Unpaired and spin-singlet paired states of a two-dimensional electron gas in a perpendicular magnetic field M. Polini1, K. Moulopoulos2, B. Davoudi1,3 and M. P. Tosi1 1NEST-INFM and Classe di Scienze, Scuola Normale Superiore, I-56126 Pisa, Italy 2Department of Physics, University of Cyprus, P.O. Box 20537, 1678 Nicosia, Cyprus 3Institute for Studies in Theoretical Physics and Mathematics, Tehran 19395-5531, Iran 2 0 0 We present a variational study of both unpaired and spin-singlet paired states induced in a 2 two-dimensional electron gas at low density by a perpendicular magnetic field. It is based on an n improved circular-cell approximation which leads to a number of closed analytical results. The a ground-stateenergyoftheWignercrystalcontainingasingleelectronpercellinthelowest Landau J level is obtained as a function of the filling factor ν: the results are in good agreement with those 6 of earlier approaches and predict νc 0.25 for the upper filling factor at which the solid-liquid ≈ 2 transition occurs. A novellocalized stateof spin-singlet electron pairsis examined and found tobe a competitor of the unpaired state for filling factor ν > 1. The corresponding phase boundary is ] quantitatively displayed in themagnetic field-electron density plane. l e - PACS number: 73.20.Dx, 73.40.Hm r t s . t a m I. INTRODUCTION - d The nature of the ground-state of a two-dimensional (2D) many-electron system in a perpendicular magnetic field n has been for a number of years a subject of intense experimental and theoretical investigation. Since the discovery o of the Integer and Fractional Quantum Hall Effects1, this has been one of the richest sources of fundamental new c physics in Condensed Matter. The transition from a liquid to a Wigner crystal (WC) state, originating from strong [ electron-electron correlations at low density in strong magnetic fields, has received considerable attention (see e.g. 1 Ref. 2 and references given therein). More recently, novel and intriguing behaviors of 2D electron transport in high v Landau levels have been reported3. In such regime of weak magnetic fields Koulakov et al.4 predicted that charge 0 density waves(CDW)wouldbreakthe translationalsymmetryinone directionandformstripes. Theyalsopredicted 9 stabilization of a novel2D-CDW state with more that one electron per unit cell. Such a “bubble” phase arising from 4 1 spin-triplet electron pair formation and 2D lattice ordering has been confirmed in systematical numerical studies of 0 Haldane et al.5 and of Shibata and Yoshioka6. 2 Inthepresentworkweintroduceanewpossiblelow-temperaturephasefora2Delectronsysteminaperpendicular 0 magnetic field, namely a localized paired-electron state consisting of a spin-singlet electron pair in each cell, and / t investigate its stability at low densities against the more conventional WC state containing a single electron per cell. a We have in mind situations where the Zeeman splitting can be neglected7. In free space, where the Land`e g-factor m of an electron is g = 2, the Zeeman splitting is exactly equal to the cyclotron splitting. It turns out, however, that - this is incorrect, for example, in a GaAs heterostructure for the following reasons: (i) the small effective mass in the d ∗ n conduction band increases the cyclotron energy by a factor of m/m 14; (ii) the effective coupling of the electron ∼ o spin to the external magnetic field is reduced by spin-orbit scattering by a factor of 5, making the effective Land`e c factor g 0.4. Thus the Zeeman energy is about 70 times smaller than the cyclotro−n energy7. : ∼− v This study has two main motivations. Firstly, from the solution of the simple problem of two-electrons moving on i a plane under parabolic confinement8 it emerges that the effective radial potential for the relative motion develops a X pronouncedminimumatsufficientlyweakfield. Theminimumoccurseveninthecaseofvanishingangularmomentum r (m=0)andisthe resultofthe competitionbetweenthe Coulombrepulsionandthe localizingeffects ofthe magnetic a field and of the confinement. This feature may hide, in the many-body context and at sufficiently low electron densities, some new phase transition to a spin-singlet (for m = 0) paired electron state in weak magnetic fields. Secondly, and following early work on a m = 0 paired state in the 3D electron gas9, such a state was found in a variational calculation10 to be a possible competitor to the conventional 3D-WC in zero field. A mixed spin state with a preference for spin pairing has been proposed to occur in some regions of the liquid phase from experiments on 2D electron systems in the ultra-quantum limit11. In our evaluation of the energetics of the paired state we introduce an improved circular-cell approximation for an interacting 2D electron gas on a uniform positive background. We first study by this approach the conventional unpaired-electron WC state in a magnetic field12. Each electron is treated as a distributed charge cloud, which can 1 be taken to a good approximation to be localized within a cell if the density is not too high and the field is not too small – the radius of the circular cell being available as an additional variational parameter13. Our results are in close agreement with those obtained in earlier approaches. We then adopt a similar method to deal with two interacting electrons within each circular cell. This yields new and more complex expressions leading to a window of thermodynamic stability for the paired-electronstate at higher values of the electron density. The layout of the paper is briefly as follows. In Section II we present our treatment of the energy of the unpaired- electron 2D-WC in a magnetic field and compare our results with earlier ones and with the energy of the Laughlin liquid state. Section III summarizes the main results for the problem of two electrons moving in a plane under an applied magnetic field. Section IV reports our treatment of the singlet-paired state and Section V gives a variational calculationoftheenergyofthisstateandadeterminationofthephaseboundarybetweentheunpairedandthesinglet- paired states. Finally, Section VI summarizes our main conclusions. Some technical points of detail are worked out in an Appendix. II. DISTRIBUTED-CHARGE APPROACH TO THE WIGNER CRYSTAL IN A MAGNETIC FIELD A 2D system of carriers in a pure semiconductor sample subject to a magnetic field B is expected to undergo a transition to an ordered triangular lattice structure at sufficiently low temperature T and filling factor ν14. Here, ν =2πn ℓ2 where n is the arealcarrierdensity andℓ=(~c/eB)1/2 isthe magnetic length. The ground-stateenergy s s ofsucha2D-WChasbeenevaluatedbyanumber ofauthorswithvariousmethods. Inthe followingweshallindicate by δǫ the values of the ground-state energy after subtraction of the energy of the lowest Landau Level (LL). The leading term of a low-density expansion is given by a classical Madelung-potential calculation15, leading to δǫ = 2.212206r−1 Ryd∗ = 0.782133ν1/2 (e2/ǫℓ) where ǫ is the dielectric constant of the host medium and rclaas∗s =−(πn )−1/2 wsith Ryd∗ a−nd a∗ the effective Rydberg and Bohr radius. Following early Hartree-Fock s B s B calculations16 and a classical-plasmasimulation for the determination of the energy of the Laughlin liquid17, a varia- tional calculation based on a correlatedmagnetophonon wave-function for the WC18 gave the result δǫ = 0.782133ν1/2+0.2410ν3/2+0.16ν5/2 (e2/ǫℓ) (1) CWC − and predicted that the crystalenergy would be lower than that of the liquid only at values of ν lower than ν 0.14. c Subsequent work by Vignale19 used Current-Density Functional Theory in the projected local-density approxi≈mation and located the upper filling factor for the liquid-solid transition at ν 0.25, in fairly good agreement with the c ≈ experimental evidence. In our approach we model the 2D interacting electron system on a uniform positive background as a collection of non-interacting disks, each having a Wigner-Seitz radius r to be eventually treated as a variationalparameter and WS containing a single electron in a trial wave function with a variational width parameter σ. Specifically, we adopt a Gaussian trial wave function normalized to unity, φ (r)=(πσ2)−1/2 exp( r2/2σ2) (2) SP − The single-particledensity thus isρ(r)=φ2 (r). Takingthe arealdensity ofthe backgroundasρ (r)=n θ(r r), SP b s WS− the total electronic charge inside the disk radius is determined by d2rφ2 (r)θ(r r)=1 exp( r2 /σ2) (3) SP WS− − − WS Z and we must require σ r in order to avoid charge leakage errors. As noted by Nagy13, the handling of r as ≪ WS WS a variational parameter ensures that the error involved in setting the electrical potential outside the disk to zero is minimized. A. Electron-background and background-background interaction energy ThepotentialenergyV (r)createdbythepositivebackgroundinsidethecircularcellisgivenintermsoftheGauss in hypergeometric function F(a,b;c;d) by13 4 1 1 r V (r)= F , ;1;r2/r2 WS Ryd∗. (4) in r 2 −2 WS r a∗ s (cid:18) (cid:19) s B 2 This result is equivalent to the more familiar expression 8 r V (r)= E(r2/r2 ) WS Ryd∗ (5) in πr WS r a∗ s s B where E(x) is the complete elliptic integral of the first kind (see Ref. 20, p. 591). The self-interaction energy of the background is thus given by 3 1 16 r ǫ (r ,r )= d2rρ (r)V (r)= WS Ryd∗, (6) bb s WS 2 b in 3πr r a∗ Z s (cid:18) s B(cid:19) having used the result 1 E(x2)xdx=2/3. Similarly, the interaction energy of the electron with the backgroundis 0 R rWS ǫ (r ,r ,σ)=2π rdrφ2 (r)V (r) eb s WS SP in Z0 = 16rWS rWS/σxE σ2x2/r2 e−x2dx Ryd∗. (7) −πr2a∗ WS s B Z0 (cid:0) (cid:1) For σ r Eq. (7) can be replaced by the approximate analytic expression ≪ WS 4 r 1 (σ/a∗)2 ǫ (r ,r ,σ)= WS + B Ryd∗. (8) eb s WS −r r a∗ r3 (r /r a∗) (cid:20) s s B s WS s B (cid:21) Thisso-calledharmonicapproximation(HA)fortheelectron-backgroundenergyisobtainedfromEq. (7)byextending the range of integration up to and by using the expansion of the elliptic integral E(x) = (π/2)(1 x/4)+o(x2) ∞ − (see 20, p. 591). An alternative way to calculate the electron-backgroundenergy is by using the expression rWS ǫ (r ,r ,σ)=2π ρ (r)V (r)rdr, (9) eb s WS b SP Z0 where V (r) is the electrical potential created by a single electron in the state (2). This can be obtained in closed SP form, 2√π V (r)= exp( r2/2σ2)I (r2/2σ2) Ryd∗ (10) SP −(σ/a∗) − 0 B where I (x) is the modified Bessel function of the n-th order (see Appendix). The evaluation of Eq. (9) can then be n carried out in closed form, with the result 2 2√π r ǫ (r ,r ,σ)= WS exp( r2 /2σ2) I (r2 /2σ2)+I (r2 /2σ2) Ryd∗ (11) eb s WS −(σ/a∗) r a∗ − WS 0 WS 1 WS B (cid:18) s B(cid:19) (cid:2) (cid:3) (see Appendix). The expression (8) is recovered from Eq. (11) for r σ by using the asymptotic expansion of WS ≫ the Bessel functions (see 20, p. 377). We shall see below that the “anharmonic corrections” implied by Eq. (11) are crucial in a comparison with earlier calculations. B. Total energy and variational procedure The total energy per electron in a perpendicular magnetic field B is ǫ =ǫ (σ,B)+ǫ (r ,r ,σ)+ǫ (r ,r ) (12) t k eb s WS bb s WS where ǫ (σ,B) is the kinetic energy, k 1 e 2 ǫ (σ,B)= d2rφ (r) p+ A φ (r) k SP 2m∗ c SP Z +∞ (cid:16) ~2 (cid:17) 1 =2π rφ (r) ∂ (r∂ )+ m∗ω2r2 φ (r)dr (13) SP −2m∗r r r 2 0 SP Z0 (cid:20) (cid:21) 3 with ω = eB/2m∗c. We have taken the vector potential in the symmetric gauge, A = B r/2, and used the fact 0 × that the wave function (2) is an isotropic state of zero angular momentum. Eq. (13) yields 2 1 σ ∗ ǫ (σ,B)= +λ Ryd (14) k (σ/a∗)2 B a∗ " B (cid:18) B(cid:19) # where λ m∗ω2a∗2/(2 Ryd∗) 3.6 10−12 (ǫm/m∗)4 (B/Tesla)2. For a GaAs heterostructure we have λ 0.4 B ≡ 0 B ≃ · B ≃ if B =10 Tesla. Equation (12), after insertion of Eqs. (6), (11) and (14) is minimized numerically with respect to the variational ∗ ∗ parameters Σ = σ/a and α = r /r a . The equilibrium values of these parameters are shown in Figure 1 as B WS s B functions of r for two values of the magnetic field. The ratio Σ¯/r α¯ =(σ/r ) should be appreciably smaller than s s WS eq unity for internal self-consistency and the extent to which this consistency criterion is satisfied is shown in Figure 2. It is evident from Figures 1 and 2 that for these values of the field the harmonic approximation is applicable for the estimation of the variational parameters whenever the consistency criterion Σ¯/r α¯ 1 is satisfied. s ≪ Figure 3 reports our results for the ground-stateenergyδǫ as a function ofthe filling factor in the lowestLL, after t subtraction of the kinetic energy 2λ1/2 Ryd∗ from the total energy ǫ and rescaling to e2/ǫℓ energy units. The left B t panel in Figure 3 shows that δǫ in these units is still weakly dependent on the field intensity at values of ν larger t thanabout0.4. The rightpanelinFigure 3 comparesour resultsfor δǫ inthe range0<ν 0.5with thoseobtained t in the correlated Wigner crystal approach of Lam and Girvin18 (see Eq. (1)). It is also se≤en that the classical limit of Bonsall and Maradudin15 is approximately recoveredfor ν 0. → ItisalsoseenfromFigure3(leftpanel)thattheground-stateenergyoftheunpairedWCasobtainedinthepresent approach crosses the energy of the Laughlin liquid as reported by Levesque et al.17 at ν 0.25. This value for the upper critical filling factor of the liquid-solid transition agrees with that reported by Vign≈ale19 and, as discussed by this author, there is strong experimental evidence supporting the fact that the crystal exists even at filling factors as large as 0.22-0.232,21. However, again as discussed by Vignale, it appears that the liquid-solid transition is not a simple crossingoftwophasesoccuringata singlevalue ofν. Rather,it showsacomplex reentrantbehavior,with the liquid phase being stable at or near the Fractional Quantum Hall Effect fractions and the solid phase being stable in between. We would also like to remark that the simple choice α=1, which ensures that the Wigner-Seitz cell is electrically neutral, implies only small changes relative to the prescription proposed by Nagy13. We in fact find that when we take α=1 the critical filling factor at which the liquid-solid transition occurs shifts from ν 0.25 to ν 0.23. Finally, we briefly comment on the issue of Landau Level Mixing (LLM), as studied in de≈tail by Zhu≈and Louie22 andbyPriceet al.23. LLMisimportantwhentheratiobetweenthemagneticenergy(oforder~ω )andthe Coulomb 0 energy (of order e2/(ǫr a∗)) becomes smaller then unity, i.e. for r < λ−1. This effect is included in our approach s B s B through the use of a trial wave function having a variational width23 and becomes unimportant as r increases and s σ approaches the value √2ℓ for which Eq. (2) becomes the exact lowest-energy eigenfunction of the single-particle eq Hamiltonian. This asymptotic behavior is clearly seen from Figure 1 (left panel). C. Analytic results in the harmonic approximation Having assessed numerically the range of validity of the harmonic approximation (HA) for the estimation of the modelvariationalparameters,weproceedtoreportanumberofanalyticresultswhichfollowfromit. Theground-state energy is given by 4α Σ2 16α3 ǫHA = Σ−2+λ Σ2 + + Ryd∗. (15) t B − r αr3 3πr (cid:18) s s s (cid:19) Minimization of Eq. (15) with respect to Σ and α yields the results shown in Figure 1. Let us consider first the choice α=1 (i.e. r =r a∗), where the equilibrium value of Σ is WS s B 3/4 r Σ¯ (r ,B)= s . (16) α=1 s (1+λ r3)1/4 B s Thus, Σ¯ (r ,B) decreaseswith increasingfield, due to increasedlocalizationofthe electroninside the circularcell, α=1 s and saturates to the value λ−1/4 corresponding to σ =(~/m∗ω )1/2. The total energy per electron becomes B eq 0 (1+λ r3)1/2 16 1 ǫHA(r ,B)= 2 B s + 4 Ryd∗ (17) t s (cid:20) rs3/2 (cid:18)3π − (cid:19)rs(cid:21) 4 and increases with the magnetic field, saturating to the value 16 1 ǫHA(r ,B )= 2λ1/2+ 4 Ryd∗. (18) t s →∞ B 3π − r (cid:20) (cid:18) (cid:19) s(cid:21) The first term in the brackets in Eq. (18) is the cyclotron zero-point energy ~ω . On the other hand, in the limit of 0 vanishing field Eq. (17) yields 16 1 2 ǫHA(r ,B =0)= 4 + Ryd∗ (19) t s (cid:20)(cid:18)3π − (cid:19)rs rs3/2(cid:21) 2.30 2 ∗ + Ryd , ≃(cid:18)− rs rs3/2(cid:19) which is the result obtained by Seidl et al.24 in their “PC model” (the correlation energy is obtained from Eq. (19) by subtracting from it the exchange energy term given by 8√2/(3πr )). The dependence of ǫHA(r ,B = 0) in Eq. − s t s (19) on the density parameter r is correct, but the numerical coefficients are somewhat different from those which s are precisely known for the WC in zero field15. It may be remarked that the expression for ǫHA(r ,B = 0) in 3D t s provides a lower bound to the energy of the WC25. Minimization of the energy in Eq. (15) with respect to α is intended to approximately correctfor the fact that the electrical potential in 2D does not vanish outside the circular cell and is expected to yield a variational lower bound for the ground-state energy13. The equilibrium value of the reduced Gaussianwidth is givenby an expressionsimilar to Eq. (16), α¯r3 1/4 Σ¯ (r ,B)= s (20) α¯ s 1+α¯λ r3 (cid:18) B s(cid:19) while α¯ converges with increasing r to the value √π/2 (see Figure 1). That is, Nagy’s result13 r = (√π/2)r a∗ s WS s B remains valid in the presence of a magnetic field at low electron density. The expression for the ground-state energy becomes (1+λ √πr3/2)1/2 4√π 1 ǫHA(r ,B)= 2√2 B s Ryd∗. (21) t s (cid:20) π1/4rs3/2 − 3 rs(cid:21) However, this simple analytic expression for the energy of the unpaired WC in a magnetic field is not sufficiently accurate on the energy scale needed for the comparisons made in Figure 3. III. MOTION OF AN ELECTRON PAIR IN A MAGNETIC FIELD InthissectionwemotivatetheintroductionofapairedstatefortheWCbydiscussingthespin-singletground-state and effective interaction potential for two electrons moving in a plane under a perpendicular magnetic field. As is known from the work of Taut8, this is an example of a quasi-exactly soluble problem with a hidden sl -algebraic 2 structure26: an exact solution exists for special values of the field. Our aim will be to use the analytic form of the wave function determined in such a case in order to make a reasonable Ansatz for the variational treatment which will be developed for the paired phase in the next Section. The Hamiltonian describing the two electrons is written in terms of the relative coordinate r = r r , of the 2 1 − center-of-mass coordinate R=(r +r )/2 and of the corresponding momenta p and P as 1 2 1 e 2 1 e 2 e2 = P+ A (R) + p+ A (r) + + (22) H 4m c CM m c rel r Hspin h i h i whereA (R)=B R,A (r)=B r/4and = (ge~/mc)(s +s )withs beingthe spins. Theadditionof a harmonCiMc confining×potentrieall merely×shifts theHvaslpuine of−the cyclotron1freq2uency8. iThe center-of-mass motion in the ground-stateisdescribedbyaGaussianwavefunctionhavingwidthgivenbythemagneticlengthℓ=(~c/eB)1/2. The wave function φ (r) for the relative motion in the spin-singlet ground-state is even under space inversion. Following rel thetreatmentdevelopedbyTaut8,ananalyticexpressionisobtainedforthe stateofzerorelativeangularmomentum (M = 0) if the magnetic field satisfies the condition γ = 1, where γ ℓ /a with ℓ = √2ℓ and a = ~2/me2. rel rel B rel B This state lies at relative energy 2~ω and apart from a normalization f≡actor is 0 φ (r) (1+x)e−x2/4 (23) rel ∝ 5 with x=r/ℓ . This is, in fact, the solution which corresponds to the lowest value of γ and to the lowest energy. rel Twomain points ofthis two-electronproblemneed emphasizing. Firstly, the effective potentialV (r) entering the eff Schr¨odinger equation for the reduced wave function f (r)=φ (r)√r at M =0 is eff rel rel 1 γ 1 V (r)= x2+ (24) eff 4 x − 4x2 and develops a minimum for γ > (16/27)1/4 (see Figure 4). This fact will motivate the introduction of the paired phase in the next Section. Secondly, in a solid semiconducting medium the parameter γ becomes ∗ ℓ m ∗ rel γ = = γ, (25) a∗ ǫm B γ being the vacuum value. Thus, using ℓ 363.5(B/Tesla)−1/2 ˚A and the values ǫ 12.4 and m∗/m = 0.067 rel ≃ ≃ for a GaAs heterostructure, we find that the minimum is present in V (r) for all values of magnetic field lower than eff B 20 Tesla. crit ≃ IV. THE PAIRED PHASE IN A MAGNETIC FIELD We return to the circular cell model to develop a variationalapproachto the total energy in the case of a localized paired phase in a spin-singlet configuration. Two electrons are placed inside each disk of radius r , the pair wave WS function being taken in the form r r r2+r2 ψκ,σ(r1,r2)=Aκ,σ 1+κ| 1−σ 2| exp − 14σ2 2 . (26) (cid:18) (cid:19) (cid:18) (cid:19) whereκandσ arevariationalparametersand isanormalizationconstant. Thesingle-paircasetreatedinSection κ,σ A III is recovered by setting σ = ℓ and κ = 1/√2. Notice that the relative motion of the two electrons is described in Eq. (26) by a Gaussian factor times a linear superposition of the Hermite polynomials of zeroth and first order: thus the Ansatz (26) allows level mixing due to the electron-electron interactions and is the natural extension of the variational-width method used for the unpaired phase in Section II. More refined wave function would include higher-order Hermite polynomials in the mixing, weighted by additional variational parameter. The main properties of the wave function (26) are as follows. Firstly, normalizationto two electrons per cell yields = √2πσ2 1+2κ√π+4κ2 1/2 −1 . (27) κ,σ A h (cid:0) (cid:1) i Secondly,themostprobablevalueoftherelativedistancebetweenthetwoelectrons,asobtainedfromtheappropriate maximum of the square of the pair wave function, is σ r r = 1+16κ2 1 . (28) 2 1 mp | − | − 2κ (cid:16)p (cid:17) This increases with κ and saturates to the value 2σ. Thirdly, the single-particle density ρ (r)= d2r′ ψ (r,r′)2 κ,σ κ,σ | | can be obtained in closed form (see Appendix), with the result R r2 r2 r2 r2 r2 ρ (r)=πσ2 2 e−3r2/4σ2 2 1+2κ2+κ2 er2/4σ2 +κ√2π (2+ )I ( )+ I ( ) . (29) κ,σ Aκ,σ σ2 σ2 0 4σ2 σ2 1 4σ2 (cid:26) (cid:18) (cid:19) (cid:20) (cid:21)(cid:27) This expression tends to the correct value ρ (r)=(πσ2)−1 exp( r2/2σ2) for κ 0. κ=0,σ − → We proceed to evaluate the total energy per electron associated with the wave function (26). We have 1 ǫ = (ǫ +ǫ +ǫ ), (30) t eb bb pair 2 where for the background-background term we can use the result in Eq. (6). The electron-background term is evaluated,asinthe calculationperformedinSectionII.A,fromtheelectricalpotentialV (r)createdbytheelectron κ,σ distribution ρ (r) in Eq. (29) according to κ,σ 6 rWS ǫ =2πn rdrV (r) eb s κ,σ Z0 ∞ 4α dx ∗ = ρ (x) J (r x/σ) Ryd (31) − r x κ,σ 1 WS s Z0 (see Appendix), where J (x) are Bessel functions. Heere n 2 1 ρ (x)= ψ e−ix·ri/σ ψ κ,σ κ,σ κ,σ 2 h | | i i=1 X e = 2 e−x2/2 1+4κ2 κ2x2 √π κex2/8 (x2 4)I (x2/8) x2I (x2/8) . (32) 1+2κ√π+4κ2 − − 2 − 0 − 1 (cid:26) (cid:27) (cid:2) (cid:3) In the limit of strong localization (σ r ), we obtain the harmonic-approximationresult ≪ WS 8α 2Σ2 2+5κ√π+12κ2 ǫHA = + Ryd∗. (33) eb − r αr3 1+2κ√π+4κ2 (cid:20) s s (cid:21) The same limiting result can be obtained from the electrical potential V (r) created by the background disk, which in is still given by Eq. (5). Finally the contribution ǫ in Eq. (30) is given by pair 1 ǫ = ψ ψ pair κ,σ κ,σ 2h |H| i 1 = 2 φ φ φ φ + φ φ φ φ , (34) 2Aκ,σ h rel| relih CM|HCM| CMi h CM| CMih rel|Hrel| reli (cid:16) (cid:17) the center-of-massandrelativemotionstatesandHamiltonians being immediately obtainedfromEqs. (22)and(26). The result is 1 2+3κ√π+8κ2 2+5κ√π+12κ2 √π1+4κ/√π+2κ2 ǫ = +2λ Σ2 + Ryd∗. (35) pair 2Σ2 1+2κ√π+4κ2 B 1+2κ√π+4κ2 Σ 1+2√πκ+4κ2 (cid:18) (cid:19) The last term in this equation arises from the electron-electroninteraction. V. TOTAL ENERGY OF THE PAIRED PHASE The totalvariationalenergyper electronin the pairedcircular-cellapproximationis obtainedfromEqs. (30),(31), ∗ ∗ (35) and (6). It depends on the three variational parameters Σ = σ/a , α = r /(r a ) and κ. The equilibrium B WS s B values of Σ and α as functions of r show the same trends as displayed in Figure 1 for the unpaired phase. However, s the asymptotic value of Σ¯ for an electron pair is reduced by a factor of √2 and that of α¯ is α¯ = π/2, increased by a factor √2 over the unpaired state as one expects from the double occupancy of the cell. Internal consistency p of the theory is ensured by the ratio (σ/r ) being smaller than about 0.5, both in the full calculation and in the WS eq harmonic approximation, for values of r >1 and of the field B > 5 Tesla. The equilibrium value of κ as a function s of r is displayed in Figure 4 for two values of the field. The two electrons in each cell are pushed closer together as s the magnetic field increases, leading from Eq. (28) to a decrease of κ¯ with increasing field as is shown in Figure 5. Figure 6 reports the energy δǫ of the paired phase (that is, after subtraction of the kinetic energy in the lowest t LL and reduction to e2/(ǫℓ) energy units) as a function of the filling factor extending up to ν = 1.5. The residual dependence ofδǫ onthe magnetic fieldextends into the high-r (low-ν)regime,as illustratedinthe insetin Figure 5 t s on a magnified scale,and a finite value is attained byδǫ as ν 0. These features are a consequence of the Coulomb t → interaction between the two electrons inside each cell. The result in Figure 6 have been fitted to the functional form δǫ =f(ν) e2/(ǫℓ), where t aν1/2+bν3/2+cν5/2+d f(ν)= . (36) eν1/2+f The values of the coefficients in this fitting formula are reported in Table I for various values of the magnetic field. 7 Figure 7 compares the energies per electron of the unpaired and paired phases as functions of r . The two curves s cross at two values of r for each value of the field, but the crossing at lower r is to be discarded as it corresponds s s to situations where charge leakage out of the circular cell is unacceptably high ((σ/r ) 1). The conclusion from WS eq ≈ the physically significant crossing at larger r , therefore, is that at each value of the magnetic field in the range from s 5 to 20 Tesla the paired phase becomes stable relative to the unpaired one as r decreases. s From the location of the physically acceptable crossing of the energies of the two phases in Figure 7 we obtain their phase boundary in the (r ,λ ) plane, which is reported in Figure 8. This shows a region of thermodynamic s B stability for the paired phase (at least relative to the unpaired one) in the left portion of the plane, before charge leakage outside the cell boundary is expected to occur in an important way so that the model becomes unreliable (to the left of the dash-dotted line in the Figure). Within numerical accuracy we find that the boundary between the two phasesis actually setby the ν =1 line in the plane27. Therefore,the presentsimple model predicts that electron clusterizationmayoccuronlyinLandaulevelsabovethelowestone. Ofcourse,theunpairedstateintheν <1region (ontherightoftheν =1line)couldbecomestableagainsttheLaughlinliquidonlyforν <0.25,asalreadydiscussed in connection with Figure 3. Weshouldemphasizethattheground-stateenergydifferencethatweestimatebetweenthepairedandtheunpaired state is quite appreciable as long as the magnetic field intensity lies in the range that we have illustrated in our calculations. The difference decreases with the field intensity, so that more refined calculations would be needed at low fields. VI. SUMMARY AND CONCLUSIONS We havepresentedavariationalstudywhichapproximatelyaccountsforthe ground-stateenergeticsofa2Dmany- electron system with Coulombinteractions at low density and in the presence of a perpendicular magnetic field. The method consists of an improved 2D version of the well-known spherical cell approximation and is applied to two distinct situations, a single-electron state and a spin-singlet paired state. For the former it gives new analytical and numerical results which are in close agreement with the state-of-the-art calculations on the 2D Wigner crystal. The evaluationofthe spin-pairedstate has beenmotivated by previousstudies andsuggeststhat this state may be stable in a range of system parameters corresponding to Landau levels above the lowest one. It remains to be seen whether such a paired state would be confirmed in more sophisticated treatments allowing in particular for inter-cell corrections. Itmaybealsoemphasizedatthispointthatwehavenotexaminedthestabilityofthespin-singletpaired state against the emergence of spin-polarized states. Evidence from measurements of Knight shift of the 71Ga nuclei in n-doped GaAs29 indicates that this quantity, which is proportionalto the spin polarization,drops precipitously on eithersideofν =1,whichisevidencethatthechargedexcitationsoftheν =1groundstatearefinite-sizeSkyrmions. ACKNOWLEDGMENTS This work was partially supported by MURST through the PRIN Initiative. One of us (M.P.) wishes to thank the Physics Department of the University of Cyprus for their hospitality during part of this work. APPENDIX. DETAILS OF ANALYTIC CALCULATIONS We report in this Appendix some details on the derivation of some analytic results given in the main text. A. Equations (10) and (11) The single-electron potential V (r) is SP +∞ ∗ V (r)= (2a /σ) dxρ (x)J (xr/σ) (37) SP − B SP 0 Z0 ∗ (in Ryd ), where e ρ (x)= d2rφ2 (r)J (xr/σ)=exp( x2/4). (38) SP SP 0 − Z e 8 Equation (10) in the main text follows by using the result +∞ exp( x2/4)J (ax)=√π exp( a2/2)I (a2/2). (39) 0 0 − − Z0 Finally, the integral in Eq. (9) is carried out with the help of the relation b b2 x exp( x2/2)I (x2/2)dx= e−b2/2 I (b2/2)+I (b2/2) . (40) 0 0 1 − 2 Z0 (cid:2) (cid:3) leading to Eq. (11) in the main text. B. Equation (29) The single-particle density ρ (r) is calculated from the expression κ,σ 2π +∞ ρ (r)=2σ2 2 exp( r2/σ2) dθ xdx(1+κx)2 exp( xrcosθ/σ x2/2). (41) κ,σ Aκ,σ − − − Z0 Z0 Eq. (29) is obtained by using the results 2π exp( xycosθ)dθ =2πI (xy) (42) 0 − Z0 and +∞ x(1+κx)2 exp( x2/2)I (xy)dx= 0 − Z0 =exp(y2/4) 1+2κ2+κ2y2 exp(y2/4)+κ π/2 (2+y2) π/2I (y2/4)+y2I (y2/4) . (43) 0 1 n(cid:0) (cid:1) p h p io C. Equation (31) The electrical potential V (r) created by the electron distribution ρ (r) in Eq. (29) is written as κ,σ κ,σ +∞ ∗ V (r)= (2a /σ) dx ρ (x)J (xr/σ) (44) κ,σ − B κ,σ 0 Z0 (in Ryd∗), where in essence ρ (x) is the Fouriertransformoftheeelectrondensity distribution (see Eq. (32)). After κ,σ insertingthisformulaintothefirstlineofEq. (31)andinterchangingtheorderofthetwointegrations,theintegration over r can be carried out with the help of the result e b b J (ax)xdx= J (ab). (45) 0 1 a Z0 This yields the second line of Eq. (31) in the main text. 1Seee.g.The Quantum Hall Effect,eds.R.E.PrangeandS.M.Girvin(Springer,NewYork1987); Perspectives inQuantum Hall Effects, eds. S. Das Sarma and A. Pinczuk (Wiley, New York1997). 2M. Shayegan, in Perspectives in Quantum Hall Effects, eds. S. Das Sarma and A. Pinczuk (Wiley, New York1997), p.343. 9 3M.P.Lilly,K.B.Cooper,J.P.Eisenstein,L.N.Pfeiffer, andK.W.West,Phys.Rev.Lett.82,394(1999); R.R.Du,D.C. Tsui, H.L. Stormer, L. N.Pfeiffer, K.W. Baldwin, and K. W. West, Solid StateCommun. 109, 389 (1999); K. B. Cooper, M. P. Lilly, J. P. Eisenstein, L. N. Pfeiffer, and K. W. West,Phys. Rev.B 60, R11285 (1999). 4A. A.Koulakov, M. M. Fogler, and B. I. Shklovskii,Phys. Rev. Lett. 76, 499 (1996); M. M. Fogler, A.A. Koulakov and B. I. Shklovskii, Phys. Rev. B 54, 1853 (1996); R. Moessner and J. T. Chalker, Phys. Rev. B 54, 5006 (1996); M. M. Fogler and A. A.Koulakov, Phys.Rev. B 55, 9326 (1997). 5F. D. M. Haldane, E. H. Rezayi, and Kun Yang,Phys. Rev.Lett. 85, 5396 (2000). 6N.Shibata and D. Yoshioka, Phys.Rev.Lett. 86, 5755 (2001). 7S. M. Girvin, The Quantum Hall Effect: Novel Excitations and Broken Symmetries, Lectures at Ecole d’Et´e Les Houches (1998). 8M.Taut,Phys.Rev.A48,3561(1993); M.Taut,J.Phys.A27,1045and4723(1994); M.Taut,J.Phys.: Condens.Matter 12, 3689 (2000). 9G. V. Shusterand A.I. Kozinskaya, Fiz. Tverd. Tela 13, 1240 (1971) [Sov. Phys.Solid State 13, 1038 (1971)]. 10K. Moulopoulos and N.W. Ashcroft, Phys. Rev.B 48, 11646 (1993). 11G. M. Summers, R. J. Warburton, J. G. Michels, R. J. Nicholas, J. J. Harris, and C. T. Foxon, Phys. Rev. Lett. 70, 2150 (1993); R.B.Dunford,E.E.Mitchell,R.G.Clark,V.A.Stadnik,F.F.Fang,R.Newbury,R.H.McKenzie,R.P.Starrett, P. J. Wang, and B. S.Meyerson, J. Phys.: Condens. Matt. 9, 1565 (1997). 12See, e.g. Appendix A in Ref. 10 for the 3D version of the cell approximation in zero magnetic field. A spherical cell is there exploited in both a point-charge and a distributed-charge approach, to deal with the WC and with an orientationally symmetric paired crystalline phase. 13I. Nagy,Phys. Rev.B 60, 4404 (1999). 14Y.E. Lozovik and V.I. Yudson,JEPT Lett. 22, 11 (1975). 15L. Bonsall and A.A. Maradudin, Phys.Rev.B 15, 1959 (1977); E. Cockayne and V. Elser, Phys.Rev.B 43, 623 (1991). 16D.YoshiokaandH.Fukuyama,J.Phys.Soc.Jpn.47,394(1979);D.YoshiokaandP.A.Lee,Phys.Rev.B27,4986(1983); K. Maki and X. Zotos, Phys. Rev.B 28, 4349 (1983). 17D. Levesque,J. J. Weis, and A. H.MacDonald, Phys. Rev.B 30, 1056 (1984). 18P. K.Lam and S.M. Girvin, Phys.Rev.B 30, 473 (1984). 19G. Vignale, Phys.Rev. B 47, 10105 (1993). 20Handbook of Mathematical Functions, eds. M. Abramowitz and I. A.Stegun (Dover, NewYork 1972). 21D. C. Tsui, H. L. Stormer and A. C. Gossard, Phys. Rev.Lett. 48, 1559 (1982); E. Y. Andrei, G. Deville, D. C. Glattli, F. I. B. Williams, E. Paris and N. Etienne, Phys. Rev. Lett. 60, 2765 (1988); R. L. Willett, H. L. Stormer, D. C. Tsui, L. N. Pfeiffer, K. W. West and K. W. Baldwin, Phys. Rev. B 38, 7881 (1988); H. W. Jiang, R. L. Willett, H. L. Stormer, D. C. Tsui,L.N.PfeifferandK.W.West,Phys.Rev.Lett.65,633(1990); H.W.Jiang, H.L.Stormer,D.C.Tsui,L.N.Pfeiffer andK.W.West,Phys.Rev.B44,8107 (1991); V.J.Goldman,M.Santos,M.ShayeganandJ.E.Cunninghan,Phys.Rev. Lett.65,2189 (1990); D.C.Glattli, G.Deville, V.Duburcq,F.I.B.Williams, E.Paris, N.EtienneandE.Y.Andrei,Surf. Sci. 229, 344 (1990); H. Buhmann, W. Joss, K. von Klitzing, I. V. Kukushkin, A. S. Plaut, G. Martinez, K. Ploog and V. Timofeev, Phys. Rev. Lett. 66, 926 (1991); F. I. B. Williams, P. A. Wright, R. G. Clark, E. Y. Andrei, G. Deville, D. C. Glattli, O. Probst, B. Etienne, C. Dorin, C. T. Foxon and J. J. Harris, Phys. Rev. Lett. 66, 3285 (1991); R. L. Willett, M. A.Paalanen, R.R.Ruel, K. W. West,L. N. Pfeiffer and D.J. Bishop, Phys.Rev.Lett. 65, 112 (1990). 22X.Zhu and S.G. Louie, Phys.Rev.Lett. 70, 335 (1993). 23R.Price, X. Zhu,S.Das Sarma, and P. M. Platzman, Phys. Rev.B 51, 2017 (1995). 24M.Seidl,J.P.Perdew,andM.Levy,Phys.Rev.A 59,51(1999); M.SeidlandJ.P.Perdew,Phys.Rev.B 50,5744(1994). 25E. H. Lieb and H.Narnhofer, J. Stat. Phys.12, 291 (1975). 26A.Turbiner, Phys. Rev.A 50, 5335 (1994). 27The lines of constant ν in the (r ,λ ) plane are described by the equation λ =1/(ν2r4), so that the line ν =1 in Figure s B B s 8 separate thelowest LL (on its right) from thehigher ones (on its left). 28S. L.Sondhi, A.Karlhede, S. A.Kivelson and E. H.Rezayi, Phys. Rev.B 47, 16419 (1993). 29S. E. Barrett, G. Dabbagh,L. N. Pfeiffer, K. W.West, and R.Tycko, Phys.Rev.Lett. 74, 5112 (1995). 10