Spontaneous electromagnetic superconductivity of vacuum in strong magnetic field:

an evidence from the Nambu–Jona-Lasinio model

M. N. Chernodub∗
CNRS, Laboratoire de Mathématiques et Physique Théorique, Université François-Rabelais Tours,
Fédération Denis Poisson, Parc de Grandmont, 37200 Tours, France and
Department of Physics and Astronomy, University of Gent, Krijgslaan 281, S9, B-9000 Gent, Belgium
(Dated: December 30, 2010)
Using an extended Nambu–Jona-Lasinio model as a low energy effective model of QCD, we sup-
port our earlier proposal that the QCD vacuum in a strong external magnetic field (stronger than
1016 Tesla) experiences a spontaneous phase transition to an electromagnetically superconduct-
ing state. The unexpected superconductivity of, basically, empty space is induced by emergence
of quark-antiquark vector condensates with quantum numbers of electrically charged rho mesons.
The superconducting phase possesses an anisotropic inhomogeneous structure similar to a periodic
arXiv:1101.0117v1 [hep-ph] 30 Dec 2010

Abrikosov lattice in a type-II superconductor. The superconducting vacuum is made of new type of
vortices which are topological defects in the charged vector condensates. The superconductivity is
realized along the axis of the magnetic field only. We argue that this effect is absent in QED.

PACS numbers: 12.38.-t, 13.40.-f, 74.90.+n

It is known that strong magnetic fields may lead to
unusual effects such as magnetic catalysis in QED and
QCD [1, 2], the chiral magnetic effect in hot quark-gluon
plasma [3], and metallic-like conductivity in a quark-
less vacuum of lattice SU(2) Yang-Mills theory [4]. Re-
cently, we suggested in Ref. [5] that an interplay between
strong and electromagnetic interactions in a background
of a sufficiently strong magnetic field may turn the cold
vacuum into an electromagnetic superconductor if the
strength of the magnetic field exceeds
Bc = m2ρ /e ≈ 1016 Tesla , (1)
where mρ = 775.5 MeV is the mass of the ρ meson and e
is the elementary electric charge.
Our idea is based on a very simple argument: the parti-
cle spectrum of QCD contains a charged vector resonance
(a spin-triplet excitation), ρ± meson, which has a large
magnetic dipole moment associated with an anomalous
gyromagnetic ratio g = 2 of the ρ meson. If one treats
the ρ meson as a free particle, then in a background of a
uniform magnetic field B its ground state energy [corre-
sponding to the Lowest Landau Level (LLL)] becomes
a decreasing function of the magnetic field strength,
Eρ2± (B) = m2ρ± − eB. The energy of ρ± vanishes when
the magnetic field reaches the value (1). As the field
strength increases further, the ground state energy Eρ±
becomes purely imaginary indicating a tachyonic insta-
bility of the ground state towards condensation of the ρ
mesons. Since ρ± are electrically charged, their conden-
sation implies an electromagnetic superconductivity of
the new ground state. Surprisingly, there is no Meissner
effect in the g = 2 case. Moreover, the strong magnetic
field makes the ρ mesons stable at QCD timescales [5].
The suggested vacuum superconductivity has at least
two other analogues in particle physics: the Nielsen-
Olesen instability of the gluonic vacuum in Yang-Mills
theory [6], and the Ambjørn–Olesen condensation of the
W -bosons induced by a strong magnetic field in the stan-
dard electroweak model [7].
In condensed matter physics, a similar phenomenon is
known as “reentrant” superconductivity [8]. Usually, an
external magnetic field suppresses the superconductivity
via pair breaking effects, so that in a strong magnetic field
the superconductivity is lost. However, there are super-
conductors which may reenter the superconducting phase
again at stronger magnetic fields. For example, in the
uranium compound URhGe the superconductivity disap-
pears at 2 Tesla, and then, unexpectedly, a stronger (pre-
sumably, spin-triplet) superconductivity reappears again
at strikingly large magnetic field of about 8 Tesla [9].
There are various material-dependent proposals to de-
scribe specific reentrant superconductors in the con-
densed matter physics. Our suggestion in QCD [5] is
close to the idea of Refs. [8, 10] that in a very strong
magnetic field the Abrikosov flux lattice of a type-II su-
perconductor may exhibit a “reentrant” quantum regime,
characterized by the LLL dominance, the absence of the
Meissner effect, a spin-triplet pairing, and a supercon-
ducting current flow along the magnetic field axis only.
In Ref. [5] we found the new superconducting phase us-
ing an effective bosonic electrodynamics of the ρ mesons
of Ref. [11]. Here we use much more general fermionic
Nambu–Jona-Lasinio (NJL) model [12] as a low energy
effective theory of QCD in order to confirm the existence
of the electromagnetic superconductivity induced by the
strong magnetic fields in the vacuum.
We consider an extended two-flavor (Nf = 2) Nambu-
Jona-Lasinio model with three colors (Nc = 3) [13]:
/ − M̂ 0 ψ + L(4)
L(ψ, ψ̄) = ψ̄ i∂/ + Q̂ A

(4)
S + LV , (2)
where the light quarks are represented by the doublet
ψ = (u, d)T and M̂ 0 = diag(m0u , m0d ) is the correspond-
 2

ing bare mass matrix. The uniform magnetic field back- Next, we calculate the effective action (7) in the strong
ground B~ = (0, 0, B) is encoded in the Abelian gauge field magnetic field background in the mean field approach.
A ≡ (A0 , A)
µ ~ = (0, −Bx2 /2, Bx1 /2, 0), and the electric We use simplified notations for the expectation values of
charges of the quarks, qu = +2e/3 and qd = −e/3, are the fields, hσi = σ etc. In the absence of the external
combined into the matrix Q̂ = diag(qu , qd ). The hat over magnetic field the expectation values of the fields ~π , V ,
a symbol indicates a 2 × 2 matrix in the flavor space. and A are zero [13], while the expectation value of σ plays
The last two terms in Eq. (2) represent the scalar and a role of the constituent quark mass, mq = σ ∼ 300 MeV.
vector four-quark interactions, respectively: In order to simplify our calculations, we notice that
the presence of the external magnetic field breaks the
(0)
(4) GS 
2
2  flavor symmetry down to its diagonal subgroup, so that
LS = ψ̄ψ + ψ̄iγ 5~τ ψ , (3) 3
2 the diagonal chiral rotations Ω = eiα5 τ γ5 can still be
(0)
G X3 h
2
2 i used to eliminate the neutral pion condensate π 0 . We
(4)
LV =− V ψ̄γµ τ i ψ + ψ̄γµ γ5 τ i ψ ,(4) also neglect the mass matrix M 0 because our calculations
2 i=0
show that the conducting properties of the vacuum are
(0) (0)
where GS and GV are corresponding bare couplings, almost independent on the bare quark masses m0u,d  σ.
and ~τ = (τ 1 , τ 2 , τ 3 ) are the Pauli matrices The operator (9) can be represented as the sum iD =
We follow the standard approach [13] and introduce iD0 + Ŵ of the tree-level operator iD0 = i∂/ + Q̂ A / −σ
the following bosonic fields corresponding to the quark- and the contribution Ŵ from the “exotic” condensates,
antiquark bilinears: one scalar field σ ∼ ψ̄ψ, the triplet
Ŵ = V/̂ µ + γ 5 A
/̂ − iγ 5 (π 1 τ 1 + π 2 τ 2 ) , (10)
of three pseudoscalar fields ~π ∼ ψ̄γ 5~τ ψ [made of the elec-
0 3
√ π ≡ π , and electrically charged,
trically neutral, π± = At low magnetic fields Ŵ ≡ hŴ i = 0. Let us assume,
(π ∓ iπ )/ 2, pions], four vector fields Vµ ∼ ψ̄γµ τ i ψ,
1 2 i
that at certain strong magnetic field B = BcNJL the ex-
and four pseudovector (axial) fields Aiµ ∼ ψ̄γ 5 γµ τ i ψ, pectation value of the condensate (10) gets nonzero. Let
√ +  us advance slightly into the new phase taking B & BcNJL ,
0

X3
i i ω√
µ + ρµ 2ρµ so that the magnitude the suspected condensate is still
V̂µ ≡ τ Vµ = , (5)
i=0 2ρ−
µ ωµ − ρµ
0
small, 0 < |Ŵ |  σ. Then the effective action (7) can
0
√ be expanded in powers of the Ŵ field, and the fact of
2a+
 
X3 f√
µ + aµ
µ ≡ τ i Aiµ = µ
. (6) the emergence of the new condensate should be seen as
i=0 2a−
µ fµ − a0µ
a tachyonic instability of the corresponding potential at
The vector-meson matrix (5) is composed of the singlet Ŵ = 0 (alternatively, we could also solve Schwinger-
(in the flavor space) vector (in the coordinate space) √ ω– Dyson gap equations and get the same result).
meson field ωµ , while ρ0µ ≡ ρ3µ and ρ± 1 2
µ = (ρµ ∓ iρµ )/ 2
The tree-level propagator S (0) ≡ D0−1 of the fermion
represent, respectively, electrically neutral and charged doublet in the strong magnetic field has the following
 (0) (0) 
components of the ρ-meson triplet. The light axial form: S (0) (x, y) = diag Su (x, y), Sd (x, y) , where Sf
mesons are encoded in the matrix (6): the fields fµ and is the propagator of the f th quark species.
(a0µ , a±µ ) represent, respectively, the singlet axial f1 me- The ρ-meson condensation and, consequently, the in-
son and the ~a1 triplet of the axial mesons, respectively. duced superconductivity are the LLL phenomena [5].
We rewrite the four-quark interactions (3) and (4) via Thus, it is natural to restrict ourselves to the LLL ap-
Gaussian integrals over the bosonic fields σ, ~π , V̂µ , µ , proximation which usually gives a dominant contribution
and integrate over the quarks in the partition function: to nonperturbative low-energy quantities in the limit of
Z R 4
Z the strong magnetic field (in particular, because of the
Z= Dψ̄Dψ ei d x L = DσDπDV DA eiS[σ,~π,V,A] , large energy gap δE ∼ (eB)1/2 between the LLL and
higher levels) [1, 2, 14]. The full solution beyond the
where the effective bosonic action is LLL approximation will be presented elsewhere.
(0)
In the LLL regime the propagator Sf factorizes into
S[σ, ~π , V, A] = Sψ [σ, ~π , V, A] (7) the B-transverse and B-longitudinal parts which depend,
separately, on the B-transverse, x⊥ = (x1 , x2 ), and B-
Z h 1 1 i
+ d4 x − (0) (σ 2 + ~π 2 ) + (Vµk V kµ + Akµ Akµ )
2GS 2GV
(0) longitudinal, xk =(x0 , x3 ), coordinates [1]:
(0),LLL k
Sψ = −iNc Tr Ln(iD) , (8) Sf (x, y) = Pf⊥ (x⊥ , y ⊥ ) Sf (xk − y k ) (11)
0 5 5
iD = i∂/ + Q̂ A
/ − M̂ + V/̂ µ + γ A
/̂ − (σ + iγ ~π~τ ) . (9) (below we omit the superscripts “(0)” and “LLL”). Here
The sources JO of the mentioned quark-antiquark bilin- |qf B| i qf Bεab xa xb − 1 |qf B|(x⊥ −y⊥ )2
Pf⊥ (x⊥ , y ⊥ ) = e2 4 , (12)
ears can be introduced via the shifts O → O − JO of the 2π
corresponding fields O = σ, ~π , V, A in the second line of is the transverse projector onto the LLL states and qf is
Eq. (7), so that hψ̄ψi = −hσ/GS i etc. the electric charge of the f th quark.
 3

The longitudinal part of the fermion propagator (11), We find that the potential (15) involves only the B-
k k transverse components of the vector and axial mesons,
Sf ≡ Ssgn(qf B) is, basically, a fermion propagator in the
1+1 dimensions (we always take eB > 0 for definiteness), 2Nc |eB| h 1 σ2
Z 
(2)
d2 x⊥ Vψ = − − ln + 1 (φ∗ ◦ Pe ◦ φ)
i 1l − if γ 1 γ 2 9π 2  µ2
k k k
Sf (kk ) = Pf , Pf = , (13) 1 σ2
γ k kk
 i
−m 2 + − ln 2 − 1 (ξ ∗ ◦ Pe ◦ ξ) + c.c. , (16)
 µ
k
and the matrix Pf (we use f = ±1 for, respectively,
where φ = (ρ+ + +
1 + iρ2 )/2 and ξ = (a1 + ia2 )/2.
+
f = u, d) is the spin projector operator onto the fermion
The B-transverse projector for unit charged particle,
states with the spin polarized along (for u quarks) or op-
Pe⊥ (x⊥ , y ⊥ ) = (9π/|eB|)Pu⊥ (x⊥ , y ⊥ )Pd⊥ (y ⊥ , x⊥ ) is given
posite (for d quarks) to the magnetic field. The operator
k by Eq. (12) with qf → e.
Pf projects the original four 3+1 fermionic states onto The unstable tachyonic mode of the potential (7),
two 1+1 dimensional fermionic states, so that fermions (16) turns out to be an inhomogeneous eigenstate of the
can move only along the axis of the magnetic field. The charge-1 projection operator Pe ,
projector (12) satisfies the relation Pf⊥ ◦Pf⊥ = Pf⊥ , where
”◦” is theRconvolution operator in the B-transverse space, (Pe ◦ φ)(x⊥ ) = φ(x⊥ ) , (17)
A ◦ B ≡ d2 y ⊥ A(. . . , y ⊥ ) B(y ⊥ , . . . ). The solution is a general Abrikosov-like configuration [16]
For a coordinate–independent condensate σ, the zero-
p
order (in powers of Ŵ ) contribution to the effective φ = φ0 K z̄/LB , LB = 2π/|eB| , (18)
(0)
action (8) gives us the potential V (σ) = Vψ (σ) + π 2 2 X+∞ 2

(0)
K(z) = e− 2 (|z| +z ) cn e−πn +2πnz , (19)
σ 2 /(2GS ) related to the action as S = − d4 x V with
R n=−∞

where φ0 and cn are arbitrary complex parameters and

(0) |eB|Nc h 2 σ 2  1  i
z = x1 + ix2 (and similarly for the axial vector field ξ).
Vψ = iNc Tr Ln iD0 = σ ln − + 1 σ2 ,
8π 2 µ2  The solution (18) represents a (periodic) flux-tube
where 1/ = 1/−γE +log 4π, and γE ≈ 0.57722 is Euler’s structure similar to the Abrikosov lattice which is realized
constant, and µ is a renormalization mass scale. In or- in a mixed state of a type-II superconductor subjected to
der to regularize the divergent contributions of the 1 + 1 a near-critical external magnetic field [16]. Generally, the
dimensional fermions we implemented the dimensional coefficients cn can be fine-tuned by a complicated min-
regularization in d = 2 − 2 dimensions. The renormal- imization procedure if the full potential is known [16].
ization of the NJL coupling constant in the MS scheme, Here we follow Ref. [5, 7] and set cn = 1 so that the so-
(0)
1/GS = 1/GS − Nc |eB|/(4π 2 ), resembles the renor- lution (18) represents a square lattice with the quantized
malization of the Gross-Neveu (GN) model [15] with the area 2π/|eB| ≡ L2B given by the magnetic length LB .
identification GS ≡ 2πGGN /(Nc |eB|), first noticed in [1]. The quadratic potential, evaluated at the solution (18),
The minimum σ = σmin of the renormalized potential, √  1 Nc |eB|  √  1 Nc |eB|  2
V (2) = 2 − |φ 0 |2
+ 2 + |ξ0 | ,
1 2 |eB|Nc  σ 2  GB 9π 2 GB 9π 2
V (σ) = σ + ln − 1 σ2 ,
2GS 8π 2 µ2 is unstable towards a spontaneous creation of the B-
transverse ρ± condensates with the tachyonic mode ρ+ 1 =
provides us with the B-dependent quark mass
iρ+
2 = φ if the strength of the magnetic field exceeds
mq (B) = σmin (B) = µ exp{−2π 2 /(GS Nc |eB|)} .(14)
9π 2 1 1 4
BcNJL = , = − . (20)
In the LLL approximation to the NJL model the scale µ e Nc GB GB GV 9GS
is not fixed as it is related to the B-longitudinal 1 + 1 (0)
motion of the quarks. Beyond the LLL approach the The quantity GB ≡ GB is a combination of the NJL
scale may perhaps be set as µ2 ∝ |eB| following Ref. [2]. couplings which is not affected by the LLL renormaliza-
The effective bosonic model [5] predicts that the pos- tion because the LLL fermions renormalize the vector
(0)
sible superconducting ground state should exhibit an in- coupling as follows: 1/GV = 1/GV − Nc |eB|/(9π 2 ).
homogeneous behavior in the B-transverse plane. Thus, Since the phenomenological values of the parameters
we assume that the exotic condensates may be x⊥ - GS,V vary in a broad region [17], we can only give an ap-
dependent, Ŵ = Ŵ (x⊥ ), and calculate the correspond- proximate estimation of the critical field: eBc ∼ 1 GeV2 .
ing quadratic contribution to the effective action (8), The quartic correction to the potential in Eq. (8),
Z
iNc 1 1 (4) |eB|Nc
(2)
Sψ = − d4 x Vψ =
(2)
Tr W W , (15) Vψ = C0 |φ0 |4 , (21)
2 iD0 iD0 2π 2 m2
allows us to find the condensate at B > BcNJL :
(despite the expected x⊥ –dependence of the condensates
(2)
1/2
Ŵ , we still call the functional Vψ as the “potential”). φ0 (B) = eiθ0 Cφ mq (B) 1 − BcNJL /B , (22)

where θ0 is a constant phase, C0 ≈ 1.2, Cφ ≈ 0.51 and
the quark mass mq is given in Eq. (14). At B < BcNJL the
condensate (22) is zero. The phase transition at B = Bc
is of the second order with the critical exponent 1/2.
Thus, the magnetic field induces the quark condensate
 x + ix 
1 2
hūγ1 di = −ihūγ2 di = ρ0 (B) K ≡ ρ(x⊥ ), (23)
LB
where ρ0 (B) = φ0 (B)/GV . Using known (see, e.g., Ref.
[16]) general properties of the function K(z), Eq. (19),
we conclude that the ground state should be given by
a periodic (in general) lattice of new type of topological
vortices which are parallel to the magnetic field. The
phase of the condensate (23) winds around the center of
each vortex where the absolute value of ρ(x⊥ ) vanishes.
The condensate (23) locks the local U (1)e.m. trans-
formations with the global O(2)rot rotations of the co-
ordinate space about the magnetic field axis [5, 18]:
U (1)e.m. × O(2)rot → Glat , where Glat is a discrete sym-
metry group of rotations of the ρ-vortex lattice.
The bosonic ρ-meson electrodynamics predicts that the
new vacuum state is superconducting [5]. In order to
check this property in the NJL model we apply a very
weak (test) electric field E~ to the new vacuum state (23)
and then we calculate the induced electric current
X
J µ (x) = qf hψ̄f γ µ ψf i ≡ −tr[γ µ Q̂S(x, x)] , (24)
f =u,d
in a linear-response approximation using retarded Green
functions. Since the B-transverse electric field may be
removed by a Lorentz boost, we consider a constant B-
~ = (0, 0, Ez ) with Ez  B. There is
longitudinal field E
no B-transverse current, J 1 =J 2 =0, because the electric
current flows along the magnetic field axis only. We aver-
age the current (24) over the B-transverse plane and we
get in the leading order in powers of the ρ condensate:
∂Q(xk ) ∂J (xk ) 2Cq 3
e B − BcNJL Ez ,


+ = 3
(25)
∂t ∂z (2π)
where Q is the plane-averaged electric charge density J 0 ,
and J is the plane-averaged current J z , and Cq ≈ 1 [19].
At B < Bc the right hand side of Eq. (25) is zero. Apart
from prefactors, the transport laws in the NJL model (25)
and in the ρ-meson electrodynamics [5] are identical.
The linear response law (25) can be rewritten in a
Lorentz-covariant form, ∂ [µ, J ν] = γ · (F, Fe)Feµν , via the
~ E)
invariants (F, Fe) = 4(B, ~ and (F, F ) = 2(B ~2 − E ~ 2 ).
αβ
Here Fµν = µναβ F /2 and γ is a function of (F, F ) [18].
e
Equation (25) is a London equation for an anisotropic
superconductivity. Thus, we have just shown that the
strong magnetic field induces the new electromagneti-
cally superconducting phase of the vacuum if B > Bc .
An empty space becomes an anisotropic superconductor.
The superconductivity of the vacuum is a new effect
which is realized at the QCD-QED interface. The de-
4
scribed mechanism should not work in QED alone since
electrically charged spin-1 bound states are absent there.
So far we have discussed the zero temperature case. On
general grounds one can expect that increase in tempera-
ture T should lead to an evaporation of the ρ condensate
with a loss of the superconductivity. The suggested low-
T part of the B − T phase diagram is shown in Fig. 1.
T Hadronic phase nsition
cti ng tra
condu
Super Phase of
electromagnetic
superconductivity
0 Bc B
FIG. 1. Low-temperature part of the QCD phase diagram.
The author is grateful to A. Niemi, M. Ruggieri and
V. I. Zakharov for useful discussions, and to P. Olesen for
interesting correspondence. The work was partially sup-
ported by the grant ANR-10-JCJC-0408 HYPERMAG.
∗
On leave from ITEP, Moscow, Russia.
[1] V.P. Gusynin, V.A. Miransky, I.A. Shovkovy, Phys. Rev.
Lett. 73, 3499 (1994); Nucl. Phys. B 462, 249 (1996).
[2] V. A. Miransky, I. A. Shovkovy, Phys. Rev. D 66, 045006
(2002).
[3] K. Fukushima, D. E. Kharzeev and H. J. Warringa, Phys.
Rev. D 78, 074033 (2008).
[4] P.V.Buividovich et al, Phys. Rev. Lett. 105, 132001 (2010).
[5] M. N. Chernodub, Phys. Rev. D 82, 085011 (2010).
[6] N. K. Nielsen, P. Olesen, Nucl. Phys. B 144, 376 (1978).
[7] J. Ambjorn, P. Olesen, Nucl. Phys. B 315, 606 (1989);
ibid. 330, 193 (1990); Phys. Lett. B 218, 67 (1989).
[8] M. Rasolt, Z. Tešanović, Rev. Mod. Phys. 64, 709 (1992).
[9] F. Lévy, I. Sheikin, B. Grenier, A. D. Huxley, Science
309, 1343 (2005); D. Aoki et al., J. Phys. Soc. Jpn. 78
(2009) 113709; arXiv:1012.1987.
[10] M. Rasolt, Phys. Rev. Lett. 58, 1482 (1987);
Z. Tešanović, M. Rasolt, L. Xing ibid., 63 2425 (1989).
[11] D. Djukanovic, M. R. Schindler, J. Gegelia and
S. Scherer, Phys. Rev. Lett. 95, 012001 (2005).
[12] Y. Nambu, G. Jona-Lasinio, Phys. Rev. 122, 345 (1961).
[13] D. Ebert, H. Reinhardt, Nucl. Phys. B 271, 188 (1986).
[14] N. Sadooghi, A.J. Salim, Phys. Rev. D 74, 085032 (2006);
S. Fayazbakhsh, N. Sadooghi, ibid. 82, 045010 (2010).
[15] D. J. Gross and A. Neveu, Phys. Rev. D 10, 3235 (1974).
[16] A.A. Abrikosov, Fundamentals of the Theory of Metals
(North Holland, Amsterdam, 1988).
[17] See, e.g., V. Bernard et al, Annals Phys. 249, 499 (1996).
[18] M. N. Chernodub, arXiv:1011.2658 [hep-ph].
[19] The presence of the derivatives in the left hand side of
the 1 + 1 dimensional equation (25) allows us to bypass
the LLL anomaly problem found in E.V. Gorbar, M. Ha-
shimoto, V.A. Miransky, Phys. Lett. B 611, 207 (2005).