WHEN ATOMS BEHAVE AS WAVES: BOSE-EINSTEIN
CONDENSATION AND THE ATOM LASER
Nobel Lecture, December 8, 2001
by
WOLFGANG KETTERLE*
Department of Physics, MIT-Harvard Center for Ultracold Atoms, and Re-
search Laboratory of Electronics, Massachusetts Institute of Technology, Cam-
bridge, Massachusetts, 02139, USA.
INTRODUCTION
The lure of lower temperatures has attracted physicists for the past century,
and with each advance towards absolute zero, new and rich physics has
emerged. Laypeople may wonder why “freezing cold” is not cold enough. But
imagine how many aspects of nature we would miss if we lived on the surface
of the sun. Without inventing refrigerators, we would only know gaseous mat-
ter and never observe liquids or solids, and miss the beauty of snowflakes.
Cooling to normal earthly temperatures reveals these dramatically different
states of matter, but this is only the beginning: many more states appear with
further cooling. The approach into the kelvin range was rewarded with the
discovery of superconductivity in 1911 and of superfluidity in helium-4 in
1938. Cooling into the millikelvin regime revealed the superfluidity of
helium-3 in 1972. The advent of laser cooling in the 1980s opened up a new
approach to ultralow temperature physics. Microkelvin samples of dilute
atom clouds were generated and used for precision measurements and
studies of ultracold collisions. Nanokelvin temperatures were necessary to ex-
plore quantum-degenerate gases, such as Bose-Einstein condensates first
realized in 1995. Each of these achievements in cooling has been a major ad-
vance, and recognized with a Nobel prize.
This paper describes the discovery and study of Bose-Einstein condensates
(BEC) in atomic gases from my personal perspective. Since 1995, this field
has grown explosively, drawing researchers from the communities of atomic
physics, quantum optics, and condensed matter physics. The trapped ultra-
cold vapor has emerged as a new quantum system that is unique in the preci-
sion and flexibility with which it can be controlled and manipulated. At least
thirty groups have now created condensates, and the publication rate on
Bose-Einstein condensation has soared following the discovery of the gaseous
condensates in 1995 (see Fig. 1).
118
* URL: http://cua.mit.edu/ketterle_group/
The phenomenon of Bose-Einstein condensation was predicted long ago,
in a 1925 paper by Albert Einstein [1] using a method introduced by
Satyendra Nath Bose to derive the black-body spectrum [2]. When a gas of
bosonic atoms is cooled below a critical temperature Tc , a large fraction of the
atoms condenses in the lowest quantum state. Atoms at temperature T and
with mass m can be regarded as quantum-mechanical wavepackets that have a
spatial extent on the order of a thermal de Broglie wavelength λdB =
(2π�2/mkBT)
1/2. The value of λdB is the position uncertainty associated with
the thermal momentum distribution and increases with decreasing tempera-
ture. When atoms are cooled to the point where λdB is comparable to the in-
teratomic separation, the atomic wavepackets “overlap” and the gas starts to
become a “quantum soup” of indistinguishable particles. Bosonic atoms un-
dergo a quantum-mechanical phase transition and form a Bose-Einstein con-
densate (Fig. 2), a cloud of atoms all occupying the same quantum mechan-
ical state at a precise temperature (which, for an ideal gas, is related to the
peak atomic density n by nλd
3
B = 2.612). If the atoms are fermions, cooling
gradually brings the gas closer to being a “Fermi sea” in which exactly one
atom occupies each low-energy state.
Creating a BEC is thus simple in principle: make a gas extremely cold until
the atomic wave packets start to overlap! However, in most cases quantum de-
generacy would simply be pre-empted by the more familiar transitions to a
liquid or solid. This more conventional condensation into a liquid and solid
can only be avoided at extremely low densities, about a hundred thousandth
the density of normal air. Under those conditions, the formation time of mo-
lecules or clusters by three-body collisions (which is proportional to the in-
verse density squared) is stretched to seconds or minutes. Since the rate of bi-
nary elastic collisions drops only proportional to the density, these collisions
are much more frequent. Therefore, thermal equilibrium of the translation-
119
Figure 1. Annual number of published papers, which have the words “Bose” and “Einstein” in
their title, abstracts or keywords. The data were obtained by searching the ISI (Institute for
Scientific Information) database.
al degree of freedom of the atomic gas is reached much faster than chemical
equilibrium, and quantum degeneracy can be achieved in an effectively
metastable gas phase. However, such ultralow density lowers the temperature
requirement for quantum degeneracy into the nano- to microkelvin range.
The achievement of Bose-Einstein condensation required first the identifi-
cation of an atomic system which would stay gaseous all the way to the BEC
transition, and second, the development of cooling and trapping techniques
to reach the required regime of temperature and density. Even around 1990,
it was not certain that nature would provide us with such a system. Indeed,
many people doubted that BEC could ever be achieved, and it was regarded
as an elusive goal. Many believed that pursuing BEC would result in new and
interesting physics, but whenever one would come close, some new pheno-
menon or technical limitation would show up. A news article in 1994 quoted
Steve Chu: “I am betting on nature to hide Bose condensation from us. The
last 15 years she’s been doing a great job” [3].
In brief, the conditions for BEC in alkali gases are reached by combining
two cooling methods. Laser cooling is used to precool the gas. The principle
of laser cooling is that scattered photons are on average blue-shifted with re-
spect to the incident laser beam. As a result, the scattered light carries away
more energy than has been absorbed by the atoms, resulting in net cooling.
Blue-shifts are caused by Doppler shifts or ac Stark shifts. The different laser
cooling schemes are described in the 1997 Nobel lectures in physics [4–6].
After the precooling, the atoms are cold enough to be confined in a magne-
tic trap. Wall-free confinement is necessary, otherwise the atoms would stick
to the surface of the container. It is noteworthy that similar magnetic con-
finement is also used for plasmas which are too hot for any material con-
tainer. After magnetically trapping the atoms, forced evaporative cooling is
applied as the second cooling stage [7–9]. In this scheme, the trap depth is
reduced, allowing the most energetic atoms to escape while the remainder
rethermalize at steadily lower temperatures. Most BEC experiments reach
quantum degeneracy between 500 nK and 2 µK, at densities between 1014 and
1015 cm-3. The largest condensates are of 100 million atoms for sodium, and a
billion for hydrogen; the smallest are just a few hundred atoms. Depending
on the magnetic trap, the shape of the condensate is either approximately
round, with a diameter of 10 to 50 µm, or cigar-shaped with about 15 µm in
diameter and 300 µm in length. The full cooling cycle that produces a con-
densate may take from a few seconds to as long as several minutes.
After this short overview, I want to provide the historical context for the
search for BEC and then describe the developments which led to the obser-
vation of BEC in sodium at MIT. Finally, some examples will illustrate the
novel physics which has been explored using Bose-Einstein condensates. A
more detailed account of the work of my group has been presented in four
comprehensive review papers [8, 10–12].
120
BEC AND CONDENSED-MATTER PHYSICS
Bose-Einstein condensation is one of the most intriguing phenomena pre-
dicted by quantum statistical mechanics. The history of the theory of BEC is
very interesting, and is nicely described in the biographies of Einstein [13]
and London [14] and reviewed by Griffin [15]. For instance, Einstein made
his predictions before quantum theory had been fully developed, and before
the differences between bosons and fermions had been revealed [16]. After
Einstein, important contributions were made by, most notably, London,
Landau, Tisza, Bogoliubov, Penrose, Onsager, Feynman, Lee, Yang, Huang,
Beliaev and Pitaevskii. An important issue has always been the relationship be-
tween BEC and superfluidity in liquid helium, an issue that was highly con-
troversial between London and Landau (see ref. [14]). Works by Bogoliubov,
Beliaev, Griffin and others showed that Bose-Einstein condensation gives the
microscopic picture behind Landau’s “quantum hydrodynamics.” BEC is
closely related to superconductivity, which can be described as being due to
Bose-Einstein condensation of Cooper pairs. Thus Bose-Einstein condensa-
tion is at the heart of several macroscopic quantum phenomena.
BEC is unique in that it is a purely quantum-statistical phase transition, i.e.,
it occurs even in the absence of interactions. Einstein described the transition
as condensation “without attractive forces” [16]. This makes BEC an impor-
tant paradigm of statistical mechanics, which has been discussed in a variety
of contexts in condensed-matter, nuclear, particle and astrophysics [17]. On
the other hand, real-life particles will always interact, and even the weakly-
interacting Bose gas behaves qualitatively differently from the ideal Bose gas
121
Figure 2. Criterion for Bose-
Einstein condensation. At high
temperatures, a weakly interact-
ing gas can be treated as a sys-
tem of “billiard balls.” In a sim-
plified quantum description, the
atoms can be regarded as
wavepackets with an extension
of their de Broglie wavelength
λdB. At the BEC transition tem-
perature, λdB becomes compar-
able to the distance between
atoms, and a Bose condensate
forms. As the temperature ap-
proaches zero, the thermal
cloud disappears, leaving a pure
Bose condensate.
[18]. It was believed for quite some time that interactions would always lead
to “ordinary” condensation (into a solid) before Bose-Einstein condensation
would happen. Liquid helium was the only counter-example, where the light
mass and concomitant large zero-point kinetic energy prevents solidification
even at zero kelvin. Erwin Schrödinger wrote in 1952 in a textbook on ther-
modynamics about BEC: “The densities are so high and the temperatures so
low – those required to exhibit a noticeable departure [from classical stati-
stics] – that the van der Waals corrections are bound to coalesce with the pos-
sible effects of degeneration, and there is little prospect of ever being able to
separate the two kinds of effect” [19]. What he didn’t consider were dilute sys-
tems in a metastable gaseous phase!
The quest to realize BEC in a dilute, weakly interacting gas was pursued in
at least three different directions: liquid helium, excitons and atomic gases.
Experimental [20, 21] and theoretical work [22] showed that the onset of su-
perfluidity for liquid helium in Vycor has features of dilute-gas Bose-Einstein
condensation. At sufficiently low coverage, the helium adsorbed on the po-
rous sponge-like glass behaved like a dilute three-dimensional gas. However,
the interpretation of these results is not unambiguous [23].
Excitons, which consist of weakly-bound electron-hole pairs, are composite
bosons. The physics of excitons in semiconductors is very rich and includes
the formation of an electron-hole liquid and biexcitons. As nicely discussed in
refs. [24, 25], there are systems where excitons form a weakly interacting gas.
However, the initial evidence for Bose-Einstein condensation in Cu2O [26]
was retracted [27]. Recent work in coupled quantum-well structures is very
promising [28]. When excitons strongly interact with light in a cavity, they
form polaritons. In such polariton systems, stimulated scattering and non-
equilibrium condensates have been observed recently [29–31].
SPIN-POLARIZED HYDROGEN
Dilute atomic gases are distinguished from the condensed-matter systems dis-
cussed above by the absence of strong interactions. Interactions at the densi-
ty of a liquid or a solid considerably modify and complicate the nature of the
phase transition. Hecht[32], and Stwalley and Nosanow [33] used the quan-
tum theory of corresponding states to conclude that spin-polarized hydrogen
would remain gaseous down to zero temperature and should be a good can-
didate to realize Bose-Einstein condensation in a dilute atomic gas. These
suggestions triggered several experimental efforts, most notably by Silvera
and Walraven in Amsterdam, by Greytak and Kleppner at MIT, and by others
at Moscow, Turku, British Columbia, Cornell, Harvard, and Kyoto. The stabi-
lization of a spin-polarized hydrogen gas [34, 35] created great excitement
about the prospects of exploring quantum-degenerate gases. Experiments
were first done by filling cryogenic cells with the spin-polarized gas and by
compressing it, and since 1985, by magnetic trapping and evaporative
cooling. BEC was finally accomplished in 1998 by Kleppner, Greytak and col-
laborators [36]. See refs. [9, 37–39] and in particular ref. [40] for a full ac-
122
count of the pursuit of Bose-Einstein condensation in atomic hydrogen. Evi-
dence for a phase transition in two dimensions was reported in 1998 [41].
The work in alkali atoms is based on the work in spin-polarized hydrogen
in several respects:
• Studies of spin-polarized hydrogen showed that systems can remain in a
metastable gaseous state close to BEC conditions. The challenge was then to
find the window in density and temperature where this metastability is suf-
ficient to realize BEC.
• Many aspects of BEC in an inhomogeneous potential [42–44], and the the-
ory of cold collision processes (see e.g. [45]) developed in the 1980s for hy-
drogen could be applied directly to the alkali systems.
• The technique of evaporative cooling was developed first for hydrogen [7,
46] and then used for alkali atoms.
LASER COOLING
Laser cooling opened a new route to ultralow temperature physics. Laser
cooling experiments, with room temperature vacuum chambers and easy op-
tical access, look very different from cryogenic cells with multi-layer thermal
shielding around them. Also, the number of atomic species that can be studi-
ed at ultralow temperatures was greatly extended from helium and hydrogen
to all of the alkali atoms, metastable rare gases, several earth-alkali atoms, and
others (the list of laser-cooled atomic species is still growing). A full account
of the relevant laser cooling techniques and their development is given in
refs. [47–49] and in the 1997 Nobel lectures of Chu, Cohen-Tannoudji and
Phillips [4–6].
Some papers and proposals written in the early and mid 1980s, before and
during the developments of the basic cooling and trapping techniques, listed
quantum degeneracy in a gas as a visionary goal for this new emerging field
[50–52]. However, major limitations of laser cooling and trapping were soon
identified. Although there is no fundamental low temperature limit, the final
temperature provided by polarization gradient cooling – about ten times the
recoil energy – was regarded as a practical limit. Sub-recoil laser cooling tech-
niques, especially in three dimensions, were harder to implement, and re-
quired long cooling times. The number and density of atoms were limited by
inelastic, light-induced collisions (leading to trap loss [53, 54]) and by ab-
sorption of scattered laser light [55], which results in an outward radiation
pressure (weakening the trapping potential and limiting the density).
Furthermore, since the lowest temperatures could not be achieved at the
highest densities [56–58], most trapping and cooling techniques reached a
maximum phase-space density of around nλd
3
B = 10
-5; and a value of 2.612 is
needed for BEC. This was the situation when the author joined the field of
cold atoms in 1990. It was only more recently that major increases in phase-
space density were achieved by laser cooling [59–61], but so far laser cooling
by itself has not been able to reach BEC.
123
THE EFFORT AT MIT 1990–1996
Improving laser cooling
When I teamed up with Dave Pritchard at MIT in 1990 as a postdoc, the initial
goal was to build an intense source of cold atoms to study cold collisions and
pure long-range molecules. However, Dave and I frequently talked about the
limitations in density and temperature of the current techniques and tried to
develop ideas on how to get around them. One limitation of magnetic traps is
that they can hold atoms only in weak-field seeking hyperfine states.
Therefore, a collision between two trapped atoms can lead to a spinflip, and
the Zeeman energy is converted into kinetic energy (dipolar relaxation). This
process has been a major limitation to the experiments in atomic hydrogen.
First, we asked ourselves if the inclusion of electric and gravitational fields
would allow the stable confinement of atoms in their lowest hyperfine states-
– but the answer was negative [62]. One loophole was time-dependent mag-
netic fields, and building on an earlier proposal [63], I designed an experi-
ment to confine sodium atoms with ac magnetic fields which looked feasible.
However, we learnt that Eric Cornell at Boulder had developed a similar idea
and experimentally implemented it [64] – so we left the idea on the drawing
board. It wasn’t the last time that Eric and I would develop similar ideas in-
dependently and almost simultaneously!
Trapping atoms in the lowest hyperfine state was not necessary to accom-
plish BEC. Already in 1986, Pritchard correctly estimated the rate constants of
elastic and inelastic collisions for alkali atoms [52]. From these estimates one
could easily predict that for alkali atoms, in contrast to hydrogen, the so-
called good collisions (elastic collisions necessary for the evaporation pro-
cess) would clearly dominate over the so-called bad collisions (inelastic two-
and three-body collisions); therefore, evaporative cooling in alkalis would
probably not be limited by intrinsic loss and heating processes. However,
there was pessimism [65] and skepticism, and the above-mentioned experi-
mental [64] and theoretical [62] work on traps for strong-field seeking atoms
has to be seen in this context.
In those years, there were some suggestions that time-dependent potentials
could lead to substantial cooling, but we showed that this was not possible
[66]. Real cooling needs an open system which allows entropy to be removed
from the system – in laser cooling in the form of scattered photons, in evapo-
rative cooling in the form of discarded atoms. Dave and I brainstormed about
novel laser cooling schemes. In 1991, at the Varenna summer school, Dave
presented a new three-level cooling scheme [67]. Inspired by these ideas, I
developed a scheme using Raman transitions. Replacing the six laser beams
in optical molasses by counterpropagating beams driving the Doppler-sensi-
tive Raman transition, we hoped to realize Doppler molasses with a linewidth
that was proportional to the optical pumping rate, and therefore adjustable.
We had started setting up radio-frequency (rf) electronics and magnetic
shields for Raman cooling when we heard that Mark Kasevich and Steve Chu
were working on Raman cooling using laser pulses [68]. For this reason, and
also because around the same time we had developed the idea for the Dark
124
SPOT (spontaneous force optical tray; see later in this section), trap, we
stopped our work on Raman cooling.
Our experimental work in those years focused first on generating a large
flux of slow atoms. In my first months at MIT when I overlapped with Kris
Helmerson and Min Xiao, we built a sodium vapor cell magneto-optical trap