APS/123-QED Measurement scheme and analysis for weak ground state hyperfine transition moments through two-pathway coherent control J. Choi1 and D. S. Elliott1,2,3 1School of Electrical and Computer Engineering, 2Department of Physics and Astronomy, and 3Purdue Quantum Center Purdue University, West Lafayette, IN 47907 (Dated: January 13, 2016) We report our detailed analysis of a table-top system for the measurement of the weak-force- 6 inducedelectricdipolemomentofagroundstatehyperfinetransitioncarriedoutinanatomicbeam 1 geometry. Wedescribeanexperimentalconfigurationofconductorsforapplicationoforthogonalr.f. 0 andstaticelectricfields,withcavityenhancementofther.f.fieldamplitude,thatallowsconfinement 2 ofther.f.fieldtoaregioninwhichthestaticfieldsareuniformandwell-characterized. Wecarryout detailednumericalsimulationsofthefieldmodes,andanalyzetheexpectedmagnitudeofstatistical n and systematic limits to the measurement of this transition amplitude in atomic cesium. The a J combination of an atomic beam with this configuration leads to strong suppression of magnetic dipole contributions to the atomic signal. The application of this technique to the measurement of 1 extremely weak transition amplitudes in other atomic systems, especially alkali metals, seems very 1 feasible. ] h PACSnumbers: p - m I. INTRODUCTION as evidenced by the many efforts underway world- wideinavarietyofsystems. Laboratoryeffortshave o t sought, or are currently underway, to determine the a Laboratory measurements of very weak atomic anapole moment of other nuclei, including Tl [24], . s transitions that violate the usual parity selection Yb [10, 25–27], Fr [28–30], Ba+ [31–33], Ra+ [34– c i rules are a means of determining the weak force 37], and Yb+ [38], and several molecular systems ys at low collision energies [1–5]. The component of as well [8, 12]. Differences between EPNC on various this electric dipole transition moment E that is hyperfinelinesforthesesystemscouldrevealthenu- h PNC p induced by the weak-force coupling between nucle- clearanapolemomentofthesesystems. Comparison [ ons has become of great interest in recent years [6– between different isotopes of the same species could 12]. These Nuclear Spin Dependent (NSD) contri- removethedependenceofthedeterminationonpre- 1 butions to E are expected to result from the cise atomic theory, subject to the ability to correct v PNC 0 nuclear anapole moment of the nucleus, with addi- forvariationsinthenuclearstructureamongtheiso- 9 tional smaller contributions from the weak neutral topes [39–43]. 6 axial-vectornucleonvectorelectron(A ,V )current, Measurements performed on a hyperfine tran- n e 2 and the combined effect of the hyperfine interaction sition between components of an atomic ground 0 and the (V ,A ) current [9, 13–16]. To date, the state present an attractive alternative to the above . n e 1 only non-zero determination of NSD contributions schemes for determining the NSD contributions to 0 to E in any element was based upon the differ- E . This moment contains only the NSD contri- PNC PNC 6 encebetweenmeasurementsofE /βinatomicce- bution, simplifying the measurement, and in many PNC 1 sium[17],whereβ isthevectorpolarizabilityforthe cases, the value of E on ground state transitions : PNC v transition, on two different hyperfine components of ispredictedtobelargerthantheweakamplitudebe- i the6s→7stransition; theF =3→F(cid:48) =4andthe tween different electronic states [11]. Of particular X F = 4 → F(cid:48) = 3 lines. E /β on these lines dif- interest is a large program on francium [7], one goal PNC r a fered by ∼5% of their average value. This NSD fac- of which is to measure EPNC on transitions between torwasmuchlargerthanwasexpected,andtheoret- hyperfine levels of the ground state of this unsta- ical efforts [3, 6, 9, 13, 18] to understand this result ble heavy element at TRIUMF. To carry out these have not been successful. Meson exchange coupling measurements, development of techniques for cool- constants of the so-called DDH model [19] derived ing and trapping these species in a magneto-optical fromthisresultdonotagreewellwithresultsderived trap (MOT) and carrying out the measurements in from measurements of the asymmetry in the high- this restricted space is necessary. energy scattering of light nuclei [3, 20–23]. While The measurement in atomic cesium that we have the applicability of the DDH model to such a large under development in our laboratory, which we de- atom is questionable, there is none-the-less strong scribe in this work, has several features in common interest in understanding the NSD of large nuclei, with those of the francium effort. As a ground state 2 transition, atomic coherences are long-lived, and we PPTL structure, and describe the field modes sup- exploit the interference between the direct transi- ported by it. Finally, we analyze the magnitudes tion driven by a radio frequency (r.f.) field and the of the dominant residual contributions to the mea- Raman process driven by a two-frequency cw laser surement of E , and consider the effects of the PNC field, in a derivation of the two-pathway coherent distribution of atomic velocities in the beam. control techniques that we have developed for sim- ilar measurements [44, 45]. Atomic cesium offers several benefits over the francium system that are II. THE COHERENT CONTROL SCHEME derived from an atomic beam geometry: that is, a greater atomic density, the capability of sequential We employ the two-pathway coherent control preparation, interaction, and detection, and a less schemeforsensitivemeasurementofweakmoments. restrictive experimental environment. Furthermore, This technique is based on the interference between thebeamgeometryallowustospatiallyseparatethe various optical interactions driven by two or more interaction regions for the different coherent fields, coherently-related fields. We developed and em- and to highly suppress the magnetic dipole contri- ployed this technique on measurements of the mag- butions to the atomic signal, a primary challenge in neticdipoletransitionmomentM onthe6s2S → ground-state measurements of weak signals. In this 1/2 7s2S transitioninatomiccesium[44,45]. TheFr work, we discuss how the two-pathway interference 1/2 collaborationbasestheirmeasurementsonthistech- method can be used to determine the ratio of the niquealso[7]. Inthissection,wedescribetheprinci- PNCamplitudetotheStarkvectorpolarizabilityβ. plesbehindthistechnique,withparticularattention While our primary interest is in atomic cesium, the paid to a transition between hyperfine components techniqueisgenerallyapplicableinanyofthestable of a ground state system, in which both states are alkali metal species. longlived. Weshowhowthismeasurementcanyield We describe in detail the measurement require- a determination of E /β, independent of the pro- PNC ments, and the capability of our technique. The op- fileoramplitudeofther.f.fieldthatdrivesthetran- timal arrangement uses r.f. and static electric fields sition. that are oriented in perpendicular directions, and We consider a sinusoidal wave of amplitude εrf the r.f. field should be confined to a space within and frequency ωrf, incident upon a two-level atom whichthestaticfieldisuniform. Theserequirements with hyperfine components ψ and ψ of the ground i f can be satisfied by a parallel plate transmission line state, of energy E and E , respectively. We choose i f (PPTL)configurationtowhichcylindricalreflectors the field to be continuous wave, but spatially vary- (to form an r.f. resonant cavity) and isolated con- ing, such that as the atoms move across the interac- ducting pads (for application of the orthogonal d.c. tion region, they effectively see a time-varying field. field) have been added. We report the results of our When the atoms are initially prepared in a single detailed numerical analysis of the electric and mag- hyperfine component ψ , and when the field com- i netic fields supported by this structure, and we use ponents are chosen so as to couple the initial state the magnitudes of the field components to estimate to a single final state ψ , the atomic system is very f the residual systematic effects that one should ex- closely described as a two-level system, and we can pect in a determination of E in atomic cesium. PNC write the state of the atoms using the time-varying This paper is organized as follows. In Sec. II, amplitudes c (t) and c (t) as i f we discuss the transition probability of a two-level atom interacting with a resonant r.f. field and a ψ(t)=ci(t)ψie−iωit+cf(t)ψfe−iωft. two-frequency optical field through a Raman inter- action. We show that, when a variable d.c. elec- The time evolution of the system is described in tric field is applied, this coherent control process al- termsoftheHamiltonianH +Vint,whereH isthe 0 0 lows one to determine E /β. We then discuss in atomic Hamiltonian and Vint describes the interac- PNC Sec. III the various transition amplitudes, including tion between the atom and the field. In this work, the magnetic dipole, Stark-induced electric dipole, we consider the weak-force induced electric dipole and weak-induced electric dipole, for the transition interaction Vint , the Stark-induced electric dipole PNC between hyperfine levels of the ground state of an interactionVint,andthemagneticdipoleinteraction St alkali metal atom. We present an estimate of the Vint of the atom with the r.f. field, plus a Raman M signalsizeinSection IV,withanestimateofthesta- interaction Vint of the atom with a two-frequency Ram tisticaluncertainty,andreviewthebenefitsofcarry- laser field, all of which we describe in more detail ing out the measurement in a standing wave cavity later, and write Vint as the sum of the individuals for suppression of magnetic dipole contributions in SectionV.Inthefollowingsection,weintroducethe Vint =Vint +Vint+Vint+Vint . PNC St M Ram 3 fields propagate in the y-direction, the d.c. electric E0 and magnetic B0 fields are oriented in the z- direction, and the electric field εrf of the r.f. field is directed in the x-direction. (Parallel propagation of the r.f. and Raman fields is necessary to maintain a uniform phase difference between interactions.) Not shown in this figure are the two components of the laser electric field that drive the Raman transition, each linearly polarized, one in the x-direction, the other in the z-direction. In this geometry, the pri- maryr.f.andRamanfieldseachindependentlydrive a ∆F = ±1, ∆m = ±1 transition, the magnetic dipole contribution on this transition is suppressed, andtheStark-inducedandthePNCinteractionsare FIG. 1: (Color online) An abbreviated energy level dia- inquadrature-phasewithoneanother. Theprimary gram showing the relevant ground state levels. We pre- contributionshere,undertheprecise(idealized)con- pare the cesium atoms one hyperfine component of the ditions specified in Fig. 2, are ground state, (F,m), where m = ±F. Through the in- teractions with the r.f. field and the optical field, some oftheatomsaretransferredtothelevel(F(cid:48),m(cid:48)). Inthis Vint =βE0εrf ei(ωrft−ky−φrf) CF(cid:48)m±1 (3) figure, we show (3,3) as the initial state, and (4,4) as St z x Fm the final state. and We illustrate these interactions schematically in Fig. 1. Vint =∓i Im{E }εrfei(ωrft−ky−φrf)CF(cid:48)m±1. When the atoms exit the interaction region, the PNC PNC x Fm (4) probability that they are in state ψ is f InEq.(3),β isthevectorpolarizabilityandCF(cid:48)m±1 Fm (cid:32)(cid:12) (cid:12)(cid:33) isafactorrelatedtotheClebsch-Gordoncoefficients, (cid:12)(cid:88) (cid:12) |c (∞)|2 =f(δ)sin2 (cid:12) Θ (cid:12) , (1) defined in detail in Ref. [47]. Note that we have f (cid:12) i(cid:12) (cid:12) i (cid:12) explicitly included the phase of the r.f. field in these expressions. wheretheΘ aretheintegratedinteractionstrengths i of any of the individual interactions In addition to these primary amplitudes, extra contributionsduetomagneticdipoletransitionsand (cid:90) ∞ Θ = Ω (t)dt. i i −∞ The Rabi frequencies of the various interactions are Ω = Vint/(cid:126), and f(δ) represents the reduction in i i amplitude when the r.f. frequency is detuned from E0 the resonant frequency by δ = ωrf −|E −E |/(cid:126). f i k f(δ) depends on the temporal shape of the ‘pulse’ y B0 as the atoms travel through the interaction region Cs inanon-trivialway,andwewilllimitourdiscussion x z εrf atomic to resonant excitation, δ =0, for which f(δ)=1. beam In an atomic beam, collisions are infrequent, and theatomstravelthroughtheinteractionregionwith FIG.2: (Coloronline)Thefieldorientationsforthemea- a constant velocity v. In this case, the interaction surementofE /βonthe∆F =±1,∆m=±1ground PNC strength can be rewritten state hyperfine transition. The static electric and mag- netic fields are oriented in the z-direction, while the po- 1(cid:90) ∞ Θ = Ω (z)dz. (2) larization of the r.f. field is in the x-direction. The po- i v i larizations of the laser field components that drive the −∞ Raman interaction, not shown, are aligned with the x- We use notation similar to that of Gilbert and and z-axes. The r.f. and Raman fields propagate paral- Wieman [46] for each of the various interactions, leltooneanother,shownasthedirectionk,asrequired and show the optimal field geometry for this mea- to maintain a uniform phase difference between interac- surement in Fig. 2. That is, the r.f. and Raman tions. 4 fieldmisalignmentscanarise. Thelargestoftheseis limitoftheRamaninteractionstrengthΘ being Ram much greater than any of the interactions driven by (cid:40) the r.f. field Θ , Θ , and Θ . Under these con- Vint = η M (cid:2)∓hrf +ihrf(cid:3)CF(cid:48)m±1 (5) St M PNC M 0 x y Fm ditions, andwiththedetuningδ =0, Eq.(1)canbe expanded to the form (cid:32)±B0+iB0CF(cid:48)mCF(cid:48)m +hrzf xB0 y Fmg F(cid:48)m±1 |cf(∞)|2 = sin2(|ΘRam|)+sin(2|ΘRam|) (7) z F(cid:48) ×sin[|Θ +Θ +Θ |cos(∆φ+δφ(E ))]. ∓B0+iB0CFm±1CF(cid:48)m±1(cid:33)(cid:41) St M PNC z + xB0 y Fm g Fm±1 ei(ωrft−ky−φrf) ∆φ = φrf − φRam is the controllable phase differ- z F ence between the r.f. field and the phase difference φRam, and δφ(E )=tan−1(E /βE0) is the phase for ∆m = ±1 transitions, where the hrf are the z PNC z i shift introduced by the quadrature combination of components of the magnetic field of the r.f. wave, E and βE0. (In writing this phase shift, we pre- M is the magnetic dipole transition moment, η = PNC z (cid:112) 0 sumethatthemagneticdipolecontributionsaresup- µ /(cid:15) =120πΩ is the impedance of vacuum, and 0 0 pressed, as we show later.) We see from this expres- g and g are the gyromagnetic ratio of the ini- F F(cid:48) sion a feature that is similar to that of the coherent tial and final states. For cesium, g is −1/4 for F controlschemeonashort-livedstate[44,45];thatis, the F = 3 level and +1/4 for the F = 4 level of that the signal consists of a d.c. term resulting from the ground state. The first terms in Eq. (5) are the the Raman interaction alone, plus a sinusoidally- magneticdipoleamplitudedrivenbythehrf andhrf x y varying contribution that varies with the phase dif- fieldcomponents, whilethelasttermsinhrf andB0 z x ference ∆φ between the Raman field and the one- or B0 arise from Zeeman mixing of the hyperfine y photon r.f. field. Furthermore, the amplitude of the componentsbythestaticmagneticfield. Toinvesti- modulating term is the magnitude of the sum of in- gatepossibleinterferencesfrom∆m=0transitions, teractionangles|Θ +Θ +Θ |≈|Θ +Θ |, St M PNC St PNC we also present the magnetic dipole transition am- where we have omitted the small magnetic dipole plitude for these transitions integrated angle in the final step. A laboratory (cid:40) measurement of this population modulation ampli- VMint = η0M hrzfCFFm(cid:48)m+(cid:88)(cid:2)∓hrxf +ihryf(cid:3) tude as a function of the d.c. electric field Ez0 yields E /β. We see this as follows. ± PNC ×(cid:34)(cid:32)∓Bx0B+z0iBy0(cid:33)CFFm(cid:48)m±g1FC(cid:48)FF(cid:48)(cid:48)mm±1 (6) |ΘSt+ΘPNC|= v1(cid:12)(cid:12)(cid:12)(cid:12)(cid:90)−∞∞[ΩSt(z)+ΩPNC(z)] dz(cid:12)(cid:12)(cid:12)(cid:12), + (cid:32)±Bx0+iBy0(cid:33)CFFmm∓1CFFm(cid:48)m∓1(cid:35)(cid:41)ei(ωrft−ky−φrf). which, using Eqs. (3) and (4) becomes B0 g In additiozn to these tranFsitions driven by the r.f. |ΘSt+ΘPNC| = (cid:126)1v (cid:12)(cid:12)βEz0∓i Im{EPNC}(cid:12)(cid:12) (cid:90) ∞ field, we consider the Raman transition of the form ×CF(cid:48)m±1 εrf(z)dz, (8) Fm x −∞ Vint =β˜εR1(εR2)∗ei(ωrft−φRam) CF(cid:48)m±1 Ram z x Fm valid when E0 is uniform in the interaction region. z where εR1 and εR2 are the electric field amplitudes Since the Stark and PNC moments add in quadra- z x of the two laser components, and ωrf = ωR1−ωR2, ture, the amplitude of the sinusoidal modulation of where ωR1 and ωR2 are the optical frequencies. The the signal scales as phase φRam is the phase difference between the (cid:113) pmhaansepsoolafrtihzaebtiwlitoycβo˜mdpeponenendtssoφnRt1h−e dφeRt2u.niTnhge∆Rao-f (cid:12)(cid:12)βEz0∓i Im{EPNC}(cid:12)(cid:12)= (βEz0)2+|EPNC|2. (9) these field components from the D transition fre- Atsmalld.c.field,themodulationamplitudeispro- 2 quency, and the Raman transition can be enhanced portional to Im{E } alone, while at large field, PNC by making ∆ small. the modulation amplitude is nearly proportional to We will analyze these r.f. transition amplitudes βE0. Bymeasuringthisamplitudeofthepopulation z later using electric and magnetic field amplitudes modulation as a function of the d.c. field, therefore, that we expect to encounter for our parallel plate one can determine the ratio E /β. PNC structure to place limits on unwanted magnetic To optimize the amplitude of the signal modula- dipolecontributionstothePNCsignal. Beforewedo tion in Eq. (7), one should adjust the amplitude of this, we return to Eq. (1), which we examine in the the Raman interaction to |Θ | = π/4. At this Ram 5 shows the state amplitudes when the interactions are π out of phase with one another. The peak Rabi frequency, center position, and beam radius are Ω = 23.9 ms−1, z = -4 cm, and w = Ram,0 c Ram 0.5cmfortheRamanbeam, andΩ =0.61ms−1, w,0 z = 0, and w = 2.5 cm for the r.f.-driven inter- c rf action. We use 270 m/s, the peak velocity of the atoms in our atomic beam for v. The duration of the interaction is w /v (cid:39) 19 µs for the Raman Ram beam,andw /v (cid:39)93µsforther.f.field. Whenthe rf amplitudes are in phase with one another, |c (z)| f growsmonotonically,whilewhentheinteractionsare out of phase, the amplitude decreases after its ini- tial preparation by the Raman beam. The value of |c (∞)| after the atoms have exited the interaction f (cid:112) region is 1/2+sin(|Θ |) for in-phase interactions w (cid:112) FIG.3: (Coloronline)Thevariationofstateamplitudes and 1/2−sin(|Θw|) for out-of-phase interactions. |c (z)| (red solid) and |c (z)| (blue dashed) versus z as When the PNC and Stark-induced terms are driven f i the atoms pass through the interaction region from left by the r.f. field, then |Θ | is |Θ +Θ |, where w St PNC to right. The atoms are prepared by the Raman beams the PNC interaction angle is in a superposition state before entering the broad r.f. field. BothfieldsareGaussianinshape,withpeakRabi Θ = (cid:16)∓iIm{E }CF(cid:48)m±1/(cid:126)v(cid:17)(cid:90) ∞ εrf(z)dz frequency and beam radii of Ω = 23.9 ms−1 and PNC PNC Fm x Ram,0 −∞ 20..55ccmmffoorrtthheeRr.fa.m-darinvebneainmt,eraancdtioΩnw.,0In=(a0).6,1thmesR−a1manand =(cid:16)∓iIm{EPNC}CFFm(cid:48)m±1/(cid:126)v(cid:17)√π wrf εrxf,0. (10) andr.f.interactionsareinphasewithoneanother,while in (b), the interactions are out of phase. In either case, Similarly, the integrated area of the Stark-induced thedurationoftheinteractionisw /v(cid:39)19µsforthe interaction angle for this Gaussian-shaped profile is Ram √ Raman beam, and wrf/v(cid:39) 93 µs for the r.f. field. ΘSt =βEz0CFFm(cid:48)m±1 π wrf εrxf,0/(cid:126)v. The term 1/2 in theexpressionsfor|c (∞)|comesfromsin2(|Θ |) √ f Ram with |Θ | = πw |Ω |/v. The weak sig- value, the factor sin(2|Θ |) is equal to 1, and Ram Ram√Ram,0 Ram nal strength is |Θ | = πw |Ω |/v in this ex- the atomic population due to the Raman interac- w rf w,0 ample is 0.10. Any interaction of the atoms with tion alone is equal to 1/2, i.e. equal probability in the r.f. field therefore is evident as a modulation of the initial and final states. Any additional interac- this signal as we vary the phase difference between tions of the atom with the r.f. field add (slightly) to the fields. We illustrate this in Fig. 4, which shows thepopulationintheψ statewhenthisinteraction f the sinusoidal modulation of the final state popula- isinphasewiththeRamaninteraction,andsubtract tion as a function of ∆φ. Here the parameters are when out-of-phase. as they were in Fig. 3, with the exception of Ω We can gain some insight into the interference w,0 which we have decreased to 0.061 ms−1 for this fig- by following the evolution of the amplitudes |c (t)| f ure. The amplitude of the modulation of |c (∞)|2 (red solid) and |c (t)| (blue dashed) as the atoms √ f i is |Θ | = πw Ω /v = 0.010, in agreement with move across the interaction region, which we show w rf w,0 thenumericaldatainthefigure. Inoursimulations, inFig.3. Forthisillustration, theatomsmovefrom theamplitudeofthemodulationscaleslinearlywith left to right, and encounter the Raman field first, the weak amplitude. centered at z = -4 cm, which prepares them in a Important conditions and features of this mea- coherent superposition state. The atoms then enter surement technique include: the broad r.f. field. We use Gaussian profiles for the r.f. and Raman fields. For the former, the peak am- 1. Mutualcoherenceofthedifferenttime-varying plitudeisεrf andbeamradiusw intheinteraction fields is required. This can be implemented x,0 rf region, in the laboratory by using non-linear mixing, injection locking of diode lasers, or frequency εrf(z)=εrf e−(z/wrf)2. modulation techniques. x x,0 We show this for two values of the phase ∆φ in 2. Thecoherentbeamsthatdrivetheinteractions Fig. 3. Fig. 3(a) shows the magnitudes of the state must propagate in the same direction in order amplitudeswhentheRamanandr.f.-driveninterac- to maintain a uniform phase difference for all tions are in phase with one another, while Fig. 3(b) atoms in the interaction region. 6 III. EXPECTED MAGNITUDES OF M, β, AND E PNC In order to design a measurement system and un- derstand the effect of stray fields and the magni- tude of unwanted contributions to the signal, we mustfirstknowtheexpectedmagnitudesofthePNC moment, E , the vector polarizability β, and the PNC magnetic dipole moment M for the transition. The PNC amplitude for this transition is calcu- lated [11] to be FIG. 4: (Color online) The sinusoidal variation of the signal as a function of the phase difference between the E =1.82×10−11iea , (11) PNC 0 r.f. and Raman interactions. The peak Rabi frequency of the r.f.-driven interaction is Ω = 0.061 ms−1 for whereeanda aretheelectronchargeandtheBohr w,0 0 this plot. Other parameters are as given in the caption radius,respectively. ThisislargerthanE forthe PNC to Fig. 3. moment on the 6s → 7s transition in cesium by a factor of 2.2. The vector polarizability has not previously been 3. TheRamanandther.f.fielddistributionneed calculated, but we can estimate its approximate not overlap one another. Since the ground magnitude using the sum-over-states expansion of state is long lived, the final level retains its Refs. [1] and [47], coherence, and the net excitation of the final statedependsontheaccumulatedeffectacross (cid:34) (cid:18) (cid:19) the interaction region. β = e (cid:88)r2 1 − 1 6(cid:126) n,1/2 ∆ ∆ 4. We control the phase difference between the n 4;n,1/2 3;n,1/2 (cid:18) (cid:19)(cid:21) transition amplitudes with r.f. devices, com- 1 1 1 + r2 − , pletely external to the interaction region. 2 n,3/2 ∆ ∆ 4;n,3/2 3;n,3/2 5. We select the particular interactions that con- where r represents the reduced dipole matrix ele- tribute to the measurement by choosing the n,j ments(cid:104)np ||r||6s (cid:105)forj =1/2or3/2,and(cid:126)∆ orientation of the various fields in the interac- j 1/2 F;n,j are the energy differences E −E for the two tion region. 6s,F npj hyperfine states F = 3 or 4 of the ground 6s2S 1/2 6. The measurement uses only modest d.c. elec- andtheexcitednp2P states. Then=6termdomi- j tric fields, (cid:46) 100 V/cm. This allows flexibility natesthissum,andthegroundstatehyperfinesplit- in the experimental configuration. ting ∆ is small compared to the energy of the 6p hfs states, so the polarizability is approximately 7. Since the interactions Ω and Ω are π/2 PNC St out of phase with one another, these ampli- (cid:34)(cid:12) (cid:12)2 tudes add in quadrature. This indicates that β (cid:39) e∆hfs (cid:12)(cid:104)6p1/2||r||6s1/2(cid:105)(cid:12) the amplitude of the modulating signal is at 6 (E6s−E6p1/2)2 a minimum when the static electric field is (cid:12) (cid:12)2(cid:35) turned off, and increases when a static field +1(cid:12)(cid:104)6p3/2||r||6s1/2(cid:105)(cid:12) . of either polarity is applied. 2 (E6s−E6p3/2)2 8. Usingdifferentfieldorientations,thiscoherent We use (cid:104)6p ||r||6s (cid:105) = 4.5062 a and control technique may be used to determine 1/2 1/2 0 (cid:104)6p ||r||6s (cid:105) = 6.3400 a [48–53] to estimate M/β. This may be a useful means of deter- 3/2 1/2 0 the vector polarizability for this transition as β (cid:39) mining the vector polarizability β, but we de- 0.00346a3. Based on these expected magnitudes of fer any further discussion of this to a future 0 β and E , the ratio E /β is about 27 V/cm; report. PNC PNC uponapplicationofastaticelectricfieldofthismag- In the following sections, we will discuss the ex- nitude, the magnitudes of the Stark-induced ampli- pected magnitudes of the different interactions, and tudeandthePNCamplitudeareequivalent. Sinceβ present an experimental assembly of conductors for is so small for this transition, we conclude that sys- such a measurement in an atomic beam configura- tematic errors due to uncontrolled electric fields in tion. Finally, we will analyze the effect of expected the interaction region, due to surface contamination magnetic dipole contributions to the measurement. and patch effects and estimated to be (cid:46)0.1 V/cm, 7 are inconsequential in these ground state measure- phase with one another (|c |2 = 1 +|Θ |), and f 2 PNC ments. This is in strong contrast to measurements N the total count of excitations when the r.f. and − of E on the 6s → 7s transition [17], for which Raman interactions are π out of phase with one an- PNC uncontrolled electric fields were of major concern. other (|c |2 = 1 −|Θ |). Then f 2 PNC In addition to these two relatively weak ampli- 1 N −N tudes driven by the r.f. field, the magnetic dipole Θ = + −. moment is active on this transition. The amplitude PNC 2 N++N− for this transition is Vint = (cid:104)6s2S F(cid:48)m(cid:48)|−µ · M 1/2 m To use this result to determine E , however, one brf|6s2S Fm(cid:105), where µ =µ (g L+g S+g I) PNC 1/2 m B L S I must also have an accurate determination of the r.f. isthemagneticmomentoftheatom,µ =e(cid:126)/2mis B beam profile and field amplitude. Alternatively, one theBohrmagneton,andbrf isthemagneticfluxden- canapplyad.c.electricfieldtotheatoms,andmea- sityofther.f.wave. L,SandIaretheusualorbital, sure the amplitude of the modulation as a function spin,andnuclearangularmomenta,andgL,gS,and of the field amplitude E0, as suggested in Eqs. (8) g are the respective gyromagnetic ratios. For the z I and (9). transition of this work, the orbital angular momen- When the precision of N and N is limited by + − tum is zero, and g is much less than g (which is I S counting statistics, then√the uncertainty in either of ≈ 2) due to the heavy mass of the nucleus. For the these counts is σ = N, where N represents ei- N ground state transition, the spatial parts of ψ and i ther N or N (which are essentially the same). ψ are the same, and using εrf/brf = c, the mag- + − √ f The uncertainty in Θ is σ = 1/ 8N, and netic dipole amplitude is M = µ g /2c (cid:39) µ /c. PNC PNC B S B to achieve a 3% measurement of Θ , one must But µ /c = ea α/2, where α (cid:39) 1/137 is the fine PNC B 0 count N = 1/8σ2 = 3 × 1012 atoms for each structure constant, so M (cid:39) ea α/2, and the ratio PNC 0 individual measurement. In a counting interval T, M/E (cid:39) 2×108. The magnetic dipole contribu- PNC the number of counts is N = 1ρ AvT, where 1 is tions to the signal must be suppressed for a success- 2 Cs 2 the average excitation probability, ρ is the num- ful measurement of E , representing the primary Cs PNC berdensityoftheatomicbeam(109 cm−3),Aisthe challenge of these measurements. The orientations cross sectional area of the atomic beam (1 mm2), ofthefieldcomponentsthatwehaveshowninFig.2 andv isthepeakvelocityoftheatomsinthebeam. areanimportantfirststepinmeetingthischallenge. The counting time T to achieve the required statis- tics is 20 seconds per data point. During the course of a measurement, one must repeat the process at IV. MAGNITUDE OF SIGNAL many different phases, not just two, and one must varythed.c.electricfieldstrengthE0andrepeatthe z In this section, we will use the results of the anal- measurement. Regardless, the estimate of the inte- ysis of Sec. II, in particular Eqs. (7) and (10), and gration time shows that the measurement is feasible thecalculatedvalueofEPNC giveninEq.(11),toes- in the beam geometry. timate the magnitude of the PNC signal, and from We conclude this section with an estimate of thistheintegrationtimerequiredtoachieveauseful the maximum value of the d.c. field amplitude E0 statistical uncertainty of the measurement. To cal- needed. As discussed in the previous section, we culatethesignalsize,wewilluse|CF(cid:48)m±1|=(cid:112)7/8, expect that the ratio E /β is approximately 27 Fm PNC εrf = 250 V/cm, and w = 2.50 cm. The value of V/cm. In carrying out the measurements, we must x,0 rf CF(cid:48)m±1 is valid for cesium ground state transitions vary the Stark-induced angle ΘSt over the range Fm from zero to ∼ ±3|Θ |. This requires a variable (F,m) = (3,±3) → (4,±4) or (4,±4) → (3,±3), PNC field strength of maximum value 3E /β ≈ ±80 and we will show in Sec. VI that the values of PNC V/cm. the peak field amplitude and radius are reasonable. Then using the cesium atomic beam peak velocity v = 270 m/s, we estimate that the interaction angle V. STANDING WAVE CAVITY for the PNC interaction is Θ =±i5.6×10−6. In the previous section, we estimated the mag- PNC nitude of the hyperfine ground state PNC coherent To measure this amplitude, one can drive the inter- controlsignal,basedonexpectedatomicparameters fering Raman and PNC interactions, and count the andreasonablefieldstrengthsthatcanbegenerated transition rate as a function of the phase difference in the laboratory. Among the latter was an r.f. field between the transitions. A minimal measurement amplitude εrf of 250 V/cm. This field amplitude x,0 may consist of N , the total count of atomic exci- canbeachievedeitherinsidearesonantpowerbuild- + tations when the r.f. and Raman interactions are in upcavity,orbyusingaverylarger.f.amplifier. Use 8 +V -V y x z d Cs w -V +V FIG. 6: (Color online) The electrode configuration that supports the standing wave r.f. field ε and the static x electric field E0. z FIG. 5: (Color online) The standing wave pattern of the r.f. electric field εrf and magnetic field hrf, with the x z atomic beam located at the node of the magnetic field. basedonaparallelplatetransmissionlinestructure, whichallowsspatialconfinementofther.f.fieldand generation of a transverse d.c. electric field. of a resonant cavity also helps to suppress the mag- neticdipolecontributionstothemeasuredsignal,as we now discuss. This approach is also discussed in VI. PARALLEL PLATE TRANSMISSION Ref. [7]. LINE STRUCTURE As we discussed earlier, the large magnetic dipole amplitude is suppressed to first order by the choice Themeasurementthatwehavedescribedpresents of orientations of the primary fields. (The hrf com- several experimental challenges. First, we must ap- z ponent drives a ∆m=0 transition, whereas the in- ply r.f. and static electric fields that are oriented terference that we have discussed takes place on a in directions that are perpendicular to one another. ∆m = ±1 transition.) Still, due to the large mag- Second, we require that the r.f. field is in a standing nitude of the ratio M/E and reasonable limits wave configuration for suppression of the magnetic PNC in the field uniformity and experimental alignment, dipole contributions. And third, we must minimize additionalmeasuresarerequiredtosuppressthisin- the unwanted field components of the r.f. field, as teraction further. This additional suppression can thesealsoleadtosystematicmagneticdipolecontri- be achieved by working in a standing wave config- butionstothesignal. Inthissection,wedescribean uration, in which the nodes of the magnetic field electrode configuration that allows us to meet these coincide with the anti-nodes of the electric field, as requirements. we illustrate in Fig. 5. At this point, the interac- In Fig. 6, we show a section of a parallel plate tionsVint andVint aremaximized,andVint ismin- transmissionline, withwavespropagatinginthe±y PNC St M imized. To take best advantage of this, one should directions, that is modified in two regards. First, (1) use a cavity geometry in which the amplitudes we have isolated several conducting pads on the top of the traveling waves propagating in the +y and andbottomconductorsforapplicationofad.c.bias, −y directions, εrf and εrf, respectively, are equal, and secondly, we have inserted cylindrical reflectors + − and (2) keep the radius b of the atomic beam small. toeithersideoftheinteractionregiontoformanr.f. Thefirstrequireseitherthatthecavityissymmetric cavity, open on the z faces, allowing power build-up (thereflectivitiesofthetwoendreflectorsareequal, ofthecavitymodeattheresonantfrequency. When and the cavity is excited by equal amplitude inputs wehavebiasedthed.c.padsprogressively,atavolt- on each side), or that one of the reflectors has unit age +V on one side to −V on the other, we can reflectivity. The choice of beam radius b is a com- generate an electric field E0 in the central region promise between large atom number, improving the between the plates that is primarily directed in the counting statistics, or small magnetic dipole ampli- ±z-direction. We capacitively couple each of the tudeforatomsattheedgeofthebeam,whichscales bias pads to the transmission line structure so that as sin(kb) = sin(2πb/λ), where λ = 3.2 cm is the they carry the a.c. components without any signifi- wavelength of the 9.2 GHz wave. For b = 0.5 mm, cantperturbation. Foratransmissionlinecharacter- this reduction factor is ∼0.1. Furthermore, the sign istic impedance Z = 50 Ω, this requires a coupling 0 of the magnetic dipole amplitude is opposite on the capacitance of C (cid:38) 30 pF. twosidesofthenode,furtherreducingthiscontribu- We can model the cavity modes that are sup- tion. WewillreturntothisreductioninSectionVII. ported by the parallel-plate structure in the region In the next section, we will discuss the design and between the cylindrical reflectors approximately us- analysis of a symmetric r.f. power build-up cavity ing the elliptical Hermite-Gaussian modes as de- 9 scribed in Yariv [54]. These modes are nearly Gaus- (cid:96) = 11.9 cm, the cavity has a resonance at the ce- c sian in shape in the z-direction, but uniform in the sium hyperfine transition frequency (9.2 GHz), its x-dimension, in the limit of an infinite beam size in free spectral range (FSR) is c/2(cid:96) = 1.26 GHz, the c this dimension. Within the cavity, the spatial mode beamradiusatthewaistis2.50cm,thebeamradius is described by the superposition of waves traveling atthereflectorsis3.53cm,andthetransversemode in the +y and −y directions, spacingis0.2487timestheFSR,orabout313MHz. Weestimatethefieldamplitudeattheinteraction εrxf(y,z)=εr+f(y,z)+εr−f(y,z), (12) regionasfollows. Wechoosethespacingbetweenthe parallel plates of the transmission line to be 1 cm, and and the conductor width 7.5 cm. These dimensions hrf(y,z)= 1 (cid:0)εrf(y,z)−εrf(y,z)(cid:1), (13) yield a characteristic impedance of the transmission z η + − line of 50 Ω, and allow for a reasonable clearance 0 of the atomic beam in the space between the con- where ductors. With a copper thickness on the reflectors of 170 nm, we calculate a reflection coefficient of (cid:114) w (cid:26) εrf(y,z) = εrf 0 exp ∓i[ky−η(y)] 0.9992. Note that this thickness is smaller than the ± 0,± w(y) skindepthδ=680nmofcopperatthisfrequency,so (cid:20) 1 ik (cid:21)(cid:27) thetransmissionlossesaresmall, butnotnegligible. −z2 + , w2(y) 2R(y) With this reflectivity, thecavity losses due toreflec- tion are of the same magnitude as the losses L due Intheseexpressions,w isthe1/e2 (intensity)beam to other mechanisms, primarily conduction losses in 0 radius at the focus, the beam profile radius a dis- the upper and lower conducting plates, and diffrac- tance y from the focus is tion losses due to the finite size of the conductor. (These results come from our numerical analysis of (cid:112) w(y)=w 1+(y/y )2, thecavitymodes,whichwediscussnext.) Foranr.f. 0 0 input power of 250 mW incident on the cavity from y0 is the confocal parameter either side, the incident voltage amplitude is 5.0 V, and the electric field of the traveling wave incident y0 =πw02/λ, on the cavity is ε+ = 5.0 V/cm. The amplitude of in the traveling wave inside the cavity is R(y) is the radius of curvature of the wavefronts R(y)=y(cid:2)1+(y /y)2(cid:3), ε+ =2 t ε+in =125 V/cm, 0 1−r2(1−L) and η(y) whereweuset=0.04forthetransmissioncoefficient 1 ofthereflectorand(1−L)=r2. Thefactor2results η(y)= tan−1(y/y ) 2 0 from symmetric inputs from the two sides. At the anti-node of the field, where the amplitudes of the istheslowphaseshift(theGuoyphase)throughthe two traveling waves inside the cavity add in phase, focal region. For a symmetric cavity constructed of thefieldamplitudeistwicethisvalue, or250V/cm. cylindrical reflectors of radius of curvature R sep- This is the value of the r.f. field amplitude that we arated by a distance (cid:96)c, the confocal parameter is usedinSec.IVtoestimatethesignalsize. Inmaking (cid:112) y0 = ((cid:96)c/2) 2R/(cid:96)c−1, the beam radius at the this estimate, we have not included the absorption center is w = (λ(cid:96) /2π)1/2(2R/(cid:96) −1)1/4, and the ofthecopperreflector,whichreducestheamplitude, 0 c c beam radius at the reflectors is w(y = ±(cid:96) /2) = or the increase of the wave amplitude as the wave c (λR/π)1/2(2R/(cid:96) −1)−1/4. The cavity mode has an propagates to the waist of the Gaussian profile. c electric field anti-node (and magnetic field node) at In order to determine more-detailed field parame- thecenterwhenthecavitylength(cid:96) isapproximately ters,wehavecarriedoutaseriesofnumericalsimula- c (n+1/2)λ, where n is an integer. The r.f. beam ra- tionsofthecavitymodeusingComsolMultiPhysics. dius w(y = ±(cid:96) /2) at the reflectors is minimized These simulations allow us to determine the effects c when the reflector spacing is confocal, i.e. (cid:96) = R. ofresistivelossesoftheparallelplates,thethickness c By adjusting the reflector slightly away from the of the reflective copper layers, and the finite width confocalspacing, onecanretainthesmallmodesize of the cavity on the cavity Q; the effect of the gaps w(±(cid:96) /2) at the reflectors, but shift the frequencies intheconductorbetweenthed.c.biaspads;andthe c of the transverse modes away from the frequency of uniformity of the static electric field in the interac- the lowest order mode, improving the selectivity of tionregion. Weshowthethreeprimarycomponents, cavity modes. We calculate that for R=12 cm and Re[εrf(y,z)], Im[hrf(y,z)], and Im[hrf(y,z)], of the x z y 10 simulated r.f. field mode in Fig. 7. We note very electric field E0, and shown that with an array of close agreement of the components εrf(y,z) with 10 bias pads and ∆V = 100 V between pads, we x the analytic result in Eq. (12) and hrf(y,z) with can generate a relatively uniform field of magnitude z Eq. (13). The component Im[hrf(y,z)] would be E0(z)∼140V/cm. Weshowthisfield,normalizedto y z negligible for a weakly focused beam, but since in its maximum value, as the red dashed line in Fig. 8. our geometry, w ∼ λ, this component survives. We also show E0(z) in the plot (black dotted line), rf x For this figure, the separation between the upper which is small in magnitude, and has an average and lower conducting planes of the PPTL and the value of zero. The non-uniform part of E0(z), seen z width of the conductors are as before, 1.0 cm and in Fig. 8 as a nearly sinusoidal modulation of am- 7.5 cm, respectively, as are the radius of curvature plitude ∼7% of the constant part, has little impact of the cylindrical reflectors R = 12.0 cm, and the onthemeasurement. Wecanseethisbyintegrating reflector separation (cid:96) = 11.9 cm. With the thick- theproductE0(z)εrf(z)acrosstheinteractionregion c z x ness of the copper reflector layers equal to 200 nm, in z. For the case of ten bias pads, as shown, the we determine a cavity Q of 9000, while for a 1.5 µm correction to the signal due to the sinusoidal modu- ∼ 2δ layer, the Q increases to 13,000. In the lat- lation is less the 0.7% of the signal. We can also see ter case, the Q is limited primarily by the resistive in this figure that the width of the Gaussian shaped losses in the conductors and diffraction losses of the r.f. field profile is somewhat less than the width of finitewidthofthereflectors. ForacavityQof9000, the d.c. field, allowing us to avoid fringe effects of the linewidth of the transmission peak of the cavity the d.c. field near the edges of the conductors. is ∆ν = ν /Q ∼ 1 MHz. We show the computed We have used these simulations of the field am- 0 Gaussian r.f. field amplitude, εrf(0,z) across the in- plitudes, andtheirvariationthroughtheinteraction x teractionregionasthesolidbluecurveinFig.8. The region, to estimate systematic contributions to the diameter of the cavity mode agrees well with 2w = PNC signal. We discuss these contributions in the 0 5.0 cm that we determined analytically earlier. next section. We used the Eigenfrequency module and fre- quency domain analysis to carry out these calcula- tions,anddeterminedthequalityfactorofthecavity VII. ESTIMATION OF MAGNETIC DIPOLE as the ratio of the energy stored inside the cavity to CONTRIBUTIONS TO THE PNC SIGNAL the diffraction and dissipation losses. We obtained the field patterns by launching a 9.2 GHz plane- In this section, we will make use of the field sim- wave-likeelectricfieldontheparallelplatetransmis- ulations of the previous section in order to estimate sion line towards the cavity, exciting a TE cavity q,n the expected systematic contributions to the PNC mode, where indices q and n label the transverse signal. Theprimarycontributionsthatmustbecon- andlongitudinalmodes. Themodespacingbetween sidered are the magnetic dipole terms, due to the the TE and TE mode agrees well with the q,n q,n+1 relatively large magnetic dipole moment M on this 1.26 GHz FSR that we determined earlier. We used transition. As we have shown, the primary mag- a trial-and-error approach to reduce the diffraction netic field components of the r.f. field are h(cid:48)(cid:48)(y,z) losses by varying the cavity parameters, such as the z and h(cid:48)(cid:48)(y,z), where we use primed (double-primed) width and height of the cavity, while maintaining y variables for the real (imaginary) part of the field the resonant mode frequency close to 9.2 GHz. quantities, and omit the superscript ‘rf’. By set- In order to calculate the r.f. field distributions in ting up the geometry of the experiment to make the a more refined manner in the interaction region, we atomic beam cross the r.f. field at the center of the added about ten thousand times more mesh points cavity, where the component h(cid:48)(cid:48)(y) is minimal, the z in the vicinity of the interaction region. Higher magneticdipolecontributionstothesignalfromany mesh point density helped to reduce errors that are individual atom can be reduced. Furthermore, the present in the interpolation schemes, without com- contributions from atoms on one side of the node promising the eigenfrequency calculations. We used are of opposite sign to those on the other side of ten bias pads, with the spacing between the pads the node, and the net magnetic dipole contribution about one tenth the width of the pads. As long as can be suppressed even further. In this section, we the transmission lines are thin (less than 0.1 mm), usethenumericalsimulationsofthefieldssupported the gaps have little impact on the r.f. fields. We by the resonant cavity to explore the magnitude of found that neither horizontal nor vertical misalign- magnetic dipole contributions to the PNC signal. mentofthecylindricalreflectorsaffectsthefieldpat- The net contribution of the h(cid:48)(cid:48)(y,z) term can be z terns or the Q factor, for misalignment less than 1 minimized by adjusting the relative position ∆y of degree. the center of the atomic beam relative to the node Wehavealsomodeledallcomponentsofthestatic of the magnetic field. (No control of the x-position