Representing unresolved moist convection in coarse‐scale climate models remains one of the main bottlenecks of current climate simulations. Many of the biases present with parameterized convection are strongly reduced when convection is explicitly resolved (i.e., in cloud resolving models at high spatial resolution approximately a kilometer or so). We here present a novel approach to convective parameterization based on machine learning, using an aquaplanet with prescribed sea surface temperatures as a proof of concept. A deep neural network is trained with a superparameterized version of a climate model in which convection is resolved by thousands of embedded 2‐D cloud resolving models. The machine learning representation of convection, which we call the Cloud Brain (CBRAIN), can skillfully predict many of the convective heating, moistening, and radiative features of superparameterization that are most important to climate simulation, although an unintended side effect is to reduce some of the superparameterization's inherent variance. Since as few as three months' high‐frequency global training data prove sufficient to provide this skill, the approach presented here opens up a new possibility for a future class of convection parameterizations in climate models that are built “top‐down,” that is, by learning salient features of convection from unusually explicit simulations.

PACS 37.10.Vz -Mechanical effects of light on atoms, molecules, and ions PACS 37.10.Gh -Atom traps and guides PACS 37.90.+j -Other topics in mechanical control of atoms, molecules, and ions Abstract -We report the transport of ultracold atoms with optical tweezers in the non-adiabatic regime, i.e. on a time scale on the order of the oscillation period. We have found a set of discrete transport durations for which the transport is not accompanied by any excitation of the centre of mass of the cloud after the transport. We show that the residual amplitude of oscillation of the dipole mode is given by the Fourier transform of the velocity profile imposed to the trap for the transport. This formalism leads to a simple interpretation of our data and simple methods for optimizing trapped particles displacement in the non-adiabatic regime.

We report on a far above saturation absorption imaging technique to investigate the characteristics of dense packets of ultracold atoms. The transparency of the cloud is controlled by the incident light intensity as a result of the non-linear response of the atoms to the probe beam. We detail our experimental procedure to calibrate the imaging system for reliable quantitative measurements, and demonstrate the use of this technique to extract the profile and its spatial extent of an optically thick atomic cloud.

Abstract. -We report the achievement of an optically guided and quasi-monomode atom laser, in all spin projection states (mF = −1, 0 and +1) of F = 1 in rubidium 87. The atom laser source is a Bose-Einstein condensate (BEC) in a crossed dipole trap, purified to any one spin projection state by a spin-distillation process applied during the evaporation to BEC. The atom laser is outcoupled by an inhomogenous magnetic field, applied along the waveguide axis. The mean excitation number in the transverse modes is n = 0.65 ± 0.05 for mF = 0 and n = 0.8 ± 0.3 for the low field seeker mF = −1. Using a simple thermodynamical model, we infer from our data the population in each excited mode.When atoms are coherently extracted from a BoseEinstein condensate (BEC) they form an atom laser, a coherent matter wave in which many atoms occupy a single quantum mode. Atom lasers are orders of magnitude brighter than thermal atom beams, and are first and second order coherent [1,2]. They are of fundamental interest, for example, for studies of atom-light entanglement, quantum correlations of massive particles [3] and quantum transport phenomena [4][5][6][7][8][9][10]. They are of practical interest for matter-wave holography through the engineering of their phase [11], and for atom interferometry because of their sensitivity to inertial fields [12].Many prospects for atom lasers depend upon a high degree of control over the internal and external degrees of freedom and over the flux. The control of the output flux in a pulsed or continuous manner has been investigated using different outcoupling schemes: short and intense radiofrequency pulses [13], gravity induced tunneling [14], optical Raman pulses [15], long and weak radiofrequency fields [16], and by decreasing the trap depth [17].The control of their internal state is intimately related to the outcoupling strategy. Atoms are either outcoupled in the magnetically insensitive (to first order) Zeeman state m F = 0 or another Zeeman state, each offering different advantages. Atom lasers in m F = 0 are ideal for precision measurement [18] because of their low magnetic sensitivity. Atoms in other Zeeman states, however, are ideal for measurements of magnetic fields because of their high magnetic sensitivity [19].The control of the external degrees of freedom has been investigated through the atom laser beam divergence while propagating downwards due to gravity [20,21]. Inhomogeneous magnetic field have been used to realize atom optical elements [22]. Recently, a guided and quasi-continuous atom laser from a magnetically trapped BEC has been reported [23].In this Letter, we report on a new approach to generate guided atom laser. This method can produce an atom laser in any Zeeman state. In addition, our nonstate changing outcoupling scheme leads to an intrinsically good transverse mode-matching, that enables the production of a quasi-monomode guided atom laser. Therefore, we achieve simultaneously a high degree of control of the internal and external degrees of freedom.The atom laser...