Quasiclassical theory of quantum defect and spectrum of highly excited rubidium atoms Ali Sanayei and Nils Schopohl∗ Institut fu¨r Theoretische Physik and CQ Center for Collective Quantum Phenomena and their Applications in LISA+, Eberhard-Karls-Universita¨t Tu¨bingen, Auf der Morgenstelle 14, D-72076 Tu¨bingen, Germany Jens Grimmel, Markus Mack, Florian Karlewski, and J´ozsef Fort´agh† Physikalisches Institut and CQ Center for Collective Quantum Phenomena and their Applications in LISA+, Eberhard-Karls-Universita¨t Tu¨bingen, Auf der Morgenstelle 14, D-72076 Tu¨bingen, Germany (Dated: August 8, 2016) 5 We report on a significant discrepancy between recently published highly accurate variational 1 calculations and precise measurements of the spectrum of Rydberg states in 87Rb on the energy 0 scale of fine splitting. Introducing a modified effective single-electron potential we determine the 2 spectrum of the outermost bound electron from a standard WKB approach. Overall very good agreement with precise spectroscopic data is obtained. n a J PACSnumbers: 31.10.+z,32.80.Ee 8 2 I. INTRODUCTION reported in [5] choosing ] h The spectrum of the outermost bound electron of an Zeff(r;l)=1+(Z−1)e−ra1(l)−re−ra2(l)[a3(l)+ra4(l)] p alkali atom like 87Rb is hydrogen like, but lacks the n2- (3) - and t degeneracy of the eigenstates labeled by the principal n a quantum number n of the pure Coulomb potential [1],[2] (cid:20) (cid:16) (cid:17)6(cid:21) 1−exp − r qu 1 V (r;l)= αc rc(l) . (4) [ En,l =−(n−δ )2. (1) pol 2 r4 l 1 A table of the parameters a1(l), a2(l), a3(l), a4(l), αc, v This effect is the well-known quantum defect δl, re- and rc(l) can be found in [5]. 8 sulting from the interaction of the outermost electron In an attempt to also describe the fine splitting of the 6 with the ionic core of the atom and the nucleus. In excitation spectrum of the outermost electron of 87Rb, 0 a refined version of the statistical Thomas-Fermi the- it has been suggested [4] to superimpose a posteriori a 7 ory [3], an effective potential determining the interac- spin-orbit term 0 tion between the outermost electron and the nucleus can . 501 atecnctuiraalteVleyffb(re;lm) oddeepleenddbinygaonspthheericdaislltyanscyemrmfertormic pthoe- V(cid:101)SO(r;j,l)= [1−VSαO2V(reff;j(,rl;)l)]2, (5) center and depending on the orbital angular momentum 1 l∈{0,1,2,...,n−1} [4, 5],[2]: onthepotentialV (r;l),whichtheninfluencesthespec- : eff v trum E on the scale of fine splitting and the orbitals n,j,l Xi V (r;l)=−2(cid:20)Zeff(r;l) +V (r;l)(cid:21) (2) ψn,j,l(r) accessible to the outermost electron. Here eff r pol r 1∂V (r;l) a V (r;j,l)=α2 eff g(j,l), (6) Here the function Z (r;l) represents a position- SO r ∂r eff dependent weight function that interpolates the value of the charge between unity for large r and charge number and α = λaBC (cid:39) 1371.036 denotes the fine-structure con- stant, and Z neartothenucleusforr →0,andV (r;l)represents pol a short-ranged interaction taking into account the static 0 if l=0, electric polarizability of the ionic core [1, 6]. Overall good agreement with spectroscopic data of al- g(j,l)= (7) kali atoms (but discarding the fine splitting) has been j(j+1)−l(l+1)−34 if l≥1, 2 where j ∈ (cid:8)l− 1,l+ 1(cid:9). To determine those orbitals 2 2 (with principal quantum number n = n + l + 1 and r ∗ [email protected] radial quantum number n ∈ N ), a normalizable solu- r 0 † [email protected] tiontotheSchr¨odingereigenvalueproblemfortheradial 2 wavefunctionU (r)=rR (r)andassociatedeigen- thusengendersasystematic(small)errorofthematrixel- n,j,l n,j,l values E <0 is required: ementscalculatedin[7]onthefine-splittingscale. When n,j,l employingsubstantiallymoreSTOsthiserrorwouldcer- (cid:20) d2 l(l+1) (cid:21) tainly become larger. With N = 500 STOs the discrep- −dr2 + r2 +V(cid:101)(r;j,l)−En,j,l Un,j,l(r)=0, ancy of these theoretical results with the high precision (8) spectroscopic data, as shown in Tables I and II, is far where too large to be corrected by simply replacing VSO(r;j,l) with V(cid:101)SO(r;j,l). Hence another explanation is required. V(cid:101)(r;j,l)=Veff(r;l)+V(cid:101)SO(r;j,l) (9) denotes the effective single-electron potential. II. QUASICLASSICAL APPROACH AND FINE Ahighlyaccuratevariationalcalculationoftheexcita- SPLITTING OF THE HIGHLY EXCITED 87RB tionspectrumoftheoutermostelectronof87Rbhasbeen carried out recently [7], in which the authors expand the In 1941 alkali atoms have already been studied in the radial wavefunction of the Schr¨odinger eigenvalue prob- context of modern quantum mechanics in the seminal lem (8) in a basis spanned by 500 Slater-type orbitals workbyGoeppertMayer[3],whoemphasizedtheexcep- (STOs). Ontheotherhand,modernhighprecisionspec- tional role of the l = 1 and l = 2 orbitals. According troscopyofRydberglevelsof87Rbhasbeenconductedre- to Goeppert Mayer, the outermost electron of an alkali cently. Millimeter-wavespectroscopyemployingselective atomisgovernedbyaneffectiver-dependentchargeterm field ionization allows for precise measurements of the energy differences between Rydberg levels [8]. An inde- Zeff(r)=1+(Z−1)F(r), (12) pendent approach is to perform purely optical measure- wherethefunctionF(r)hasbeendeterminedbyemploy- ments on absolute Rydberg level energies by observing ing the semi-classical statistical Thomas-Fermi approach electromagnetically induced transparency (EIT) [9, 10]. to the many-electron-atom problem, posing the bound- However,thereisasystematicdiscrepancybetweenvari- aryconditionsaslim F(r)=1andlim F (r)=0. ational calculations and the spectroscopic measurements r→0 r→∞ As discussed by Schwinger [12], this approach ceases to of the fine splitting be valid in the inner shell region Z−1 < r < Z−31 of ∆E =E −E (10) the atom. Therefore, taking into account the fine split- n,l n,l−1,l n,l+1,l 2 2 ting in the spectrum of the outermost electron of alkali as shown in Tables I and II. Given the fact that the er- atoms a posteriori by simply adding the phenomenolog- rorbarsoftheindependentexperiments[8,10]arebelow ical spin-orbit term (5) to (2), resulting in the effective 1.1MHz down to 20kHz, and on the other hand con- single-electronpotential(9),seemstobequestionableon sidering the high accuracy of the numerical calculations general grounds in that inner shell region. presented in [7], such a discrepancy between experiment On a more fundamental level, the treatment of rela- and theory is indeed significant. tivistic effects in multi-electron-atom spectra requires an So, what could be the reason for the reported discrep- apriori microscopicdescriptionbasedonthewell-known ancies? First, it should be pointed out that in the vari- Breit-Pauli Hamiltonian [13, 14] ational calculations [7] a slightly different potential was H =H +H +H . (13) used, that is, nr rs fs Here H is the ordinary nonrelativistic many-electron V (r;j,l)=V (r;l)+V (r;j,l). (11) nr eff SO Hamiltonian, while the relativistic corrections are repre- sented by the perturbation operators H and H . The Certainly, within the first-order perturbation theory rs fs perturbation term H contains all the relativistic per- thereexistsnonoticeablediscrepancyinthespectrumof rs turbations like mass correction, one- and two-body Dar- the outermost electron on the fine-splitting scale, when win terms, and further the spin-spin contact and orbit- takingintoaccountthespin-orbitforceswithV (r;j,l) SO orbit terms, which all commute with the total angular instead of working with V(cid:101)SO(r;j,l). This is due to the momentum L and total spin S, thus effectuating only differences being negligible for r > Zα2. However, since smallshifts ofthespectrumofthenonrelativisticHamil- VSO(r;j,l)eventuallydominateseventhecontributionof tonianH . TheperturbationoperatorH ontheother nr fs the centrifugal barrier term l(l+1) within the tiny region hand breaks the rotational symmetry. It consists of the r2 0 < r (cid:46) α2Z, a subtle problem with a non-normalizable standardnuclear spin-orbit,thespin-other-orbit,andthe radialwavefunctionU (r)emergeswhenattemptingto spin-spin dipole interaction terms, which all commute n,j,l solve the Schr¨odinger eigenvalue problem for any l > 0 withJ=L+S,butnotwithLorwithSseparately,thus with the potential V (r;j,l). Such a problem is absent inducingthefinesplittingofthenonrelativisticspectrum. SO when one works with V(cid:101)SO(r;j,l) [4]. Although the proposed functional form of the poten- A variational calculation with the potential (11) em- tial (11) is highly plausible on physical grounds out- ploying N = 500 normalizable STOs as basis functions side the inner core region r > Z−31, prima facie it ap- 3 TABLE I. Fine splitting ∆En,l=1 for P states in [MHz]. State n,l=1 Exp. [11] Exp. [8] Theory [7] Theory (this work) | (cid:105) 8P 565.1(4) 103 N/A 602.00 103 567.75 103 × × × 10P 219.1(4) 103 N/A 231.87 103 218.77 103 × × × 30P N/A 4246.30(5) 4500.50 4246.46 35P N/A 2566.41(32) 2717.41 2566.28 45P N/A 1144.09(13) 1217.24 1143.95 55P N/A 605.77(7) 644.81 605.68 60P N/A 460.76(5) 480.32 460.68 TABLE II. Fine splitting ∆En,l=2 for D states in [MHz]. State n,l=2 Exp. [11] Exp. [8] Exp. [10] Theory [7] Theory (this work) | (cid:105) 8D 30.4(4) 103 N/A N/A 113.17 103 36.42 103 × × × 10D 14.9(2) 103 N/A N/A 52.05 103 16.56 103 × × × 30D N/A 452.42(18) 452.5(11) 1447.53 456.13 35D N/A 279.65(10) 280.4(11) 894.84 281.52 45D N/A 128.33(4) 127.8(11) 407.64 128.98 55D N/A 69.17(2) 69.4(11) 223.71 69.47 57D N/A 61.98(2) 62.2(11) 197.39 62.24 pearstobeinconsistenttolumptheaforementionedrela- fectivesingle-electronpotentialV (r;j,l)describingthe eff tivisticmany-bodyforcesintoaneffectivesingle-electron fine splitting of the spectrum of the outermost electron potential of the functional form (11), so that it pro- in the alkali atoms, we introduce a cutoff at a distance vides an accurate description also for small distances rso(l) with Z−1 < rso(l) < Z−13 so that the effective Z−1 <r <Z−13. single-electron potential is now described by the follow- Intheabsenceofabettermicroscopictheoryforanef- ing modified potential: V (r;l) if 0≤r ≤r (l), eff so V(cid:101)mod(r;j,l)= (14) V (r;l)+V (r;j,l) if r >r (l). eff SO so The choice [2] of the atom. Because the higher the orbital angular mo- mentum quantum number l, the lower the probability of r (l=1)=0.029483×r (l=1)=0.0442825, so c the electron being located near to the center, it is clear (15) that the quantum defect decreases rapidly with increas- r (l=2)=0.051262×r (l=2)=0.2495720, ing orbital angular momentum l. Therefore, ∆ is only so c j,l notably different from zero for l=0,1,2. gives a surprisingly accurate description of the fine split- Writing∆ =δ +η withη (cid:28)δ ,thefinesplitting ting in the spectroscopic data for all principal quantum j,l l j,l j,l l to leading order in α2 is: numbers n, see Tables I and II. The calculation of the spectrum of the outermost ηl−1,l−ηl+1,l ∆E =2 2 2 (17) bound electron is then reduced to solving the radial n,l (n−δ )3 l Schr¨odinger equation (8) with the modified potential √ The quasiclassical momentum p≡ −Q of the bound V(cid:101)mod(r;j,l). The resulting spectrum is actually hydro- electron with orbital angular momentum l>0, total an- gen like, that is, gular momentum j =l± 1, and taking into account the 1 2 E =− , (16) Langershiftl(l+1)→(cid:0)l+ 1(cid:1)2 inthecentrifugalbarrier n,j,l (n−∆ )2 2 j,l [15, 16], is then given by where∆ denotesaquantumdefectcomprisingalsothe j,l (cid:0)l+ 1(cid:1)2 fianreedsupclittitoinngo.fItnheacntuumalbfearctofthneodqeusanntuomftdheefercatddiaelscwraibvees- Q(r;j,l,E)= r22 +V(cid:101)mod(r;j,l)−E. (18) r function for l = 0,1,2 as a result of the short-range in- For l =0 the centrifugal barrier term and the spin-orbit teraction of the outermost electron with the ionic core potential are absent. 4 100 20 80 15 ) E 60 Q j,l, √−10 ( ν 40 5 20 0 0 0 20 40 60 80 100 0 0.2 0.4 0.6 0.8 1 1.2 1/√ E r/r (l) − c FIG. 1. (Color online) The action integral ν(j,l,E) associ- FIG. 2. (Color online) The quasiclassical momentum avtse.dscwailtehd ethneeregffyec√t−1ivEe fsoinrglle=-e0lec(tbrloune lpinotee)n,tlia=l 1V(cid:101)m(roedd(rl;inj,el)), (cid:112)bla−ckQ),(rl;=j,l1,E(g)revesn.),scla=led2 (driesdta),ncfoerrEcr(l=) fEorn,jl,l=co0rr(edspasohnedd- l=2 (green line), all for j =l+ 1. The curves for j =l 1 ing to principal quantum number n = 57 and j = l + 1. 2 − 2 2 only differ by a tiny shift proportional to α2. The main contribution to the quantum defect values in (24) originates from the inner core region r<rc(l). Considering high excitation energies E < 0 of the where ν(j,l,E) denotes the action integral boundoutermostelectron,i.e. aprincipalquantumnum- ber n(cid:29)1, the respective positions of the turning points 1 (cid:90) r(+) (cid:112) r(±) are given approximately by ν(j,l,E)= dr −Q(r;j,l,E) π r(−) 1 (cid:73) = drp(r;j,l,E). (21) (cid:0)l+ 1(cid:1)2 2π r(−) = 2 if l≥3, (cid:113) 1+ 1+(cid:0)l+ 1(cid:1)2E Plotting the function ν(j,l,E) versus √1 for l = 2 −E 0,1,2 clearly reveals a linear dependence of the form (19) ν(j,l,E)= √1 +c(j,l), see Fig. 1. (cid:115) −E 1 (cid:18) 1(cid:19)2 According to [6], for A,B,C,D ∈R, with A>0, B > r(+) (cid:39) −E 1+ 1+ l+ 2 E if l≥1, 0, C >0, and |D|(cid:28)C the following equality holds: 1 (cid:73) (cid:114) 2B C D B √ BD dr −A+ − + = √ − C+ √ where 0 < l (cid:28) √1 . Of course for l = 0 only a single 2π r r2 r3 A 2C C −E (22) (large)turningpointr(+) = 2 existsduetotheabsence −E For a pure Coulomb potential A ≡ −E, B ≡ 1, C ≡ of the centrifugal barrier. However, the lower turning (cid:0)l+ 1(cid:1)2 and D ≡ α2g(j,l). The corresponding action points r(−) are strongly modified for l = 1,2 compared 2 integral then reads to the pure Coulomb potential case taking into account 0th.0e2c×orercp(ol)larhiozladtsio;nt.haFtorisl,=r(1−,)2(lth=e1r)ela(cid:39)tio0n.0r3(4−7)2(la)n(cid:39)d √−1E if l=0, r(−)(l=2)(cid:39)0.12827 [2]. Since the cutoff r (l) in (15) ν(C)(j,l,E)= so ipsosinutbssrta(−n)ti(al)ll,yaaqbuoavseicltahsossiceavlaclaulceusloaftiothneoflotwheerfintuer-snpinligt √−1E −(cid:0)l+ 12(cid:1)+ 2α(2lg+(j12,)l)3 if l>0. (23) spectrum of the bound outermost electron is reliable. Forachosenradialquantumnumbern ,theassociated It is thus found from WKB theory that the quan- r eigenvaluesE =E <0oftheoutermostelectronnow tum defect associated with the single-electron potential n,j,l follow from the WKB patching condition [17, 18]: V(cid:101)mod(r;j,l) is: (cid:104) (cid:105) ∆j,l = lim ν(j,l,E)−ν(C)(j,l,E) (24) nr+1 if l=0, E→0− ! ν(j,l,E)= (20) Ignoring spin-orbit coupling, i.e. for α = 0 , one has n + 1 if l>0, ∆ ≡ δ , the standard quantum defect. For l = 0 the r 2 j,l l 5 TABLE III. The values of quantum defect ∆j,l associated with the Rydberg level n=57 for l=0,1,2. Quantum defect ∆j,l Exp. [8] Exp. [10] Theory [7] Theory (this work) ∆ 3.1312419(10) 3.13125(2) 3.12791 3.13095 1/2,0 ∆ 2.6549831(10) N/A 2.65795 2.65197 1/2,1 ∆ 2.6417735(10) N/A 2.64399 2.63876 3/2,1 ∆ ∆ 0.0132096(14) N/A 0.01396 0.01321 1/2,1− 3/2,1 ∆ 1.3478971(4) 1.34789(2) 1.35145 1.34851 3/2,2 ∆ 1.3462733(3) 1.34626(2) 1.34628 1.34688 5/2,2 ∆ ∆ 0.0016238(5) 0.00163(3) 0.00517 0.00163 3/2,2− 5/2,2 centrifugal barrier and the spin-orbit coupling term (6) of the radial Schr¨odinger equation (8) [21]. are zero, so ∆ →∆ ≡δ . j,l 1,0 0 2 The dependence of the quasiclassical momentum III. CONCLUSIONS (cid:112)−Q(r;j,l,E) on the scaled distance r is shown for rc(l) l = 0,1,2 in Fig. 2. Clearly, it is the inner core re- In this work we reported a significant discrepancy be- gion r(−)(l) < r < rc(l) that provides the main con- tween experiment [8, 10] and highly accurate variational tribution to the quantum defect values. We find, for calculations[7]ofthespectrumofRydbergstatesof87Rb l = 0,2, that changing the fitting parameter a3(l) in on the energy scale of the fine splitting. We discussed (3) from its tabulated value in [5] according to the thattheusuala posteriori addingoftherelativisticspin- scaling prescription a3(l=0) → 0.814×a3(l=0) and orbit potential to the effective single electron potential a3(l=2) → 0.914×a3(l=2), leads to a slight down- governing the outermost electron of alkali atoms is in- ward constant shift of the WKB-quantum defect. As a deed inconsistent inside the inner atomic core region. result of this change, the calculated WKB-quantum de- In the absence of a full microscopic theory that lumps fect ∆l±1,l then agrees well with the spectroscopic data, all many-body interactions together with the relativistic 2 see Table III. Such a change of a3(l) does not affect the corrections into an effective single-electron potential in fine splitting values ∆En,l though. We also find that the a consistent manner, we suggested a modified effective dependence of the fine splitting ∆En,l on the principal single-electron potential, see (14), that enables a cor- quantumnumberniswelldescribedby(17)foralln≥8, rectdescriptionofthespectrumofRydbergstatesonthe see Tables I and II. fine splitting scale in terms of a simple WKB-action in- tegral for all principal quantum numbers n≥8. Modern Inactualfact,forr(+) (cid:29)r(−),whichisacriterionthat √ precision spectroscopy of highly excited Rydberg states isalwaysmetforhighexcitationenergies −E (cid:39)0ofthe thusenablestheprobingofthemulti-electroncorrelation outermostelectron, theuniformLanger-WKBwavefunc- problem of the ionic core of alkali atoms. This is cer- tionU(WKB)(r)[19,20],withr(+) consideredastheonly n,j,l tainly a fascinating perspective for further experiments turning point, describes the numerical solution Un,j,l(r) and theoretical studies. to the radial differential equation (8) under the influ- ence of the effective modified single-electron potential (14) rather accurately [21]. Only very near to the sec- ACKNOWLEDGMENTS ondturningpointr(−),atadistancesmallerthanr (l), so the Langer-WKB wavefunction U(WKB)(r) ceases to be ThisworkwasfinanciallysupportedbytheFET-Open n,j,l agoodapproximationtothenumericalsolutionU (r) Xtrack Project HAIRS and the Carl Zeiss Stiftung. n,j,l [1] T. F. Gallagher, Rydberg Atoms, 1st ed. (Cambridge [6] M. Born, Vorlesungen u¨ber Atommechanik (Springer, Univ. Press, Cambridge, 1994). Berlin, 1925) 27, 28 and II. Anhang. § § [2] We use scaled variables so that length is measured [7] M. Pawlak, N. Moiseyev, and H. R. Sadeghpour, Phys. in units of the Bohr radius aB = m(cid:125)2e4|πe|ε20 (cid:39) [8] WRe.vL.iA,I8.9M,o0u4r2a5c0h6ko(2,0M14.)W. .Noel, andT.F.Gallagher, 5.2918 10−11mandenergyismeasuredinunitsofRy- dberg,×Ry= m8εe20|he|24 (cid:39)13.605eV. [9] APh.yKs..MRoevh.apAat6r7a,,T05.2R5.0J2a(c2k0s0o3n),.andC.S.Adams,Phys. [3] M. Goeppert Mayer, Phys. Rev. 60, 184 (1941). Rev. Lett. 98, 113003 (2007). [4] C. H. Greene and M. Aymar, Phys. Rev. A 44, 1773 [10] M. Mack, F. Karlewski, H. Hattermann, S. Ho¨ckh, (1991). F. Jessen, D. Cano, and J. Forta´gh, Phys. Rev. A 83, [5] M. Marinescu, H. R. Sadeghpour, and A. Dalgarno, 052515 (2011). Phys. Rev. A 49, 982 (1994). 6 [11] J. E. Sansonetti, J. Phys. Chem. Ref. Data 35, 301 [17] A. B. Migdal, Qualitative Methods in Quantum Theory (2006). (Addison-Wesley, 1977). [12] J.Schwinger,QuantumMechanics: SymbolismofAtomic [18] B.M.KarnakovandV.P.Krainov,WKBApproximation Measurements, 1st ed. (Springer, Berlin, 2001). in Atomic Physics (Springer, Berlin-Heidelberg, 2013). [13] H. A. Bethe and E. E. Salpeter, Quantum mechanics of [19] R. E. Langer, Bull. Am. Math. Soc. 40, 545 (1934). one- and two-electron atoms (Springer, Berlin, 1957). [20] C. M. Bender and S. A. Orszag, Advanced mathemati- [14] C. Froese-Fischer, T. Brage, and P. Jo¨nsson, Compu- cal methods for scientists and engineers (McGraw-Hill, tational atomic structure: An MCHF approach (IOP Singapore, 1978). Physics, Bristol and Philadelphia, 1997). [21] A. Sanayei and N. Schopohl, “Unpublished”. [15] R. E. Langer, Phys. Rev. 51, 669 (1937). [16] M.V.BerryandK.E.Mount,Rep.Prog.Phys.35,315 (1972).