Typical experimental setup

Now that all the relevant concepts have been introduced, I present the components of a typical experimental setup and the values of the different parameters involved.

Vacuum

The aim of atom cooling and trapping is to replace ensembles in warm vapour cell, where their collisions – with the walls or other atoms – fundamentally limit their coherence time to∼10 µs−100 µs [Horsley2013]. To prevent such collisions with a warm

background, the preparation of cold atoms is performed in vacuum, at residual ambient pressures of the order of 10−10Torr. For the detail procedure to reach this pressure level,

I refer the reader to the previous theses at LKB [Giner2013PhD, Gouraud2016PhD].

Dispensers

The atoms used to form the MOT are provided by dispensers, also known as getters[1].

Their principle relies on the heating up of a alkali-metal wire to dispense a vapour of a given element in a vacuum background.

Dispensers from SAES (SAES CS/N F/3.9/12F T 10) are used at LKB. They contain 5 mg of caesium and should last 6 months each, if used on average 10h per day. In practice during the course of my PhD, we have switched dispensers twice in 4 years.

At ANU, where the experiment uses rubidium, we are also using dispensers, from Alfavakuo, formerly known as Alvatec, with similar specifications. Rubidium having two stable isotopes, the dispensers provide a mixture of these, with their natural abundance. However as their atomic transitions differ, only one isotope is captured in the MOT, the other one only contributing to background pressure.

Magnetic trapping

A typical magneto-optical trap uses a 10-G cm−1 magnetic gradient over 10 cm, most

commonly achieved with a pair of coils in an anti-Helmholtz configuration, in two or all three directions of space, whether it is a 2D or 3D MOT. The anti-Helmholtz configura- tion ensures the cancellation of the magnetic field in the centre and a uniform gradient over the overlap region of the cooling beams, with minimal next-order contribution off the anti-symmetry axis.

[1]

The termgetter usually refers to the use of the component to react and condense,i.e. “get”, residual gas inside a sealed vacuum system.

2.3 Large cold atomic ensembles 47

Wire radius The radius of the coil wire limits the maximum current they can sustain

with no other heat dissipation than standard air flow, possibly helped by placing radiators on the coils thermally-conductive structure. Although some realisations involve the use of water cooling, a standard MOT setup usually simply relies on using wires with a large-enough radius to decrease their resistance. The compromise is made on the volume taken by a coils with the tens if not hundreds of turns required to achieve the desired gradient. Secondarily, a thicker wire is less pliable, which makes the winding more challenging: often the practicality of tightly winding the coil calls for using the thinnest wire capable of supporting the required current. In practice, typical wires have a 1 mm−2 mm radius.

Current source The typical current used during a MOT loading sequence ranges

between 5 A−10 A, with coils which have a resistance of a 1 Ω−10 Ω. The simplest way

to provide this current is with a power supply, of about 1 kW and capable of delivering the relatively high current. A commercial solution adopted on both experiments I worked on is the Delta Elektronika SM 52-30. Usually such a source cannot sink the current injected in the high inductance circuit. This task needs to rely on another component, which can in this case be a solid-state relay like the Power-IO HDD-1V50. The original setup at ANU consisted of having the relay in series with the coils, and relying on its internal resistance when switched off to dissipate the current, effectively giving a 1 ms current decay time, on the same order as the decay of the eddy currents induced in nearby conductors by the varying magnetic field. The relays, especially newer units, can wear off quickly with this operation, however, as it does not correspond to their intended use. Shorter decay times and a more sustainable solution would be to use a capacitor and resistor in parallel to the coils, in order to form a critically damped RLC circuit when switching off the relay. This solution is currently being investigated.

Previously at LKB, a current source was developed by our electronic engineers Jean- Pierre Okpisz and Brigitte Delamour to commute the coils to a discharge circuit when we want the magnetic gradient turned off. The discharge circuit was made of a few 1 mF capacitors capable of withstanding the potential difference and speeding up the decrease of the current by effectively applying an opposite voltage to the loading one, yielding decay times around 50 µs−100 µs.

Sources of light for optical trapping

A general rule with optical trapping is that the more available power, the better. Indeed it is always desirable to increase the trapping detuning to reduce the scattering on the atoms, and this can be done without compromising on the number of captured atoms by increasing the intensity of the trapping beams.

A typical experiment uses trapping beams with total power around 100 mW−500 mW,

and diameters of 25 or 50 millimetres. Laser diode light amplified by a tapered amplifier typically provides around 1 W of continuous light power. Frequency control and gating with AOMs and coupling in single-mode fibres reduces this figure to the required power range. This light is then divided in several beams to provide the six-directional optical trapping. A widespread method to roughly divide by two the amount of power needed is to retro reflect each of three beams. This however limits the flexibility of independently

fine tuning the power in each direction, and the retro-reflected beams always experience some losses due to the optics reflecting them.