Collisional losses from a light-force atom trap.

Spontaneous-force light traps' have provided a way to obtain relatively deep static traps for neutral atoms. These allow one to produce samples containing large numbers of very cold atoms. In this paper we present an experimental study of the collisions which eject atoms from such a trap. These collisions are of considerable interest because the temperatures of the trapped atoms (10 K) are far lower than in usual atomic collision experiments. The theory of such low-energy collisions and their novel features have been discussed by several authors. Perhaps the most notable feature is that the collision times are very long, and the collision dynamics are dominated by long-range interactions and spontaneous emission. These collisions also have important implications with regard to potential uses of optically trapped atoms. For many applications the maximum density that can be obtained is a critical parameter, and these collisions limit the attainable density. There have been two experimental studies of collisions in optical traps. Gould et al. measured the cross section for associative ionization of sodium. However, there is no evidence that this process is significant in limiting trapped-atom densities, and for some atoms, including cesium, it is energetically forbidden. Prentiss et al. studied the collisional losses which limited the density of sodium atoms which were held in a spontaneous-force trap. Their surprising and unexplained results were a direct stimulus for our work. In particular, they observed no dependence of the loss rate on the intensity of the trapping light. This was quite surprising, because a ground and an excited atom interact at long range via the strong 1/r resonant dipole interaction, and a portion of the excited-state energy can be converted into sufficient kinetic energy to allow the atoms to escape from the trap. By comparison, two atoms in their ground states interact only through a much weaker short-range 1/r Van der Waals attraction, and even when such collisions occur, they may not produce significant kinetic energy to cause trap loss. This implies that the dominant collisional loss mechanism would involve the excited atomic states, and thus depend on the intensity of the light which causes such excitations. In this paper we present measurements which show that, in contrast with Ref. 9, the collisional loss rate has a marked dependence on the trap laser intensity. We will present strong circumstantial evidence that the dependence at very low intensities is due to hyperfinechanging collisions between ground-state atoms. We believe that the loss rates at higher intensities are associated with collisions involving excited states, and are the type discussed by Gallagher and Pritchard. As discussed in Ref. 7, these collisions are very different from normal ground-excited-state atomic collisions in which two initially distant atoms, A and A approach, collide, and separate in a time much less than the radiative lifetime of the excited state. In contrast, for these very-low-temperature collisions the absorption and emission of radiation in the midst of the collision drastically alter the motion. In particular, if the excitation takes place when the two atoms are far apart (R) 1000 A) they will reradiate before being pulled into the small-R region where energy transfer occurs. However, if they are sufficiently close when excited, they can be pulled close enough together for substantial potential energy to be transferred into kinetic energy before decaying. The two dominant transfer processes are excited-state fine-structure changes and radiative redistribution. In the first, A changes its fine-structure state in the collision and the pair acquire a fine-structureinterval worth of kinetic energy. The second process, radiative redistribution, refers to A Areemitting a-photon which, because of the A-A* attractive potential, has substantially less energy than that of the photon which was initially absorbed. This energy difference is transferred to the subsequent kinetic energy of the ground-state atoms. The trap loss rate depends on the probability of exciting such "close" A-A pairs, and this probability is determined by the frequency and intensity of the exciting radiation. Light which is tuned to the red of the atomic resonance frequency, vo, excites pairs which are closer together (and shifted in energy) and thus is more eN'ective at causing trap loss than light which is at vo. We tested this hypothesis by examining how the loss