes,
by Hosking and Schekochihin [61].
2
(
1 B ξM Let us see how the analysis with the Hosking inte-
ρ+p η , (ReM = 1, viscous regimes)
v= 1 B2
(60) gral better describes the results of existing numerical
ρ+p αξM , (ReM = 1, dragged regimes)
MHD simulations compared with the one by Banerjee
Using these expressions of v for each regime, we may and Jedamzik [54]. See the plot in Fig. 2. By parametriz-
express ing time dependence of each quantity as [57]

B2 B2

PrM v 2 , (ReM = 1, viscous regimes)
=
PrM rdiss v 2 . (ReM = 1, dragged regimes)
(61) ∝ τ −pM , ξM ∝ τ qM , v 2 ∝ τ −pK , ξK ∝ τ qK , (63)
ρ+p ρ+p
 7

TABLE I: Summary of the decay laws of non-helical and magnetically dominant regimes.

RegimesDissipation Condition Decay laws of the magnetic field velocity field Decay time scale
Sec. III A Shear viscosity Eqs. (30) Eqs. (25) and (26) or frozen
Nonlinear Eqs. (28) τSP = σξM ξK
Sec. III B Drag force Eqs. (37) Eqs. (34) and (35) or frozen
η 
Sec. III C Shear viscosity Eqs. (48) Eqs. (44) and (45) or frozen Eqs. (39) and (41) τeddy ξK
Linear α =
Sec. III D Drag force Eqs. (58) Eqs. (54) and (55) or frozen Eqs. (39) and (51) τeddy v

we compare the analytic formula, which is first studied in the magnetic coherence length is safely short, ξM,ini =
Ref. [61] and summarized in Sec. III A of the present let- 10−17.4 Mpc, at the electroweak symmetry breaking tem-
ter, and the numerical results in the literature [57, 58, 65]. perature ∼ 100 GeV. Note that if we assume that the
The five runs that can be interpreted to be magnetically primordial magnetic field is generated before the elec-
dominant and non-helical from Ref. [58], the non-helical troweak symmetry breaking, such a choice of the ini-
one in Ref. [57], and the three with the standard magnetic tial condition is motivated by the constraint that comes
dissipation term in Ref. [65] are employed. Plots near from the big-bang nucleosynthesis [70]. A stronger or
the origin indicate that the theory well describes the nu- longer-ranged U(1)Y magnetic field unavoidably gener-
merical calculations. The dots representing our analysis ates baryon isocurvature perturbation through the chiral
scatter around the origin2 , while the crosses representing anomaly at the electroweak symmetry breaking, leading
the analysis in Ref. [54] are off the origin rightward. Note to the deuterium overproduction at the big-bang nucle-
that we take values of the parameters, pM/K and qM/K , osynthesis to be inconsistent with the present universe.
up to their fluctuations in time (which may be roughly The initial condition chosen here almost saturates this
. 0.1), from the results of numerical simulations in the constraint.
literature, and they are not so accurate. Nevertheless, With this initial condition of the magnetic field, the re-
this plot strongly suggests that the new analysis with connection is driven by the Sweet–Parker mechanism at
the Hosking integral better describes the reality, com- the non-linear stage, and thus the analysis in the present
pared with the one in Banerjee and Jedamzik [54], which letter applies. At first, the system is in the nonlinear
tends to predict too fast decay of both the magnetic and
the kinetic energy.
Now we turn to the implication of the main topic of
this letter, the cosmological evolution of the magnetic
field that involves not only the non-linear regime but also
the linear regime with dissipation due to the shear vis-
cosity as well as dissipation due to the drag force. As
a demonstration, we investigate the cosmological evolu-
tion of the magnetic field generated at the time of elec-
troweak symmetry breaking. In this case, we need to take
into account the shear viscosity due to electron-electron,
electron-neutrino, proton-proton, electron-photon, and
hydrogen-hydrogen collisions and the drag force due to
the free-streaming neutrinos and photons, when their
mean free path is larger than the relevant length scale
of the system [54, 69]. The Hubble friction in the matter
domination and the ambipolar drag [54] also serve as the
drag forces. These phenomena lead to the change of the
regimes, where we can smoothly connect the evolution of
magnetic field by using Table I.
Figures 3 and 4 show an example of the evolution his-
tory of magnetic field strength and coherence length, re-
spectively. The initial condition is set such that the en- FIG. 2: Comparison between theoretical analyses
ergy density of the magnetic field is several orders below (dots: the analysis in Sec. III A; crosses: Banerjee and
the critical energy density, Bini = 10−10.3 G, and that Jedamzik [54]) and the numerical results in the
literature [57, 58, 65]. Since the power indices read off
from numerical simulations may well contain errors
around ±0.1, one should conclude that the theoretical
2 The apparent outlier (the red point in the above-right of the model reproduces the numerical simulation well if the
origin) is the run “K60D1c” in Zhou et al [65]. We do not have
a clear explanation about the origin of the deviation, but could
points are located within a circle of a radius ∼ 0.14
be originated from the insufficient resolution. centered at the origin.
 8

Noting that the neutrino drag αν ∝ τ −4 , the magnetic

coherence length increases following Eq. (35). Therefore,
rdiss decreases and the system enters the nonlinear regime
with the electron viscosity again. At around the QCD
phase transition, the system becomes frozen due to the
sudden change of the viscosity, which becomes dominated
by the proton-proton collision. The scaling evolution re-
sumes when electrons become massive and the photon
drag coefficient decreases. Then rdiss drops below unity,
and the linear regime with proton viscosity begins. After
the recombination, the system becomes frozen again.
Let us demonstrate how our analysis is different from
what is commonly accepted. According to the analysis
by Banerjee and Jedamzik [54], the evolution in the tur-
bulent regime is B ∝ τ −5/7 and ξM ∝ τ 2/7 , assuming
the Batchelor spectrum. This scaling behavior is often
FIG. 3: The evolution of the strength of the magnetic
just extrapolated throughout the history before the re-
field. Solid lines are based on our analysis, while the
combination, to connect initial conditions and resultant
dashed line is based on the naive extrapolation of the
configurations. We plot these naive extrapolations with
analysis by Banerjee and Jedamzik [54]. Colored bars
dashed lines in Figs. 3 and 4.
indicate which regime determines the scaling evolution
at each temperature range. The black shaded region One can see that the evolution history is quite differ-
corresponds to over-closure of the universe. ent. First, based on our analysis, the resultant strength
is much stronger and the coherence length is much longer
than the previous expectations. This is because the con-
served quantity, the Hosking integral, leads to a slower
decay of the magnetic energy with the inverse cascade.
Note that the resultant properties of magnetic field are
not likely to account for the long-range intergalactic
magnetic field suggested by the blazar observations [13]
although the magnetic field strength becomes stronger
than the previous estimate.
Second, it is impossible to approximate the evolution
throughout the history before the recombination by a sin-
gle power law. This is because the time-dependent dissi-
pation coefficients, σ, η, and α, play critical roles even in
the nonlinear regime.
It is essential that the magnetic field is often frozen
in its evolution history. When it is frozen, the initial
conditions and the decay time scale at that time are in-
sufficient to determine the properties at that time. To
FIG. 4: The evolution of the coherence length of the connect arbitrary initial conditions and the correspond-
magnetic field. Solid lines are based on our analysis, ing final configurations, one should draw the evolution
while the dashed line is based on the naive history like Figs. 3 and 4 for every initial condition, since
extrapolation of the analysis by Banerjee and Jedamzik the evolution histories are highly diverse depending on
[54]. Colored bars indicate which regime determines the the initial conditions. Since the system experiences mul-
scaling evolution at each temperature range. The tiple regimes in general, integration of regime-dependent
dotted line indicates the Hubble horizon. analysis given in Sec. IV is inevitable to follow this evo-
lution.
In summary, we have clarified decisive elements to de-
regimes for a while. Since the dimensionful Fermi con- termine the evolution history of the cosmological mag-
stant GF is involved in the electron-neutrino collision, netic field in the case it is non-helical and its energy sur-
the dominant contributor to the shear viscosity eventu- passes the kinetic energy. As a result, the evolution may
ally changes from electron-electron collisions to electron- be classified into the four possible regimes summarized
neutrino collisions. Since the neutrino viscosity evolves in Table I.
as ην ∝ τ 4 , which compensates the τ dependence in Cases with helical magnetic field and kinetically dom-
Eqs. (25) and (26), the time evolution is accidentally inant regimes require some extension of the analysis.
frozen. When the neutrinos start free-streaming, neu- Also, we have only considered the magnetic reconnec-
trino drag begins to dissipate the energy of the system. tion described by the Sweet–Parker model [63, 64]. Dif-
 9

ferent mechanisms of magnetic reconnection [66] should ACKNOWLEDGMENTS

also be taken into account for full description [62]. Our
forthcoming paper [71] will give a more general descrip- The work of FU was supported by the Forefront
tion about the evolution of the cosmological magneto- Physics and Mathematics Program to Drive Trans-
hydrodynamic system. formation (FoPM). The work of MF was supported
by the Collaborative Research Center SFB1258 and
by the Deutsche Forschungsgemeinschaft (DFG, Ger-
man Research Foundation) under Germany’s Excel-
lence Strategy - EXC-2094 - 390783311, JSPS Core-
to-Core Program (No.JPJSCCA20200002). This work
was partially supported by JSPS KAKENHI, Grant-in-
Aid for Scientific Research Nos. (C)JP19K03842(KK),
(S)20H05639(JY), the Grant-in-Aid for Innovative Areas
Nos. 18H05542(MF) and 20H05248(JY).

[1] A. Neronov and I. Vovk. Evidence for strong extra- on Magnetic Fields in Intergalactic Voids from Long-
galactic magnetic fields from Fermi observations of TeV term GeV-TeV Light Curves of the Blazar Mrk 421.
blazars. Science, 328:73–75, 2010. Astrophys. J. Lett., 771:L42, 2013.
[2] F. Tavecchio, G. Ghisellini, L. Foschini, G. Bon- [11] Justin D. Finke, Luis C. Reyes, Markos Georganopoulos,
noli, G. Ghirlanda, and P. Coppi. The intergalactic Kaeleigh Reynolds, Marco Ajello, Stephen J. Fegan, and
magnetic field constrained by Fermi/Large Area Tele- Kevin McCann. Constraints on the Intergalactic Mag-
scope observations of the TeV blazar 1ES0229+200. netic Field with Gamma-Ray Observations of Blazars.
Monthly Notices of the Royal Astronomical Society: Letters, Astrophys. J., 814(1):20, November 2015.
406(1):L70–L74, July 2010. [12] M. Ackermann et al. The Search for Spatial Extension in
[3] Shin’ichiro Ando and Alexander Kusenko. Evidence for High-latitude Sources Detected by the Fermi Large Area
Gamma-ray Halos Around Active Galactic Nuclei and Telescope. ApJS, 237(2):32, 2018.
the First Measurement of Intergalactic Magnetic Fields. [13] Rafael Alves Batista and Andrey Saveliev. Mul-
The Astrophysical Journal Letters, 722(1):L39–L44, Oc- timessenger Constraints on Intergalactic Mag-
tober 2010. netic Fields from the Flare of TXS 0506+056.
[4] F. Tavecchio, G. Ghisellini, G. Bonnoli, and L. Foschini. The Astrophysical Journal Letters, 902(1):L11, Oc-
Extreme TeV blazars and the intergalactic magnetic field. tober 2020.
Monthly Notices of the Royal Astronomical Society: Letters,[14] Michael S. Turner and Lawrence M. Widrow. Inflation-
414(4):3566–3576, July 2011. produced, large-scale magnetic fields. Phys. Rev. D,
[5] K. Dolag, M. Kachelriess, S. Ostapchenko, and 37(10):2743–2754, May 1988.
R. Tomàs. Lower Limit on the Strength and [15] Bharat Ratra. Cosmological “Seed” Magnetic Field from
Filling Factor of Extragalactic Magnetic Fields. Inflation. The Astrophysical Journal Letters, 391:L1,
The Astrophysical Journal Letters, 727(1):L4, January May 1992.
2011. [16] W. Daniel Garretson, George B. Field, and Sean M. Car-
[6] Warren Essey, Shin’ichiro Ando, and Alexander Kusenko. roll. Primordial magnetic fields from pseudo Goldstone
Determination of intergalactic magnetic fields from bosons. Phys. Rev. D, 46(12):5346–5351, December 1992.
gamma ray data. Astroparticle Physics, 35(3):135–139, [17] Francisco D. Mazzitelli and Federico M. Spedalieri.
October 2011. Scalar electrodynamics and primordial magnetic fields.
[7] A. M. Taylor, I. Vovk, and A. Neronov. Extragalac- Phys. Rev. D, 52(12):6694–6699, December 1995.
tic magnetic fields constraints from simultaneous GeV- [18] L. Campanelli, P. Cea, G. L. Fogli, and L. Tedesco.
TeV observations of blazars. Astronomy & Astrophysics, Inflation-produced magnetic fields in Rn F 2 and IF 2
529:A144, May 2011. models. Phys. Rev. D, 77:123002, Jun 2008.
[8] Charles D. Dermer, Massimo Cavadini, Soebur Raz- [19] Kazuharu Bamba and Sergei D. Odintsov. Inflation and
zaque, Justin D. Finke, James Chiang, and Benoit late-time cosmic acceleration in non-minimal Maxwell-
Lott. Time Delay of Cascade Radiation for TeV Blazars F(R) gravity and the generation of large-scale magnetic
and the Measurement of the Intergalactic Magnetic fields. Journal of Cosmology and Astroparticle Physics,
Field. The Astrophysical Journal Letters, 733(2):L21, 2008(4):024, April 2008.
June 2011. [20] David Lemoine and Martin Lemoine. Primordial
[9] Ie. Vovk, A. M. Taylor, D. Semikoz, and A. Neronov. magnetic fields in string cosmology. Phys. Rev. D,
Fermi/LAT Observations of 1ES 0229+200: Implica- 52(4):1955–1962, August 1995.
tions for Extragalactic Magnetic Fields and Background [21] M. Gasperini, M. Giovannini, and G. Veneziano. Primor-
Light. The Astrophysical Journal Letters, 747(1):L14, dial magnetic fields from string cosmology. Phys. Rev.
March 2012. Lett., 75(21):3796–3799, November 1995.
[10] Keitaro Takahashi, Masaki Mori, Kiyotomo Ichiki, [22] Kazuharu Bamba and J. Yokoyama. Large-scale mag-
Susumu Inoue, and Hajime Takami. Lower Bounds netic fields from inflation in dilaton electromagnetism.
 10

Phys. Rev. D, 69(4):043507, February 2004. Journal of Cosmology and Astroparticle Physics,
[23] Kazuharu Bamba and Misao Sasaki. Large- 2016(10):039, October 2016.
scale magnetic fields in the inflationary universe. [40] Craig J. Hogan. Magnetohydrodynamic Effects of a First-
Journal of Cosmology and Astroparticle Physics, Order Cosmological Phase Transition. Phys. Rev. Lett.,
2007(2):030, February 2007. 51:1488–1491, Oct 1983.
[24] Jérôme Martin and Jun’ichi Yokoyama. Generation of [41] Jean M. Quashnock, Abraham Loeb, and
large scale magnetic fields in single-field inflation. Journal David N. Spergel. Magnetic Field Generation
of Cosmology and Astroparticle Physics, 2008(01):025, during the Cosmological QCD Phase Transition.
2008. The Astrophysical Journal Letters, 344:L49, September
[25] Vittoria Demozzi, Viatcheslav Mukhanov, and Hec- 1989.
tor Rubinstein. Magnetic fields from inflation? [42] Baolian Cheng and Angela V. Olinto. Primordial mag-
Journal of Cosmology and Astroparticle Physics, netic fields generated in the quark-hadron transition.
2009(8):025, August 2009. Phys. Rev. D, 50:2421–2424, Aug 1994.
[26] Sugumi Kanno, Jiro Soda, and Masa-aki Watanabe. Cos- [43] Tanmay Vachaspati. Magnetic fields from cosmological
mological magnetic fields from inflation and backreac- phase transitions. Physics Letters B, 265(3-4):258–261,
tion. Journal of Cosmology and Astroparticle Physics, August 1991.
2009(12):009, December 2009. [44] K. Enqvist and P. Olesen. On primordial magnetic fields
[27] Takeshi Kobayashi. Primordial magnetic of electroweak origin. Physics Letters B, 319(1-3):178–
fields from the post-inflationary universe. 185, December 1993.
Journal of Cosmology and Astroparticle Physics, [45] Mark Hindmarsh and Allen Everett. Magnetic fields from
2014(5):040, May 2014. phase transitions. Phys. Rev. D, 58:103505, Oct 1998.
[28] Esteban A. Calzetta, Alejandra Kandus, and Fran- [46] Gordon Baym, Dietrich Bödeker, and Larry McLerran.
cisco D. Mazzitelli. Primordial magnetic fields in- Magnetic fields produced by phase transition bubbles in
duced by cosmological particle creation. Phys. Rev. D, the electroweak phase transition. Phys. Rev. D, 53:662–
57(12):7139–7144, June 1998. 667, Jan 1996.
[29] M. Giovannini and M. Shaposhnikov. Primordial mag- [47] Axel Brandenburg, Kari Enqvist, and Poul Olesen.
netic fields from inflation? Phys. Rev. D, 62(10):103512, Large-scale magnetic fields from hydromagnetic turbu-
November 2000. lence in the very early universe. Phys. Rev. D, 54:1291–
[30] A. C. Davis, K. Dimopoulos, T. Prokopec, and 1300, Jul 1996.
O. Törnkvist. Primordial spectrum of gauge fields from [48] Günter Sigl, Angela V. Olinto, and Karsten Jedamzik.
inflation. Physics Letters B, 501(3-4):165–172, March Primordial magnetic fields from cosmological first order
2001. phase transitions. Phys. Rev. D, 55:4582–4590, Apr 1997.
[31] K. Dimopoulos, T. Prokopec, O. Törnkvist, and A. C. [49] Jarkko Ahonen and Kari Enqvist. Magnetic field gen-
Davis. Natural magnetogenesis from inflation. Phys. Rev. eration in first order phase transition bubble collisions.
D, 65(6):063505, March 2002. Phys. Rev. D, 57:664–673, Jan 1998.
[32] A. D. Dolgov. Breaking of conformal invariance and elec- [50] John M. Cornwall. Speculations on primordial magnetic
tromagnetic field generation in the Universe. Phys. Rev. helicity. Phys. Rev. D, 56:6146–6154, Nov 1997.
D, 48(6):2499–2501, September 1993. [51] M. Joyce and M. Shaposhnikov. Primordial Magnetic
[33] O. Bertolami and D. F. Mota. Primordial magnetic fields Fields, Right Electrons, and the Abelian Anomaly. Phys.
via spontaneous breaking of Lorentz invariance. Physics Rev. Lett., 79:1193–1196, Aug 1997.
Letters B, 455(1-4):96–103, May 1999. [52] George B. Field and Sean M. Carroll. Cosmological
[34] Amjad Ashoorioon and Robert B. Mann. Generation magnetic fields from primordial helicity. Phys. Rev. D,
of cosmological seed magnetic fields from inflation with 62:103008, Oct 2000.
cutoff. Phys. Rev. D, 71(10):103509, May 2005. [53] Tanmay Vachaspati. Estimate of the Primordial Mag-
[35] Andrés Dı́az-Gil, Juan Garcı́a-Bellido, Margarita Garcı́a netic Field Helicity. Phys. Rev. Lett., 87:251302, Nov
Pérez, and Antonio González-Arroyo. Magnetic Field 2001.
Production during Preheating at the Electroweak Scale. [54] R. Banerjee and K. Jedamzik. The Evolution of cosmic
Phys. Rev. Lett., 100(24):241301, June 2008. magnetic fields: From the very early universe, to recom-
[36] Mohamed M. Anber and Lorenzo bination, to the present. Phys. Rev. D, 70:123003, 2004.
Sorbo. N-flationary magnetic fields. [55] Jonathan Zrake. Inverse Cascade of Nonheli-
Journal of Cosmology and Astroparticle Physics, cal Magnetic Turbulence in a Relativistic Fluid.
2006(10):018, October 2006. The Astrophysical Journal Letters, 794(2):L26, October
[37] Ruth Durrer, Lukas Hollenstein, and Rajeev Kumar Jain. 2014.
Can slow roll inflation induce relevant helical magnetic [56] Axel Brandenburg, Tina Kahniashvili, and Alexander G.
fields? Journal of Cosmology and Astroparticle Physics, Tevzadze. Nonhelical Inverse Transfer of a Decaying Tur-
2011(3):037, March 2011. bulent Magnetic Field. Phys. Rev. Lett., 114(7):075001,
[38] Tomohiro Fujita, Ryo Namba, Yuichiro Tada, Naoyuki February 2015.
Takeda, and Hiroyuki Tashiro. Consistent gen- [57] Axel Brandenburg and Tina Kahniashvili. Classes of hy-
eration of magnetic fields in axion inflation mod- drodynamic and magnetohydrodynamic turbulent decay.
els. Journal of Cosmology and Astroparticle Physics, Phys. Rev. Lett., 118(5):055102, 2017.
2015(5):054–054, May 2015. [58] A. Brandenburg, T. Kahniashvili, S. Mandal, A. R. Pol,
[39] Peter Adshead, Jr. Giblin, John T., Timo- A. G Tevzadze, and T. Vachaspati. Evolution of hydro-
thy R. Scully, and Evangelos I. Sfakianakis. magnetic turbulence from the electroweak phase transi-
Magnetogenesis from axion inflation. tion. Physical Review D, 96(12):123528, 2017.
 11

[59] Johannes Reppin and Robi Banerjee. Nonhelical turbu- dominated turbulence. arXiv preprint arXiv:2206.07513,
lence and the inverse transfer of energy: A parameter 2022.
study. Phys. Rev. E, 96(5):053105, November 2017. [66] Hantao Ji and William Daughton. Phase diagram
[60] Kiwan Park. On the inverse transfer for magnetic reconnection in heliophysical, astrophys-
of (non-)helical magnetic energy in a de- ical, and laboratory plasmas. Physics of Plasmas,
caying magnetohydrodynamic turbulence. 18(11):111207, 2011.
Monthly Notices of the Royal Astronomical Society: Letters,[67] W. Park, D. A. Monticello, and R. B. White. Recon-
472(2):1628–1640, December 2017. nection rates of magnetic fields including the effects of
[61] D. N. Hosking and A. A. Schekochihin. Reconnection- viscosity. The Physics of fluids, 27(1):137–149, 1984.
Controlled Decay of Magnetohydrodynamic Turbulence [68] R. Durrer and A. Neronov. Cosmological magnetic
and the Role of Invariants. Phys. Rev. X, 11:041005, fields: their generation, evolution and observation. The
2021. Astronomy and Astrophysics Review, 21(1):1–109, 2013.
[62] D. N. Hosking and A. A. Schekochihin. Cosmic-void ob- [69] P. Arnold, G. D. Moore, and L. G. Yaffe. Trans-
servations reconciled with primordial magnetogenesis. 3 port coefficients in high temperature gauge theories (I):
2022. leading-log results. Journal of High Energy Physics,
[63] P. A. Sweet. The neutral point theory of solar 2000(11):001, 2000.
flares. Electromagnetic phenomena in cosmical physics, [70] K. Kamada, F. Uchida, and J. Yokoyama. Baryon
6(1):123, 1958. isocurvature constraints on the primordial hypermag-
[64] E. N. Parker. Sweet’s Mechanism for Merging Magnetic netic fields. Journal of Cosmology and Astroparticle
Fields in Conducting Fluids. Journal of Geophysical Physics, 2021(04):034, 2021.
Research, 62(4):509–520, 1957. [71] F. Uchida, M. Fujiwara, K. Kamada, and J. Yokoyama.
[65] H. Zhou, R. Sharma, and A. Brandenburg. Scaling of in preperation.
the Saffman helicity integral in decaying magnetically-