Measurement of Vacuum Pressure with a Magneto-Optical Trap: a Pressure-Rise Method

0.66 MB·

by  Rowan W. G. Moore

Measurement of Vacuum Pressure with a Magneto-Optical Trap: a
Pressure-Rise Method
Rowan W. G. Moore,1 Lucie A. Lee,1 Elizabeth A. Findlay,1 Lara Torralbo-Campo,1 Graham D. Bruce,1 and
Donatella Cassettari1,a)
SUPA School of Physics and Astronomy, University of St Andrews, North Haugh, St Andrews, Fife KY16 9SS,
United Kingdom
(Dated: 30 January 2015)
The lifetime of an atom trap is often limited by the presence of residual background gases in the vacuum
5 chamber. Thisleadstothelifetimebeinginverselyproportionaltothe pressure. Hereweusethisdependence
1 to estimate the pressure and to obtain pressure rate-of-rise curves, which are commonly used in vacuum
0 science to evaluate the performance of a system. We observe different rates of pressure increase in response
2
todifferentlevelsofoutgassinginoursystem. Thereforewesuggestthatthis isasensitivemethodwhichwill

n find useful applications in cold atom systems, in particular where the inclusion of a standard vacuum gauge
a is impractical.
J

9
2 There is a trend of making cold atom experiments conductance.

simpler and more portable in view of taking them
]
h outside the laboratory1–8, where they can be used for
p applications such as precise inertial sensors9–13. In a
- compact apparatus, it is not always practical to include Rubidium
m Dispenser
a vacuum gauge, and therefore alternative methods of
Photodiode
o estimating the backgroundpressureare desirable. Given Titanium
t Sublimation
a that pressure is in many cases the dominant factor 55 l/s Pump
Ion
. affecting the lifetime of a trapped sample, the lifetime
s Pump
c can in turn be used to estimate the pressure, effectively
si using the atom trap as a vacuum gauge. MOT GClaeslls Electrical
y In Ref. 14 this idea was developed into a quantitative Valve Feedthrough
h method, which we further extend in the present paper
p by using a Magneto-Optical Trap (MOT) to acquire
[ FIG. 1. Vacuum system: the MOT is created in the glass
pressure rate-of-rise curves. These are useful diagnostic cell and the trapped atoms are monitored by collecting their
2 tools in vacuum science, and they are taken by turning fluorescence on a photodiode.
v off the pump after the base pressure of the system has
9 been achieved and monitoring the subsequent pressure
4 Our experiment is a vapour cell 87Rb MOT. Because
increase. The pressure evolution will then indicate
9 the MOT selectively loads rubidium atoms, but loses
whether a realleakis present,inwhichcasethe pressure
7 atoms to collisions with untrapped fast rubidium atoms
increaseslinearlywith time leading tothe determination
. and with other background gases, MOT measurements
1 of the leak size. Or, in absence of real leaks, the
0 can be used to extract two distinct contributions to the
pressure as a function of time may reach a plateau,
4 pressure: that of the rubidium vapour, and that of any
which indicates that an element inside the chamber is
1 other background gas. To separate these contributions,
outgassing or that a virtual leak (i.e. a small volume
: wefirstcharacteriseourMOTatbasepressure(i.e. with
v of trapped gas) is present. Because the gas released in
Xi the chamber in those cases is limited, an equilibrium is pumpson)byusinganNeq-τ plot: weacquireMOTload-
ingcurvesandmeasuretheequilibriumnumberofatoms
reached and the pressure will not increase indefinitely15.
r N andthe1/eloadingtimeτ. Byrepeatingthesemea-
a Therefore the pressure-rise method can help establish eq
surements for different levels of rubidium pressure, we
whetherthebasepressureinasystemislimitedbyareal
gain information on three parameters that characterise
leak or by internally-released gas. While pressure–rise
theMOT:thetrappingcrosssection,thelossratedueto
curves are commonly measured with a vacuum gauge,in
collisions with non-Rb background gases, and the loss-
this paper we take the new approach of using the effect
rate coefficient for the collisions with Rb background.
of the pressure increase on the MOT. This offers the
These measurements fully characteriseour MOT. To ac-
further advantage that the pressure is measured locally,
quirepressure–risecurves,wethenturnofftheionpump
rather than at a separate location of the vacuum system
and monitor the MOT over a period of hours, while the
wherethepressuremaysignificantlydifferduetolimited
pressure in the system slowly rises. The MOT parame-
tersdeterminedfromtheinitialcharacterisationarethen
usedtoconvertthesedataintoquantitativeevolutionsof
the Rb pressure and of the non-Rb pressure.
a)[email protected] Our six-beam MOT is created in a pyrex cell with
2
30 mW of optical power and a magnetic field gradient The solution of (1) is
of 18 G/cm. The trapped atoms are detected by collect-
ing fluorescence with a photodiode. The vacuum system N(t)=Neq(1−exp(−t/τ)), (2)
isshowninFig. 1andalsocomprisesanisotopicallypure
where the equilibrium number of atoms in the MOT is
87Rb dispenser from Alvatec, a 55 L/s ion pump and a
titanium sublimation pump. After assembly, the system
N =αP τ, (3)
◦ eq Rb
was baked at 220 C and a base pressure of the order of
2×10−10 Torr was obtained in the ion pump region. and the MOT loading time is

τ =1/(βP +γ), (4)
Rb
16
which coincides with the trap lifetime14. Combining (3)
(a)
70) and (4) eliminates PRb, giving
ms (x 112 Neq = αβ(1−γτ), (5)
ato 8
of . which relates the two easily measurable quantities Neq
o. and τ. Plotting (5) experimentally provides the initial
N 4
characterisation of the MOT. For this purpose, a large
amount of rubidium is released into the chamber, after
0 which the Rb source is switched off. A sequence of load-
0 10 20 30 40
ing curves is taken as the Rb partial pressure gradually
Tim e (s)
decays, while the non-Rb partial pressure remains con-
stant. This is continued until a data set spanning a suf-
ficiently large range of N and τ is obtained as shown
eq
15 in Fig. 2. Fitting these data with (5) gives γ = (0.11 ±
70) 0.01) s−1 and α/β = (19.6 ± 0.3)×107.
x 112
N (eq 16

9
12
)
7
6 (b) 0
1
x 8
(
2 3 4 5 6 q .
t (s) Ne
4
FIG. 2. Construction of the Neq-τ plot with (a) the MOT 0
loading with different levels of Rb pressure, and (b) the re-
0 1 2 3 4 5 6
sultant Neq-τ plot where the data shown in (a) have been t (s)
encircled. The data in (b) are fittedwith (5).
FIG.3. TheN -τ plotsmeasuredwiththedispensercurrent
eq
ForaMOTloadedfrombackgroundvapour,theMOT reduced to 4 A and the ion pump turned off. The solid line
dynamicscanbewellapproximatedbythefollowingrate is the original Neq-τ plot from Fig. 2(b). The square and
equation14,16: triangulardatapointsarenewNeq-τ plotsobtainedafterthe
ion pump has been turned off, starting at two different Neq
dN(t) values.
=αP −(βP +γ)N(t). (1)
Rb Rb
dt
Physically, the value of α/β represents the largest
This describes the balance between the rates at which MOT that can be obtained in our system, while 1/γ is
atomsareaddedtoandlostfromthetrappedpopulation the theoretical upper limit for the loading time as the
N. The first term on the right-hand side is the rate at Rb pressure tends to zero, i.e. the longest possible trap
which atoms are captured; the constantα representsthe lifetime in our system. This is a useful technique for
MOT trapping cross-section while P is the partial Rb MOT characterisation that we have previously applied
Rb
pressure. The second set of terms represents the losses to the study of MOT loading enhanced by Rb pressure
fromthetrap. Thefirstoftheseterms,βP N,describes modulation17. Inthefollowingthismethodisfurtherap-
Rb
losses due to collisions with background Rb atoms. The plied to measuring pressure-rise curves in our system to
second term, γN, describes losses due to collisions with measure vacuum quality and distinguish between differ-
non-Rb background. ent levels of outgassing.
3
1.0 (a) 20 Non-R Torr) 5 (a) 5.0 A5.5 A 6.0 A 5.5 A
0.8 16 b P -810 4
-1g (s) 0.0001...0246 4812 ressure (x10Torr) -9 87Non-Rb Pressure (x 0123 .
2.2R
b (b)
-1b P (s)Rb 000...000789 (b) 112...680 Pressure (x10 To -9 -9essure (x10 Torr) 34
1.4rr) Pr
b
0 40 80 120 160 200 240 R
Time from turning pump off (min) 87
2
FIG.4. Pressureevolutionasafunctionoftimeaftertheion 0 30 60 90 120 150
Time from turning pump off (minutes)
pump has been switched off, extracted from the square data
pointsinFig. 3. (a)Thenon-Rbgasesinthesystemshowthe
expectedpressurerise. Thelinearfitallowsthedetermination FIG. 5. Pressure evolution after turning off the ion pump,
of the gas load. (b) The Rb pressure decreases because the withpulsesappliedtothedispensercurrent. Theverticallines
dispenser current has been lowered at t=0. showthedurationandcurrentofeachpulse. (a)Pressurerise
ofthenon-Rbbackgroundshowingtemporaryoutgassing. (b)
EvolutionoftheRbpressureshowingaclearincreaseafterthe
6 A pulse.
The value of γ taken from the linear fit is directly
each time t is calculated by rearranging (5):
proportional to the non-Rb pressure P in the system.
On the assumption that this is mostly due to molecular 1 βN
hydrogen, we use the conversion factor γ/P = 4.9×107 γ = 1− eq . (6)
τ (cid:18) α (cid:19)
Torr−1s−1 given in Ref. 14. Combined with γ = 0.11
s−1 asobtainedfromFig. 2,we estimate abase pressure Once again, γ is converted to pressure and this pres-
of2.2×10−9 Torr. Thisestimate ishigherthanthe value sureisplottedasafunctionoftimeinFig. 4(a),givinga
quoted above and the discrepancy can be explained by pressurerisecurvewhichislinearintime. Fromthemea-
the limited conductance in our system. suredrateofriseandthe volumeofthe chamber(∼1L),
The partial Rb pressure may also be calculated by us- weestimateagasloadof1.1×10−12TorrL/s. Weobtain
ingtheconversionfactorβ =4.4×107Torr−1s−1asgiven comparable values for the gas load from both the square
in Ref. 14. Thus α can be determined, and hence the and the triangular data points in Fig. 3, confirming the
rubidium pressure P = N /(ατ) (using (3)). This robustness of this method. We take this gas load as a
Rb eq
pressure varies over the course of the measurements but baseline for the subsequent comparative measurements
a typical value is in the 10−9 Torr regime. of outgas rates, and note that this gas load is very low.
To obtain pressure–rise curves the procedure is sim- By comparison, previously baked stainless steel (which
ilar to that for MOT characterisation, but with the ion constitutes most of the surface in our system) outgasses
pump switched off at an initial time t = 0 to allow the at a rate of 10−12 Torr L/s cm2, which corresponds to
non-Rb pressure to rise. Before t = 0 the dispenser cur- > 10−10 Torr L/s for the surface of our system18. We
rent is set at 5-6 A to trap large numbers of atoms in attribute the observed low rate of pressure rise to the
the MOT, and then is lowered to 4 A at t = 0 to lower presence of an active titanium layer pumping gas in our
the Rb pressure and prevent overloading the chamber. system.
MOT–loading curves are taken for up to four hours and Our method is capable of discriminating between Rb
N isagainplottedasafunctionofτ asshowninFig. 3. and non-Rbpressure. The partialRb pressure is plotted
eq
Both N and τ decrease over time as the quality of the as a function of time in Fig. 4(b), using the conversion
eq
vacuum deteriorates. This measurement is shown twice, outlined above. Due to the reduction in dispenser cur-
startingfromdifferentinitialrubidiumpressures,andthe rent at t = 0, the Rb pressure actually falls over time
separate evolutions of N (τ) are shown to converge. while the non-Rb pressure is rising. Rubidium pumping
eq
Using α/β obtained from Fig. 2, the value of γ for isalwaysdominatedbyhighadsorptiontothesteelwalls
4
ofthe vacuum chamber19 andtherefore switching offthe 1S. Knappe, P. D. D. Schwindt, V. Gerginov, V. Shah, L. Liew,
ion pump has little effect on the Rb pressure. J. Moreland, H. G. Robinson, L. Hollberg, and J. Kitching, J.
Opt.A:PureAppl.Opt.8,S318(2006).
Totestthesensitivityofourmethodtooutgassing,we
2F.Sorrentino,K.Bongs,P.Bouyer,L.Cacciapuoti,M.deAnge-
investigate the effect of repeatedly pulsing the dispenser
lis, H. Dittus, W. Ertmer, A. Giorgini, J. Hartwig, M. Hauth,
to higher currents, as shown in Fig. 5. The current is S. Herrmann, M. Inguscio, E. Kajari, T. T. K¨onemann,
kept at 4 A between pulses. After the 6 A pulse the C.L¨ammerzahl,A.Landragin,G.Modugno,F.PereiradosSan-
Rb pressure clearly increases. However we also see that tos,A.Peters,M.Prevedelli,E.Rasel,W.Schleich,M.Schmidt,
A. Senger, K. Sengstock, G. Stern, G. Tino, and R. Walser,
there is a temporarily increased rise in non-Rb pressure,
MicrogravitySci.Technol.22,551(2010).
i.e. an increased gas load. The rate of pressure rise then
3E. A. Salim, J. DeNatale, D. M. Farkas, K. M. Hudek, S. E.
returns to the pre-pulse levelwhich is comparableto the
McBride, J. Michalchuk, R. Mihailovich, and D. Z. Anderson,
base gas load shown in Figure 4(a). This is indicative of QuantumInf.Process.10,975(2011).
atemporaryoutgassingeitherfromwithinthedispenser, 4M.Schmidt,A.Senger,M.Hauth,C.Freier,V.Schkolnik, and
or from a region of the chamber that is being heated up A.Peters,Gyrosc.Navig.2,170(2011).
5M.deAngelis,M.Angonin,Q.Beaufils,C.Becker,A.Bertoldi,
by proximity to the dispenser. By measuring the rate
K. Bongs, T. Bourdel, P. Bouyer, V. Boyer, S. D¨orscher,
increase in Fig. 5(a), we estimate a gas load from the H.Duncker,W.Ertmer,T.Fernholz,T.M.Fromhold,W.Herr,
6 A pulse of 6.7×10−12 Torr L/s, which is small and P. Kru¨ger, C. Ku¨rbis, C. Mellor, F. P. D. Santos, A. Peters,
compatible with UHV operation. N.Poli,M.Popp,M.Prevedelli,E.Rasel,J.Rudolph,F.Schreck,
K.Sengstock,F.Sorrentino,S.Stellmer,G.Tino,T.Valenzuela,
In conclusion, we have used an N -τ plot to charac-
eq T.Wendrich,A.Wicht,P.Windpassinger, andP.Wolf,Procedia
terise our MOT, and used the MOT effectively as a vac- Comput.Sci.7,334 (2011).
uum gauge to acquire pressure-rise curves, which quan- 6B. Barrett, P.-A. Gominet, E. Cantin, L. Antoni-Micollier,
tifyoutgassinginourvacuumsystem. Thesmallchanges A.Bertoldi,B.Battelier,P.Bouyer,J.Lautier, andA.Landra-
in gas load that we have detected demonstrate the sen- gin, in Proceedings of the International School of Physics “En-
rico Fermi”, Vol. 188: Atom Interferometry (IOS Press, 2014)
sitivity of the method. More generally, it should be pos-
pp.493–555.
sible to use this approach to check for leaks in a system 7T. Farah, C. Guerlin, A. Landragin, P. Bouyer, S. Gaffet,
and to discriminate between real and virtual leaks. F. Pereira Dos Santos, and S. Merlet, Gyrosc. Navig. 5, 266
One particular advantage that this method has over a (2014).
8J. Rushton, M. Aldous, and M. Himsworth, “The Feasibility
standard vacuum gauge is that the pressure is measured
of a Fully Miniaturized Magneto-Optical Trap for Portable Ul-
directlyintheMOTregion;thisismorerelevantforcold
tracoldQuantumTechnology,” (2014), http://arxiv.org/abs/
atoms experiments than the pressure at another point 1405.3148.
in the system. Moreover, we are able to separate the 9J. P. Dowling and G. J. Milburn, Phil. Trans. R. Soc. Lond. A
contributiontothepressureoftherubidiumbackground, 361,1655(2003).
10S. Bize, P. Laurent, M. Abgrall, H. Marion, I. Maksimovic,
whichcanbemonitoredforthepurposeofcharacterising
L. Cacciapuoti, J. Gru¨nert, C. Vian, F. Pereira dos Santos,
the dispenser output.
P. Rosenbusch, P. Lemonde, G. Santarelli, P. Wolf, A. Clairon,
We expect this method to find broad applicability to A.Luiten,M.Tobar, andC.Salomon,J.Phys.B:At.Mol.Opt.
cold atom experiments and to be of particular interest Phys.38,S449(2005).
11D.BudkerandM.Romalis,Nat.Phys.3,227(2007).
for applications that require a miniaturised vacuum
12A.D.Cronin,J.Schmiedmayer, andD.E.Pritchard,Rev.Mod.
system. Phys.81,1051(2009).
13J.Kitching,S.Knappe, andE.A.Donley,IEEESensorsJournal
11,1749(2011).
14T.Arpornthip, C. A.Sackett, and K.J. Hughes, Phys. Rev. A
85,033420 (2012).
ACKNOWLEDGMENTS 15BasicVacuumPractice (VarianAssociates,Incorporated,1986).
16A.M.Steane,M.Chowdhury, andC.J.Foot,J.Opt.Soc.Am.
B9,2142(1992).
R.M. and L.L. contributed equally to this work. 17L. Torralbo-Campo, G. D. Bruce, G. Smirne, and D. Cas-
We thank Robert Nyman for useful discussions. This settari, “Light-induced atomic desorption in a compact sys-
research was supported by UK EPSRC, IOP Scotland tem for ultracold atoms,” (2014), submitted to Phys. Rev. A
http://arxiv.org/abs/1312.6442.
and the Leverhulme Trust Research Project Grant
18N.Yoshimura,Vacuum Technology (ChapmanandHall,Tokyo,
RPG-2013-074.
2008).
19C. Wieman, G. Flowers, and S. Gilbert, Am. J. Phys. 63, 317
(1995).