Vol 453|5 June 2008

Q&A
CONDENSED-MATTER PHYSICS

Optical lattices
Markus Greiner and Simon Fölling
Optical lattices have rapidly become a favoured tool of atomic and condensed-matter physicists.
These crystals made of light can be used to trap atoms at very low temperatures, creating a workshop
in which to pore over and tinker with fundamental properties of matter.

Why make an optical lattice? for. Disorder and the inevitable effects of with the number of particles involved — one
Imagine you are trying to understand a complex vibrations of the crystal lattice add to the intri- of the problems confronting theorists today.
quantum-physical phenomenon such as cacy. It is not desirable, and usually impossible, Using results from optical lattices, theorists
high-temperature superconductivity, just to take account of all these effects simulta- want to test which of their models they can
one of many intriguing effects that occur in neously. The theorists’ approach is to con- best rely on to, say, construct a phase diagram
the ‘quantum gas’ of electrons that pervades a struct highly simplified models that focus on of a condensed-matter system, or assess the
solid crystal. The crystal lattice of a solid has particular aspects of the system. But often even evolution of a physical parameter. In addition,
tiny dimensions — atoms are spaced about a these cannot be solved reliably. That is particu- optical lattices provide ways to control vari-
tenth of a nanometre apart — and it would be larly true of models involving fermions — the ous factors, such as the strength of interatomic
great to have an enlarged version to enable us group of particles to which electrons, protons, interactions, band structure, spin composition
to look at its physics more closely. An opti- neutrons and many atoms belong. And so a and levels of disorder, more easily than in real
cal lattice affords just that possibility: it is a divide has opened up: on the one hand, real crystals — often even dynamically during the
crystal formed by interfering laser beams, with materials cannot be sufficiently described by course of a single experiment.
a typical dimension about 1,000 times larger theorists; on the other, experimentalists cannot
than that of a conventional crystal. Ultracold test the (sometimes contradictory) theoretical Are there limits to what you can
atoms in the lattice play the part of electrons in results obtained from simple models. simulate with an optical lattice?
the solid; they tunnel quantum-mechanically We can’t realistically use them to simulate very
between lattice sites just as single or paired How can optical lattices help? complicated crystals, such as superconduct-
electrons (Cooper pairs) tunnel through the They provide a way of realizing the simpli- ing ceramics, in their entirety. But that’s not
periodic potential wells created by positive fied models of condensed-matter theory in the goal. To understand phenomena such as
ions in crystalline materials (Fig. 1). experimental practice. Optical lattices thus high-temperature superconductivity, we want
implement Richard Feynman’s pioneering to find the simplest possible system that shows
Why not just use a real crystal? idea of ‘quantum simulation’ — using one the effect. Consequently, it is not a bad idea to
Real solid materials are incredibly complex. quantum system to investigate another. Feyn- start from simple models and add complexity
They have an involved band structure of man introduced the concept because a correct step by step.
allowed energy states, and the Coulomb inter- numerical description of a quantum system
actions of electrons are difficult to account requires resources that scale exponentially How exactly do you make an optical
lattice?
Figure 1 | Crystal simulation. Lattice potentials are created by making
Ultracold atoms in an optical lattice two coherent laser beams propagating in
can simulate condensed-matter opposite directions interfere with each other.
a Optical lattice phenomena that usually occur only The result is a standing wave with a periodic
in the ‘electron gas’ of a solid-state pattern of dark and bright stripes. The light
crystal. In an optical lattice (a), atoms
are trapped in a sinusoidal potential
induces an electric dipole moment in the
well (grey) created by a standing-wave atoms of the ultracold gas, modifying their
laser beam. The atoms’ wavefunctions energy. Depending on the frequency of the
(blue) correspond to those of light, atoms are pulled towards either the
b Real crystal valence electrons in a real crystal bright or the dark regions and are therefore
(b). Here, the periodic potential is confined to small areas in space. By using addi-
– – – – – caused by the attractive electrostatic tional lasers from different directions, two- or
force between the electrons (–) and even three-dimensional lattice structures can
the ions (+) forming the crystal. be constructed (Fig. 2).
The motion and interaction of the
particles, whether ultracold
atoms or electrons, determine the Can you make a lattice for any
+ + + + + physics of the material. Thus, for kind of atom?
example, superfluidity in a gas of Pretty much, although alkali atoms from group I
ultracold atoms corresponds to of the periodic table are generally easiest to work
superconductivity in an electron gas. with — their single valence electron gives them
736
NATURE|Vol 453|5 June 2008 NEWS & VIEWS Q&A

particularly suitable optical properties. At the

very low energy scales (temperatures) at which
a
optical-lattice models operate, most atoms are a Superfluid
Laser beam actually very similar. The most important prop-
erty is whether they are bosons (with integer
spin) or fermions (with half-integer spin). Many
bosons can exist in the same quantum state, but
quantum mechanics absolutely forbids this for
fermions, according to the rule known as the
Pauli exclusion principle. Whether a particular
atomic isotope is bosonic or fermionic depends
Potential well
on the number of its (fermionic) constituents:
protons, neutrons and electrons. If this number
is even, the total spin is an integer, and the atom
b is a boson; if it is odd, the total spin is a half-
Laser standing wave
integer, and the atom is a fermion. It’s easier b Mott insulator
to work with bosons: one consequence of the
Pauli exclusion principle is that fermions are
tougher to cool. Today, optical-lattice experi-
ments are being pursued with many different
atoms — bosonic rubidium-87, sodium-23,
potassium-39 and caesium-133; and fermionic
potassium-40, lithium-6 and strontium-87.

How do you put the atoms

c into the lattice?
Usually, by approaching the problem from
the other direction: you put the lattice into
the atoms. First, you create a Bose–Einstein
condensate — an ensemble of bosons all in Figure 3 | Superfluids and insulators. In a
the same quantum state — or a cold gas of quantum simulation of a bosonic Hubbard
fermionic atoms. Then you slowly ramp up the model in an optical lattice, two distinct ground
lasers to create the periodic lattice potential, states exist. a, For weak interaction strengths
and the atoms reorder to adapt to their new the ground state is a superfluid. The atoms
environment. Similarly, the whole ‘crystal’ can have a common macroscopic wavefunction
d (purple wave), with particles fully delocalized
be removed from the atoms simply by ramping
throughout the available space. After the atoms
down the lasers, thus liberating the atoms into are released from the lattice, the coherent
free space once more. wave nature becomes evident in the fact that
a distinct diffraction pattern arises (inset).
How did the work with optical b, If the interactions become very strong, the
lattices start? ground state is a ‘Mott insulator’ in which each
The first generation of optical lattices came particle becomes localized to one specific lattice
on the scene in the 1990s. They were mainly site. In that state, the number of atoms in each
used for the laser-cooling of atoms, as the site is fixed, and phase coherence, as well as
lattice potential increases the efficiency of superfluidity, disappears.
some methods of optical cooling. Still, the tem-
peratures at that point were too high and the states, relevant for simulating condensed-
occupation numbers too low (many lattice sites matter systems, started only around 2000,
were left unfilled) for many-body quantum first with Bose–Einstein condensates in three-
Figure 2 | Potential lattice. Laser light creates physics to feature. Nevertheless, fundamental dimensional lattices, and later with ultracold
the peculiar potential landscapes of optical phenomena such as Bloch oscillations of atoms Fermi gases.
lattices. a, Laser light creates a repulsive or an within the lattice and the Bragg scattering of
attractive potential that is proportional to the
laser’s intensity along its axis of propagation.
light on the atoms could already be observed. What are three-dimensional
b, By allowing two counter-propagating laser lattices good for?
beams to interfere, a sinusoidal standing wave When did people start using lattices to This geometry turns out to be a virtually per-
can be formed. Ultracold atoms can be trapped investigate many-body physics? fect rendering of the Hubbard model, a basic
in the potential minima that occur every half- A major boost came with the creation of the model that describes particles in a crystalline
wavelength, separated from the neighbouring first Bose–Einstein condensates in the mid- lattice. It takes into account the atoms’ mutual
minimum by a potential ‘wall’; this is the basis 1990s. This meant the availability of a dense repulsion or attraction and the rate at which
of a one-dimensional optical lattice. c, d, Adding and extremely cold sample of atoms on which they hop from site to site. Experiments with
more laser beams at right angles to the first one a lattice could simply be superimposed. It bosonic atoms in optical lattices confirmed
creates a two-dimensional (c) and finally a three-
became possible to study superfluid behaviour results arrived at using models, in particular
dimensional cubic ‘crystal’ in which atoms (d)
are trapped in the minima of the lattice potential. in the lattice, emulating for example the physics that ultracold bosons undergo a transition
Real optical lattices can have millions of sites, of arrays of Josephson junctions — tunnel junc- between two very distinct quantum states. The
and by changing the angles and wavelength of tions in superconductors that have important first is the Bose–Einstein condensate, a super-
the laser beams, different lattice geometries can applications in quantum-mechanical circuits. fluid state in which the gas can move without
be created. But experiments with complex many-body friction and particles are delocalized over
737
NEWS & VIEWS Q&A NATURE|Vol 453|5 June 2008

ground state that interact through electric

dipole–dipole forces. And then there are
a Ferromagnet b Antiferromagnet
efforts to improve how we can analyse the
quantum state, for example by directly read-
ing out every single lattice site or by probing
characteristic quantum fluctuations of the
whole ensemble.

Are there applications for optical

lattices beyond condensed-matter
simulations?
Because as few as one or two atoms can be
Figure 4 | Magnetic effects. When atoms in two different spin states (purple or red) are
simultaneously present in the lattice, they can either favour or disfavour identical spins on
isolated in each of the small traps, the traps
neighbouring sites. a, In the first case, they distribute in such a way that compact areas of identical can also act as miniature ‘test tubes’ in which
spins form — magnetic ‘domains’, which are the hallmark of ferromagnetism in solids. b, Conversely, atomic interactions can be studied with high
in the antiferromagnetic state, a chessboard pattern of alternating spins develops. The underlying precision. Even ‘micro-chemistry’ can be
force driving the system into one of these states might be direct, long-range interactions, for example performed, with the ability, for example, to
through the magnetic dipole of the atoms. Surprisingly, it can also arise from the tunnelling itself in precisely control the formation of a molecule
combination with the peculiar quantum statistical properties of bosons and fermions. This ‘quantum from two atoms. Equally, with just one atom in
magnetism’ does not require any direct magnetic interaction between the particles, and should be each site, interactions can be completely sup-
observable in an optical lattice implementing a simple fermionic Hubbard model. pressed. That is seen as a way of improving the
precision of optical atomic clocks still further.
the whole lattice. When the lattice ‘depth’ (the a second, unrelated class of material. Despite These create highly accurate time references
height of the potential barriers) is increased, the 20 years of effort, there is still no consensus through the extremely fast oscillations that
role of interactions between particles increases as to what causes either the formation of the occur between the energy states of the atoms.
compared with that of the tunnelling. Eventu- ‘d-wave’ Cooper pairs that seem to lie behind Those energy states are disturbed by atom–
ally, the interactions bring the system into what the superconductivity, or other anomalous atom interactions that can be suppressed in
is known as a Mott insulator state, in which free properties of these materials. The Hubbard an optical lattice. And last but not least, there
movement becomes impossible and each parti- model could be a good place to start: high- are intriguing proposals to use atoms trapped
cle is localized to one lattice site only (Fig. 3). temperature superconducting phases seem at single lattice sites as ‘qubits’ for a so-called
always to be found near phases with antiferro- universal quantum computer.
Can optical lattices go beyond magnetic order, in which particles of opposing
merely confirming theory? spin states arrange themselves in a chessboard How would a universal quantum
So far, experimenters have mostly studied the pattern (Fig. 4). This is a typical ground state computer work?
Bose–Hubbard model: thanks to numerical of a fermionic Hubbard model. The first step would be to create large-scale
simulations, we understand this model well, so entanglement between all atoms in a lattice,
it’s easy to compare experimental and numeri- So the Hubbard model holds the key to by letting each atom interact only with its next
cal results. But there’s a lot of physics in boson high-temperature superconductivity? neighbour in a way that depends on its state.
interactions that isn’t covered by this ‘plain There are many reasons to believe that the This technique has already been demonstrated
vanilla’ model, and that is not so well under- Hubbard model contains most — but not all experimentally. In principle, a universal quan-
stood theoretically. For example, introducing — of the ingredients necessary for under- tum computer can be realized just by reading
more than one atomic spin state leads to mag- standing high-temperature superconductiv- out and manipulating these single entangled
netic effects. As different types of interactions ity in the cuprates. A full explanation is likely qubits following such an entanglement step.
compete, each favouring a different type of to require additional phenomena, such as the The challenge is to do all the steps — entan-
magnetic ordering, exciting new states such as interactions between electrons and lattice glement, manipulation, readout and error
‘spin liquids’ are expected to arise. Then there vibrations (phonons), and inhomogeneities correction — with high enough fidelity. We
are fermionic Hubbard models. As mentioned at the mesoscopic level, although there is no might have to wait years to see a useful univer-
before, fermions behave fundamentally differ- consensus about which, if any, of these ingre- sal quantum computer. But in the meantime,
ently from bosons because of the Pauli exclu- dients are important. But that is just where the quantum simulators in the spirit of Feynman
sion principle. To form a superfluid, which remarkable control over ultracold atoms in an are already a specialized type of quantum com-
requires all the particles to be in the same, optical lattice comes into its own — we expect puter: simulated condensed-matter models in
coherent quantum state, fermions must ‘pre- that we can approach the problem by realizing optical lattices look set to become the first
tend’ to be bosons by, for example, teaming up the bare Hubbard model first and then adding practically useful application of a quantum-
to form Cooper pairs. Several laboratories are the other ingredients in a controlled way. This computational system. ■
working on reproducing the behaviour of fer- really would be a quantum simulator as Feyn- Markus Greiner and Simon Fölling are in the
mions in an optical lattice. Promising results, man first envisaged it. Department of Physics, Harvard University,
such as the creation of insulating and superfluid 17 Oxford Street, Cambridge, Massachusetts
behaviours, are already being reported. How might we add to this quantum- 02138, USA.
simulation tool-box in the future? e-mails: [email protected];
Why should we care about the A current focus is on extending the technique [email protected]
fermionic Hubbard model? to further atomic species, each with different
FURTHER READING
One reason is what it might tell us about high- interaction properties, as well as on decreas- Feynman, R. P. Int. J. Theor. Phys. 21, 467–488
temperature superconductivity. This puzzling ing the temperature of the trapped atoms still (1982).
phenomenon — electrical conduction without further through new cooling techniques. An Bloch, I. Nature Phys. 1, 23–30 (2005).
resistance at temperatures of up to 130 kelvin intense line of research is also to introduce long- Auerbach, A. Interacting Electrons and Quantum
Magnetism (Springer, Berlin, 2006).
— occurs in complex cuprate (copper oxide) range interactions into optical lattices by trap- Bloch, I., Dalibard, J. & Zwerger, W. Preprint at
materials, and has recently been reported in ping, not atoms, but molecules with a dipolar http://arxiv.org/pdf/0704.3011 (2007).

738