Persistent currents in a toroidal trap Juha Javanainen Department of Physics, University of Connecticut, Storrs, Connecticut 06269-3046 Sun Mok Paik and Sung Mi Yoo Department of Physics, Kangwon National University, Chunchon, Kangwon-do 200-701, Korea Received 14 August 1997 Using elementary microscopic methods, we theoretically study persistent flow of an alkali-metal vapor Bose-Einstein condensate around a tight toroidal trap. The angular velocity of a persistent current must be smaller than the angular frequency of the lowest condensate excitation. A supercurrent may be excited by rotating a perturbing potential that is strong enough to cut the toroidal condensate. S1050-29479807907-4 PACS numbers: 03.75.Fi, 05.30.-d, 32.80.Pj Experimental studies of Bose-Einstein condensates BEC in alkali-metal vapors are now well under way 1. While experimental data are not yet available, the potential of su- perfluidity and persistent currents in these systems has been recognized from the outset. Theoretical effort 2has focused on vortex states of a weakly interacting Bose gas bound to a harmonic trap, states in which the entire condensate rotates in accordance with a quantized value of angular momentum. Studies of excitations of such vortices 3,4have led to ar- guments that a vortex cannot be stable 4in a trap with the minimum of the confining potential at the center. Unimpeded by the no-go rule 4, a persistent current may be stable in other types of traps that can pin the vortex. We consider a rather extreme case, a trap that confines a conden- sate to a torus. With the additional assumption that trans- verse confinement is tight, the motion of the condensate along the torus is amenable to a simple microscopic treat- ment. We point out that superfluid flow may be stable as long as the angular velocity at which the condensate circu- lates around the torus is lower than the angular frequencies of the elementary excitations of the condensate. Creating a persistent current also proves to be a nontrivial task. We develop an approach whereby the condensate is stirred up by a rotating potential strong enough to cut the torus. Specifically, we take the radius of the condensate torus R to be much larger than the transverse dimensions of a cut across the condensate. Two simplifications ensue from such an assumption. First, the frequencies of the excitations in- volving the transverse coordinates, call them y and z , tend to be much higher than the excitation frequencies of the motion in the direction along the ring coordinate x . In what follows we assume that the transverse motion is frozen to a wave function ( y , z ). Second, as has been anticipated in our no- tation already, in our mathematics we straighten the torus and treat the motion along the ring as linear translation. As a vestige of the original topology we impose periodic bound- ary conditions over the circumference of the torus 2 R . For atom-atom interactions we adopt the conventional function pair potential characterized by the s wave scattering length a . As it comes to the interactions, the only relevant parameter of the transverse motion is the length scale l de- fined by l -2 =dydz | ( y , z ) | 4 . We embody atom-atom in- teractions into the dimensionless parameter =2 aR / l 2 . It will frequently prove convenient to discuss the fluid in a rotating coordinate system in which a stationary condensate would rotate like a wheel at the angular velocity -. We employ the dimensionless parameter =mR 2 / for the an- gular velocity. Without restricting the generality, in the fol- lowing we assume that 0. Finally, we use R as the unit of length, the atomic mass m as the unit of mass, and 2 /( mR 2 ) as the unit of energy. All told, the atoms move in the interval x -, with periodic boundary conditions. In the basis of the plane waves u k ( x ) =1/2 e ikx with k =0,1, . . . , the second- quantized many-body Hamiltonian in the rotating frame reads H = k k 2 2 -k b k b k + p , q V ˜ p -q b p b q + 1 2 k , p , q b k +q b p -q b p b k . 1 We allow for a potential V ( x ) in the direction of the torus, and V ˜ ( k ) =(1/2) - dxe -ikx V ( x ) are the Fourier coeffi- cients of the potential. The corresponding Gross-Pitaevskii equation 5,6GPEfor a system of N atoms is - 1 2 2 x 2 +i x +V +2 N | | 2 = . 2 In the absence of the potential V , the plane waves u k are still the eigenstates of the GPE, though the energy chemical po- tentialdepends on both rotation and atom-atom interactions: k =k 2 /2-k +N . 3 It is a peculiarity of the transformation to the rotating frame that atoms in the rotating-frame eigenstate u k still have the velocity v =k with respect to the stationary frame. Let us momentarily ignore both atom-atom interactions and the potential V ( x ), and work in the nonrotating labora- tory frame with =0. A state with all the N atoms in any one-particle state u k is evidently an eigenstate of the Hamil- tonian 1. By the translational symmetry, such a state should be a good first approximation to an eigenstate of the Hamil- tonian 1even in the presence of atom-atom interactions. PHYSICAL REVIEW A JULY 1998 VOLUME 58, NUMBER 1 PRA 58 1050-2947/98/581/5804/$15.00 580 © 1998 The American Physical Society