Atomic physics: When ultracold is not cold enough

Journal name:
Date published:
Published online

A technique for cooling ultracold atoms in optical lattices has been demonstrated. This advance should allow the physics of strongly correlated systems, including that of quantum magnetism, to be explored. See Letter p.500

Experiments with ultracold neutral atoms routinely reach nanokelvin temperatures. When combined with optical lattices — light 'crystals' created with standing waves of light — ultracold atoms are an almost ideal quantum many-body system1, 2. Thus, they are an excellent platform for simulating the physics of solid materials. Lattice-trapped atoms can simulate theoretical model systems that are relevant to understanding strongly correlated materials — systems in which electrons interact strongly. However, for some model systems, even nanokelvin temperatures can be too hot for simulating the relevant phenomena3.

On page 500 of this issue, Bakr et al.4 demonstrate a technique for cooling quantum atomic gases in optical lattices that may allow much lower temperatures to be reached than those currently attainable. This opens up the possibility of studying a number of outstanding problems in many-body physics, such as quantum magnetism and high-temperature superconductivity1, 2, 3. In addition, the control achievable with this technique may provide a way of producing logic gates based on two quantum bits (qubits) and creating quantum registers — the working memory needed for quantum computing — using ultracold atoms.

Bakr and colleagues' technique for cooling atomic gases relies on atom blockade. Blockade phenomena arise from the interactions between tightly confined particles. If the interaction energy is sufficiently high, it is much harder to add another particle to the system, because of the increased amount of energy required. Blockade phenomena are used in many systems. For example, Coulomb blockade of electron charges can be used to make transistors based on single electrons5. Blockade effects have also been proposed as a way to produce qubits6, 7, 8.

In ultracold-atom experiments, atom blockade occurs as a result of repulsive interactions between the atoms. When trapped in an optical lattice, ultracold atoms develop an energy-band structure just like that of electrons in a solid. The higher the number of atoms in a given lattice site, the higher the interaction energy, creating a barrier to the addition of a further atom. If the optical lattice's sites are sufficiently deep, these interactions give rise to insulating behaviour, analogous to the insulating behaviour of electrons in some solids. As a result, there are a fixed number of atoms per site and tunnelling of atoms between different lattice sites is inhibited.

In their study, Bakr et al.4 show that, in addition to this transport blockade, a blockade can occur if atoms are excited to different energy bands within a single lattice site. The authors introduce the phenomenon of orbital exchange blockade (OEB), which allows the entropy (and thus the temperature) of the system to be reduced.

Bakr et al. demonstrated OEB by first preparing a two-dimensional gas of rubidium-87 atoms in a square optical lattice created using two perpendicular standing waves of light. Initially, they arranged the system such that there was a known number of atoms per individual site, and they prepared the atoms in the same quantum state — the ground state, or band, of the lattice. If the neighbouring sites were already occupied, the repulsive interactions between the atoms suppressed tunnelling between lattice sites.

Next, by changing the intensity of one of the standing waves, the authors modulated the depth of the lattice sites. If the frequency of this modulation was tuned to exactly match the separation between the ground band and an excited band of the lattice, atoms could be excited to the higher band. In a similar situation to that for suppression of tunnelling between sites, if an atom was already excited to the higher band, the excitation of a second atom in the same site was suppressed as a result of the interactions. Because of this atom-number-sensitive OEB, the frequency needed to excite atoms depends on the number of atoms in the ground and excited bands. Atoms excited to the higher band could then be removed from the system. Bakr et al. demonstrated that OEB allows spectroscopic differentiation of sites with different numbers of atoms and control over the final number of atoms in a given site.

Having a lattice-trapped atomic system at finite temperature creates 'defects' in it. Instead of a uniform filling, defects exist in lattice sites at which either no atoms are present (holes) or there are extra atoms. OEB cannot be used to directly fill the holes, but by using atom-number-sensitive OEB, sites with extra atoms can be corrected, removing entropy from the system and so cooling it.

Bakr et al. achieved such atom removal in two experiments. In the first one, they initially loaded the lattice with atoms such that the system was in an insulating state — the ground state for the system — with a known number of between one and four atoms per site. Next, by sweeping the modulation frequency, they deterministically reduced the lattice occupation number to a single atom per site.

In the second experiment, the authors4 loaded the lattice such that the system was not in the ground state. Instead, they loaded a random number of atoms per site, and by sweeping the modulation frequency, they were able to remove all of the 'extra' atoms. After the frequency sweep, they adjusted the depth of the lattice sites so that the final state, an insulating state with a single atom per site, was close to the ground state. This cooling 'algorithm' is analogous to heat-bath algorithmic cooling, in which entropy is pumped out of a system into a heat bath. Heat-bath algorithmic cooling has previously been demonstrated using nuclear magnetic resonance of solid-state qubits9.

Bakr and colleagues' technique allows entropy to be removed from the system. But it also allows individual lattice sites to be targeted, as has been shown previously10. The combination of these two features could facilitate the creation of a quantum register consisting of thousands of atoms, and provide a roadmap to scalable quantum computing using lattice-trapped atoms. Achieving picokelvin temperatures (1 picokelvin is 10−12 kelvin) for lattice-trapped atoms remains an important goal, which the authors' technique could make achievable4.

In Bakr and colleagues' experiments, the final entropy (and thus temperature) was limited by heating caused by the optical-lattice light beams, as well as by inefficiencies in the excitation of atoms to higher bands. Its value was 0.27kB per particle (where kB is the Boltzmann constant), which is comparable to previously reported values10. These limitations will need to be overcome if optimal cooling is to be achieved. But even without additional cooling, the use of OEB to deterministically control the number of atoms in individual sites will be a valuable tool for future experiments.


  1. Lewenstein, M. et al. Adv. Phys. 56, 243379 (2007).
  2. Bloch, I., Dalibard, J. & Zwerger, W. Rev. Mod. Phys. 80, 885964 (2008).
  3. McKay, D. C. & DeMarco, B. Rep. Prog. Phys. 74, 054401 (2011).
  4. Bakr, W. S. et al. Nature 480, 500503 (2011).
  5. Kastner, M. A. Rev. Mod. Phys. 64, 849858 (1992).
  6. Hanson, R., Kouwenhoven, L. P., Petta, J. R., Tarucha, S. & Vandersypen, L. M. K. Rev. Mod. Phys. 79, 12171265 (2007).
  7. Isenhower, L. et al. Phys. Rev. Lett. 104, 010503 (2010).
  8. Wilk, T. et al. Phys. Rev. Lett. 104, 010502 (2010).
  9. Baugh, J., Moussa, O., Ryan, C. A., Nayak, A. & Laflamme, R. Nature 438, 470473 (2005).
  10. Bakr, W. S. et al. Science 329, 547550 (2010).

Download references

Author information


  1. Gretchen K. Campbell is at the Joint Quantum Institute, National Institute of Standards and Technology, and the University of Maryland, Gaithersburg, Maryland 20899-8424, USA.

Corresponding author

Correspondence to:

Author details


Subscribe to comments

Additional data