This paper examines two collisional mechanisms by which cold alkali-metal atoms can escape from a neutral-atom trap, due to collison of excited- and ground-state atoms. One is fine-structure-changing collisions, by which the atoms are heated by the amount of the fine-structure splitting. The other produces hot ground-state atoms following emission of a red-shifted photon during the course of a collision. A rate expression is obtained that applies to both normal and ultracold temperatures (1 mK). This expression assumes a canonical distribution of initial states, low-intensity excitation, and a semiclassical treatment of the survival probability relative to excited-state decay during the long time of the ultracold collision, and uses fine-structure-changing probabilities found by quantum scattering calculations. The known properties of the attractive molecular states of the alkaline-metal dimers are used to identify and calculate the probabilities for the specific mechanisms. We conclude that the earlier semiclassical analysis by Dashevskaya [Opt. Spectrosk. 46, 423 (1979)] of the fine-structure-changing mechanisms for the various alkali-metal species is qualitatively correct. We present quantum-mechanical calculations of the rate coefficient for fine-structure-changing collisions between ground- and excited-state Cs atoms from 1000 K to 100 \ensuremath{\mu}K. Collision-rate coefficients for the trap-loss processes are calculated for pairs of Li, Na, K, Rb, and Cs atoms at low temperature. There is a wide variation of predicted loss rate among alkali-metal species from the fastest for K and the slowest for Li. Retardation corrections to molecular lifetimes must be taken into account to predict the correct rate coefficients for Na, K, and Rb. Good agreement is obtained with the observed trap loss rate in a Cs trap.
Grazing collisions with background gas are the major cause of trap loss and trap heating in atom traps. To first order, these effects do not depend on the trap density. In collisionally opaque trapped atom clouds, however, scattered atoms with an energy E larger than the effective trap depth
Eeff, which are destined to escape from the atom cloud, will have a finite probability for a secondary collision. This results in a contribution to the heating rate that depends on the column density
〈nl〉 of the trapped atoms, i.e., the product of density and characteristic size of the trap. For alkali-metal atom traps, secondary collisions are quite important due to the strong long-range interaction with like atoms. We derive a simple analytical expression for the secondary heating rate, showing a dependency proportional to
〈nl〉Eeff1/2. When extrapolating to a vanishing column density, only primary collisions with the background gas will contribute to the heating rate. This contribution is rather small, due to the weak long-range interaction of the usual background gas species in an ultrahigh-vacuum system---He, Ne, or Ar---with the trapped alkali-metal atoms. We conclude that the transition between trap-loss collisions and heating collisions is determined by a cutoff energy $200 \ensuremath{\mu}\mathrm{K}<~{\mathcal{E}}_{\mathrm{eff}}<~400 \ensuremath{\mu}\mathrm{K},$ much smaller than the actual trap depth
E in most magnetic traps. Atoms with an energy
Eeff<E<E escape into the Oort cloud: a mechanism of effective traploss in the microkelvin range of trap temperatures. We present results of secondary heating rates for the alkali-metal atoms Li through Cs as a function of the effective trap depth, the column density of the trap, and the species in the background gas. The predictions of our model are in good agreement with the experimental data of Myatt for heating rates in high-density
87Rb-atom magnetic traps at JILA, including the effect of the rf shield and the composition of the background gas. It is shown that collisions with atoms from the Oort cloud also contribute to the heating rate. For
85Rb the calculated heating rate is below the experimentally observed value at JILA, supporting the idea that inelastic collisions in the trap are the major source of heating.
Measurements of the trap loss rate in ultra-cold magneto-optically trapped
sp6Li and
sp7Li are presented and compared. Clear evidence is presented for two different trap loss mechanisms involving inelastic collisions between one ground state atom and one excited state atom. The fine-structure changing (FSC) mechanism, in which the excited state atom changes fine-structure level during a collision, is seen to dominate the rate at low trap laser intensities. However, as the intensity is increased, the trap becomes sufficiently deep to be able to contain the products of FSC collisions and this mechanism no longer contributes to trap loss. It is a unique aspect of Li that the energy imparted to an atom in a FSC collision is in a range that may be recaptured with experimentally attainable trap parameters. When the products of all FSC collisions remain trapped, only the radiative escape (RE) trap loss mechanism, due to the emission of a less energetic photon than the initial excitation photon, contributes to the trap loss rate. At small detunings, the rate of the RE trap loss mechanism is seen to be over two orders-of-magnitude smaller than the FSC rate. The FSC trap loss rate for
sp6Li and
sp7Li were found to be largest at smaller detunings and of comparable magnitude in
sp6Li and
sp7Li. The RE trap loss rate in
sp6Li was seen to be roughly four times the RE trap loss rate in
sp7Li. To understand better the dynamics of trap loss and magneto-optical traps in general, a sophisticated model of magneto-optical trap kinetics has been developed. This model has demonstrated that most critical factor determining the maximum velocity an atom may have and yet remained trapped is the initial atom's frame detuning,
Delta - k v, where
Delta is the trap laser detuning, k is the propagation vector of the trap beam most nearly anti-parallel to v, the atom's velocity. All else being equal, minimizing the magnitude of this quantity maximizes the velocity that a trap will retain. More detailed results of this model are also presented.
We develop an optical-Bloch-equation approach for modeling ultracold-atom collisions in optical traps. The method incorporates a molecular picture of the atomic collision, laser-field dressing of the molecular states participating in the dynamics, and decay of the population and polarization due to spontaneous emission of excited states. The last is important to incorporate because the duration of the cold collisions is longer than the excited-state lifetimes. The relative motion of the atoms during the course of the collision is treated semiclassically with corrections for the time-dependent relative motion of the atoms in the various channels. An application of the method to Cs trap loss due to fine-structure-changing collisions is presented. Good agreement with experiment is obtained.
We show through close-coupled quantum-scattering calculations that the cross section for fine-structure-changing collisions between excited
2P3/2 and ground
2S1/2 A atoms of the same species, where A=Rb or Cs, is very sensitive to the molecular potentials and spin-orbit matrix elements of the alkali-metal-dimer species. Spectroscopic studies of these species are needed to extract accurate parameters. New high-resolution photoassociation spectroscopy of trapped atoms could be used for this purpose.
Universal collision rate constants are calculated for ultracold collisions of two like bosonic or fermionic heteronuclear alkali-metal dimers involving the species Li, Na, K, Rb, or Cs. Universal collisions are those for which the short range probability of a reactive or quenching collision is unity such that a collision removes a pair of molecules from the sample. In this case, the collision rates are determined by universal quantum dynamics at very long range compared to the chemical bond length. We calculate the universal rate constants for reaction of the reactive dimers in their ground vibrational state v = 0 and for vibrational quenching of non-reactive dimers with v ≥ 1. Using the known dipole moments and estimated van der Waals coefficients of each species, we calculate electric field dependent loss rate constants for collisions of molecules tightly confined to quasi-two-dimensional geometry by a one-dimensional optical lattice. A simple scaling relation of the quasi-two-dimensional loss rate constants with dipole strength, trap frequency and collision energy is given for like bosons or like fermions. It should be possible to stabilize ultracold dimers of any of these species against destructive collisions by confining them in a lattice and orienting them with an electric field of less than 20 kV cm(-1).
We calculate the light-induced collisional loss of laser-cooled and trapped magnesium atoms for detunings up to 50 atomic linewidths to the red of the ${}^{1}{S}_{0}{\ensuremath{-}}^{1}{P}_{1}$ cooling transition. We evaluate loss rate coefficients due to both radiative and nonradiative state-changing mechanisms for temperatures at and below the Doppler-cooling temperature. We solve the Schr\"odinger equation with a complex potential to represent spontaneous decay, but also give analytic models for various limits. Vibrational structure due to molecular photoassociation is present in the trap loss spectrum. Relatively broad structure due to absorption to the ${\mathrm{Mg}}_{2}{ }^{1}{\ensuremath{\Sigma}}_{u}$ state occurs for detunings larger than about ten atomic linewidths. A much sharper structure, especially evident at low temperature, occurs even at smaller detunings due to ${\mathrm{Mg}}_{2}{ }^{1}{\ensuremath{\Pi}}_{g}$ absorption, which is weakly allowed due to relativistic retardation corrections to the forbidden dipole transition strength. We also perform model studies for the other alkaline-earth-metal species Ca, Sr, and Ba, and for Yb, and we find similar qualitative behavior as for Mg.