You are on page 1of 4

commentary

A new frontier
for superconductivity
Ivan Bozovic and Charles Ahn
Monolayer films of iron selenide deposited on strontium titanate display signatures of superconductivity
at temperatures as high as 109K. These recent developments may herald a flurry of exciting findings
concerning superconductivity at interfaces.

ince the discovery of superconductivity


in mercury a century ago, scientists
have searched for new and better
superconducting materials. Better can
be gauged by several criteria, but for a
superconductor the key figure of merit is
usually the critical temperature (TC) up
to which it superconducts the higher
thebetter.
As has been pointed out by others1,
simple materials such as elemental metals
and stoichiometric binary compounds
have already been explored extensively.
However, there is much more room left
for exploration as a result of the virtually
inexhaustible parameter space of complex
compounds and solid solutions allowed
by the periodic table. Cuprates are
perhaps the best known among complex
superconducting materials, because of
their high-temperature superconductivity
(HTS). With a record TCof133K in
HgBa2Ca2Cu3O9 (at ambient pressure),
cuprates currently outshine all other
materials discovered. Although TC
a

Bulk superconductor

Interfacial superconductor

Cuprate

140

140

120

120

100

100

80
Pnictide

60
MgB2

40
20
0

FeSe

Tc (K)

Tc (K)

is not the only important factor for


technological success2, cuprates have
also been at the forefront of research and
development aimed at both high-power
applications (cables, magnets, motors,
fault current limiters and so on) and
low-power devices (SQUIDs and other
superconductiveelectronics).
From a fundamental point of view, the
yearning to understand the mechanism of
HTS in cuprates has been an apparently
inexhaustible source of inquiry and
speculation. Most researchers believe that
the standard phonon-based Bardeen
CopperSchrieffer (BCS) theory which
assumes that electrons form pairs due to
a weakly attractive effective interaction
that originates from electronphonon
coupling does not apply, but this is
where the consensus ends. An alternative
complete explanation is still being sought,
and nearly 30years after the discovery of
HTS in cuprates, its origin remains one of
the foremost open problems in condensedmatterphysics.

FeSe
SrTiO3

80

La2CuO4+

60

La2xSrxCuO4

40
20
0

LaAIO3
SrTIO3

Figure 1 | a,b, The critical temperature (TC) of superconducting materials discussed in the text.
Single-phase bulk compounds (a) and bilayer heterostructures (b) showing interface superconductivity.
The values shown for bulk pnictides and cuprates are the maximal values of TC attained so far within the
two respective families of compounds, at normal pressure. The TC values shown for heterostructures are
taken from ref.9 for LaAlO3/SrTiO3, from refs5 and 6 for La2CuO4+/La2xSrxCuO4, and from ref.10 for
monolayer FeSe/SrTiO3.
892

Iron-based superconductors represent


another important family of compounds
that have been making headlines in the past
decade. Many of these compounds contain
arsenic, which is one of the pnictogen
elements, and thus these compounds are
referred to as pnictides. Many of them
are superconductors, and the highest TC
has been observed in Sr0.5Sm0.5FeAsF, at
56K(ref.3).
A key question is whether pnictides share
the same mechanism of superconductivity
as cuprates. Here, opinion is divided. One
faction points to the many similarities of
copper- and iron-based superconductors:
they all have a layered structure, are
relatively bad metals, and are prone
to other instabilities such as spin- and
charge-density waves (SDW,CDW), as
well as antiferromagnetic ordering. In
addition, they are both considered to be
exotic superconductors, characterized
by a non-trivial symmetry of the
superconducting order parameter. In
contrast to simple superconductors in
which the order parameter is spherically
symmetric (reminiscent of the s-orbital
of an atom), cuprates display d-wave
symmetry, and pnictides are generally
believed to have s symmetry (which
may occur if two electron bands cross
the Fermilevel and the phase of the
superconducting order parameter changes
sign between two concentric Fermi surface
sheets). The other camp argues that
pnictides do not derive from an insulator
parent compound and are not strongly
correlated. Most importantly, the maximum
TC observed in pnictides is nearly three
times lower than in the cuprates (Fig.1a)
and just slightly higher than in MgB2, which,
like the pnictides, is a multiband metal but
is generally believed to be a conventional
phonon-based superconductor. As we will
see below, this last argument may fall flat in
the face of newdiscoveries.

NATURE PHYSICS | VOL 10 | DECEMBER 2014 | www.nature.com/naturephysics

2014 Macmillan Publishers Limited. All rights reserved

commentary

Monolayer iron selenide

Two years ago, researchers in China reported


the synthesis and observation of HTS in
monolayer FeSe/STO (ref.11). Striking
scanning tunnelling microscopy (STM)
images testified to meticulous STOsubstrate
preparation using a novel method whereby
it was cleaned by being heated to 950C
and then exposed to a flux of selenium.

b
Se

KBaFe2As2

Fe
O

Energy

Ti

Momentum

Sr

ky (/a)

Half a century ago, VitalyGinzburg


proposed4 that in an otherwise insulating
material a near-surface layer can
acquire metallic character and become
superconducting if the electrons in surface
states partially fill the surface bands.
Alternatively, in metallic samples, the
pairing (attractive) interaction between
carriers may occur only near the surface,
due to surface phonons, variation in
screening or other factors, even if in the
bulk the interaction is repulsive. In the past
decade, these ideas have found experimental
validation, through discoveries of interface
superconductivity in certain oxide
heterostructures. In one such demonstration,
a bilayer film comprised of lanthanum
cuprate was fabricated with one layer
doped with Sr and the other without Sr but
containing some extra (interstitial) oxygen5.
These displayed a TC of 52K, which was25%
higher than had been previously observed in
the lanthanum cuprate system. Later it was
shown that this enhanced superconductivity
originates from a single CuO2 plane near
the interface6,7 and that TC scales linearly
with the lattice expansion along the
c-axis8, but the detailed mechanism of
interfacial enhancement remained obscure.
Interface superconductivity has also been
observed in lanthanum cuprate bilayers
in which neither constituent material was
superconducting perse one was metallic
and the other insulating but in this case
TC was somewhat lower (3540K; ref. 6).
Subsequently, interface superconductivity
was also discovered9 at the interface of
epitaxial LaAlO3 and SrTiO3. In this case,
both constituent materials are insulators.
However, a high-mobility two-dimensional
electron gas forms at the interface and
becomes superconducting belowTC=0.3K.
It is with this backdrop that one must
consider the recent results on FeSe films
with a thickness of one unit cell grown on a
single-crystal SrTiO3 substrate (FeSe/STO),
which, it is now claimed, have a remarkable
TC of 109K (ref.10). Assuming this claim
stands the test of time, this represents a
record TC for interface superconductivity
with any combination of materials (Fig.1b).
As we argue below, it could also become a
game-changer.

0.5

One-unit-cell FeSe
Energy

Interface superconductivity

0
Momentum

0.5
kx (/a)

Figure 2 | a, A ball-and-stick model of an FeSe/SrTiO3 heterostructure. b, A sketch of the electron


band structure near the Fermi energy in a typical pnictide, KBaFe2As2 (top), as compared with that in
monolayer FeSe/SrTiO3 (bottom). Adapted from ref.24. c, The Fermi surface of monolayer FeSe/SrTiO3,
as determined by angle-resolved photoemission spectroscopy. Adapted from refs 14 and 18.

The films also appeared atomically


smooth under STM, except for occasional
monolayer-high steps, which originate
from a slight miscut of the STO polished
surface. Measurements of resistance as a
function of temperature showed the onset of
superconductivity at about 53K. This result
is striking because bulk FeSe only becomes
superconducting below 8K (although TC
can be increased to 35K by applying a large
hydrostaticpressure).

Basic characterization

The structure of monolayer FeSe/STO


heterostructure is illustrated schematically
in Fig.2a. The in-plane lattice constant of
FeSe (a0=3.77) is significantly smaller
(3.6%) than that of STO (a0=3.905), so
the film is under substantial tensile strain.
However, this strain relaxes quickly with
increasing FeSe film thickness, and thicker
films are almost completely relaxed.
Ex situ transport measurements revealed
a broad superconducting transition
beginning at 53K, but only reaching
zero resistance at 20K. This latter value
is consistent with subsequent magnetic
measurements that employed the two-coil
mutual inductance technique12 and showed
diamagnetic screening below 21K. Without
entering the semantic debate about which
of the two values should be properly called
the critical temperature, we simply note
that in any case these data show a dramatic
interfacial enhancement of TC with respect
to bulkFeSe.
Beyond these basic facts, almost all we
know about FeSe/STO heterostructures
comes from STM and angle-resolved
photoemission spectroscopy (ARPES)
studies1318. This is an interesting reversal of

NATURE PHYSICS | VOL 10 | DECEMBER 2014 | www.nature.com/naturephysics

2014 Macmillan Publishers Limited. All rights reserved

what happened with cuprates and pnictides.


Given the demanding experimental
requirements for ARPES and STM, these
measurements are usually accomplished
only after extensive transport and
magnetization studies have been carried out.
Even then, only select materials that can be
easily cleaved to prepare a pristine surface,
such as Bi2Sr2CaCu2O8, are amenable to
study. FeSe is very air sensitive, hampering
standard ex situ transport measurements.
Attempts to cover and protect the surface
have not been successful. As soon as FeSe is
capped with a protective layer, the TC drops
substantially. If one deposits additional
FeSe layers, they turn out to be neither
superconducting nor metallic11. Hence, the
path forward to confirming higher TC in
monolayer FeSe/STO requires insitustudies.
The technical solution to this challenge
has been to integrate a molecular beam
epitaxy (MBE) synthesis chamber with
ARPES and STM chambers, so that the
samples can be transferred from one
system to another without breaking ultrahigh vacuum conditions. This advance
has enabled the study of the electronic
structure of pristine FeSe surfaces by
ARPES and STM, providing a wealth of
valuable information. We have learned1315
that the FeSe layer is doped by electron
accumulation (about 0.1 electrons per iron
atom), apparently by charge transfer from
localized electron states generated by oxygen
vacancies in STO. The superconducting
FeSe layer shows just a single electron band
crossing the Fermi level (Fig.2b) and a
simple Fermi surface with just one circular
electron pocket centred at the zone edge
(Fig.2c). The superconducting gap seems
to be essentially isotropic, that is, with
893

commentary
conventional s-wave symmetry, without
any nodes. For bulk FeSe, ARPES shows
a cross-like feature centred at the point
(the Brillouin zone centre), which has
been attributed to the formation of a SDW.
This feature is completely suppressed in
monolayer-thick FeSe films, but emerges
at two monolayers, and becomes more
pronounced for films thicker than three
monolayers, before eventually reaching
the bulk limit 15. There appears to be a
correlation between the evolution of this
feature with strain relaxation as the film
thickness increases. Another unusual
observation is band replication18, which
has been attributed to strong electron
phonon coupling due to a flat oxygen bondstretching phonon mode in STO, with an
energy of about 100meV.

the superconducting gap. An important


implication, though, was that if FeSe/STO
can support superconductivity at 77K or
higher, one would need to perform transport
or magnetic measurements insitu, without
exposing the film to air. This is a nontrivial
technical challenge, but two groups took it
up and are now claimingsuccess.

The most recent breakthrough

Motivated by the promise of higher critical


temperatures, Jinfeng Jia and colleagues
have developed an important technical
capability 10 in the form of an insitu fourpoint contact probe that can measure
resistance as a function of temperature
and magnetic fields up to 11T (Fig.3a).
Using four microscopic metallic probes
separated by 10100m, currentvoltage
(IV) characteristics measured at 3K show
a large critical current density (jC) of about
107Acm2, which in a magnetic field of
11T decreases by about a factor of two. It
also decreases with increasing temperature,
fitting well to the GinzburgLandau
dependence jC(T)=jC(0)(1T/TC)3/2, with
a TC value of1114K. The resistance drops
sharply to zero at 109K, as reproduced
inFig.3b.
As exciting as these results may sound,
some potential pitfalls are immediately
apparent to the trained eye. In particular,
the resistance does not change when the
contact separation is increased tenfold,
and at 109K it drops so quickly that there
are no data points between the transition
onset and zero resistance. It is also notable
that such dramatic resistance drops have
been observed only in some, and not all,
monolayer FeSe/STO samples, and then
only at some locations on the surface.
This is worrisome, and at the very least
requires a plausible explanation. The claim
of ultra-high TC based on these data alone,

Indirect indications

Already in the first report 11, Xue and


co-workers speculated that TC could be
raised to even higher temperatures because
STM showed a superconducting gap of
20.1meV at 4.2K much larger than in
bulk FeSe (2.2meV). Subsequently, the
gap was shown by ARPES to persist up
to about 65K. Moreover, when FeSe was
subjected to an even larger tensile strain by
depositing it on a KTaO3 substrate (which
has the lattice constant a0=3.94, which
is 4.5% larger than in FeSe), covered with
a thin pseudomorphic STO buffer layer,
the observed gap persisted up to 70K
(ref.17). Most recently, this temperature
was raised further 18 to 75K by depositing
a monolayer of FeSe on Nb-doped BaTiO3,
which has an even larger lattice constant
(a0=3.99). Nevertheless, because
exsitu transport measurements showed
no evidence of superconductivity above
50K, one could legitimately question
whether this gap indeed corresponds to
b

C1423
C1234

Resistance (m)

SrTiO3 substrate

Although the signatures of HTS in


monolayer FeSe/STO are tantalizing, the
history of the field is strewn with so-called
USOs (unidentified superconducting
objects), and it is important to remain
cautious. In particular, more compelling
transport data and independent
confirmation from other groups would be
especially desirable at this stage. Further
enhancement of insitu measurement
capabilities (including magnetic
measurements) as well as finding a
capping material that does not reduce TC
will be critical to ensure progress. The latter
advancement would also be a prerequisite
for any newapplications.
Nevertheless, a significant interfacial
enhancement of TC in monolayer FeSe/STO
seems to have been established beyond
doubt, and this begs for a theoretical
explanation. It clearly bears a great
fundamental interest and with some luck,
understanding this system may even
provide clues to deciphering the mysterious
HTS in cuprates. Already at this early
stage, the existing data seem to rule out a

c
4

FeSe film

Open questions

Magnetization (106 emu)

without corroborating magnetization


measurements, is bound to raiseeyebrows.
A number of groups in China have
joined forces to address these issues, and
magnetization data have now been obtained19
(Fig.3c). Once again, a critic would remark
that a major diamagnetic change occurs only
at a low temperature, below 21K, with only a
slow and gradual decrease in magnetization
(perhaps originating from fluctuations)
that starts at about 85K. However, it should
be noted that in this case the monolayer
FeSe/STO heterostructure was covered with
a protective FeTe layer, and as discussed
earlier, this capping procedure is known to
diminishTC.

2
1
0

-1

20

40

60

80

100

120

140

T (K)

C1234

2
4
One-unit-cell FeSe / ten-unit-cell FeTe
6
ZFC
FC
1000 Oe

10

50

100

150

200

250

300

T (K)

Figure 3 | a, A sketch of the in-situ probe for four-point-contact resistance measurements used in ref.10. b, The dependence of resistance on temperature, for
two different probe configurations, measured insitu on an FeSe/SrTiO3 sample. Reproduced with permission from ref.10. c, The dependence of magnetization
on temperature, zero-field cooled (ZFC) and cooled in a magnetic field H=1000 oersted (FC), measured ex situ on a monolayer-thick FeSe film covered with a
ten-monolayer-thick protective layer of FeTe. Reproduced with permission from ref.19.
894

NATURE PHYSICS | VOL 10 | DECEMBER 2014 | www.nature.com/naturephysics

2014 Macmillan Publishers Limited. All rights reserved

commentary
class of theoretical scenarios: multiband
superconductivity, Fano resonances and
the like seem not to be necessary for HTS.
(Note that a similarly high TC is seen in
some cuprates like HgBa2CuO6, which
also features a single electron band and a
simple Fermi surface topology). Monolayer
FeSe/STO is n-type (electron doped) at
odds with the belief that hole doping is
crucial. The shape of its Fermi surface in
essence, a perfect circle rules out nesting.
The gap data are even more informative.
Much of the thinking about HTS in cuprates
has focused on the d-wave symmetry of
the gap, which is taken as an indication
that strong electron correlations are at play
(since nodes in the order parameter help
reduce Coulomb repulsion) and that the
pairing is mediated by antiferromagnetic
spin fluctuations. Most frequently, these
ideas have been formalized in terms of
hole-doped, 2D square lattice, spin-1/2
Heisenberg and Hubbard models. This
line of pursuit apparently does not carry
over easily to FeSe/STO, which shows an
isotropic s-wave gap, as seen in conventional
superconductors with phonon-mediated
pairing. Another major challenge for these
models is how to account for the tenfold
increase in TC when FeSe is brought into
the proximity of STO. Altogether, if HTS
above 100K indeed occurs in monolayer
FeSe/STO, this would leave us with two
choices: either each member of the HTS club
requires its own theorynot a very alluring
prospector we need to rethink thebasics.
To make progress, perhaps the most
important step would be to obtain detailed
information on the atomic-scale structure
near the interface not just the lattice
constants but also the actual atomic
positions. In principle, such information

can be obtained by synchrotron-based


X-ray diffraction phase-retrieval techniques,
such as the Coherent Bragg Rod Analysis
(COBRA) method8. In practice, this
approach may require an MBE system to
be coupled to a synchrotron beam line, and
equipped with the capability to directly
measure TCinsitu.
It will also be important to understand
why TC drops when FeSe/STO is covered
with a layer of another material, or even
with just more FeSe itself. The possibilities
that come to mind include electron
deconfinement, a decrease in local
charge density (the charge induced by
the interfacial mismatch in the chemical
potentials), relaxation of strain (either
Poisson or Madelung 20), a change in the
electron or phonon spectrum in the relevant
interfacial layer 21,22, and a reconstruction
of the atomic structure. It seems possible
to distinguish between these scenarios by
performing a series of experiments in which
one parameter is changed in a controlled
way and the effect on TC documented.
One would like to study in parallel the
evolution of the CDW/SDW, as this may
be the key competing instability that is
perhaps suppressed in FeSe/STO by the
effect of strain on the FeSeFe bond
angle (in pnictides, small variations in
FeAsFe bond angle have been shown to
affect TCdramatically 23).

Outlook

Signatures of interface superconductivity


have been observed at temperatures up
to 109K in monolayer FeSe/STO, which
is more than twice as high as in any other
known iron-containing superconductor
and close to the record TC in cuprates.
The data presented so far, however, are

NATURE PHYSICS | VOL 10 | DECEMBER 2014 | www.nature.com/naturephysics

2014 Macmillan Publishers Limited. All rights reserved

not quite conclusive, and it is essential


that this discovery is reproduced and
confirmed by other groups. Further
exploration of interface superconductivity is
a promising avenue for exciting discoveries,
as well as for shedding new light on the
mechanismofHTS.

Ivan Bozovic is at Brookhaven National Laboratory,


Upton, New York 11973, USAand in the Department
of Applied Physics at Yale University, New Haven,
Connecticut 06511, USA. Charles Ahn is in the
Department of Applied Physics at YaleUniversity,
New Haven, Connecticut 06511, USA.
References

1. Canfield, P.C. Nature Mater. 10, 259261 (2011).


2. Gurevich, A. Nature Mater. 10, 255259 (2011).
3. Wu, G. etal. J.Phys. Condens. Matter 21, 142203 (2009).
4. Ginzburg, V.L. Phys. Lett. 13, 101102 (1964).
5. Bozovic, I. etal. Phys. Rev. Lett. 89, 107001 (2002).
6. Gozar, A. etal. Nature 455, 782785 (2008).
7. Logvenov, G., Gozar, A. & Bozovic, I. Science 326, 699702 (2009).
8. Zhou, H. etal. Proc. Natl Acad. Sci. 107, 81038107 (2010).
9. Reyren, N. etal. Science 317, 11961199 (2007).
10. Ge, JF. etal. Nature Mater. http://dx.doi.org/10.1038/
nmat4153(2014).
11. Wang, QY. etal. Chinese Phys. Lett. 29, 037402 (2012).
12. Zhang, WH. etal. Chinese Phys. Lett. 31, 017401 (2014).
13. Liu, D. etal. Nature Commun. 3, 931 (2012).
14. He, S. etal. Nature Mater. 12, 605610 (2013).
15. Tan, S. etal. Nature Mater. 12, 634640 (2013).
16. Peng, R. etal. Phys. Rev. Lett. 112, 107001 (2014).
17. Peng, R. etal. Nature Commun. 5, 5044 (2012).
18. Lee, J.J. etal. Nature 515, 245248 (2014).
19. Sun, Y. etal. Sci. Rep. 4, 6040 (2014).
20. Butko, V. etal. Adv. Mater. 21, 36443688 (2009).
21. Xiang, YY. etal. Phys. Rev. B 86, 134508 (2012).
22. Coh, S., Cohen, M.L. & Louie, S.G. Preprint at http://arxiv.org/
abs/1407.5657 (2014).
23. Okabe, H., Takeshita, N., Horigane, K., Muranaka, T. &
Akimitsu, J. Phys. Rev. B 81, 205119 (2010).
24. Borisenko, S. Nature Mater. 12, 600601 (2013).

Acknowledgments

I.B. is supported by the US Department of Energy, Office of


Basic Energy Sciences, Materials Sciences and Engineering
Division. C.A. acknowledges support from NSF under
NSF-DMR MRSEC 1119826.

895

You might also like