arXiv:cond-mat/0101446v2 [cond-mat.supr-con] 18 Feb 2001
Superconductivity of metallic boron in MgB
2
J. Kortus
†
, I.I. Mazin
†
, K.D. Belashchenko,
‡
V.P. Antropov
‡
, L.L. Boyer
†
†
Center for Computational Materials Science, Code 6390, Naval Research Laboratory, Washington, DC 20375
‡
Ames Laboratory, ISU, Ames, IA, 50011
(February 1, 2008)
Boron in MgB
2
forms layers of honeycomb lattices with magnesium as a space filler. Band struc-
ture calculations indicate that Mg is substantially ionized, and the bands at the Fermi level derive
mainly from B orbitals. Strong bonding with an ionic component and considerable metallic density
of states yield a sizeable electron-phonon coupling. Using the rigid atomic sphere approximation
and an analogy to Al, we estimate the coupling constant λ to be of order 1. Together with high
phonon frequencies, which we estimate via zone-center frozen phonon calculations to be between
300 and 700 cm
−1
, this produces a high critical temperature, consistent with recent experiments.
Thus MgB
2
can be viewed as an analog of the long sought, but still hypothetical, superconducting
metallic hydrogen.
Before the discovery of high-temperature supercon-
ductors much effort was devoted to boosting the criti-
cal temperature of conventional, BCS-Eliashberg super-
conductors [1]. An exo tic and appealing idea go ing
back to the early 60’s, was that of metallic hydrogen
[2]. The arguments were very simple: due to the light
mass, the phonon frequencies in metallic hydrogen would
be very high, of the o rder of several thousand K, and
the prefactor in the BCS formula would be very lar ge,
so that even a moderate coupling constant would pro-
vide a size able T
c
. This idea can be quantified as fol-
lows: for monoatomic solids, the electr on-phonon cou-
pling (EPC) constant, λ, which enters the BCS equation,
can be wr itten in the so-called McMillan-Hopfield form
[3], λ = N(0)
I
2
/M
ω
2
, where N(0) is the density
of states (DOS) at the Fermi level per spin per atom,
I
2
is the pr operly averaged electron-ion ma trix ele-
ment squared, M is the atomic mass and
ω
2
is (again,
properly averaged) the phonon frequency. The product
M
ω
2
does not depend on the mass, but on the force
constants only [1], while η = N(0)
I
2
, also known as
the Hopfield factor, is a purely elec tronic property. Cor-
respondingly, light elements, everything else being the
same, are beneficial for superconductivity.
Lacking metallic hydrogen, attention was focused upon
compounds with light elements: car bides , nitrides (ar-
guably, the superconductivity in fullerenes was a discov-
ery along this road). Indeed, many of them were ‘high-T
c
supe rconductors’ on the contemporary scale: 10-15 K. It
was pointed out [4] tha t
I
2
is rather large in these ma-
terials due to the relatively high ionicity (although not as
high as in MgB
2
), but N (0) is rather small. This led to
the suggestion of cubic MoN, which would have a larger
N(0) than existing nitrides and carbides, as a hypothet-
ical superconductor with T
c
> 30 K [5].
The recently discovered medium-T
c
supe rconductor
MgB
2
[6] with T
c
>
∼
39 K is clearly a c ontinuation of the
same idea. The main component, B, is even lighter than
C and N. Furthermore, electronic structure c alculations
show tha t the co mpound is not only quite ionic with a
reasonable DOS, but als o has strong covalent B-B bond-
ing (the bonding-antibonding splitting due to in-plane
B-B hopping is about 6 eV) and thus exhibits strong
electron-phonon interactions. Interestingly, unlike car-
bides and nitrides , and similar to metallic hydrogen, elec-
trons at the Fermi level (and below) are predominantly B-
like. Mg s-states are pushed up by the B p
z
orbitals and
fully donate their electrons to the boron-derived conduc-
tion bands. In the following we will describe the physics
of such ‘metallic’ boron in detail, pr e sent an estimate of
the EPC constant, and propose some routes for optimiz-
ing T
c
in this kind o f compound.
MgB
2
occurs in the so-called AlB
2
structure. Borons
form a primitive honeycomb lattice, consisting of
graphite-type s hee ts stacked with no displacement. The
borons form hexagonal pr isms with the base diameter
of 3.5
˚
A nearly equal to the height. This creates large,
nearly spherical pores for Mg. As in graphite, the intra-
planar B-B bonds are much shorter than the distance be-
tween the planes, and hence the B-B bonding is strongly
anisotropic. However, the interplane bonds are only twice
as long a s the intraplane ones, as compared to the ratio
of 2.4 in graphite, allowing for a significa nt interplane
hopping.
We have calculated the electronic structure of MgB
2
using a general potential LAPW code [7]. For the
Rigid Atomic Spheres calculations we used the Stuttgart
LMTO-TB code [8]. For the exchange-correlation poten-
tial, the Generalized Gradient Approximation of Ref. [9]
was employed. Despite the rather simple crystal struc-
ture, very few electronic structure calculations for MgB
2
have been rep orted (a model TB calculation of Burdett
and Miller [10] and a recent full-potential LMTO study
[11]), and these have concentrated mainly on chemical
bonding, paying hardly any attention to transport and
electronic prop e rties. The results of our LAPW calcula-
tions ar e shown in Figs. 1 and 2. We note first that there
is almost no valence charge inside the Mg MT sphere (less
than 0.2 e). About half of the total valence charge resides
inside the B spheres, and about the same amount in the
1