Resonant control of spin dynamics in ultracold quantum gases by microwave dressing

0.27 MB·English

by  Fabrice Gerbier

Resonant control of spin dynamics in ultracold quantum gases by microwave dressing
Fabrice Gerbier,∗ Artur Widera, Simon F¨olling, Olaf Mandel, and Immanuel Bloch
Institut fu¨r Physik, Johannes Gutenberg-Universit¨at, 55099 Mainz, Germany.
(Dated: February 6, 2008)
We study experimentally interaction-driven spin oscillations in optical lattices in thepresence of
an off-resonant microwave field. We show that the energy shift induced by this microwave field
6 can be used to control the spin oscillations by tuning the system either into resonance to achieve
0 near-unity contrast or far away from resonance to suppress the oscillations. Finally, we propose a
0 schemebased on thistechniqueto createa flat sample with eithersingly- or doubly-occupiedsites,
2 starting from an inhomogeneous Mott insulator, where singly- and doubly-occupied sites coexist.

n
PACSnumbers: 03.75.Mn,03.75.Lm,03.75.Qk
a
J

8
The fact that coupling an atomic system to a strong m=-1 m=0 m=+1
] radiation field can modify its level structure is well- a b
er known,leadingtosucheffectsastheAutler-Townesdou- F=2
h blet or the Bloch-Siegert shift (see, e.g., [1]). Today, (1)
z
t this phenomenon is often explained in a dressed-atom H
o G
t. picture, where the bare atomic states are replaced by 6.8 wµ
a eigenstates of the total atom-plus-field system. For cold ~ ω (2)
m atoms, this picture helps to understand how magnetic F=1
m=-1
- trapping potentials can be engineered with off-resonant m=0 m=+1 (3)
d
radio-frequency or microwave radiation [2], allowing for
n
example to create nearly two-dimensional [3] or double-
o
c well trapping potentials [4, 5].
[ Here, we show how an off-resonant microwave field FIG. 1: Anoff-resonant microwave field with approximately

1 (hereafter called dressing field) can be used for coher- π-polarizationandfrequencyωµw ((a))isusedtocontrolthe
detuningforcoherentspinoscillationsinopticallattices. Case
v ent control of the dynamics of a spinor quantum gas.
(b1) corresponds to zero or weak microwave intensity. With
1 Ultracold gases with internal (spin) degrees of freedom
properly chosen microwave parameters, one can compensate
5
[6] are novel quantum systems where atom-atom inter- thisdetuningandachievefullresonance(b2)orovercompen-
1
actions play a key role. In an optically trapped gas of sate to suppress thespin-changing collisions (b3).)
1
0 spin-f = 1 atoms, spin-dependent interactions can pro-
6 mote anatompair fromaninitial two-atomstate ,where
0 both atoms are initially in the m=0 magnetic sublevel, the topic of spin oscillations in optical lattices. Control-
/
t to a final state with one atom in m = +1 and the other ling the spin oscillations enables one to use them as a
a
inm= 1[7,8,9,10,11]. Thecoherenceofthisprocess tool for quantum state detection, preparation or manip-
m −
was demonstrated in [12, 13] for atom pairs in optical ulation. Ina separatepaper,we have describedhow this
-
d lattices and in [14, 15] for a seeded Bose-Einstein con- technique can be used to probe atom number statistics
n densate. In the former case, the dynamics is controlled across the superfluid to Mott insulator (MI) transition
o by the energy mismatch, or detuning, between the ini- [16]. We propose here to use it in order to create an
c
tial and final spin states of the colliding atom pair. This homogeneous MI with singly-occupied sites out of a in-
:
v detuningoriginatesnotonlyfromdifferentquadraticZee- homogeneous system where singly- and doubly-occupied
Xi manshiftsintheinitialorfinaltwo-atomstates,butalso sites coexist.
fromdifferentinteractionenergies. Ifpositive,asin87Rb In our case, the microwave is relatively weak, so that
r
a [10,13,15],this“residual”interactiondetuningprevents the dressed states almost coincide with the bare atomic
to reach the resonance for spin oscillations by magnetic states and the level shifts can be calculated perturba-
means. tively [1]. We consider a 87Rb atom in a constant and
Inthispaper,weshowhowtotunethespinoscillations homogeneous magnetic field B, plus an arbitrary po-
inandoutofresonanceusingtheenergyshiftinducedby larized microwave field with amplitudes B(µw) , with
q q
an off-resonant dressing field (see Fig.1). We apply this q = 0, 1 for π and σ± polarizations, re{spectiv}ely. Far
±
technique to an ensemble of atom pairs in a deep optical from any hyperfine resonance, we find the level shift for
lattice and initially polarized in the m=0 sublevel. We the state F =1,m as
| i
showthattheconversiontoanensembleofm= 1pairs
canbe achievedwith nearunit efficiency, evenin±a finite ~Ω2 µ B
∆E πf x= B . (1)
magneticfield. Thesignificanceofthisworkgoesbeyond m ≈ 4∆ m(cid:18) ~∆ (cid:19)
2
1 . After these pulses, the magnetic field is eventu-
y − i
sit0,15 ally ramped down to a final value (typically B = 0.4G
n
a bde0,10 for the experiments described here). Subsequently, colli-
al0,05 sional spin oscillations take place at each individual lat-
c
pti0,00 tice site where an atom pair is present. As discussed in
O
-600 -200 0 200 600 [12,13],theatompairpreparedwithbothatomsinitially
Position(µm)
in m = 0 behaves as a two-level system, and undergoes
Rabi-likeoscillationsat aneffective oscillationfrequency
FIG. 2: (a) Absorption image of an expanded cloud after
tosc = 0 ms (upper image) and tosc = 15 ms (lower image).
Thedashedlineindicatestheboxused tofitthem=0com- ~Ωeff = (~Ω0)2+(∆ǫ+Us)2. (3)
ponent. (b) Horizontal cut through the lower image in (a) p
(black line). Subtracting a Gaussian fit (red dashed line) to Here the coupling strength ~Ω0 = 2√2Us and the in-
the m=0 component yields a new image (blue line) used to teraction part of the detuning are both fixed by a spin-
count separately thepopulation in m=±1by integration. dependent interaction energy U = U˜(a a )/3, with
s 2 0
U˜ an on-site interaction energy per Bo−hr radius a
B
[13] and a the scattering length characterizing scatter-
In Eq. (1), ∆ = ω ω denotes the microwave de- F
µw − hf ing in the total spin-F channel [6, 12]. The quantity
tuning,whereω isthemicrowavefrequencyandwhere
ωhf isthehyperfiµwnefrequencysplittinginthe5S1/2 man- ∆enǫce=inǫ+Z1ee+maǫ−n1e−ner2gǫi0es∝ǫ Bb2ectowrereensptohnedisnittoialthaenddififfnear-l
ifold of Rubidium 87. We have introduced the Rabi m
states. Without dressing field, ∆ǫ = (µ B)2/2~ω is
frequency Ω = µ B(µw)/~ for the π-polarized transi- B hf
π B π determined by the magnetic field. In the presence of the
tion, with µ the Bohr magneton, and assumed that
B dressing field, the Zeeman sublevels shift according to
x=µ B/~∆ 1. The function
B ≪ Eq. (1), so that one obtains
3C I 1
fm(x)=Xq 4m,q Iπq 1−(m+ q2)x, (2) ∆ǫ+Us =~×α(Ω2π −Ω2res), (4)
accounts for different microwave polarizations and the whereα=(f+1(x)+f−1(x)−2f0(x))/4∆andwherethe
resonance Rabi frequency Ω solves
multi-level structure. Here C = F = 1,m;1,q F = res
m,q
|h |
2,m + q 2 are square moduli of Clebsch-Gordan coef- (µ B)2
cfircoiewnatvse, cia|onmdpIoqneisntt.hWe einctaenlisbirtayteofthtehemiqc−ropwoalavreizinedtenmsii-- ~αΩ2res+ 2~Bωhf +Us =0. (5)
tiesforeachpolarizationq bydrivingfast,single-particle When normalized to the total atom number, the oscilla-
Rabi flopping between the hyperfine states F = 1,m = tion amplitude for the lattice as a whole is reduced by a
|
1 and F = 2,m = 1+q . From this, we obtain factor n , corresponding to the fraction of atom pairs in
− i | − i 2
Iσ−/Iπ ≈ 0.33 and Iσ+/Iπ ≈ 0.02 for our experimental the lattice [12, 13]. Forourtrappingparametersandour
parameters. The peak Rabi frequency observed in our atom number, approximately half of the atoms are in a
experiment is Ωπ =2π×82(1) kHz. central region with two atoms per site (n2 ≈0.5) [12].
To observe the microwave-induced shift, a MI of We have investigated the resonantly dressed spin dy-
around 2 105 Rb atoms is first created in the F = namics at a magnetic field of 0.4 G, by varying the mi-
× |
1,m= 1 Zeemansublevel. This MI is heldina purely crowave power (or equivalently the dressing frequency),
− i
optical lattice potential at a magnetic field of 1.2 G. In andrecordingseveralperiodsofspinoscillationsforeach
this work, we use a typical lattice depth of 40Er, with power. The microwavefieldwas detuned by severalhun-
Er = h2/2Mλ2 and λ = 840nm the lattice laser wave- dred MHz to the red of the hyperfine resonances to sup-
length. Afteranexperimentalphasedescribedbelowand press population transfer to F = 2. Three representa-
a 12-ms-long free expansion including a magnetic gradi- tive examples are shown in Fig. 3a, corresponding to
ent pulse, we detect the population in each Zeeman sub- the cases illustrated in Fig. 1: the unperturbed case
level as shown in Fig. 2 [12, 13]. We find the population (Ω Ω ,squares),thenear-resonantcase(Ω Ω ,
π res π res
in the m=0 sublevelby fitting a Gaussianfunction to a circle≪s) where the microwave dressing compensa≈tes ex-
restricted area, by substracting it from the data, and by actlythe“bare”detuning,andtheovercompensatedcase
integrating over the cloud areas to find the populations (Ω Ω , triangles), where the dynamics is blocked
π res
in m= 1 (see Fig. 2). due ≫to the large detuning induced by the microwave
±
The main topic of this paper is the study of spin dy- dressing. We have found empirically that the data were
namics in a dressing field. In our experiments, the spin well described by a fitting function of the form
dynamics is initialized as described in [12, 13] by first
transferring the sample to F = 1,m = 0 by apply- N t
| i 0 =A 1 e−γofft +Bcos 2π e−γosct. (6)
ing two microwave π pulses starting from F = 1,m = N − (cid:18) T (cid:19)
− | (cid:0) (cid:1) osc
3
1
0.5
0.8
s
N 0.4 a ne
N)/-1 0.3 2 1 ationsianstat 0.6
+ 0.2 ulm 0.4
1 pe
+ oe
N PZ 0.2
( 0.1
0 3 0
0 100 200 300
Time(ms)
0 10 20 30 40 50 60 70 80
Time(ms)
FIG. 4: Spin oscillations recorded for a magnetic field B =
30
) 1.2G.FilledcirclesdenotethepopulationintheZeemansub-
ms 25 b statem=0,andhollowcirclesthesumofthepopulationsin
2
( substates m=±1.
d 20 3
o
eri 15 1
P
10 expected behavior. The fit parameters include the same
0 5 10 15 20 25 30 35 40 parameters α,Ω as in the previous case, plus an ad-
RabiFrequencyΩπ (kHz) ditional parametreesr n corresponding to the fraction of
2
atoms participating in the oscillation. The values for α
0.4
ne andΩ areconsistentwiththosededucedfromthefitto
od 0.3 c res
cillatimplitu 0.2 itshselimghetalsyusrmedaplleerritohda.nTthheeofrvaecrtailolnamn2pl=itu0d.5eonf2a≈tom0.4p0a(i3rs)
s
Oa 0.1 expectedintheMottphase. Wehaveperformedanother
measurementwhere the atomic sample is held at a mag-
0
0 5 10 15 20 25 30 35 40 netic field of 1.2G using the dressing field on resonance
RabiFrequencyΩπ (kHz) accordingtoEq.(5). Thespinoscillationcurveshownin
Fig. 4 yields an amplitude of n 0.47(2), much closer
2
≈
to the expectation than for the B = 0.4G case. We at-
FIG. 3: Spin oscillations at a magnetic field B = 0.4 G. (a) tribute part of this 15 % reduction in contrast to a
Spin oscillations without microwave dressing (squares), for a finite ramp time of t∼he magnetic field in the latter case,
resonantdressing(circles)andforanovercompensateddress-
inducing a small amount of adiabatic population trans-
ing (triangles). (b) and (c), period and amplitude of the
fer at the beginning of the oscillation [13]. From the
oscillation (circles), together with a fit to the expected Rabi
datashowninFig.4,wehavealsodeterminedthedamp-
dependance.
ing times γ ,γ and the offset value A, as defined in
osc off
Eq. 6. Within our experimental accuracy, we find that
The second term describes the spin oscillations with pe- they are independent of microwave power and magnetic
riod T , including a damping term with rate γ . The field. The average offset value is A 0.6, and the av-
osc osc
first term describes a slowly varying offset approaching erage damping times γ 140s−1 a≈nd γ 160s−1.
osc off
the value A with a time constant γ−1. Although the origin of the≈damping is not enti≈rely clear,
off
The measured period of the oscillation is plotted in the fact that γ γ indicates a correlation between
osc off
≈
Fig. 3b as a function of the Rabi frequency Ω of the the damping of the oscillations and the rise of the offset.
π
dressing field. The period is fitted to the expected Possible damping mechanisms can be either local (i.e.
form of the Rabi frequency given in Eq. (3,4) treat- involving only on-site processes) or non-local (involving
ing U , Ω and α as free parameters. We find good tunneling to a neighboring site, particularly effective if
s res
agreement between the expected 2πα = 0.069s, Ω = atoms are transferredto excited Bloch bands by heating
res
2π 23.4kHzandthe measuredvalues 2πα=0.066(1)s, processes). On-site damping mechanisms could possibly
×
Ω = 2π 22.8(1)kHz. The “bare” Rabi frequency be off-resonant Raman transitions between the m = 1
res
× ±
U /h = 10.5(1)Hz is in excellent agreement with an sublevels induced by the lattice beams. We believe the
s
independent measurement of the spin-dependent colli- limitations of the present experiment to be of technical
sionalcoupling,whichweredonewithoutadressingfield origin. In principle, even longer coherence times could
[12, 13]. be achieved.
In Fig. 3c, we show the oscillation amplitude versus Because of these long coherence times, the technique
microwave Rabi frequency Ω , together with a fit to the presented here could find applications beyond the study
π
4
singly- doubly- tem with only singly-occupied sites in the ground state.
occupied occupied This would help to remove local defects with two atoms
F=2
onasite,thusprovidinganovelcoolingtechniqueunique
F=1
to the MI state.
We acknowledgesupportfromthe DFG,fromAFOSR
and from the EU under the OLAQUI program.
∗
email: [email protected]
remove
[1] C. Cohen-Tannoudji, J. Dupont-Roc, and G. Grynberg
(Wiley VCH, 1997).
[2] R.J.C.Spreeuw,C.Gerz,L.S.Goldner,W.D.Phillips,
S. L. Rolston, C. I. Westbrook, M. W. Reynolds, and
I. F. Silvera, Phys. Rev.Lett. 71, 3162 (1994).
time m =-1 0 +1 [3] Y. Colombe, E. Knyazchyan, O. Morizot, B. Mercier,
F
V.Lorent,andH.Perrin,Europhys.Lett.67,593(2004).
[4] P. Treutlein, P. Hommelhoff, T. Steinmetz, T. W.
FIG. 5: Illustration of the proposed scheme to create a ho-
H¨ansch, and J. Reichel, Phys. Rev. Lett. 92, 203005
mogeneousMIwithonlysingly-occupiedsitesstartingfroma
(2004).
sample with singly- and doubly-occupied Mott plateaus. Af-
[5] T. Schumm, S. Hofferberth, L. M. Andersson, S. Wil-
ter half a spin oscillation period, atoms in singly-occupied
dermuth,S.Groth,I.Bar-Joseph,J.Schmiedmayer,and
sites are transferred into the upper hyperfine level. An-
P. Kru¨ger, NaturePhysics 1, 57 (2005).
other half-spin oscillation brings back atoms in doubly oc-
[6] T.-L. Ho, Phys.Rev.Lett. 81, 742 (1998).
cupied sites to |F = 1,m = 0i, while a second microwave
[7] D. M. Stamper-Kurn, H.-J. Miesner, A. P. Chikkatur,
pulse is used to transfer atoms in singly-occupied sites to
S. Inouye,J. Stenger, and W.Ketterle, Phys.Rev.Lett.
|F = 1,m = −1i. Another spin oscillation nearly creates
83, 661 (1999).
the homogeneous n = 1 system, the only difference being a
[8] C. K. Law, H. Pu, and N. P. Bigelow, Phys. Rev. Lett.
supplementary atom in |F =1,m=+1i in doubly occupied
81, 5257 (1998).
sites. This can be corrected by transferring these atoms to
[9] H. Schmaljohann, M. Erhard, J. Kronj¨ager, M. Kottke,
F = 2, where a resonant laser beam can be used to remove
S. van Staa, L. Cacciapuoti, J. J. Arlt, K. Bongs, , and
them out of the trap.
K. Sengstock, Phys. Rev.Lett. 92, 040402 (2004).
[10] M.-S. Chang, C. D. Hamley,M. D.Barrett, J. A.Sauer,
K. M. Fortier, W. Zhang, L. You, and M. S. Chapman,
Phys. Rev.Lett. 92, 140403 (2004).
of spin oscillations. For example, Refs. [17, 18, 19] dis-
[11] T. Kuwamoto, K. Araki, T. Eno, and T. Hirano, Phys.
cuss how spin squeezing experiments could benefit from
Rev. A 69, 063604 (2004).
such a control. In Fig.5, we show how a homogeneous
[12] A.Widera,F.Gerbier,S.F¨olling,O.Mandel,T.Gericke,
MI with one atom per site only can be created, starting and I.Bloch, Phys. Rev.Lett. 95, 190405 (2005).
from of an inhomogeneous system where Mott plateaus [13] A.Widera,F.Gerbier,S.F¨olling,O.Mandel,T.Gericke,
with n = 1,2 coexist. This could be important for the and I.Bloch, in preparation (2005).
implementationofparallelcollisionalquantumgatesina [14] J. Kronj¨ager, C. Becker, M. Brinkmann, R. Walser,
P. Navez, K. Bongs, and K. Sengstock, cond-
lattice [20, 21], where an inhomogeneous filling degrades
mat/0509083 (2005).
the fidelity of the process (see also [22]). Similarly, one
[15] M.-S.Chang,Q.Qin,W.Zhang,L.You,andM.S.Chap-
canproduceanarrayofdoubly-occupiedsitesinm= 1,
man, Nature Physics 1, 111 (2005).
−
as first proposed in [23] (see also [12, 24]). [16] F. Gerbier, A. Widera, S. F¨olling, O. Mandel, and
In conclusion, we have demonstrated resonant tuning I. Bloch, cond-mat/0511080 (2005).
[17] H.PuandP.Meystre,Phys.Rev.Lett.85,3987(2000).
ofspinoscillationsinopticallatticesusinganoff-resonant
[18] L.-M.Duan,A.Sørensen,J.I.Cirac,andP.Zoller,Phys.
microwave field. This provides dynamical control over
Rev. Lett.85, 3991 (2000).
the spinoscillations,whichtogetherwiththe longcoher-
[19] A. Sorensen, L.-M. Duan, J. I. Cirac, and P. Zoller, Na-
ence times enables to use them as a tool for detection, ture 409, 63 (2001).
preparation and manipulation of the quantum state of [20] D.Jaksch,H.J.Briegel,J.I.Cirac,C.W.Gardiner,and
the many-atom system. As an illustration, we propose a P. Zoller, Phys. Rev.Lett. 82, 1975 (1999).
filteringtechniqueproducingaflatMIwithasingleatom [21] O. Mandel, M. Greiner, A. Widera, T. Rom, T. W.
H¨ansch, and I. Bloch, Nature425, 937 (2003).
per site, starting fromasystem withsingly- anddoubly-
[22] P. Rabl, A. J. Daley, P. O. Fedichev, J. I. Cirac, and
occupied sites. From a thermodynamical point of view,
P. Zoller, Phys. Rev.Lett. 91, 110403 (2003).
this flat MI would be metastable with a lifetime on the
[23] A. Widera, O. Mandel, M. Greiner, S. Kreim, T. W.
order of the tunneling time in the system. This method H¨ansch,andI.Bloch,Phys.Rev.Lett.92,160406(2004).
couldalsobe appliedtoathermodynamicallystablesys- [24] G. Thalhammer, K. Winkler, F. Lang, S. Schmid,
5
R. Grimm, and J. HeckerDenschlag, cond-mat/0510755
(2005).