Using lasers to "cool" atoms is a subject that has interested researchers for some time.  Laser cooling is the process of using the momentum of photons to lower the kinetic energy of an atom.  Since kinetic energy and temperature are directly related, decreasing the kinetic energy of a sample of atoms decreases its temperature.

Laser cooling is accomplished by taking advantage of hyperfine atomic transitions and the absorption that causes these transitions.  Since each transition takes a certain amount of energy, this corresponds to a certain frequency of light.  This frequency is called the "resonant frequency."  After the atom absorbs this photon and becomes excited, it will soon decay from this excited state by emitting another photon.  In the case of laser cooling, where the light source is collimated in a single direction, all the absorbed photons will have momentum in the same direction.  However, since the direction of the
emitted photon is completely random, the emitted momentum will sum to zero.  Thus, there is a net change in momentum of the atom.

To actually cool a sample of atoms, however, it is necessary to detune the laser from the resonant frequency.  When the laser is detuned to the red, atoms with a velocity component moving towards the laser source will see the laser Doppler shifted towards resonance.  These atoms will therefore have a much higher probability of absorbing photons.  By reflecting the laser back through the sample, it is possible to cool a sample of atoms in that dimension.

In this experiment, the material being cooled is gallium-69, a naturally occurring isotope of gallium.  To take advantage of the strongest atomic transition, it was necessary to use light of a wavelength of 295 nanometers (nm), which is in the ultra-violet spectrum.  The gallium was only cooled in one dimension transverse to the motion of the atom beam.

To image the gallium beam, a small portion of the original cooling laser is siphoned off to become a "probe beam."  This probe beam is then sent through the atom beam "downstream" from the cooling region.  Although it is much less
intense than the cooling laser, it will still cause laser-induced fluorescence in the atom beam. This fluorescence can be captured by a CCD camera and used to interpret the overall success of the cooling process. However, capturing this
fluorescence is not a trivial task for a number of reasons: 1) it is very dim; 2) it is still UV light; 3) this fluorescence takes place in a vacuum chamber.

In order to actually "see" the fluorescence, a liquid nitrogen cooled CCD camera with computer controller was used.  However, due to the nature of the liquid nitrogen dewar, the CCD chip could not be tilted from its vertical position.
Due to the nature to the interaction between the laser and atomic beam, it was necessary to image from above the probe region.  Therefore it was necessary to design and build an optical system to get the image to the camera.

The optical system to transfer the image of the laser-induced fluorescence to the camera had to consist of a lens to gather and focus the light and a mirror to turn the image 90 degrees to send to the camera.  This system had to be constructed into and on the pre-existing structure of the vacuum system.  The window into the vacuum chamber is located at the bottom of a well approximately three inches deep with a three inch diameter.  The actual interaction of the probe beam and the gallium takes place a further three inches below the window in the vacuum chamber.

In designing the optical system, it was necessary to balance the need to gather as much of the fluorescence as possible (since there was not a great deal of light to begin with) with the constraints of cost and availability.  Availability was a concern as there is not a great demand for optics optimized for light near 295 nm and lead times for specialized optics range from four to eight weeks.  Also, magnification of the image was important as the actual size of the laser-induced fluorescence was ~7 mm by ~1 mm.

Once a design had be decided on, it was also necessary to estimate the amount of light reach the CCD itself.  Using data from a photomultiplier and information about the CCD chip itself, it was possible to calculate the number of photons reaching each pixel.  The final design was composed of a 34.1 mm diameter plano-convex fused-silica lens with a 75 mm. focal length and a flat round mirror with a 50.8 mm diameter (see Figure 1.1).
 
 

To assemble this system, it was necessary to design specialized parts and then assemble them in the machine shop.  A standard lens mount and mirror mount (along with 45-degree adapter) were purchased and mounted into the parts made
in the machine shop.  The lens mount itself was mounted into a ring that was lowered into the well.  The thickness of the ring put the lens 100 mm from the fluorescence, forming an image (magnified 3 times) 300 mm from the lens.  A housing was built atop a special platform (machined to fit on the top of the well).  Inside this housing the mirror mount and 45-degree angle adapter were mounted upside down.  From there the light was to travel to the CCD chip.  Between the lens and mirror housing and between the mirror housing and CCD shutter, two aluminum pipes were used to make the system more "light-tight."  To minimize noise, all of the pieces were spray-painted black with an ultra-flat black paint.  A full schematic of the optical system, with all parts identified, is shown below in Figure 1.2 and a photograph of the light tight optical system and CCD camera is shown in Figure 1.3.

Once the system was assembled, fast-drying epoxy was applied to several points at the base of the housing, gluing it to the top of the well.  However, besides that glue, the system was fully self-contained and could be assembled and
disassembled in about a minute.  After the initial alignment, it was possible to remove the system, insert other instruments, then reassemble the system to full working order without being forced to realign the mirror.  Finally, due to the fact that the fluorescence was very dim, it was necessary to wrap the entire system in black cloth to minimize noise.  The system is shown with the back cover removed, exposing the interior components, in Figures 1.4 and 1.5.

                               

The main limiting factor on the use of the imaging system was the CCD camera used to capture the images.  The camera itself was on loan from another research group for only one week, so there was an absolute deadline on the length of time over which images could be collected.  Once the optical system was in place, it was aligned at first by eye (a business card was inserted into the chamber and the probe laser was sent through it, causing visible fluorescence) before the camera was put into place.  Once the camera was in place and the image located (the CCD chip is a mere 2.6 cm by 0.9 cm), the
camera position was varied to in order to place the CCD chip itself as close to the focal plane of the lens as possible.  Once this was accomplished, the business card was removed and the vacuum system was pumped down to ~1x10^(-6) torr.

To take data with the camera, it was necessary to use a computer controller to set shutter times and display images.  The computer would display these pictures in delayed real time.  After a one second exposure there would be a short pause while the image was transferred to the CCD controller, another exposure would take place, and as the second exposure was being transferred to the CCD controller the first exposure would be displayed on the computer screen.  In this manner it was possible to make fine adjustments to the optical system alignment while watching the outcome of these changes on the computer.  The experimental setup with computer control is shown in Figure 1.6.

To collect the most accurate data possible, as variables were changed one picture of the normal gallium beam was always taken either directly before or after a picture of a modified gallium beam was taken.  This was done to allow the laser as little time as possible to drift from the frequency being tested (and thus changing the properties of the probe beam fluorescence).  The issue of removing the laser frequency drift is addressed below in Section II. Frequency Stabilization of Coherent 699 Dye Laser  Also, to be able to make accurate comparisons between two different images, it is necessary to make sure the auto-scaled false-color images are on the same scale.  Data was taken for a variety of different circumstances (laser detuned either blue or red by various amounts, retro-reflection on or off, etc.)

Below are false-colored images of the laser-induced fluorescence taken by the CCD camera.

---

---

Image 1.7: atom beam cooled in the transverse direction

Image 1.8: upstream laser blocked - atom beam uncooled in the transverse direction

---

Image 1.9: atom beam heated in the transverse direction

Image 1.10: upstream laser blocked - atom beam unheated in the transverse direction

---

 

II. Frequency Stabilization of Coherent 699 Dye Laser

As a separate project, it was necessary to develop a frequency stabilization mechanism for the Coherent 699 dye laser which is used to cool the gallium.  The frequency of this laser must be kept steady over an extended period of time (30 minutes or more) in order to cool the gallium evenly.  This project was an extension of an REU by Kristin Bjornsen from the University of Colorado completed here in the summer of 1999.  Kristin designed a built a frequency locking scheme for the Coherent 599 dye laser and was successful in locking the laser to within 3% of the original, unlocked, drift.  The first step was to reproduce the results of Kristin's 1999 experiment.

The system design is relatively simple (see Figure 2.1).  Is it based around a cell filled with iodine gas (a diatomic molecule).  First a beam is picked off from the main laser output and sent to the locking system.  This beam is further split into a strong "saturation" beam and two weaker "probe" beams.  The saturation beam is then reflected through the iodine cell.  The two probe beams are also sent through the sent, only in the opposite direction.  One of the two probe beams must overlap the saturation beam.  Once through the cell, the two probe beams must be reflected into photodiodes, whose voltages are then subtracted.

The output from the photodiodes will yield the hyperfine absorption spectrum of iodine gas due to the fact that as the laser frequency is scanned across the absorption frequencies, laser light will be absorbed by the iodine.  It will absorbed in both the probe and saturation beams; however, the saturation beam is so much stronger than the probe that it overlaps that it saturates the region by causing most of the iodine molecules to be excited out of the energy level where the transition takes place.  The result is that one probe beam will be absorbed and the other will remain (for the most part) unaltered.  This difference will change depending on the amount of laser light absorbed by the iodine and in turn, displays the iodine absorption spectrum.

Unfortunately, this signal is very weak in comparison to the random noise due to power fluctuation in the laser itself.  The signal can be found, however, by using a combination of a "chopper" and a lock-in amplifier.  The "chopper" is a small wheel with alternating holes and spokes.  It is placed in the saturation beam and spins such that the beam is alternatively blocked and allowed to pass at a ~1 KHz rate.  This ~1 KHz reference signal is combined with the output from the photodiodes at the lock-in amplifier.  The lock-in amplifier takes the reference signal then picks out anything from the input (in this case the output from the photodiodes) that matches the reference input, then amplifies it by the chosen amount.

Once this is done, the output from the lock-in amplifier will be the hyperfine absorption spectrum of the diatomic iodine molecule.  However, it is necessary to have the derivative of the hyperfine spectrum in order to frequency lock the laser.  The derivative must be used, as it is then possible to use the steep slope near zero to create negative feedback to the laser.  That is, since the signal crosses zero, it is possible to tell which direction the laser frequency is drifting.

The derivative function is produced by using a function generator to create a frequency modulation in the laser.  A function generator and attenuator are hooked up to the "tweeter," a piezo-electric component of the laser that can change the laser cavity length periodically.  The function generator, using a variety of frequencies, from 15 to 25 KHz, also has output sent to the reference channel in the lock-in amplifier.  This is an additional benefit, as the higher frequencies (when compared to the chopper's 1 KHz) reduce the noise in the final output.

The final component of the system is a lock box connected to the laser reference cavity.  The lock box detects when the iodine cell output has drifted from zero (corresponding to the laser frequency drifting away from the iodine resonance) and sends a signal to the reference cavity to correct the drift.  The reference cavity sees this signal as if it had internally detected a drift and moves the laser according to the signal.

The first scans matched those taken by Kristin in 1999 (see Figure 2.2).  However, after making some additional calculations, it was decided that the peak Kristin used (the second triplet) was actually around 1 GHz away from the frequency used for gallium cooling.  Research into commercially available acousto-optic modulators (which are capable of frequency shifting) found that although it would be possible to frequency shift light from the Coherent 699, it would be prohibitively expensive.

However, iodine, as a molecule, has many different absorption lines.  The absorptions targeted by Kristin are electronic transitions.  These are the strongest transitions.  However, further to the red (and closer to the gallium transition) there are another set of vibrational transitions (see Figure 2.3).  These are much smaller, but it is still possible to resolve them.  The first step, however, was to lock to the same peak as used by Kristin in order to test the system.  Still using the Coherent 599, locking was achieved for a sufficiently long period of time.  Analysis showed that the locked laser had a total peak-to-peak drift of ~10 KHz, less than 1% of the unlocked drift.  This locking far exceeds the necessary requirements for the gallium cooling experiment, therefore it was decided to proceed with attempting to lock to the smaller, redder absorption lines.

Locking to the smaller vibrational transitions proved to be more challenging.  These peaks follow the same pattern as the larger, electronic transitions, so it was decided to attempt locking on the second triplet of these peaks as well.  The Coherent 599 dye laser is an older laser and is thus much more prone to drifting or even jumping in frequency.  It was necessary to turn off the air conditioner and lambda meter to lessen the noise in the laser.  It was even possible to see the laser frequency move slightly if someone spoke too loudly.  The Coherent 599 also had the tendency to jump anywhere from 1 to 3 GHz if the output gain on the lock box was set too high.  In essence, this situation could be interpreted as the laser being locked "too tight," as the feedback signal from the lock box would cause drastic overcompensation by the laser, making it unstable and likely to jump in frequency.  However, it was possible to lock the laser for up to seven and a half minutes to within 500 KHz.

Once it was shown to be possible to lock to these smaller peaks, the laser input was switched to the Coherent 699 dye laser.  Once the system was realigned and calibrated, there was a noticeable difference in the amount of background noise caused by fluctuations in power and frequency.  The resolution was improved to the point where work was immediately moved to the smaller peaks.  With little adjustment, it was possible to lock the Coherent 699 to the same small peak as the Coherent 599 for over 15 minutes and to within 100 KHz.

The next step was to compare the absorption spectrum of gallium to that of iodine gas.  This was accomplished by running the cooling laser (which is formed by an external frequency doubling cavity) at very low power through the gallium beam and using a photo-multiplier to create an output signal.  By taking scans of first the gallium and then the iodine, without changing the laser settings, it was possible to compare the peaks of both.  Using this method, it was found that the reddest peak of the reddest set of three peaks of the iodine vibrational transitions was approximately 220 MHz to the blue of the desired gallium absorption peak (see Figure 2.4).  This means that the closest peak that we could lock the laser too was still 220 MHz away from where it needed to be.

Future work on this project would be to use an acousto-optic modulator (AOM) to shift the frequency of the laser light from the desired frequency for the gallium transition to the closest iodine absorption line, then locking the laser with that signal.  On hand, there is a AOM that has a center frequency shift of 110 MHz and a tunable bandwidth of 50 MHz.  By utilizing a double pass through the AOM this results in a shift of 220 MHz +/- 100 MHz, which will be an almost perfect match for this system.  Unfortunately, at peak efficiency, the AOM can only shift 85% of the incoming beam power.  This fact, coupled with the need to use three lens, two waveplates, an iris, and a Faraday isolator in order to perform a double pass through the AOM, translates to an output of 28-33% of the input power.  This beam must then be put through the iodine system, resulting in further power degradation before reaching the iodine cell.

These problems, however, are merely short-term setbacks that could be solved by optimizing system alignment and/or simply picking off more power from the laser.  This should not overly perturb downstream system performance, as the current picked off beam is only ~1% of the overall output.  Doubling this pick off to 2% would be bring the iodine system power back to nearly the same amount as prior to addition of the AOM setup.

In conclusion, it has been shown that is possible to lock a Coherent 699 laser to the smaller absorption peaks of the iodine spectrum and that one such peak exists at a frequency shift that can easily be reached by existing equipment.  These results will allow, with some added work, the ability to frequency lock a Coherent 699 laser to well under 1 MHz of drift over an extended period of time.  This will enable more accurate laser-cooling and focusing of a beam of gallium atoms.
 
 

Within two weeks of Henry Cook's departure from the Lasers Lab, the locking system recommendations he makes above were implemented.

First, the 110 MHz AOM was inserted into the Iodine stabilization beam to shift the location of the peaks.  By doing this, the gallium peaks and the iodine peaks were made to overlap (see Figures 3.1 and 3.2)

Once the peaks were overlapped, the 699 laser was frequency locked to the midpoint of the middle hyperfine peak, where the slope is quite steep.  In this way, a derivative signal, as described above, is not necessary.  In Figure 3.3, the laser was manually placed at the midpoint of the peak (the "lock point") and was then allowed to drift as the laser frequency changes (Unlocked signal) and was also locked using the feedback electronics (locked signal).  The time duration of the scan is about 13 minutes.  Note the excellent stability in both long term drift and short term jitter of the locked signal compared to the unlocked.

Using this frequency stabilization scheme, and the tunability of the AOM which shifts the iodine peaks, we will now be able to park the laser at any arbitrary point on the gallium fluorescence peak and expect it to stay there for the duration of the cooling experiment.  This modification significantly improves the performance of our experiment.

Any opinions, findings and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation (NSF).
 

Link to the National Science Foundation web page
 
 
 

Go back to Lasers Group main page