Intro
As already seen in the section explaining how to Bose-Einstein condense,
a certain number of physical paramaters have to be controlled and tuned during the experiment. Here
we'll briefly explore the requirements that have to be fulfilled and the practical and technological
solutions adopted.
Vacuum
The basic condition to work with cold atom clouds is the disponibility of a good vacuum. At normal atmospheric
pressure a lot of gas molecules moves around at a mean velocity of some hundred meters per second; every time
one of such molecules hits our cold cloud of rubidium atoms, it warms it up, expelling also a bunch of atoms from the magnetic trap.
To protect the cloud and make possible its same existence, we have to evacuate a suitable chamber to a sufficiently low pressure.
Considering the energy of gas molecules at room temperature and collision statistics it can be shown that to allow the existence of
the condensate for about 20 seconds (more or less the duration of the whole experimental cycle) we need a pressure of less than
10-11 mBar, or 1/100.000.000.000.000 the value of atmospheric pressure.
To reach such a low value, a stepwise evacuation is performed, with different types of pumps; also, all equipment needs to
be properly cleaned: a single fingerprint inside the chamber can prejudicate the vacuum conditions, outgassing
water molecules, oily coumpounds and other molecules.As first step we clean everything with different solvants, using where posible ultrasonic cleaning to remove dirth deeply.
Then we start to evacuate; an oil-free prepump (lower in the image) allows to reach a pressure of 0.001mBar, in some minutes; we can then switch on the turbo pump (above), directly connected to the chamber; a special turbine rotates at 1500rpm, dropping the pressure to 10-8mBar.
At such low densities the same chamber starts to perceivebly release gas and water molecules it has adsorbed at its surface. This outgassing process can go on for months, limiting the vacuum quality. To overcome this problem, we speed up the process. We first apply heating elements to the chamber, homogeneously spaced to prevent localized strains, then we isolate the whole chamber with some aluminium foil, and finally
we slowly increase the temperature, up to 250C, for some days. At this point the valves to the pumps are closed, the
pumps switched off and the temperature decreased. Back to room temperatue we have gained 2 to 3 orders of magnitude
in the pressure value.To keep this ultrahigh vacuum (UHV) two more pumps are then used.
The first is the ionic pump: it ionizes the gas atoms and molecules still present in the chamber and traps them thanks to high electric fields; this pump is always on.
The second and last one is the titanium sublimation pump: a current heats up and sublimates some Ti, that sticks to the gas and then to the chamber walls, a sort of gas-glue. This one is used once a week.
Once reached this UHV condition, it can be maintained without much effort for years.
The LASERs
To cool atoms we use light; to make measurements we take snapshots of the atoms, usually shooting a light beam
in direction of a camera and observing the shadow they produce (absorbing light out of the beam; a technique called absorption
imaging); during the experiment we also may want to change the quantum state of our cold rubidium atoms
(for example the spin) or to modulate the potential they feel with light.For all this purposes, normal light is not suitable; laser light is our tool. Differently to, for example, all the light sources we can find at home, laser light is light emitted coherently by all the active material in the laser; that means that all the active material in the lasing medium emits light of the same color (or frequency, or wavelength) and phase (the 'position' in a periodic wave, in respect to its minimum and maximum). Another important characteristic is its collimation; a laser beam can travel for kilometers without increasing significantly its diameter, and can be focused in spots of the the order of the µm.
Particularly important is the precision in the color of the laser. We are working with Rubidium in the near infrared (wavelength λ=780 nm, frequency ν=384 THz); at the transition used, the linewidth is about 30 MHz, that means that the atoms can absorb light if it has the frequency of 384·1012 ±30·306 Hz; but we want to specify the laser frequency with a precision of about 1/10 of linewidth; we clearly see that we need an absolute precision of about one part in 100 millions (equivalent to 27 bits); special techniques are adopted for this purpose; in short we let part of the laser beam pass through a cell containing Rubidium gas
and obtain an electric feedback signal observing with a photodiode its absorption (Doppler-free spectroscopy): the
absorption line of the rubidium in the cell is used as a natural reference point; to change the laser wavelength
we can act on the current passing in the laser diode (similar to those used in CD players) or on the geometric
dimensions of the laser cavity (piezoelectric elements can change it of some micrometers, scanning frequency ranges
of some GHz). Other lasers can then be locked to a different but relatively close frequency; to do that we superpose
part of the two lasers and observe with a photodiode the beating frequency (up to values of some GHz); this signal is
then fed back to the control electronics.The laser light is then possibly processed in other optic elements that can further change its intensity and frequency (acusto-optic modulators, AOM, in the small figure beside; shutters; ...), its polarization state (λ/2, λ/4 plates, polarizing beam splitters, polarizators, electro-optic modulators EOM...), or the beam shape and beam number (telescopes, beamsplitters) and finally delivered to the physical chamber.
Part of the optics modifying the laser properties.
|
The light power of all lasers lies below 0.5W, and for those realized inhouse it's around 50mW.
The fields
To realize the magnetic fields required by the experiments we have different power supplies; to minimize the
noise in the coils and wires (heating up the BEC and limiting its lifetime and/or the precision of the measurements) they
drive the output defining its voltage (not the current). As time passes, the currents flowing (up to 60A) heat up coils and wires, increasing
their resistances; this in turn decreases the currents (remember: V at the output) and thus the magnetic fields.To limit such drifts active feedback stabilization (see more on this in the downloadable documentation) has been added to the most critical lines. After about 30' of turning on the experiment, currents are stabilized to within 1mA (on a full scale value of 60A); noise level has also been reduced in the most critical frequency range (below about 2kHz) thanks to the use of special high dynamic range current detection and careful circuit design.
The BEC trap minimum depends on the value of the residual magnetic field along the trap direction, and the endpoint of the RF evaporative cooling has to be defined in respect to this field; it's evident that to avoid drifts and unwanted oscillations in the BEC atom number the fields have to be stable. It's then natural to define the bottom magnetic field (Ioffe field) in units of the controlled RF frequency; during a whole day of experiments we observe a variation of the trap bottom on the order of 5kHz, while shot to shot variations lie below the 100Hz limit, corresponding to a current variation not measurable directly (BECs are very good tools for precision measurements))
The chamber
In the Vacuum section I wrote about a chamber, but we still don't know how does it look like; actually it's not a single
chamber, but it's a double chamber setup.
In the lower part, about 20cm in diameter, we have a first magneto-optical trap (MOT), where we first collect Rb atoms (inside
the red circle in the figure below). The pressure here is not so low as said before, because we deliberately
generate a 'thick' Rb cloud to have enough atoms; we do this by heating up a piece of metal with rubidium salts in it and
causing it to evaporate. Atoms are cooled and captured thanks to 3 retroreflected laser beams (passing through
glass windows) and to magnetic fields generated by small coils.The cold 87Rb atoms are then sent with a thin laser beam (yellow arrow) to the upper part of the chamber (inside the green circle). Here they are trapped and collected by other beams and magnetic fields (copper coils are partially visible).
In the upper part of the chamber we have wider windows, to allow a better optical access for imaging; then there's the 'atom chip', so called not because it has structures the size of atoms, but because it is used to address and manipulate the BEC atoms. We'll explain in next section what this atom chip is; here it's important to understand that it is mounted on a ceramic-copper structure, that provides mechanical stability and electrical contacts to the external world. The copper structures underneath the chip, shaped in form of 'Z' or 'U' (the 'H' structure) will be used as conductors to generate locally variable magnetic fields; typical currents flowing are in the range 1-60 A. The copper structures and the special ceramic (high thermal conductivity and UHV compliant) are also vital to remove from the chip area the heat produced by the high current densities; all the mechanical parts are produced inhouse with great accuracy, to maximize thermal contact.
Inside the chamber the chip holder is mounted upside-down; this allows to observe the evolution of the condensate in free-fall situation.
Picture above, left shows the chip holder; the chip (about 35x25 mm) will be glued on top of it and bonded to the side pins; on the right a schematic view of the main laser beams in reference to the chip and to the atoms position. Two horizontal counterpropagating beams cool and trap the atoms along that direction; two others impinge on the chip at 45 and are reflected at its surface, acting effectively as 2+2 counterpropagating beams and cooling along the two other spatial directions. The atoms, trapped in a cigar-like trap along the central section of the 'Z' wire structure are observed with an extra imaging beam (in light blue in the picture) hitting a low noise, high quantum efficiency CCD camera (the shadow of the atoms is seen).
The chip
Our chips are used both as a mirror to reflect trapping laser beams and as
support for gold wires defined lithographically; the first characteristic is achieved covering
almost the whole chip surface (after proper passivation) with a 250nm thick gold layer. The second
is a bit more critical, because of the current densities necessary to trap atoms.
Typical trapping currents range from 10mA to about 1A; to have tight traps, it's also important
to have wires with a small transversal extension (in an ideally infinitely thin wire the magnetic
field gradient grows going at small distances r from the wire as 1/r). Special lithographic
processes have and are being developed to achieve clean wires, with a height up to 3µm
and widths as small as 5µm; it's also vital to have the better definition possible in the
realization of the wires'borders, as also a tiny deviation or irregularity (as small as 100nm)
can create local unwanted inhomogeneities in the trapping magnetic field. Typical current densities
of 105A/mm2 require a tuned and fully characterized thermal conductivity,
to remove the ohmic heating from the wires region and avoid destructive melting.
Bonding is also very important; depending on the current flowing, up to 20 bonds are realized between chip and connecting pins.Below you can see how the chip looks like after glueing and bonding.
After production, the chip can be further processed; using a technique called FIB milling (Focused Ion Beam), structures much smaller than the lithograpic resolution can be defined; in the images below you can see two examples: a sharp tip (in red) to create localized electric fields allowing spin-independent addressing of atoms, and two small (200X500nm, 2.7µm high) notches for local controlled modulations of the magnetic field (obtained through current flow shaping). Finite size elements simulations have to be performed for study and optimization of the shapes.
Experimental control
To control the experiment we need to address a certain number of currents/fields and laser
intensities/frequencies; their values have to be tunable in time with a certain speed and precision.
We have then switches/shutters and triggers whose value is a simple boolean (on/off).
To achieve this we have a specific computer with I/O cards poviding 16 analog outputs at 16 bits resolution,
64 digital outputs/inputs and 8 analog inputs to read back and record important experimental
data. Every channel value can be synchronously updated every 12.5µs, thanks to highly performance-optimized
real-time attentive coding. Inhouse developed
electronics (USB-programmable) allow to update special analog and digital channels with frequencies up
to 8MHz. A second computer
hosts the program running the GUI and preprocessing the values for the experiment cycle; the analog
channel value, for example, can be defined in its respective interface as a sequence of ramps
and splines; variables and calculated variables (dinamically changing their value in dependence
of other parameters/variables) can be used either as time durations or amplitudes for the channels, to
allow for more flexibility. A 'film' mode allows to run sequences of experiments changing automatically and
in a freely definable way variables' values.
 Click on the image to view a full size pict of the interface. |
Synchronization signals and files are generated on other PCs too, for example to pass important parameters to the programs acquiring and analyzing camera images. The programs controlling the experiment are continuously being developed, to follow the requests coming from the experiments to be performed; care is taken to have consistent versions of the programs running in the different experiments in the group, to optimize administration and debugging. The actual linecount of the main program amounts to about 15.000 lines of code.
The whole thing
The whole experiment is contained in a temperature-stabilized, darkened laboratory. The lasers and the
chamber are mounted on an optical table, isolated from floor vibrations with pneumatic
suspensions. In the figure below most of it can be seen.
-
A The dedicated PC controlling the experiment.
B the light-, temperature- and air flows- isolated box containing all the lasers and the spectroscopy elements to check their wavelengths.
C The electronics controlling the lasers.
D The section where all optics is mounted, with protection against dust and air flows.
E The main power supplies.
F The physical chamber, on the left of the image, in a darkened box.
G The air-cushion floating table.
H The steel frame where all electronics and mechanical stuff (shutters) is hung; any possible vibration (even the tiny one of the fans) has to be decoupled from the optical table
Needless to say, powerful and noise-generating elements have to be electrically separated, to avoid spikes and interferences on more sensitive devices; different phases are used for the vacuum pumps, the power supplies, the lasers and the PCs.
An experimental cycle lasts about 20 seconds; above you can see an image of the atoms (about 108) trapped in the magneto-optical trap (MOT); a mirrored image of the cloud is visible in the chip: for low view angles the cloud-to-cloud distance is twice the distance between cloud and chip. BEC production is very reliable, and atoms number oscillate a maximum of about 10% between different shots; depending on the experiment performed, more or less complex atom distributions can be observed; below, two 'nice' outcomes (pixel size:3x3µm).
