Clean, Robust Alkali Sources by Intercalation within Highly-Oriented Pyrolytic Graphite
Rudolph N. Kohn, Matthew S. Bigelow, Mary Spanjers, Benjamin K. Stuhl, Brian L. Kasch, Spencer E. Olson, Eric A. Imhof, David A. Hostutler, Matthew B. Squires
aa r X i v : . [ phy s i c s . a t o m - ph ] M a r Clean, Robust Alkali Sources by Intercalation within Highly-Oriented PyrolyticGraphite
Rudolph N. Kohn, Jr., a) Matthew S. Bigelow, Mary Spanjers, Benjamin K. Stuhl, Brian L. Kasch, Spencer E. Olson, Eric A. Imhof, David A. Hostutler, and MatthewB. Squires Space Dynamics Laboratory, Albuquerque, New Mexico 87106,USA Applied Technology Associates, Albuquerque, New Mexico 87123,USA Air Force Research Laboratory, Kirtland AFB, New Mexico 87117,USA
We report the fabrication, characterization, and use of rubidium vapor dispensersbased on highly-oriented pyrolytic graphite (HOPG) intercalated with metallic ru-bidium. Compared to commercial chromate salt dispensers, these intercalated HOPG(IHOPG) dispensers hold an order of magnitude more rubidium in a similar volume,require less than one-fourth the heating power, and emit less than one-half as manyimpurities. Appropriate processing permits exposure of the IHOPG to atmospherefor over ninety minutes without any adverse effects. Intercalation of cesium, potas-sium, and lithium into HOPG have also been demonstrated in the literature, whichsuggests that IHOPG dispensers may also be be made for those metals. a) Electronic mail: [email protected] . INTRODUCTION Alkali metals serve as the atomic backbone for a wide variety of physics experiments.Alkalis’ simple electronic structure and strong transitions facilitate laser cooling and makethem very attractive candidates for atomic sensors. In experiments using alkali metals, theon-demand production of a clean, dilute vapor is often the first crucial step. The stringentrequirements of modern cold atom experiments mean that significant improvements in thisfirst step can positively impact the rest of the experiment. This work introduces a newarchitecture for producing extremely pure alkali vapors using intercalated graphite, andcompares this new architecture to commercially available dispensers.Many methods exist to produce pure, dilute atomic vapors. However, experimental pa-rameters often limit which methods are viable. For example, an atomic beam experimentmight need a source with high flux and directionality, but stationary or slow-moving coldatom experiments often depend upon extremely pure, dilute vapors, so that the atoms canbe trapped and cooled with minimal interference from background gas. The desired back-ground pressures are often orders of magnitude below the room temperature vapor pressuresof the metals, impeding the use of pure metallic sources. Other characteristics, such as ca-pacity, activation temperature, ease of handling, and total vapor produced frequently placeother restrictions on the source.For the production of pure, dilute alkali vapors, there are two general architectures incommon use. The first uses ovens, usually emitting effusive beams, which offer high purityand capacity (grams) and low activation temperature ( ∼ ℃ for rubidium). However, sim-ple ovens produce enough vapor to have deleterious effects on pumps and vacuum quality. Negative effects on vacuum can be mitigated, but generally at the expense of increased com-plexity. The large quantities of pure alkali used can present difficulties for safe handling anddisposal, as well. The second general architecture uses chromate salts and non-evaporable getter (NEG)material to produce a reasonably pure alkali vapor.
Commercial chromate salt dispensersare compact and easy to handle, but they have relatively low capacity ( ∼
10 mg for a 30 mmlong dispenser), much higher activation temperatures ( ∼ ℃ for rubidium chromate), andcan emit significant quantities of unwanted gas under certain conditions. A typical chromatesalt dispenser contains NEG material to inhibit the release of unwanted gases. However,2rolonged periods at room temperature can collect unwanted gases in the NEG material or onthe nearby chamber walls, and improperly degassing before activation can contaminate thesource. We have observed that extended periods at room temperature produce a measurablepressure spike when the chromate dispenser is next heated, which can be traced to eithergas adsorbed onto the steel container of the dispenser or absorbed into its NEG material,or adsorbed onto chamber surfaces near the dispenser, which are heated during activation.We present an alternative source of alkali vapor which compares favorably to the chro-mate salt architecture in size, ease of handling, and ease of activation while improving onthe purity of their output and increasing their capacity to ∼
100 mg in a similar volume.Highly-oriented pyrolytic graphite (HOPG), essentially graphite with a high degree of inter-nal order, can absorb relatively large amounts of foreign chemicals between its graphene-likelayers. This behavior, known as intercalation, has been studied since at least the 1940s.
Graphite and HOPG have frequently been used as getters for alkali vapors because of thisbehavior.
The characteristics of intercalated HOPG (IHOPG) as a dispenser have beenexamined in the context of vapor deposition, but, to the best of our knowledge, no exami-nation of its compatibility with ultra-high vacuum (UHV) or modern cold-atom experimentshas been made. Many materials are known to intercalate into HOPG, but of special interesthere are the four alkalis known to intercalate relatively easily: lithium, potassium, rubid-ium, and cesium. Studies have shown that sodium is much more difficult to intercalateinto graphite, therefore it seems unlikely that reliable dispensers can be made for sodium.Although examples of cesium, potassium, and lithium intercalation are present in the liter-ature, the work described here uses rubidium exclusively. We devised a method to reliablyintercalate HOPG with rubidium, make the IHOPG relatively stable in atmosphere, andcontrollably dispense rubidium vapor under vacuum.We describe the methods for producing rubidium IHOPGs in Section II. The apparatusused to compare the dispensers is detailed in Section III, and then, two comparisons aremade. The first, in Section IV, compares the purity of the output vapor of both dispensersin a steady-state configuration. The second, in Section V, compares the undesired gas whichaccumulates on, in, or near the dispensers after extended periods at room temperature. Inboth cases, the IHOPG dispenser produced less undesired gas than the chromate dispenser.In general, heating dispensers emits waste gases, either from the dispenser itself or fromgases adsorbed onto nearby walls. Emitted rubidium interacts with background gas, caus-3ng a getter effect. Over the range of rubidium emission rates we tested, the waste gasesproduced by heating the chromate dispenser consistently overwhelmed this getter effect.However, at sufficiently high rubidium output rates the waste gases produced by heatingthe IHOPG were so inconsequential that the pressure in the chamber actually decreased.Aside from the IHOPG dispenser used for these tests, another IHOPG dispenser has beensuccessfully integrated into a cold atom experiment loading grating magneto-optical traps(MOTs). Another system in current daily use contains an IHOPG for dispensing rubidium,and regularly produces Bose-Einstein condensates.
II. FABRICATION AND OPERATION
Alkalis are intercalated into HOPG by placing a heated sample of HOPG in close prox-imity to vapor at high enough ( & mTorr) pressure. The heat allows alkali atoms to diffusebetween its graphene-like layers. Specific temperatures are given for rubidium. Cesium hasa very similar melting point and vapor pressure, so the temperatures required may also bevery similar. Potassium and lithium, however, melt at much higher temperatures and havea much lower vapor pressure, so higher temperatures will almost certainly be required.The procedure detailed below reliably produces rubidium IHOPG dispensers with about1 mg of rubidium per mm of dispenser. The initial samples of HOPG were 7 x 7 x 1 mm,but successful intercalation increased their volume by a factor between 2 and 3. Severaloptional steps, detailed in Subsection II A, improve ease of handling.The structure of the original HOPG affected how reliably they could be intercalated.HOPG is typically graded by crystallographic order, characterized by two parameters: mo-saic angle and grain size. Mosaic angle is a measure of the dispersion of the angles ofcrystallites in the sample. The grain size is typically measured in microns or millime-ters and describes how far apart, on average, grain boundaries can be found. In the earlystages of our work, we experimented with samples with different levels of order and foundthat samples with higher crystallographic order loaded and dispensed more reliably. Thedispensers described here were all produced from HOPG samples with mosaic angles of 0.8 ± ℃ under high vacuum to eliminate surface impuritiesand degas the sample. 48-72 hours at this temperature, with a turbomolecular pump to4aintain vacuum, is sufficient to bake out a reasonably clean HOPG sample, removed fromits packing material and handled with gloves. After the prebake, we remove the sample to adry nitrogen atmosphere and add elemental rubidium into the chamber. The glove box weuse for this purpose is fed dry nitrogen from a dewar source, but any inert gas (e.g. argon)would likely work just as well. In practice, we usually pour a few drops of molten rubidiumonto the HOPG, but other delivery methods, such as small pieces of solid rubidium droppedinto the chamber should work just as well, since the intercalation process is driven by heatingthe HOPG and exposing it to vapor, which will be present when the chamber is heatedin either case. Before heating, we attach the chamber to an oil-free roughing pump, reducingthe background pressure to the milliTorr range. The rough vacuum removes unwanted gasesin the chamber, which can impair intercalation. We then seal off the chamber containingthe HOPG and rubidium under rough vacuum and heat the chamber to 125-150 ℃ for atleast 48 hours to intercalate.After 48 hours, we turn off the heaters and remove the cooled chamber to a dry nitrogenatmosphere to examine the HOPG. Successfully intercalated HOPGs dramatically increasein thickness, with a 1 mm thick HOPG swelling to 2 or 3 mm after intercalation. Thisexpansion is strong evidence that the process creates high stresses in the HOPG sample. Wehave observed samples that broke into several pieces during intercalation, and an attempt toload rubidium into graphene foam reduced the sample to dust. Based on these observations,we infer that larger grain size inhibits structural damage to the HOPG.The structure and dimensions of IHOPG are fairly well known . IHOPG has severaldifferent stable structures with different stoichiometric ratios. The structure with the mostintercalated material has the formula XC , corresponding to 0.89 g of rubidium for each gramof carbon. However, our loaded samples usually gain more mass than this, with a 110 mgHOPG gaining 100-220 mg of rubidium. The expansion in size is also larger than predictedby the expected structure: the known thickness difference between pure and maximallyintercalated HOPG is 68%–much less than the 100-200% increases observed in our samples.In addition, IHOPGs exposed to air for long periods tend to expand further, with layerssplitting apart as the rubidium oxidizes. These observations strongly suggest that additionalrubidium is making its way between the layers of the HOPG, pushing them further apartand adding more rubidium to the dispensers than expected in a pure intercalation.Once the HOPGs are loaded, they are placed in a vacuum chamber and heated in order5o liberate the intercalated rubidium. The IHOPGs begin emitting rubidium vapor whenheated over some activation temperature, which varies somewhat from sample to sample.Typical activation temperatures range from 125-160 ℃ , and seems to roughly correlate withthe amount of intercalated rubidium. Additional heating over the activation temperatureincreases the emission rate. Dispensers rapidly plate rubidium onto nearby glass at temper-atures of 250 ℃ . A newly loaded dispenser held at 250 ℃ depleted itself in about 72 hours,implying a rate of 2-3 milligrams per hour at that temperature. A different IHOPG sampleemitted no measurable rubidium over a 12 hour period at 150 ℃ , but at 170 ℃ , it producedenough rubidium to observe laser-induced fluorescence in a small chamber after about 10minutes, suggesting an output rate of about 0.5 nanograms per hour, taking into accountthe laser power and the sensitivity of the infrared scope. Our attempts to measure rubidiumvapor emitted from IHOPGs at room temperature under vacuum have all been below thedetection limit.We attempted to heat the IHOPGs above the activation temperature with several meth-ods. Of these methods, we had success with two: conductive heating from outside thechamber and inductive heating across a glass wall. For conductive heating, a resistive tapeoutside the chamber heats the IHOPG through the chamber wall. This method allows easymonitoring of the IHOPG temperature with a thermocouple, which is useful for character-izing a sample’s activation temperature. Conductive heating might also be accomplishedby attaching the IHOPG to a heating element inside the chamber. For inductive heating, awire coil outside the chamber, oriented in the same plane as the layers of the IHOPG andcarrying a large, rapidly oscillating current, heats the sample by inducing eddy currents inthe graphite. In a glass chamber, the IHOPG can be epoxied to a chamber wall or sim-ply rest on the chamber bottom. The distance between the coil and the IHOPG controlsthe temperature and emission rate, though measuring the absolute temperature using thismethod is difficult. A. Improving Ease of Handling
Although the dispenser can be used immediately after loading, it usually leaves the in-tercalation chamber coated in a layer of metallic rubidium. Failing to remove the surfacerubidium results in rapid reactions with air and moisture upon exposure, and can cause6tructural damage to the IHOPG after only a few minutes. A few optional steps greatlyease handling in atmosphere. First, we transfer the IHOPG into a clean glass vacuum cham-ber and heat it with a heater tape to between 100 and 120 ℃ , with a turbomolecular pumpmaintaining vacuum. We maintain the temperature below the activation temperature toremove surface rubidium without dispensing from between the layers. Depending on theamount of surface rubidium, this takes 24-48 hours. We gauge the depletion of surface ru-bidium by observing laser fluorescence on the D2 line at 780 nm. Heated surface rubidiumproduces enough vapor to observe laser-induced fluorescence when a near-resonant laser ispassed through the chamber. When this fluorescence disappears, the surface rubidium hasbeen depleted and it is safe to move on to the next step.After removing the surface rubidium, we slowly raise the temperature of the IHOPG tofind the activation temperature. A thermocouple provides temperature data, and a resonantlaser beam passing through the tube produces fluorescence when the IHOPG begins to emitrubidium. Once we observe laser-induced fluorescence, we note the current temperature asthe activation temperature and raise the temperature an additional 40 ℃ for 8 hours. Thegoal of this 8 hour period is to deplete the rubidium at the edges of the IHOPG, whichwe suspect reduces its availability to react with local atmosphere. Samples treated in thisway have been handled in air for 90 minutes or more with no visible changes. Withoutthis second step, the sample turns grey as a coating of rubidium hydroxide forms on thesurface. Samples with this coating have still been used to load a MOT. However, thesereactions are usually best avoided, as they may cause physical damage, making heating theIHOPG more difficult. For example, a crack could impair inductive heating, or disconnectthe dispenser from its heater.
III. APPARATUS
In sections IV and V, we will discuss two comparisons between IHOPG dispensers andcommercial rubidium chromate dispensers. This section describes the apparatus used toperform both of those comparisons. A schematic view of the chamber is pictured in Figure1. On the left, a rectangular glass chamber provides optical access for a 3D MOT (redlaser beams). The IHOPG and a chromate dispenser were located near each other, in thecylindrical glass neck between the glass and steel parts. A sputter-ion pump and a residual7
IG. 1. The chamber used for the experiments in Sections IV and V. The two different dispensersrest near each other in the glass neck at the top of the figure. The rubidium fluorescence is measuredat the crossing of the three beams, pictured in red. The retro-reflecting optics for the 3D MOT,the magnetic quadrupole coils, and the inductive heating coils are not shown. gas analyzer (RGA) were attached to the steel part of the test chamber. The RGA measuredpartial pressures using a quadrupole mass analyzer, but measured the total pressure usinga separate ion gauge filament. The unit used in these experiments was new, and for ameasured total pressure of 1 . × − Torr the sum of the pressure peaks from the massspectrometer was 9 × − Torr. The RGA measured mass/charge ratios out to 90 AMU/e.The two types of dispensers were oriented in perpendicular planes to permit selectiveinductive heating. The power transferred by an inductive heater is highly dependent on thespatial orientation of the coils and the object to be heated. The inductive heater coil forthe chromate dispensers was wrapped around the cylindrical glass chamber and moved backand forth along the cylindrical section, changing its distance from the chromate dispenserto control its temperature. A different coil was placed under the IHOPG and moved upand down to control its temperature. The linear translation stages used to move the coilshad 8 mm of travel. Due to space constraints, only one coil was in position at any giventime. The two dispensers were only a few millimeters apart in the chamber, but based onthe effective ranges of the inductive heaters (less than 8 mm), and on how strongly theorientation of the heaters affects the level of heating, we are confident that the heating ofeach individual dispenser did not significantly heat the other. The orientation and size ofthe coil for the chromate dispenser makes the induced fields over the volume of the IHOPGvery small. Similarly, the coil for heating the IHOPG was several millimeters from the8hromate dispenser and was much smaller than the chromate dispenser, meaning that it wasextremely unlikely to induce significant currents in the chromate dispenser. In addition, thedifferent activation temperatures make it highly unlikely that the chromate dispensers wouldbe anywhere near activation when the IHOPG was heated. When the chromate dispenserswere heated, the heating of the glass in the area of the IHOPG was minimal. The lack oflocal heating is attributed to the use of Litz wire for the chromate heater coil, which helpedreduce the resistivity of the coil at the high frequencies produced by the inductive heater.The rectangular glass chamber admitted three perpendicular, circularly-polarized laserbeams. These three primary beams all contained light 13 MHz red-detuned from the 5 S / , F = 2 to 5 P / , F = 3 line in Rb. A repumping beam, tuned to the transitionbetween the 5 S / , F = 1 and 5 P / , F = 2 resonance, was aligned with the beam goingfrom right to left in Figure 1, and could be turned on and off independently of the primarybeams. The three primary beams had a total power of approximately 13 mW, and eachbeam had a waist of about 7 mm. The repump beam had 1.8 mW of light in a similarlysized beam. Each beam was retro-reflected through a quarter wave plate to produce thelight fields necessary for a 3D MOT. Magnetic coils outside the vacuum chamber producedthe necessary quadrupole field at the intersection of the three beams. The MOT ensuredthat the laser was locked to the same frequency for each run, and gave an early indicator ofrubidium vapor output.The sputter-ion pump maintained the chamber pressure near 1 . × − Torr. The cham-ber was baked out for 5 days at 125 ℃ using a turbomolecular pump to minimize impurities,but the Viton gasket seal on the exit valve limited the overall quality of the vacuum. TheRGA measured a partial pressure of water vapor of 1 . × − Torr, with additional peaksof hydrogen (2 . × − Torr) and nitrogen and carbon monoxide (1 . × − Torr, at thesame mass value). Other, smaller peaks were also present at several common mass numbers(e.g. carbon dioxide), but they did not significantly change upon heating of either dispenser.Neither dispenser, when heated, introduced new significant peaks to the RGA traces. Ru-bidium vapor was strongly attenuated by the chamber walls, never reaching the RGA inmeasurable concentrations. However, the dispensers were run for periods of several hoursper day for several days before taking data, in order to align the laser beams and test andoptimize the fluorescence measurements.Before taking data, each dispenser was individually degassed over the course of several9
IG. 2. Changes in total pressure and relevant partial pressures at steady-state with each dispensertype. The rubidium vapor density is approximated from fluorescence measurments. The vapordensity is in turn determined by dispenser temperature and serves as an indicator of the emissionrate. hours. Each dispenser was slowly heated while the RGA monitored the released gases.The MOT beams provided feedback to determine when rubidium output began. As thetemperature was increased, occasional bursts of output gas were measured on the RGAand were allowed to dissipate before further increasing the temperature. This process wascontinued until rubidium output was observed. The dispenser being used was then cooledto room temperature, and the background pressure in the chamber was allowed to stabilize,as measured on the RGA, before starting a run.
IV. STEADY-STATE OUTPUT COMPARISON
Each dispenser was heated individually to a steady-state to measure the output gases.With the small amounts of rubidium used, the steel chamber walls were never saturated withrubidium, so the rubidium vapor was not detected by the RGA. Instead, rubidium density10as measured by observing the fluorescence of the atoms illuminated by the MOT coolingbeams, imaged onto a photodiode. The measurements used in Figures 2 and 4 were takenwith the quadrupole field and repump light turned off, as day-to-day variations in MOTshape resulted in significant variation in measured MOT fluorescence. In contrast, the lightfrom the untrapped atoms was much more repeatable from day to day, and the imagingsystem had a much clearer global maximum during alignment. Between the cylindrical neckwhere the dispensers were and the chamber where the fluorescence was measured, therewas a glass plate with an aperture. Neither dispenser had direct line-of-sight access to thefluorescence chamber. Rather, each dispenser produced a local cloud of rubidium, and afraction of the atoms were diffusely reflected by the walls into the fluorescence chamber.Since the steel half of the chamber effectively functioned as a pump for rubidium, the localrubidium density near the dispensers should be roughly proportional to the dispenser’semission rate. The mean free path in the system is expected to be much larger than the sizeof the system, so the atoms that are directed at the crossed laser beams should reach themrelatively unimpeded.Calculations assuming a room-temperature distribution of rubidium and using the mea-sured powers and detunings of the beams provided a rough estimate of local rubidium pres-sure as a function of observed fluorescence. For each of the six laser beams, we calculatedthe total fluorescence F in photons per second: F = n Z ∞−∞ d ~v Z . − . d ~x f ( ~v ) Γ( ~x, ~v ) , (1)where ~x and ~v are position and velocity, respectively, n is the Rb density, f ( ~v ) is the 3DMaxwell-Boltzmann distribution f ( ~v ) = (cid:18) m πk B T (cid:19) / exp( − m | v | / k B T ) , (2)and Γ( ~x, ~v ) is the rate of spontaneous emissionΓ( ~x, ~v ) = γ I ( ~x ) /I sat I ( ~x ) /I sat + (2( δ − ~k · ~v ) /γ ) . (3) I ( ~x ) is the intensity profile of the beam in question, assumed to be a Gaussian in twodimensions and constant in the third direction, as the beams were collimated. γ is the excitedstate linewidth (2 π × . × s − ), δ is the detuning ( − π × × s − ), ~k is the beam’swave vector, m is the mass of Rb, T is 300 K, I sat is the saturation intensity (1.67 mW/cm ),11 IG. 3. An example of the RGA data used to produce one of the points in Figure 2. These dataare from a run of the chromate dispenser, and correspond to total pressure measured by the RGA.When the dispenser is heated, the pressure spikes, then falls to an equilibrium pressure greaterthan measured before heating. The pressure difference between the measurement before turning onthe dispenser and at steady state is plotted in Figure 2 against the rubidium fluorescence observedat steady-state. and k B is Boltzmann’s constant. The limits of the integral for position are determinedby the field of view for the detector and its optics. The total number of photons wascorrected for solid angle, efficiency of the detector, and amplification factor of the detectorelectronics. These calculations led to a rough estimate of the fluorescence emitted versus thenumber density (pressure) of the rubidium, corresponding to about 1 × − Torr of Rbper millivolt of fluorescence. Since the six beams technically formed a three-dimensionalmolasses configuration, similar calculations estimated the additional contribution from slowatoms, assuming that all atoms with a velocity lower than 15 m/s were slowed to the recoilvelocity, thereby increasing their time in the beams and emitted fluorescence. It should benoted that no effort was made to zero the magnetic fields, which will weaken the effect of themolasses. However, this overestimate leads to additional fluorescence from the slowed atomsthat was less than that from the rest of the thermal distribution by about three orders ofmagnitude. The effect of the molasses is therefore negligible.Figure 2 details the observed pressure changes as a function of rubidium output for thechromate and IHOPG dispensers. Each point represents a run in which one of the dispensers12as heated until the output gases measured by the RGA reached a steady state. In the caseof the chromate dispenser, a typical run started with a large spike of output gases, followed bya taper over several minutes to reach a final plateau (see Figure 3). The chromate dispenserstypically reached a steady-state after approximately ten minutes, and the IHOPG responsetimes were somewhat slower. For low output rates, the IHOPGs acted in much the sameway, but as the IHOPG temperature and rubidium output increased, a qualitative changein the IHOPG behavior occurred due to the background water vapor in the chamber.During runs with the IHOPG at temperatures above a threshold corresponding to about1 mV of rubidium fluorescence, the pressure curves measured by the RGA eventually fellbelow those measured before heating. In these cases, the pressure decrease continued veryslowly over a fairly long period. The IHOPG data shown in Figure 2 were limited to 40minutes of run time per point. The 40 minute duration was chosen to capture the majorityof the effect while constraining experimental time. A few tests of the IHOPGs at longertimes, up to two hours, showed that the 40 minute time limit captured the vast majority( > O → . (4)The chamber used in this experiment had a significant background including water vapor.However, in a UHV chamber, the getter effect of the rubidium would not be present. Thewell-known characteristics of the RGA measurement make it possible to approximate thepressure change in a chamber without background water vapor. We take the total pressuredata and add the lost water vapor back to each individual point. Figure 4 shows the dataafter this correction. Instead of simply adding back the lost water vapor, the data arecorrected for the measured sensitivity differences between the mass filter and ion gaugefilament, as well as for the measured ratio of the correlated 17 AMU/e peak. Correctingfor the lost water vapor, the total pressure change is still well below that observed fromthe chromate dispenser over the measured range. As noted above, rubidium vapor wasattenuated by the chamber walls before reaching the RGA. Therefore, the rubidium vaporpressure does not appear in the RGA total pressure levels.The chromate dispensers did not produce a measurable decrease in water vapor despite13he emission of similar amounts of rubidium to the IHOPG dispenser. We attribute this tothe fact that the activation temperature for the chromate dispenser is much higher, resultingin greater sympathetic heating of nearby chamber walls and desorption of water vapor fromthe walls. Even though the rubidium from the chromate dispenser was reacting with similaramounts of water vapor, the amount of water vapor released by heating the chamber wallsexceeded the amount reacting with the rubidium. A careful examination of Figure 2 showsthat the water vapor output when heating the chromate dispenser has a weak negative slopeas the rubidium output increased. The amount of additional desorption as the dispensertemperature was raised was less than the getter effect of the additional rubidium, but thehigh activation temperature meant that there would be significant heating of the chambereven at very low rubidium output rates.In a chamber without background water vapor, the pressure decrease shown in Figure 2would likely not appear, but the corrected data in Figure 4 suggest that the pressure increasedue to undesired gas emission from the IHOPG is extremely small. In comparison, thechromate dispensers increased total background pressure by about 1 . × − Torr regardlessof the level of rubidium output. In the worst case, when the IHOPG was producing almostno rubidium, the total pressure increase in the steady state was 7 × − Torr. These resultsshow that in the configuration used here, the IHOPGs produce, at most, about one-half ofthe waste gases of the chromate dispensers, and compare more favorably at higher outputrates. Some of the observed waste gases were almost certainly contributed by desorptionof gases on nearby walls, but this effect is difficult to engineer away, as the heat requiredfor activation, especially for the chromate dispensers, is quite large. In other words, thereseems to be a significant advantage to be gained by using dispensers that require less heatto activate, as generating less heat mitigates desorption and out-gassing from other nearbyvacuum components.
V. GAS ABSORPTION COMPARISON
UHV systems have extremely stringent requirements on their contents. Even extremelyclean materials can outgas enough to have a measurable effect on vacuum quality. Back-ground gases adhere to the surfaces of the chamber or diffuse through chamber materials,creating a persistent gas load. Heating the chamber or its contents, e.g. a dispenser, re-14
IG. 4. Total pressure comparison, with the graphite dispenser data corrected for reactions betweenwater vapor and rubidium. The chromate dispenser data are unchanged from Fig. 2, and are shownfor reference only. The pressure loss due to dispensed rubidium reacting with background watervapor has been added back to the IHOPG data, using best estimates for the sensitivity differencesbetween the mass spectrometer and ion filament, as well as the measured ratio of the 17 AMU peakto further refine the correction. These data provide approximate values for the expected pressuredifferences that would be observed in a chamber without background water vapor. Of note is thateven with the water vapor losses added back to the IHOPG data, the total pressure increase isnear zero, and still much less than the observed pressure increases from the chromate dispenser,even without correcting those data for the water vapor reaction. leases even more gas. We have observed significant gas loads produced by commercialchromate dispensers when heating them, especially after leaving them at room temperaturefor extended periods. For example, in one chamber, a SAES NEXTorr D 100-5 sputterion/non-evaporable getter (NEG) hybrid pump normally runs currents at the lower limit ofobservation, between 0 and 1 nA, corresponding to pressure at or below 1 . × − Torr.After 72 hours with the dispenser at room temperature, heating the chromate dispenser sig-nificantly increases the ion pump current. The increase depends on how rapidly the dispenseris heated, but currents of 5-6 nA (7.7 - 9.2 × − Torr) during the first hour of heatingare common. Longer periods of inactivity seem to result in larger currents. Whether thisis caused by adsorption of waste gases onto the steel portion of the dispenser or reversibleabsorption into the getter material is not certain, but it is reasonable to suppose that a15
IG. 5. Measurements of absorbed gas in the different dispensers or adsorbed to nearby surfacesheated along with the dispenser. Each point is derived by integrating the area under a differencecurve calculated by subtracting the pressure curve of a one-hour off-time from the chosen, longeroff-time, and then multiplying by the pumping rate. See Figure 6 for a set of example data. dispenser without dedicated internal getter material might collect less gas over time. Theexperiment described below tests this hypothesis, comparing adsorbed/absorbed gases fromIHOPG and chromate dispensers after various periods left at room temperature.In order to measure absorbed gas, the dispensers were left at room temperature for a rangeof times between 2 and 72 hours, and then heated to a constant temperature, correspondingto about 1.8 mV of fluorescence, or about 2 × − Torr of Rb in the glass chamber.During the heating process, the RGA measured various gas pressures every few seconds.The resulting curves were compared to a curve produced by the dispenser after 1 hour atroom temperature (i.e., when almost no gas should have been absorbed). The area differencebetween the curves is proportional to the total waste gas released, as illustrated in Figure 6.The total gas released is approximated by multiplying the area between the curves by thepump rate of the ion pump. Figure 5 shows the results of these tests and calculations.The results show that both dispensers and the sympathetically-heated parts of the cham-ber collect similar amounts of gas up to about 24 hours at room temperature. At longer16
IG. 6. Example of pressure data used to produce the 40 hour point for IHOPG gas absorptionmeasurement. The lower curve shows the total pressure as a function of time when the IHOPGwas heated after 1 hour at room temperature. The upper curve shows pressure data collected whenthe IHOPG was similarly heated after 40 hours at room temperature. The area difference betweenthe two curves is used to determine the amount of gas absorbed by the dispenser during the 40hour period. Each point in Figure 5 comes from one such comparison. times, the chromate dispenser and the chamber around it continue to absorb, but the IHOPGlevels off, except for hydrogen. This is unsurprising because rubidium intercalated graphiteis known to act as a getter for hydrogen at room temperature. These results suggest thatthe IHOPG dispensers attract less total waste gas than the chromate dispensers when leftcold for an extended period, though the amount of the effect which can be attributed to thedispenser rather than nearby chamber walls is difficult to determine. However, the data sug-gest that the IHOPG may have additional utility in experiments where long (days or longer)periods of inactivity are expected, in addition to their overall lower level of undesired gasoutput.
VI. CONCLUSIONS AND OUTLOOK
Experimental data and observations show that IHOPG is a suitable source for clean ru-bidium vapor in a cold atom experiment. IHOPGs consist of relatively inexpensive and easilyavailable materials, and their production requires equipment typically available to atomic17hysics laboratories. They are operated with similar equipment to chromate dispensers, butrequire less power to operate and produce much less waste gas. Handling time in air canbe increased to over 90 minutes with simple post-processing. Under most circumstances,this should be enough time to mount the IHOPG and bring the chamber down to roughvacuum. Another potential issue with IHOPGs is the low maximum baking temperature,which is limited to the activation temperature of the IHOPG, between 125 and 150 ℃ , butthese temperatures are often enough to bake out a chamber, albeit over a longer period.Since the expected temperatures needed to intercalate lithium and potassium are higher, itis possible that IHOPG dispensers using those metals will have higher activation tempera-tures, which would require more power to dispense gas but allow higher maximum bake outtemperatures.The increased capacity per mass and volume of IHOPGs is an advantage in experimentswhere long-term operation without service is necessary, such as space applications. The highpurity of the output vapor reduces load on the vacuum pumps, and minimizes the increasein background pressure, which may be especially useful in experiments requiring a compactform factor, where, for instance, differential pumping schemes might not be feasible.Since cesium, potassium, and lithium are known to intercalate into HOPG with relativeease, they are excellent candidates for dispenser production as well. Cesium’s applicablethermodynamic characteristics are very similar to that of rubidium, so loading and dispens-ing may work at very similar temperatures. Potassium and lithium have a much lower vaporpressure for a given temperature, so higher temperatures will almost certainly be requiredto load and activate the dispenser, but these temperatures may still be lower than thoserequired for a potassium or lithium chromate dispenser.The production of IHOPGs is simple and inexpensive. They can be produced withequipment readily available to most atomic physics labs. No highly technical skills arerequired to produce them. Therefore, they may find applications in undergraduate-levelatomic physics experiments, where budgets are limited and students have not yet developedextensive technical skill. As an undergraduate-level experiment, the production of IHOPGscould be used to instruct students on the use of a glove box, basic vacuum protocol, andsafe handling of alkali metals.IHOPGs have already been used in-house to load a 2D grating MOT. The atom numbersobserved in the 2D and 3D grating MOTs were very favorable compared to other similar18xperiments, despite using less overall laser power than other grating-based systems.An IHOPG is currently in use on an experimental apparatus that regularly produces Bose-Einstein condensates. The vacuum chamber was changed over from one with a chromatedispenser to one with an IHOPG and has experienced a noticeable improvement in thelifetime of atoms in the magnetic trap, from around 2 seconds to 5 seconds. The lifetimein the IHOPG chamber has remained at this level after about 1 year of nearly continuousoperation. This, in addition to the data described here, suggest that the integration ofIHOPG dispensers into systems is practical and can improve their behavior, compared tochromate dispensers. Our experimental apparatus currently under construction are movingover to IHOPG from chromate dispensers whenever feasible.
VII. ACKNOWLEDGMENTS
This work was funded by the Air Force Research Laboratory. We wish to thank Dr. GregPitz and Joshua Key of AFRL for additional help in producing IHOPGs.
REFERENCES Y. J. Lin, A. R. Perry, R. L. Compton, I. B. Spielman, and J. V. Porto,Phys. Rev. A , 063631 (2009). M. R. Walkiewicz, P. J. Fox, and R. E. Scholten, Rev. Sci. Instrum. , 3342 (2000). SAES Getters, “Alkali Metal Dispensers,” (2007). Various commercial equipment, instruments, and materials, as well as their suppliers areidentified in this paper to foster understanding and reproducibility. These identificationsdo not imply endorsement by the Air Force Research Laboratory, or that other materialsare unsuited to these purposes. J. Fortagh, A. Grossmann, T. W. H¨ansch, and C. Zimmermann,J. Appl. Phys. , 6499 (1998). W. R¨udorff and H. Schulz, Z. Anorg. Allg. Chem. , 121 (1940). R. C. Croft and R. G. Thomas, Nature , 32 (1951).19
S. R. Jefferts, J. Shirley, T. E. Parker, T. P. Heavner, D. M. Meekhof, C. Nelson, F. Levi,G. Costanzo, A. De Marchi, R. Drullinger, L. Hollberg, W. D. Lee, and F. L. Walls,Metrologia , 321 (2002). F. Levi, L. Lorini, D. Calonico, and A. Godone,IEEE Trans. Instrum. Meas. , 267 (2003). J. A. Baumann, D. G. Brock, M. A. Kuck, and G. K. Miller,“Graphite intercalated alkali metal vapour sources,” European Patent 0130803A2 (1985). D. Guerard and A. Herold, Carbon , 337 (1975). Z. X. Shu, R. S. McMillan, and J. J. Murray, J. Electrochem. Soc. , 922 (1993). B. Jungblut and E. Hoinkis, Phys. Rev. B , 10810 (1989). F. J. Salzano and S. Aronson, J. Chem. Phys. , 149 (1965). F. J. Salzano, J. Chem. Phys. , 4551 (1966). S. Aronson, F. J. Salzano, and D. Bellafiore, J. Chem. Phys. , 434 (1968). R. Asher, J. Inorg.Nucl. Chem. , 238 (1959). E. Imhof, B. K. Stuhl, B. Kasch, B. Kroese, S. E. Olson, and M. B. Squires, Phys. Rev.A , 033636 (2017). M. Ohler, M. del Rio, A. Tuffanelli, M. Gambaccini, A. Taibi, A. Fantini, and G. Pareschi,J. Appl. Crystallogr. , 1023 (2000). H. Bremer, F.J. and Bleichert, in
Concise Encyclopedia of Materials Characterization ,edited by E. Cahn, R.W. and Lifshitz (Pergamon Press, Inc., 1993) p. 165. L. Stout, G.H. and Jensen,
X-ray structure determination: a practical guide (Macmillan,New York, 1968) p. 24. F. J. Salzano and S. Aronson, J. Chem. Phys. , 1323 (1965). D. E. Nixon and G. S. Parry, J. Phys. D: Appl. Phys. , 291 (1968). K. Ichimura and M. Sano, J. Vac. Sci. Technol., A , 543 (1992). C. Wieman, G. Flowers, and S. Gilbert, Am. J. Phys. , 317 (1995). C. C. Nshii, M. Vangeleyn, J. P. Cotter, P. Griffin, E. A. Hinds, C. N. Ironside, P. See,A. G. Sinclair, E. Riis, and A. S. Arnold, Nat. Nanotechnol. , 321 (2013). M. Vangeleyn, P. F. Griffin, E. Riis, and A. S. Arnold, Opt. Lett. , 3453 (2010). J. Esteve, Nat. Nanotechnol.8