Production of long-lived atomic vapor inside high-density buffer gas

0.37 MB·English

by  A. O. Sushkov

Production of long-lived atomic vapor inside high-density buffer gas
A. O. Sushkov1,∗ and D. Budker1,2
1Department of Physics, University of California at Berkeley, Berkeley, California 94720-7300
2Nuclear Science Division, Lawrence Berkeley National Laboratory, Berkeley, California 94720
(Dated: February 1, 2008)
Atomic vapor of four different paramagnetic species: gold, silver, lithium, and rubidium, is pro-
ducedandstudiedinsideseveralbuffergases: helium,nitrogen,neon,andargon. Theparamagnetic
atoms are injected into the buffer gas using laser ablation. Wires with diameters 25 µm, 50 µm,
8 and 100 µm are used as ablation targets for gold and silver, bulk targets are used for lithium and
0 rubidium. Thebuffergascoolsandconfinestheablatedatoms,slowingdowntheirtransporttothe
0 cell walls. Buffergas temperaturesbetween 20 K and 295 K,and densities between 1016 cm−3 and
2 2×1019 cm−3 areexplored. Peakparamagneticatomdensitiesof1011 cm−3 areroutinelyachieved.
The longest observed paramagnetic vapor density decay times are 110 ms for silver at 20 K and
n
4 ms for lithium at 32 K. The candidates for the principal paramagnetic-atom loss mechanism are
a
J impurities in the buffergas, dimerformation and atom loss on sputtered clusters.

8
PACSnumbers: 32.10.-f,34.50.-s,33.57.+c,37.20.+j
2

]
h I. INTRODUCTION reason for this is the insufficient saturated vapor pres-
p sureofthealkalisbelowroomtemperature,whichmakes
- experiments with sealed vapor cells extremely hard.
m The ability to achieve long-lived coherences in an en-
There are, however, compelling theoretical reasons
o sembleofatomsormoleculesisatthecoreofmanyofto- to believe that buffer-gas collisional relaxation cross-
t day’s atomic physics experiments. Such experiments are
a sections should drop quickly as the temperature is low-
at the forefront of both technological applications, such
. ered. Thisrelaxationoccursduetothespin-rotationcou-
s as atomic magnetometers [1, 2], spin-exchange optical
c pling (which arises from spin-orbit interaction [14, 15,
pumping of noble-gas nuclei [3], and quantum computa-
i 16]):
s tion [4], and more fundamental research,such as tests of
y
h Lorentzinvariance[5,6]andsearchesforviolationofthe H =γ(R)S N, (1)
p discrete symmetries of nature [7]. In room-temperature ·
[ experiments involving ground-state electron-spin coher-
where γ is the interactionstrength,R is the distance be-

2 ences, alkali atoms are most often used inside sealed va- tween the colliding alkali and buffer-gas atoms, S is the
v porcells, whichareeithercoatedwithananti-relaxation alkalielectronspin, and N is the rotationalangular mo-
3 material, such as paraffin [8, 9], or filled with a buffer mentumoftheirrelativemotion. Inthepresentwork,he-
4 gas, such as helium [3]. An alkali atom inside such a lium is used as the buffer gas at cryogenictemperatures,
7 cell is prepared in a coherent state by optical pumping,
and several other buffer gases are used at higher tem-
1 the coherences here are between ground-state magnetic
peratures also. Formation of Van der Waals molecules is
8. sublevels. In an anti-relaxation coated cell, it can then neglected [17], and the spin-relaxation cross-section can
0 experiencethousandsofvelocity-changingcollisionswith
beevaluatedinthesemi-classicalbinarycollisionapprox-
7 the cell walls before de-cohering. Coherence times of
imation:
0 500ms havebeendemonstratedinsuchcells,see,forex-
v: ample,Ref.[10]. Inabuffer-gasfilledcell,depolarization 8πM2 ∞ ∞ γ(R)dR 2
Xi datuoemtsoncoelllo-nwgaelrltcroalvliesliobnaslliisstiscuapllpyr,ebsusetdh,asvienctoedthiffeuaselkianl-i σ(E)= 3h¯4 Z0 b3db(cid:12)(cid:12)(cid:12)Zr0 (1−b2/R2)−V(R)/E(cid:12)(cid:12)(cid:12) ,
r side the buffer gas, whose density ranges typically from (cid:12) p (2(cid:12))
a (cid:12) (cid:12)
1016 cm−3 up to 1019 cm−3. However, as the buffer-gas whereV(R)is theinter-atomicpotential,bisthe impact
density is increased, slowing down alkali diffusion, col- parameter, E is the collision energy, r0 is the distance
lisional de-coherence starts to dominate. This refers to of closest approach, and M is the reduced mass of the
spin-relaxation of the alkali atom in a collision with a collidingatoms. Thekeyinputintothisexpressionisthe
buffer-gas atom due to spin-orbit interaction [11]. Co- R-dependence of γ(R). It is possible to estimate γ(R)
herence times of 20 ms have been achieved in buffer-gas by approximating the alkali valence electron wavefunc-
filled cells [12, 13]. Collisional relaxation cross-sections tionat largedistancesr fromthe coreby the hydrogenic
of the alkali atoms have been studied at room temper- expression (in atomic units)
ature and above, but not for lower temperatures. The
φ(r) rν−1e−r/ν, (3)
∝
where ν is the effective principal quantum number of
∗Electronicaddress: [email protected] the ground-state valence electron. Using the results of
2
Ref. [15], γ(R) can be estimated with logarithmic preci- will limit the achievable relaxation times, such as spin-
sion, resulting in the exponential R-dependence: destruction and spin-exchange in collisions between the
alkali atoms themselves. It is much harder to estimate
γ(R) e−2R/ν. (4) the behavior of the corresponding cross-sections at low
∝
temperatures,andthe regionbetween4K and300K re-
This is in good agreement with the results of numerical mains experimentally unexplored. Data are available at
calculations presented in Ref. [15]. It is already possible higher temperatures for a variety of other buffer gases
to argue why the spin-relaxation cross-section decreases (such as N , 3He, Xe), a discussion of the the rele-
2
rapidly at low temperatures. As the temperature is low- vant temperature dependences is given in Ref. [20]. We
ered,thecollisionenergydecreases,andthealkali-helium note,however,thatoursimplemodelisnotapplicablein
closest approach distance increases due to repulsion be- thesecases,whereothereffects,suchasformationofVan
tween the He and the valence electron. Since the spin- der Waals molecules, contribute to the spin-destruction
rotation interaction of Eq. (4) drops exponentially, the cross-section.
corresponding spin-flip cross-section also decreases. We
As mentioned above, the reasonfor scarcity of experi-
make no attempt at calculating the magnitude of the
mentaldatabelowroomtemperatureisthevanishingsat-
relaxation cross-section, instead we estimate its temper-
uratedvaporpressureofthe alkalis. Therehas,however,
ature dependence, using Eq. (4) and the interatomic po-
been some recent experimental effort in this direction.
tential of the form
One technique that has been used to study alkali atoms
inhelium-filledcellsatcryogenicconditions,isloadingby
C
V(R)=AR2ν−2e−2R/ν 6, (5) light-induced atomic desorption (LIAD) from the liquid-
− R6
helium film that covers the cell walls below the helium
where the first (repulsion) term is taken to be propor- superfluid-transition temperature [21, 22]. These exper-
tional to the probability density of the alkali electron at iments were performed at 1.85 K, and spin-relaxation
the location of the helium atom, and the second term is times of minutes and even longer can be inferred from
the VanderWaals attraction. TheVander Waalscoeffi- thedata. Thisalreadysupportsthetheoreticalestimates
cientC isknownfromexperiment,andtheparameterA outlinedabove. Anothertechniquethathasrecentlybeen
6
fittedtoreproducetheexperimentallyobservedpotential welldevelopedisproductionofcoldatomicandmolecular
minimum. The result ofnumericalintegrationof Eq.(2) beams by laser ablation inside helium buffer gas at tem-
is plotted inFig. 1 forRb-He collisions. The interatomic peratures on the order of 5 K [23]. The same technique
potential parameters are obtained from Refs. [18, 19] has also been used for loading atoms and molecules into
magnetic traps [24, 25]. The experiments described in
the present paper fall somewhere in the middle between
1.0
these techniques. We use laser ablation to produce high
densities of atomic species inside helium buffer gas in a
n
axationits) 0.8 raadndgiteioonf ttoemcopoelriantgutrhesebabetlwateeedna2t0omKs,atnhde3b0u0ffeKr,gbausta,lsino
elu confines them by slowing down transport to the walls,
nal rarb. 0.6 where the atoms condense.
s collisiosection ( 0.4 II. APPARATUS
uffer-gacross- 0.2 The experimental setup is schematically shown in
B Fig. 2. We use a Janis model DT SuperVariTemp
pumped helium cryostat. Optical access is provided via
0.0
fusedquartzwindows,theinnermostwindowshavingthe
0 100 200 300 400 500
′′ ′′
diameter of 1 . A 2.5 -tall cylindrical copper cell, with
Temperature (K)
′′
inner diameter of 1.25 is mounted inside the cryostat,
in line with the windows. The cell prevents the ablated
FIG.1: TheestimatedscalingofRb-Hespin-relaxationcross- speciesfrombeingdepositedontheinsidewallsandwin-
section with temperature. dows of the cryostat sample space, which are hard to
′′
clean. The coppercell hasfour 1 -diameterfused quartz
If indeed, as shown in Fig. 1, the buffer-gas collisional windows, and several holes at the top and bottom to
relaxation cross-section drops rapidly with decreasing make sure that the buffer gas pressure inside the cell is
temperature, extremely long ground-state electron spin- equaltothatinsidethecryostatsamplespace. Theabla-
relaxation times can be achieved in alkali atoms inside tiontargetismountedonaholderinsidethe cell. Atthe
high-densityheliumbuffergasatlowtemperature. There top of the cryostat, the sample space is connected to a
′′
are, of course, other spin-relaxation mechanisms that Stokes Pennwalt rotary vane pump via a 3/4 -diameter
3
III. SILVER AND GOLD
Probe
light
source Four atomic species have been studied in our experi-
Buffer gas ment: lithium, rubidium, silver and gold. We start with
thediscussionofablationofsilverandgold. Theseatoms
Cu cell
Beam-shaping were chosen, on the one hand, because of their ground-
optics state electronic configurations, consisting of a single s
electron outside a filled shell. They have S ground
1/2
states, and their spectra are similar to those of alkali
atoms. On the other hand, unlike the alkalis, silver and
Cryostat gold are ductile, and thin wires (diameter as small as
25 µm)arereadilyavailable. Wedecidedtousesuchthin
Ablation target
wires as ablation targets for the following reasons. The
Interference filter Nd:YAG amount of material that has to be vaporized to create
laser an atomic density of 1011 cm−3 inside the cell of volume
Photodetector 20 cm3 is only 2 1012 atoms. The energy required to
×
evaporatethatamountofmaterialisapproximately1µJ.
The rest of the energy delivered to the target by the ab-
FIG. 2: A schematic top view of theexperimental setup.
lation pulse goes into the buffer-gas shock wave, ejects
macroscopic pieces of the target, or is conducted away
into the bulk of the target, heating it up, and heating
the buffer gas with it. Using the tip of a sub-millimeter
diameterwire asthe ablationtargetallowsthe heatcon-
duction away from the ablation spot to be minimized,
pumping arm. Also at the top, there is a connection becausethewire’scross-sectionalareathroughwhichthe
to an MKS Baratron Type 220B pressure gauge and a heat escapes is small. In addition, the thin wire has a
valve for letting in buffer gas froma pressurizedcylinder muchsmallersurfaceareathanaflattarget,sotheprob-
(99.995% purity). The temperature is monitored with a ability of ablated atoms to diffuse back and stick to the
LakeShore silicon-diode temperature sensor, model DT- target surface is minimized.
470-CU-13-1.4L, mounted on the top of the copper cell. Silver and gold wires of 50 µm diameter (purchased
LakeShoremodel321controller,combinedwithawound fromAlfaAesar)wereusedasablationtargets. Measure-
heater mounted next to the temperature sensor, allows ments were also taken using wires of 25 µm and 100 µm
temperature control during the experiment. diameter; wire thickness did not significantly affect ex-
perimental results in this range. The wire was mounted
vertically in the middle of the cell inside the cryostat
AQ-switchedSpectra-PhysicsDCR-11Nd:YAG laser, sample space, so that the tip of the wire was just below
operating at the fundamental wavelength (1064 nm), is the centerofthe cellwindows. Aiming the ablationlaser
usedtoablatethetargets. Thelaserbeampassesthrough focusatthetipofthewirewasdoneasfollows. Thelaser
a set of lenses to expand it (this is done to avoid dam- intensity was reducedby a neutraldensity filter to avoid
agingthe cryostatwindowsby the focusedbeam), andis ablation. When the wire was in the focus, it blocked
then focused to a spot on the ablation targetby a large- part of the beam, and a shadow was clearly visible on
numerical-aperture lens. The focus (the ablation spot) the beamstop,where the laserbeamwasprojectedafter
canbemovedaroundthesurfaceofthetargetbymoving it exited the cryostat. Once the laser was aimed directly
this lens, which is mounted on a three-coordinate trans- at the wire tip, the neutral density filter was removed
lation stage. Several different ablation-laser operating and an ablation shot was fired. As the material at the
regimes have been explored. It was found that the abla- wiretipwasevaporated,theablationspotwasmovedup
tionyieldisindependentofpulseenergywhenitisabove by translating the focusing lens.
30 mJ, but drops when the pulse energy is lower. Dou- Ablatedatomsweredetectedbymeasuringtheabsorp-
bling the frequency of the ablation pulse and operating tion of a light beam resonantwith the D1 atomic transi-
at 532nm hadno noticeable effect onthe ablationyield. tion. This lightbeam was producedby a hollowcathode
FiringthelaserwiththeQ-switchdisabledgavea“long” lamp(HCL) andfocusedwith a lens. A narrow-bandin-
100-µslightpulse. When sucha pulse wasusedforabla- terferencefilterwasusedtoselecttheD1atomicline. For
tion, the resulting atomic yield was at least an order of thesilverHCLa338-nmfilterwasused,andforthegold
magnitudelowerthanthatobtainedwiththeQ-switched HCLa270-nmfilterwasused. Ineachcasethetransmit-
laser. After exploring these ablation regimes,the follow- ted spectrum was measured with a Czerny-Turner grat-
ingparameterswerechosenfordatataking: pulseenergy ing spectrometer, confirming that only a single spectral
of50mJ,fundamentalwavelength(1064nm),Q-switched line was present. We also checked that the observed sig-
(10-ns light pulse). nal was indeed resonant atomic absorption rather than
4
30 10 the diffusion time td can be written as:
9
25 -3cm) 78 td ≃ 6RD2, (6)
cay time (ms) 1250 10Au density (10 123456 wtoDofhAe≃trhueeλaRvtc/oe≃3mlls≃1,w.5σavtlc/lms3is,σttiashnneHtdiher.etDrdHainseitrssaepnotλchreteisfcrrdotoihmffsesu-tssmhieoceentaipnorcnoof,breevffieeicbspieeatanhtmthe:
de 00 20 40 60 80 100 rms relative velocity, and nHe is the helium density. The
sity 10 Time (ms) diffusion time can be expressed as
n
de R2σ
u t tn . (7)
A 5 d ≃ 2v He
Indeed the diffusion time grows linearly with increas-
0 ing helium density, and the diffusion coefficient D
AuHe
0 2 4 6 8 10 12
for gold atoms in helium can be extracted from the
He density (1018 cm-3) slope. The result for the temperature of 295 K and at-
mospheric pressure is: D 0.5 cm2/s (which cor-
AuHe
≃
FIG. 3: Gold-atom density decay time as a function of he- responds to a reasonable value of the transport cross-
lium density at 295 K. The straight line is a linear fit to the section, σ 10−15 cm2). No published value for this
t
low-helium-density data, where diffusion dominates the gold diffusion coe≈fficient has been found in literature.
atom loss. The inset shows the evolution of gold-atom va- If diffusion to the walls were the only loss mechanism,
pordensity after a single ablation shot,for helium densityof
Au vapor lifetime would continue to grow linearly with
7×1018 cm−3. The line is an exponential fit to the density
increasing helium density, and reach about a second at
decay. the density of 3 1019 cm−3 . As evident from Fig. 3,
however, at he≈lium×density of about 1018 cm−3 another
Aulossmechanismstartstodominate,andtheAuvapor
some broad-band scattering: when a helium-neon laser lifetime decreases with increasing helium pressure. One
beam was used as the probe in place of the HCL, no candidate for this loss is dimer formation: Au + Au
→
absorption was detected. Au2. When such a dimer is formed, the Au atoms are
lost, since they are no longer resonant with the probe
The inset in Fig. 3 shows the time-evolution of gold-
laser. A collision between two gold atoms in vacuum
atom density after a single ablation shot is fired into the
cannot forma dimer, since both energyand momentum
tip of the gold wire. Gold atoms fill the cell after the
can not be conserved in such a process. A third body
ablation laser pulse. The photo-detector bandwidth is
is needed, and a helium atom can fulfill this role if it is
not wide enough to make an accurate measurement of
close enough to the colliding Au atoms. Dimer forma-
the filling time. This time is faster than 100 µs, and, in
tion rate then grows with growing helium density, and
the entire experimental range of buffer gas densities, is
this can explain the decrease of gold lifetime as helium
always much faster than the time it would take for Au
densityincreases. Simple estimatespredictdimerforma-
atomstodiffusefromthewiretothelocationoftheprobe
tionratesthatareaboutanorderofmagnitudetoosmall
beam, which is 5 mm away. The filling process must be
toexplaintheobservedgoldatomloss,butmoreaccurate
non-diffusive a shock wave propagates in the helium
− calculations are needed to rule out this mechanism.
gas after the ablation pulse, filling the cell with ablated
Another possible gold atomloss mechanism is capture
Auatoms. TypicalAudensitiesachievedareontheorder
by clusters: Au + Au Au . For this process there
of 1011 cm−3. After a maximum is reached, the atomic n → n+1
is no need to have a helium atom nearby, since excess
density starts to decrease. Exponential decay fits the
kinetic energy can be converted into vibrational energy
data well, and the loss of Au atoms from the buffer gas
of the cluster and subsequently carried away by colli-
can be characterized by the corresponding decay time.
sions of the cluster with helium atoms. Hence there is
The mechanismscausingthis Auatomdensity decayare
no dependence of the cross-section on the helium den-
discussed below.
sity. The interaction between a gold atom and a cluster
The evolution of Au atom density after an ablation is welldescribedbythe VanderWaalspotential, and,as
shot at each helium density and room temperature were a consequence of high cluster polarizabilities, the cross-
recorded, and the resulting decay times are plotted in sections are extremely large, on the order of 10−13 cm2,
Fig. 3. At small helium density, the Au loss time grows for clusters with n 10 [26]. If we take this value as the
≃
linearly with helium density. In this regime, Au loss due gold-atom capture cross-section, the number density of
to diffusion to the cell walls dominates. The atoms are clusters required to produce gold atom loss on the time
lost at the walls since the temperature is too low to sup- scale of 10 ms is 1011 cm−3. Where would these clusters
port any significant Au vapor pressure. An estimate for come from? As argued previously, dimer formation is
5
the temperature of 20 K [29]. The peak silver atomic
110 densitieswereontheorderof1011 cm−3. Datawerealso
100
taken at 80 K and 295 K with three other buffer gases:
90 nitrogen,neon,andargon. Itwasfoundthattheatomic-
s)
m 80 vapor decay time at a given buffer gas density does not
me ( 70 noticeablydepend onwhichofthese buffer gasesis used.
y ti 60
a
dec 50 T = 295 K
y T = 80 K
nsit 40 T = 20 K IV. LITHIUM AND RUBIDIUM
de 30
g
A 20 Studyingsilverandgoldatomsisdifficultbecausetheir
10 atomic transition frequencies lie in the UV: the D1 lines
0 are at the wavelengths of 338 nm for silver and 268 nm
0 5 10 15 20
for gold. The hollow-cathodelamps are sufficient for ab-
He density ( 1018 cm-3 )
sorption measurements, but for optical pumping a more
intense light source is needed. Such light sources (diode
FIG.4: Silvervapordensitydecaytimeasafunctionofhelium
lasers)arereadilyavailableforalkaliatoms. Wetherefore
buffergas density at different temperatures.
undertookastudyofalkali-atomablationwithourexper-
imental setup. Experiments with lithium and rubidium
were performed. For lithium the ablation target was in
too slow to lead to nucleation of an appreciable number
the form of 0.75-mm-thick and 19-mm-wide lithium foil
of clusters. The ablation process itself, however,is likely
(purchasedfromAlfaAesar),andforrubidiumthetarget
to produce such clusters [27, 28]. As the ablation laser
wasaroughly5-mm-thickand10-mm-widepieceofpure
hits the target, it locally heats the material, creating an
rubidium metal. Alkali metals are extremely reactive,
explosion. The explosion forms a shock wave, which ex-
andtheyoxidizequicklywhenexposedtoair. Therefore,
pands out into the helium gas. Some of the high-density
ablationtargetswerepreparedina gloveboxwith argon
gold vapor created by the explosion condenses into clus-
atmosphere. The target was mounted inside the copper
ters in the low-pressure region behind the shock. The
cell,levelwiththe cellwindows. The glovebox wasthen
higher the helium density, the greater the efficiency of
opened and the cell was quickly inserted into the cryo-
this condensation, the more clusters are produced, and
stat, which was immediately pumped out. In this way
the faster the rate of gold vapor loss. This may explain
the exposure of the alkali targets to air was minimized.
the trend observed in Fig. 3.
Ablatedatomsweredetectedbymeasuringtheabsorp-
The third possible loss mechanism is capture of a gold
tion of a laser beam resonant with one of the D1 or D2
atomby animpurity (suchas a hydrocarbonoranoxide
atomic transitions. Lithium atoms were probedwith the
molecule) that is either present in the buffer gas to start
light from a 671-nm laser diode (TOLD 9221M) inside
with, or gets ablated off the target surface. Such an im-
purity can easily have a large capture cross-section, and
the impurity density of 1011 cm−3 or more is not at all
implausible. ms) 4
A study of the temperature dependence of the atomic 4 me ( 3
vaporlifetimewasperformedusingsilveratoms. Thesil- y ti
bvlaeusrfferevraapgboalarst-daiotenntshoirtfeyaed5tee0cmaµypmetr-idamtiauemrseesat.reeTrphsilelovttetrreednwdiisrneaFirneigsh.iem4liiuflaomrr me (ms) 3 nsity deca 12
totheonesinFig.3,wheregoldvaporresultsareshown. ay ti Rb de 00 1 2
At small helium density, the silver vapor lifetime grows dec 2 He density (1018 cm-3)
linearly, and the main atom loss mechanism is diffusion y T = 295 K
sit T = 120 K
to the cell walls. The diffusion coefficient D for sil- n
ver atoms in helium can be extracted from AthgeHeslope of Li de 1 T = 32 K
thelineardependenceofdiffusiontimeonheliumdensity,
using Eq. (6). The result for the temperature of 295 K
and atmospheric pressure is: D 0.4 cm2/s. Silver 0
vapor lifetime reaches a maximAugmHea≃t n 1018 cm−3 0 1 2 3 4 5 6
He ≃ He density (1018 cm-3)
anddecreasesforlargerheliumdensities. Thesilveratom
lossmechanismsarelikelythesameasthoseforgold,de-
scribed above. It can be seen in Fig. 4 that at low tem- FIG. 5: Lithium vapor-density decay time as a function of
peratures,longeratomicvaporlifetimescanbe achieved. heliumbuffergasdensityatdifferenttemperatures. Theinset
The longest observedsilver vapor lifetime was 113ms at shows rubidium data at 295 K.
6
a temperature-controlled mount (Thorlabs TCLDM9) V. CONCLUSIONS AND OUTLOOK
paired with a TED-200 temperature controller and a
LDC-201-ULN current controller. A 750-MHz free spec-
Inconclusion,wehaveusedlaserablationtocreateand
tral range confocal Fabry-Perot spectrum analyzer (F-
study atomic vapors of silver, gold, lithium, and rubid-
P) was used for spectral diagnostics of the laser light.
ium inside helium andother buffer gases ina wide range
Tuning to the 7Li (92.5% abundance) D1 resonance line
of temperatures and buffer-gas pressures. Vapor densi-
at 671 nm was achieved by splitting off a part of the ties of 1011 cm−3 have been achieved. The longest mea-
beam, modulating the intensity with a chopper wheel
sured atomic vapor lifetimes are on the order of 110 ms
at 350 Hz, and aiming it into the cathode of a lithium
for silver [29] and 4 ms for lithium. For buffer-gas den-
hollow-cathode lamp (HCL). When the laser was tuned sity n 1018 cm−3 the atomic vapor loss is dominated
to the lithium resonance, the HCL current was modu- ≪
by diffusion to the cell walls, where the atoms condense.
latedatthechopperfrequency(opto-galvaniceffect)[30]. Increasing the buffer gas density above 1018 cm−3, how-
This modulation was detected with a lock-in amplifier,
ever,doesnotincreasetheatomicvaporlifetime,instead
referenced to the chopper-wheel frequency. The setup
the lifetime gets shorter. The possible loss mechanisms
for rubidium atoms was similar, with the 671-nm laser
that can cause this behavior are dimer formation, atom
diode replaced with a 780-nm laser diode, and the hol-
loss on clusters created during the ablation process, or
low cathode lamp replaced with a rubidium vapor cell,
atom loss on impurities in the buffer gas.
whichwasusedtotune the lasertothe RbD2transition
Toputourresultsincontext,letusconsiderthefigure
by detecting atomic fluorescence.
of merit of an atomic magnetometer, or any shot-noise
Thetime-evolutionoflithiumandrubidiumatomden-
limited precision measurement with an ensemble of N
sity after an ablation shot is very similar to that shown
atoms with coherence time τ [2, 32]:
on the inset in Fig. 3 for gold. The ablated atoms fill
the cell very quickly after the laser pulse, a maximum is
reached, and then the atomic density decreases. Expo- FOM √Nτ. (8)
∝
nential decay fits the data well, and the decay times are
plotted in Fig. 5 for lithium in a range of temperatures. The coherence times have not been measured in our ex-
Rubidium data at room temperature are shown in the periments so far (these measurements are currently in
inset. The trends are the same as those seen for gold progress). However, as argued at the beginning of this
and silver atoms. At low helium density the vapor loss paper, the buffer-gas collisional relaxation rates at cryo-
time increaseslinearly, in agreementwith Eq. (7), which genic temperatures are expected to be on the order of
describesdiffusiontothewalls. Butthemaximuminthe millihertz. If atom loss as measured in our experiments
decaytimenowoccursatlowerheliumdensity,ofapprox- proves to be the dominant mechanism for coherence re-
imately0.2 1018 cm−3. Athigherdensitiesanotherloss laxation, then coherence relaxation rates on the order of
×
mechanismdominates,keepingthedecaytimesshort. As several hertz are expected. With atomic vapor densities
discussed above, the possible candidates for this mecha- of 1011 cm−3 already demonstrated, the figure of merit
nism are dimer formation, atom loss on sputtered clus- for our system is competitive with the leading atomic
ters, and atom loss on impurities in the buffer gas. For magnetometers today [2]. Our setup, however, has the
an unknown reason, this mechanism is more efficient for advantage of operation in a wide range of temperatures:
alkali atoms than for silver and gold [31]. This leads to from 4 K to 300 K, while the vapor-cell based measure-
shorter vapor-loss times for the alkalis, on the order of ments can only be performed at room temperature and
4 ms at the maximum. Data were also taken at 295 K above.
with nitrogen as the buffer gas. It was found that the The authors thank Valeriy Yashchuk and Max Zolo-
lithium vapor decay time at a given buffer gas density torevformanyusefuldiscussions. Thisresearchhasbeen
does not noticeably depend on whether the buffer gas is supported by the National Science Foundation through
helium or nitrogen. grant 0554813.
[1] D. Budker, W. Gawlik, D. F. Kimball, S. M. Rochester, cm.
V.V.Yashchuk,andA.Weis,ReviewsofModernPhysics [5] C.J.Berglund,L.R.Hunter,J.D.Krause,E.O.Prigge,
74, 1153 (2002). M. S. Ronfeldt, and S. K. Lamoreaux, Physical Review
[2] D.BudkerandM.Romalis,NaturePhysics3,227(2007). Letters 75, 1879 (1995).
[3] W. Happer,Reviews of Modern Physics 44, 169 (1972). [6] T.W.KornackandM.V.Romalis,PhysicalReviewLet-
[4] D. Bouwmeester, A. K. Ekert, and A. Zeilinger, The ters 89, 253002 (2002).
physicsofquantuminformation: quantumcryptography, [7] I. B. Khriplovich and S. K. Lamoreaux, CP violation
quantum teleportation, quantum computation (Springer, without strangeness (Springer-Verlag, 1997).
Berlin;NewYork,2000),1sted.,dirkBouwmeester,Ar- [8] M. A.Bouchiat andJ.Brossel, Physical Review 147,41
tur K. Ekert, Anton Zeilinger, eds. ill. (some col.) ; 25 (1966).
7
[9] E. B. Alexandrov,M. V. Balabas, A. S. Pasgalev, A. K. J. S. Helton, S. V. Nguyen, D. Patterson, J. Petricka,
Vershovskii, and N. N. Yakobson, Laser Physics 6, 244 D. DeMille, and J. M. Doyle, Physical Review Letters
(1996). 95, 173201 (2005).
[10] D.Budker,V.Yashchuk,andM.Zolotorev,PhysicalRe- [24] J.Kim,B.Friedrich,D.P.Katz,D.Patterson,J.D.We-
view Letters 81, 5788 (1998). instein,R.DeCarvalho,andJ.M.Doyle,PhysicalReview
[11] R. A. Bernheim, The Journal of Chemical Physics 36, Letters 78, 3665 (1997).
135 (1962). [25] R. deCarvalho, J. M. Doyle, B. Friedrich, T. Guillet,
[12] S. Brandt, A. Nagel, R. Wynands, and D. Meschede, J. Kim, D. Patterson, and J. D. Weinstein, European
Physical Review A 56, R1063 (1997). Physical Journal D 7, 289 (1999).
[13] M. Erhard and H. Helm, Physical Review A 63, 043813 [26] V. V. Kresin, G. Tikhonov, V. Kasperovich, K. Wong,
(2001). andP.Brockhaus,JournalofChemicalPhysics108,6660
[14] Z. Wu, T. G. Walker, and W. Happer, Physical Review (1998).
Letters 54, 1921 (1985). [27] W.T.Nichols,G.Malyavanatham,D.E.Henneke,D.T.
[15] T.G.Walker,J.H.Thywissen,andW.Happer,Physical O’Brien, M. F. Becker, and J. W. Keto, Journal of
ReviewA 56, 2090 (1997). Nanoparticle Research 4, 423 (2002).
[16] E. I. Dashevskaya and E. A. Kobzeva, Optika i Spek- [28] S.I.AnisimovandB.S.Luk’yanchuk,Physics-Uspekhi
troskopiya30, 807 (1971). 45, 293 (2002).
[17] T. G. Walker, Physical Review A 40, 4959 (1989). [29] After this article was completed, measurements at
[18] C. Zhu, A. Dalgarno, S. G. Porsev, and A. Derevianko, slightly different experimentalconditions (lower temper-
Physical Review A 70, 032722 (2004). atures, for example) showed evenlonger silver atom life-
[19] G. Herzberg, Molecular spectra and molecular structure. times, up to 300 ms.
I. Spectra of diatomic molecules. (Van Nostrand, New [30] D.S.KingandP.K.Schenck,LaserFocus14,50(1978).
York,1950). [31] This may be due to the fact that only bulk targets were
[20] W. C. Chen, T. R. Gentile, T. G. Walker, and E. Bab- used for alkali ablation, while sub-millimeter diameter
cock,PhysicalReviewA(Atomic,Molecular,andOptical wires were the targets for silver and gold. On the other
Physics) 75, 013416 (2007). hand, as mentioned in Section III,the wire diameter (in
[21] A. Hatakeyama, K. Oe, K. Ota, S. Hara, J. Arai, the range 25 to 100 microns) did not noticeably affect
T. Yabuzaki, and A. R. Young, Physical Review Letters experimental results for silver and gold.
84, 1407 (2000). [32] M.Auzinsh,D.Budker,D.F.Kimball,S.M.Rochester,
[22] A. Hatakeyama, K. Enomoto, N. Sugimoto, and J. E. Stalnaker, A. O. Sushkov, and V. V. Yashchuk,
T. Yabuzaki,Physical ReviewA 65, 022904/1 (2002). Physical Review Letters 93, 173002/1 (2004).
[23] S. E. Maxwell, N. Brahms, R. deCarvalho, D. R. Glenn,