Selected for a Viewpoint in Physics

week ending
PRL 108, 210401 (2012) PHYSICAL REVIEW LETTERS 25 MAY 2012

Bose-Einstein Condensation of Erbium

K. Aikawa,1 A. Frisch,1 M. Mark,1 S. Baier,1 A. Rietzler,1 R. Grimm,1,2 and F. Ferlaino1
1
Institut für Experimentalphysik and Zentrum für Quantenphysik, Universität Innsbruck, Technikerstraße 25, 6020 Innsbruck, Austria
2
Institut für Quantenoptik und Quanteninformation, Österreichische Akademie der Wissenschaften, 6020 Innsbruck, Austria
(Received 6 April 2012; published 21 May 2012)
We report on the achievement of Bose-Einstein condensation of erbium atoms and on the observation of
magnetic Feshbach resonances at low magnetic fields. By means of evaporative cooling in an optical dipole
trap, we produce pure condensates of 168 Er, containing up to 7
104 atoms. Feshbach spectroscopy reveals
an extraordinary rich loss spectrum with six loss resonances already in a narrow magnetic-field range up to
3 G. Finally, we demonstrate the application of a low-field Feshbach resonance to produce a tunable dipolar
Bose-Einstein condensate and we observe its characteristic d-wave collapse.

DOI: 10.1103/PhysRevLett.108.210401 PACS numbers: 03.75.Nt, 37.10.De, 51.60.+a, 67.85.Hj

Ultracold quantum gases have proven to be ideal quantum gas experiments, Feshbach resonances and scat-
systems for observing spectacular many- and few-body tering lengths are a priori unknown. Magnetic lanthanides
quantum effects. The large majority of these phenomena such as erbium with their large magnetic moments and
rely on the high degree of control over the interparticle their non-S electronic ground states present a completely
interaction achieved with ultracold atoms. In the widely unexplored terrain in ultracold scattering physics. Here the
used alkalis, ultracold atoms interact isotropically via a anisotropic interaction is expected to give rise to novel
short-range contact potential. A novel exciting frontier in scattering scenarios, which are not accessible with alkali
quantum gas experiments is to access unexplored physical atoms [18,19].
scenarios based on the anisotropic and long-range dipole- In this Letter, we report on the attainment of
dipole interaction (DDI) [1,2]. A dipolar quantum gas is Bose-Einstein condensation of erbium atoms and on the
expected to exhibit fascinating phenomena, including observation of Feshbach resonances in the region of low
novel many-body quantum phases [3–6]. The DDI acts magnetic fields. We obtain pure optically trapped BECs of
in systems having sizable electric or magnetic dipole 168
Er containing 7
104 atoms. The remarkably high
moments [1]. efficiency of evaporative cooling in a standard optical
In the context of ultracold atomic quantum gases, dipole trap indicates favorable scattering properties of the
pioneering experimental work on strong DDI has been 168 Er isotope. In addition, the magnetic Feshbach spectros-

carried out with chromium atoms [7–9]. Magnetic lantha- copy provides first valuable information on the scattering
nides offer new possibilities for dipolar physics. In such behavior of submerged-shell atoms at ultralow tempera-
systems, the combination of a large magnetic moment and tures. Moreover, we demonstrate low-field Feshbach tun-
a large atomic mass leads to a particularly strong dipolar ing of the contact interaction in our strongly dipolar BEC.
character. The demonstration of the first magneto-optical Our experimental procedure to create a BEC of Er
trap of erbium atoms [10] stimulated growing interest in follows a simple and straightforward scheme, inspired by
such species for quantum gas experiments. Very recently, a work on Yb atoms [20,21]. Our starting point is the
Bose-Einstein condensate (BEC) and a degenerate Fermi narrow-line yellow magneto-optical trap (MOT) described
gas of dysprosium have been produced [11,12]. We choose in our very recent work [22]; it operates on the 583 nm line
erbium as a promising candidate for experiments on dipo- (natural linewidth 190 kHz). We choose this approach
lar quantum gases. This species has a number of very because narrow-line MOTs permit us to obtain samples
appealing features, including a large magnetic moment
with a large number of atoms at temperatures in the lower
of 7 times the Bohr magneton, several stable isotopes, a microkelvin region. This allows a direct and efficient
rich energy level scheme [13] with a non-S electronic transfer of atoms into optical dipole traps without the
ground state [14], and interesting cold collisional need for additional cooling stages [20,21]. Our MOT gives
phenomena [15,16]. about 108 atoms at a temperature of 15
K [23].
In strongly magnetic atoms, the competition between the An additional very advantageous feature of our approach
DDI and the contact interaction is very important and gives is that the MOT light automatically pumps the atoms into
rise to many intriguing phenomena. The contact interaction the lowest Zeeman sublevel mJ ¼ 6, where mJ is the
is determined by the s-wave scattering length a and can be projection quantum number of the total electronic angular
often tuned with external magnetic fields via Feshbach momentum J ¼ 6. This effect results from the interplay
resonances [17]. Tuning of a also controls the balance of between gravity and weak radiation pressure, which leads
these two interactions. In the case of a novel species in to a spatial down shift with respect to the zero of the

0031-9007=12=108(21)=210401(5) 210401-1 Ó 2012 American Physical Society

 week ending
PRL 108, 210401 (2012) PHYSICAL REVIEW LETTERS 25 MAY 2012

magnetic quadrupole field [24] and thus to a preferential (a)

absorption of the vertical MOT beam with  polarization T=1100 nK
[22]. The polarization of the sample is confirmed by
Stern-Gerlach-type measurements.
Our optical dipole trap geometry follows the concepts
originally applied in experiments with Yb BEC [20]. The
trap is created by crossing a tightly confining horizontal
beam (y axis) and a less focused vertical beam (z axis). The (b)
basic idea is that initially the atoms are predominantly 408 nK
trapped by the horizontal beam, whereas the vertical
beam provides confinement relevant in the final stage of

linear density (108 m-1)

evaporation. The horizontal beam is derived from a 100 W
broadband Yb fiber laser operating at 1075 nm and has an
initial power of 10 W. The beam has an elliptic cross
section with a waist of 30ð40Þ
m along the vertical (c)
(horizontal) direction. The vertical beam is produced by
222 nK
a 10 W Yb fiber laser source at 1064 nm and has an initial
power of 8 W. The beam profile is elliptic with a waist of
55ð110Þ
m along (perpendicular to) the axis of the hori-
zontal beam.
We load the dipole trap during the MOT compression
0.4
phase. We observe that the time period in which the com- (d)
pressed MOT and dipole trap coexist is crucial for efficient 0.3
loading. The number of atoms in the optical dipole trap
exponentially approaches its maximum value with a time 0.2

constant of about 150 ms. After 600 ms of loading, we turn

0.1
off the MOT beams and the quadrupole magnetic field, and
we switch on a weak homogeneous magnetic field of about 0
300 mG along the vertical direction to preserve the spin 1mm position (µm)
polarization of the sample. We obtain 2:6
106 atoms at a
temperature of 42
K in the optical dipole trap. At this FIG. 1 (color online). Absorption images and integrated
point, the atoms are mainly trapped by the horizontal density profiles showing the BEC phase transition for different
beam. We measure oscillation frequencies ðx ; y ; z Þ ¼ evaporation times. The absorption images are an average of four
ð1:3; 0:016; 1:95Þ kHz; the potential depth is estimated to images taken after 24 ms of expansion. The color bar shows the
optical density. The solid lines are fits to the data using
be 560
K. The peak density and the peak phase-space
Gaussian (a), bimodal (b) and (c), and Thomas-Fermi (d)
density are 1:7
1013 cm3 and 1:6
104 , respectively. distribution. The dotted lines represent the Gaussian part of
These are our starting conditions for the evaporative the bimodal fit, describing the thermal atoms. From the fit
cooling process. we extract: N ¼ 3:9
105 , T ¼ 1100 nK (a), N ¼ 2:1
105 ,
Forced evaporative cooling is performed by reducing the T ¼ 408 nK (b), N ¼ 1:6
105 , T ¼ 222 nK (c), N ¼
power of the trapping beams in a nearly exponential ramp. 6:8
104 (d), where N is the total atom number. For (b) and
The overall evaporation sequence has a duration of 5.5 s (c), we extract a condensate fraction of 5% and 20%, respectively.
[25]. We then turn off the trapping beams and let the atomic
cloud expand before applying standard absorption imag- By cooling the atomic sample below the critical tempera-
ing. For imaging, we illuminate the atomic cloud with a ture, we clearly observe that the atomic density distribution
50-
s probe beam pulse [26]. The probe beam propagates has a bimodal profile with a narrower and denser peak at the
horizontally at an angle of 14 with respect to the propa- center, which represents the BEC (b). By further evaporating
gation axis (y axis) of the horizontal trapping beam. the sample, the BEC fraction continuously increases
The phase transition from a thermal cloud to BEC mani- (c) until the thermal component is not anymore discernible
fests itself in a textbooklike bimodal distribution in the time- and an essentially pure BEC is formed with 7
104
of-flight absorption images. Figure 1 shows the absorption atoms (d).
images and the corresponding linear density profiles for To analyze our data, we fit a bimodal distribution to the
different final temperatures, i.e., for different stages of the integrated time-of-flight absorption images. This distribu-
evaporation. At higher temperatures the atomic distribution tion consists of a Gaussian function, which accounts for the
is thermal with the expected Gaussian profile resulting thermal atoms, and an inverted integrated parabolic func-
from the Maxwell-Boltzmann distribution; see Fig. 1(a). tion for the BEC component in the Thomas-Fermi limit.

210401-2
 week ending
PRL 108, 210401 (2012) PHYSICAL REVIEW LETTERS 25 MAY 2012

Just after the onset of quantum degeneracy (BEC fraction

5%), we measure trap frequencies of ðx ; y ; z Þ ¼
ð208; 70; 299Þ Hz, atom number of N ¼ 2:1
105 , and a
temperature of T ¼ 408 nK. The critical temperature of
417 nK as calculated from standard BEC theory (without
interaction shift) is consistent with this observation.
The evaporation efficiency is found to be remarkably
high as 3.5 orders of magnitude in phase-space density are
gained by losing a factor of 10 in atom number. This
observation already points to favorable scattering parame-
ters of the 168 Er isotope. First evaporative cooling experi-
ments on the most abundant 168 Er isotope reveal a lower
efficiency in the final stage of evaporation, suggesting that
a different strategy might be needed to reach BEC.
To gain insight into the ultracold collisional properties
of erbium we perform Feshbach spectroscopy [17] at low
magnetic fields. This measurement is done in a way that FIG. 2 (color online). Observation of Feshbach resonances in
allows us to identify both the poles and zero crossings of Er-Er collisions. The measured temperature (a) and atom
number (b) are plotted as a function of the magnetic field. The
Feshbach resonances [27,28]. The basic idea here is to minima in the atom number indicate the Feshbach resonance
prepare the system at a variable target value of the mag- poles, marked by the thin vertical lines. The maxima in the
netic field and then to rapidly (50 ms) decrease the depth of temperatures to the right of the three stronger loss features
the optical dipole trap by almost a factor of 2. The sample (arrows) are attributed to the respective zero crossings of the
stays near thermal equilibrium with an effective tempera- scattering length. The varying background in the atom number is
ture of 2:2
K but features a truncated energy distribution. presumably due to ramping effects caused by the sweep of the
We then let the system evolve at a constant trap depth magnetic field across the resonances.
for 250 ms, during which plain evaporative cooling and
inelastic losses can occur depending on the scattering the scattering length. For Er, a minimum cross section
length. We finally switch off the trap and take time-of- dip ¼ 8ð30a0 Þ2 results from universal dipolar scatter-
flight images to determine the temperature and number of ing, where a0 is the Bohr radius. The fact that we observe
the atoms. The measurement is then repeated for variable temperature maxima near the zero crossings suggests a
magnetic-field values. Such a Feshbach scan shows reso- dominant role of s-wave scattering and not of dipolar
nance poles as loss features and zero crossings as tempera- scattering. Preliminary cross-dimensional thermalization
ture maxima. measurements indeed point to a scattering length between
Figure 2 shows the loss spectrum and the corresponding 150 and 200a0 .
temperatures in the low magnetic-field range up to 3.2 G The existence of Feshbach resonances at low magnetic
[29]. Already in this narrow magnetic-field range, the loss fields makes the manipulation of the contact interaction in
spectrum is very rich. We identify six pronounced resonant the Er BEC very convenient and straightforward. As a
minima in the atom number that we interpret as being proof-of-principle experiment, we explore the controlled
caused by Feshbach resonances. For convenience, we de- d-wave collapse of the BEC, following the procedure
termine the resonance positions with Gaussian fits, yield- successfully applied by the Stuttgart group [7]. We first
ing 0.72, 0.91, 1.51, 2.16, 2.48, and 2.85 G. The loss produce a pure BEC by evaporative cooling at 1.2 G, which
features show different strengths and widths. For the three is above the position of the first broad Feshbach resonance
broader resonances at 0.91, 2.16, and 2.48 G, we also (0.91 G). Here we obtain 3
104 atoms in the BEC,
observe the appearance of temperature maxima to the right indicating that forced evaporation at this magnetic field is
of the loss minima (arrows in Fig. 2). These temperature slightly less efficient. We then ramp down the magnetic
maxima mark the zero crossings of the scattering length. field within 2 ms to a variable target value and let the
The other loss features are too narrow to provide clear sample evolve for 2 ms before switching off the trap. The
signatures of the zero crossing. From the difference in magnetic field is kept constant at its target value during the
positions between the minima in the atom number and first stage of the expansion (15 ms), where the main
the maxima in temperature we estimate the widths
of dynamics happens. We then set the magnetic field along
the resonances. We find
¼ 65, 60, and 180 mG for the the y axis and we image the atomic cloud after an addi-
resonances at 0.91, 2.16, and 2.48 G, respectively. tional 11 ms of expansion. Our results are summarized in
In a strongly dipolar atomic gas, universal dipolar scat- Fig. 3, where we show time-of-flight absorption images for
tering is present [30–32], so that the total cross section for different values of the target magnetic field. We observe a
elastic scattering does not vanish at the zero crossings of dramatic change in the shape of the condensate when the

210401-3
 week ending
PRL 108, 210401 (2012) PHYSICAL REVIEW LETTERS 25 MAY 2012

(a) (b) (c) 0.4 [4] B. Damski, L. Santos, E. Tiemann, M. Lewenstein, S.

Kotochigova, P. Julienne, and P. Zoller, Phys. Rev. Lett.
90, 110401 (2003).
0.3 [5] G. Pupillo, A. Micheli, H. P. Büchler, and P. Zoller, in
Cold Molecules: Theory, Experiment, Applications, edited
by R. V. Krems, W. C. Stwalley, and B. Friedrich (CRC
(d) (e) (f) 0.2
Press, Boca Raton, 2009), Chap. 12.
[6] D. S. Jin and J. Ye, Phys. Today 64, 27 (2011).
0.1 [7] T. Lahaye, J. Metz, B. Fröhlich, T. Koch, M. Meister, A.
Griesmaier, T. Pfau, H. Saito, Y. Kawaguchi, and M. Ueda,
Phys. Rev. Lett. 101, 080401 (2008).
0 [8] T. Lahaye, T. Koch, B. Fröhlich, M. Fattori, J. Metz, A.
Griesmaier, S. Giovanazzi, and T. Pfau, Nature (London)
FIG. 3 (color online). Absorption images showing the d-wave 448, 672 (2007).
collapse of the BEC. The field of view is 290
m
290
m. [9] B. Pasquiou, G. Bismut, E. Maréchal, P. Pedri, L. Vernac,
The images are an average of eight pictures. The color bar shows O. Gorceix, and B. Laburthe-Tolra, Phys. Rev. Lett. 106,
the optical density. The images are taken for different target 015301 (2011).
values of the magnetic field: 1.208 G (a), 0.963 G (b), 0.947 (c), [10] J. J. McClelland and J. L. Hanssen, Phys. Rev. Lett. 96,
0.942 G (d), 0.939 G (e), and 0.934 (f). We note that the actual 143005 (2006).
magnetic-field value might be slightly higher (10 mG) because [11] M. Lu, N. Q. Burdick, S. H. Youn, and B. L. Lev, Phys.
of ramping issues and eddy currents [7]. Rev. Lett. 107, 190401 (2011).
[12] M. Lu, N. Q. Burdick, and B. L. Lev, arXiv:1202.4444
[Phys. Rev. Lett. (to be published)].
magnetic field is reduced towards the zero crossing of [13] H. Y. Ban, M. Jacka, J. L. Hanssen, J. Readerand, and J. J.
the scattering length. At the magnetic field of evaporation, McClelland, Opt. Express 13, 3185 (2005).
the aspect-ratio of the cloud is close to the one observed at [14] The Er electronic configuration is [Xe] 4f12 6s2 . The
zero magnetic field; see Fig. 3(a). By changing the mag- electronic ground state is a 3 H6 state, meaning L ¼ 5
netic field to lower target values, the BEC shows a more (H state), S ¼ 1, and J ¼ 6. L, S, J denote the quantum
and more anisotropic shape (b) and (c). Below a critical numbers for the orbital angular momentum, for the spin,
magnetic-field value, the BEC develops a complicated and for the total electronic angular momentum, respec-
cloverleaf pattern (d)–(f) which is the striking signature tively. 168 Er has zero nuclear spin (I ¼ 0) and conse-
of the d-wave collapse in a dipolar BEC [7]. quently does not exhibit a hyperfine structure.
In conclusion, we have demonstrated the first BEC of [15] R. V. Krems, G. C. Groenenboom, and A. Dalgarno, J.
Phys. Chem. A 108, 8941 (2004).
Er atoms and the tunability of its interparticle interaction
[16] C. B. Connolly, Y. S. Au, S. C. Doret, W. Ketterle, and
via Feshbach resonances. Our scattering data provide first J. M. Doyle, Phys. Rev. A 81, 010702 (2010).
sensitive input to understand the complex collisional [17] C. Chin, R. Grimm, P S. Julienne, and E. Tiesinga, Rev.
behavior of submerged-shell atoms. The observation of Mod. Phys. 82, 1225 (2010).
the d-wave collapse is very encouraging in view of future [18] S. Kotochigova and A. Petrov, Phys. Chem. Chem. Phys.
experiments dedicated to the rich phenomena expected in 13, 19165 (2011).
strongly dipolar quantum gases. [19] A. Petrov, E. Tiesinga, and S. Kotochigova, arXiv:1203.4172.
We are grateful to F. Schreck, S. Stellmer, and [20] Y. Takasu, K. Maki, K. Komori, T. Takano, K. Honda, M.
B. Pasquiou (Innsbruck Sr team) for their continuous sup- Kumakura, T. Yabuzaki, and Y. Takahashi, Phys. Rev.
port. We thank the Yb teams in Kyoto and Tokyo for their Lett. 91, 040404 (2003).
[21] T. Fukuhara, S. Sugawa, and Y. Takahashi, Phys. Rev. A
advice, M. Springer for technical support, and E. Zupanič
76, 051604(R) (2007).
for contributions in an earlier stage of the experiments. [22] A. Frisch, K. Aikawa, M. Mark, A. Rietzler, J. Schindler,
This work is supported by the Austrian Ministry of Science E. Zupanic, R. Grimm, and F. Ferlaino, arXiv:1203.1460
and Research (BMWF) and the Austrian Science Fund [Phys. Rev. A (to be published)].
(FWF) through a START grant under Project No. Y479- [23] The MOT stage includes a 15 s loading phase, in which the
N20 and by the European Research Council under Project light is red detuned by 9:5 MHz, and a compression
No. 259435. phase, in which we reduce the detuning to 0:8 MHz and
the MOT beam intensity by about a factor of 400. The
compression phase is used to obtain colder and denser
samples and it is optimized for loading efficiency into the
dipole trap.
[1] T. Lahaye, C. Menotti, L. Santos, M. Lewenstein, and T. [24] H. Katori, T. Ido, Y. Isoya, and M. Kuwata-Gonokami,
Pfau, Rep. Prog. Phys. 72, 126401 (2009). Phys. Rev. Lett. 82, 1116 (1999).
[2] M. A. Baranov, Phys. Rep. 464, 71 (2008). [25] Initially only the power of the horizontal beam is reduced
[3] K. Goral, K. Rzazewski, and T. Pfau, Phys. Rev. A 61, to its final value of 230 mW, while the vertical beam power
051601 (2000). is reduced during the last 600 ms of evaporation to its final

210401-4
 week ending
PRL 108, 210401 (2012) PHYSICAL REVIEW LETTERS 25 MAY 2012

value of 1.5 W. estimate a short-term stability of 1 mG and a long-term

[26] We use the strong blue transition at 401 nm for imaging. stability of about 10 mG.
[27] S. Jochim, M. Bartenstein, G. Hendl, J. H. Denschlag, R. [30] V. Roudnev and M. Cavagnero, Phys. Rev. A 79, 014701
Grimm, A. Mosk, and W. Weidemüller, Phys. Rev. Lett. (2009).
89, 273202 (2002). [31] J. L. Bohn, M. Cavagnero, and C. Ticknor, New J. Phys.
[28] M. Zaccanti, C. D’Errico, F. Ferlaino, G. Roati, M. 11, 055039 (2009).
Inguscio, and G. Modugno, Phys. Rev. A 74, 041605 [32] K. K. Ni, S. Ospelkaus, D. Wang, G. Quemener, B.
(2006). Neyenhuis, M. H. G. de Miranda, J. L. Bohn, J. Ye, and
[29] The magnetic field is calibrated via rf-spectroscopy D. S. Jin, Nature (London) 464, 1324 (2010).
between Zeeman sublevels. From the calibration we

210401-5