Laser cooling delivers a Bose–Einstein condensate

It’s been more than two decades since Carl Wieman, Eric Cornell, and their co-workers created the first Bose–Einstein condensate (BEC), confirming that a macroscopic population of integer-spin particles will pile into a single quantum ground state if cooled below some critical temperature. (See Physics Today, August 1995, page 17 .) After all those years, the recipe for creating the condensates remains virtually unchanged: Laser cooling chills the cloud of atoms as close to the critical temperature as it can, and when that technique can go no further, evaporative cooling does the rest.

Experimenters have long sought to bypass the evaporative cooling step, a slow process that jettisons most of a cloud’s atoms in order to cool the remaining few. The process can take seconds, sometimes more than a minute, to unfold. Afterward, typically less than 1% of the original atoms remain.

Now MIT researchers led by Vladan Vuletić have used a laser technique known as Raman sideband cooling to take a cloud of rubidium atoms all the way to the condensation threshold—no evaporation needed. ¹ From an initial gas of 2000 atoms, they can generate a BEC of more than 500 in just 300 ms—about the time it takes to blink an eye.

The Doppler limit

The threshold for Bose–Einstein condensation is often expressed in terms of a critical temperature. But a more fundamental quantity is the dimensionless phase-space density, the peak occupation per quantum state. Roughly speaking, that quantity describes the extent to which atoms’ wavefunctions overlap. It grows as a cloud of atoms becomes colder and denser. As it surpasses a value of one, a BEC forms.

At the outset of their experiment, the MIT researchers’ vapor of rubidium atoms has a phase-space density near 10⁻²⁰. To boost that value, the team captures the atoms in a magneto-optical trap and applies a standard laser cooling technique: Doppler cooling.

First demonstrated in the 1970s, Doppler cooling exploits the frequency shifts that an atom sees as it moves toward or away from a light source. When two counterpropagating lasers are tuned just below an electron resonance, the atom will preferentially absorb light—and momentum—from the laser it’s traveling toward, since that light is slightly blueshifted to higher energy. The net effect, regardless of which source the atom moves toward, is to slow the atom’s momentum. If three such sets of counterpropagating beams are arranged orthogonal to one another, they brake the atom’s motion in every direction.

But there’s a catch. As absorbed photons are reemitted, the atoms suffer random recoils. Below a certain temperature, typically a few hundred microkelvin, those recoils are large enough to offset the cooling effect of the lasers, and Doppler cooling is rendered ineffective. After just a millisecond or so of cooling, Vuletić and his colleagues reach that Doppler limit, with their cloud’s phase-space density still seven or eight orders of magnitude below the condensation threshold.

That’s when conventional experiments would switch to evaporative cooling. Instead, Vuletić and his colleagues turn off the Doppler cooling beams, release their atoms from the magneto-optical trap, and recapture them in a two-dimensional optical lattice—a square array of potential wells formed by the interference of two orthogonal retroreflected IR beams. Now, instead of sharing one large trap, the team’s atom cloud is dispersed among hundreds of tiny traps. The tight spatial confinement creates large spacings between the atoms’ vibrational energy levels. And that sets the stage for Raman sideband cooling.

Cool, compress, repeat

Developed in the 1990s by Stanford University’s Steven Chu and Mark Kasevich, Raman cooling siphons away a trapped atom’s vibrational energy by converting it to magnetic energy and then to light. ² Figure 1 illustrates the basic scheme. An applied magnetic field is tuned to split the energies of two of the atom’s magnetic hyperfine sublevels m by an amount precisely equal to a vibrational quantum. In this case the energy of the m = −1 sublevel is raised above that of the m = −2 sublevel.

Figure 1.

Then, when an atom with n vibrational quanta in the m = −2 state is transferred to the degenerate m = −1 state, it arrives with n − 1 vibrational quanta; that step, driven by a two-photon interaction with the trapping light, is known as a Raman transition. A shorter-wavelength pump beam then transfers the atom to an excited state with m = −2. When the atom decays to the electronic ground state, it does so with n − 1 vibrational quanta.

Raman sideband cooling can easily chill an isolated atom to the ground state. But two issues that arise in dense gases of atoms have foiled previous attempts to use the technique to create BECs. First, in atomic clouds, photons emitted by one atom can scatter off neighboring atoms and generate heat-inducing recoils. Second, pairs of nearby atoms can be excited to repulsive molecular states that, upon decaying, give each atom a jolt of kinetic energy. Both effects produce unwanted heat that can make the ground state inaccessible.

In 2000 a group at the University of California, Berkeley, led by David Weiss showed that by detuning the pump beam from the atomic transition, one can suppress scattering and all but eliminate recoil heating. ³ But it wasn’t obvious how to solve the molecular-excitation problem. Near the atomic resonance frequency, the molecular-excitation rate scales identically to the rate of the atomic excitation used in the cooling cycle. Although detuning could in theory reduce the rate of heat-generating molecular excitations, it would slow the cooling rate by just as much. Vuletić eventually realized that if he red-detuned the pump beam far from the atomic resonance, he could find gaps in the molecular spectrum—special detunings where atomic excitations win out over molecular ones.

In their experiment, the MIT researchers set the pump beam to one of those special detunings. After just 100 ms of sideband cooling, the temperature of the rubidium cloud falls to 10 µK—cold enough that nearly all 2000 atoms are in their vibrational ground states. Even still, the phase-space density only reaches 0.02. To create a BEC, the researchers must compress the cloud.

To do so, they borrow a trick developed by Weiss: ⁴ They turn off one of the trapping beams, transforming the square array of 2D potential wells to a row of 1D wells, as illustrated in figure 2. Because the light that forms those wells is more intense near the beam axis, the wells slope downward toward the axis. As a result, atoms that had been trapped in the outer wells of the 2D trap fall inward in the newly 1D trap. If the 2D lattice is restored at just the right moment, those atoms can be recaptured in the low-lying potential wells near the lattice’s center.

Figure 2.

Unfortunately, that “release-and-retrap” trick doesn’t increase phase-space density; the compression-induced warming negates the rise in spatial density. So after each compression, Vuletić and his coworkers must sideband cool the atoms anew.

With two rounds of compression—one in each trap direction—the MIT researchers boost the cloud’s density nearly 50-fold. And near the end of the third 100 ms round of cooling, the phase-space density nudges past one. Barely 300 ms after the start of their experiment, Vuletić and his colleagues have their BEC.

A matter of time

The MIT team isn’t the first to laser-cool atoms to the condensation threshold. Five years ago an Austrian team led by Florian Schreck created a BEC of strontium atoms in just 100 ms using only Doppler cooling. ⁵ But their approach relied on an extraordinarily narrow-linewidth transition that gives strontium an unusually low Doppler limit; it can’t be translated to most atoms of interest in quantum experiments.

The MIT group’s method, by contrast, doesn’t use anything special about the rubidium atom. For JILA’s Jun Ye, who has helped pioneer the development of optical-lattice atomic clocks, that’s welcome news.

Atomic clocks keep time by syncing an oscillator with the frequency of some reference atomic transition. To maximize precision, one wants to measure that transition in large atomic ensembles and to repeat the measurement as many times as possible. Last October, Ye’s group reported that they could keep time to a record precision of better than one part in 10¹⁸ by measuring transitions of a quantum degenerate gas loaded in an optical lattice. ⁶ But because the researchers used evaporative cooling to generate their quantum gases—and because the protocol calls for creating a new gas for each measurement—they could make just a few measurements per minute. Says Ye of Vuletić’s sideband-cooling approach, “We could benefit from the much shorter preparation time.”

Vuletić thinks the shorter preparation times will also benefit quantum-gas microscopy, where high-resolution optics are used to image tunneling and other phenomena in lattice-confined gases designed to mimic strongly correlated condensed matter. (See Physics Today, October 2010, page 18 , and August 2017, page 17 .) But his ultimate dream is to send the technique into orbit, where BECs have been proposed for use in space-based gravitational-wave detectors, among other applications. There, the method’s chief virtue would be its economy. One wouldn’t need the powerful magnetic coils used for evaporative cooling, Vuletić explains: “You could use modest magnets and standard laser beams.”