High-Vacuum Pumps in Mass Spectrometers

S P E C T R O M E T R Y F O R U MHigh-Vacuum Pumps in Mass SpectrometersK ENNETHL. BUSCHThe last regular installment of this columncovered basic vacuum conceptsrelevant to mass spectrometry. Wereviewed basic gas parameters, suchas particle density and mean freepath, and the details of the curve that resultsfrom the application of Paschen’slaw to describe the maintenance of electricpotentials under vacuum. We dividedthe realms of vacuum pressures into variousranges and described how the properoperation of a mass spectrometer requiredthe correct vacuum in the rightplace at the right time.In this final get-your-fingernails-dirtytechnical installment, we describe the operationof the diffusion and turbomolecularpumps (1, 2) often used to achieve thehigh vacuum necessary for operatingmany mass spectrometers. These pumpsare used with a rough pump (or forepump)to move gas molecules from insidea vacuum chamber (a mass spectrometer)to outside the system. Although thepressures achieved by these two pumpingsystems are similar, the cost of equipment,the operating costs, and the proceduresand speed of pumpdown and ventcycles are quite different.In the past, a customer purchasing alarge mass spectrometer could specifyone pumping system option or the other,choosing either diffusion pumps or turbomolecularpumps. Modern instrumentsare more thoroughly integrated, and customerchoices in vacuum systems areusually no longer possible. However, inassembling new custom-built systemsand in designing new instruments, differencesin the two major pumping systemsused to achieve high vacuum should beunderstood and appreciated.At any pressure, gas molecules movein random directions, changing directionsof motion only through collision. Themean free path — the average distanceFigure 1. Components of a diffusion pump.between collisions — of gas moleculescan be calculated, and this parameterserves to define the conditions of viscousand molecular flow. When the mean freepath of the gas molecules exceeds the dimensionsof the vacuum container, thesystem is under molecular flow conditions.Under such conditions, the residualgas molecules move without collidingwith other gas molecules, instead ultimatelycolliding with the surfaces anddevices within the vacuum chamber.Pumping is accomplished when a net(nonrandom) direction to the movementof residual gas molecules in the vacuumchamber is attained.Diffusion pumps and turbomolecularpumps accomplish this same goal differently.In a diffusion pump, a net directionis achieved by collision of the residualgas molecules with a directed and confinedstream of gas-phase molecules ofthe pump’s working fluid. In a turbomolecularpump, the residual gas moleculescollide with the angled spinning rotorson a turbine shaft. In both cases, thenet direction imparted to the residual gasmolecules is into a region of higher pressureand toward the exhaust of the highvacuumpump. The exhaust of the highvacuumpump is connected through theforeline to a rough pump that accomplishesthe transport of the residual gasmolecules to the final exhaust at atmosphericpressure.The diffusion pump is a relatively simpledevice that has been used in vacuumpumping systems for many years followingits initial description by Langmuir in1916 (3, 4). Figure 1 illustrates the basiccomponents of the diffusion pump. Thediffusion pump contains a working fluidthat is evaporated by the boiler at the bottomof the pump. (Early literature aboutdiffusion pumps touted the advantage ofelectric heaters versus gas heaters for theboiler; no earlier references to shovelingcoal or stacking firewood could befound.) The working fluid vapor risesthrough the tower (or chimney) and exitsfrom the tower through a series of annularjets that direct the flow downward andtoward the walls of the pump, which areair- or water-cooled. The diffusion pumpstarts to exhibit a significant pumpingspeed at a pressure (about 50 mTorr) atwhich the mean free path of the workingfluid molecules is large enough that theyreach the cooler pump walls. There, thefluid condenses and returns under gravityflow downward into the boiler. Localpressure in the vicinity of the annular jetsis high, and collisions of the working fluidmolecules with the residual gas moleculesfrom the system impart a net directionof movement to the residual gas mol-14 SPECTROSCOPY 16(5) MAY 2001 www.spectroscopyonline.com

............................. MASS SPECTROMETRY FORUM .............................Figure 2. Components of a turbomolecularpump.ecules. Repeated collisions move theresidual gas molecules toward the regionof higher pressure at the bottom of thediffusion pump. At the ejector jet, thepressure (300–500 mTorr) is such thatthe rough pump attached to the forelinecan capture and exhaust the residual gasmolecules to the atmosphere.The working fluid used in diffusionpumps is a low-vapor-pressure hydrocarbonfluid such as Octoil, Convoil, or Santovac5 (all proprietary product names),or a silicone oil fluid such as DC 702, DC704, or DC 705 (also proprietary). Therequisite characteristics of the workingfluid are an appropriate vapor pressureand chemical inertness. The ultimate vacuumpressure attainable with a diffusionpump is directly related to the vapor pressureof the working fluid; the lower thevapor pressure, the lower the ultimatevacuum. Of course, a lower vapor pressuremeans that a higher boiler temperatureis needed to evaporate the fluid andachieve a pumping action.The chemical inertness is important ongeneral principles, but also especiallyduring the vent and pumpdown cyclesequences, and especially if the propersequences are not followed (5).Hydrocarbon-based oils will crack and oxidizeto tar if exposed to air when hot.The resultant tar coating the boiler of adiffusion pump and clogging the annularjets is difficult to remove. An abused silicone-basedoil will crystallize in theboiler; the crystals are striking in appearance,but the residue is often impossibleto remove, and the entire pump may needto be replaced.Backstreaming is a term that refers tothe migration of working fluid out of thediffusion pump into the main vacuumchamber. Backstreaming occurs whenthe operating pressure of the diffusionpump is too high, and the net downwardmotion of the working fluid moleculescannot be retained. The gas molecule motionsbecome randomized, and the workingfluid molecules diffuse back into themain vacuum chamber, coating the relativelycool metal surfaces that they eventuallyencounter. A vacuum systemdisassembled for inspection will often exhibita thin, greasy-looking film on interiormetal surfaces; this is the result ofbackstreaming of a hydrocarbon-basedworking fluid. Backstreaming occurs to asmall degree even in normal operationbut is minimized by the use of cold capsor cold traps.The aftermath of a vacuum accident —the unintentional or too-rapid venting of a“hot” system — will be a thorough dispersalof working fluid throughout the massspectrometer, and therefore the need fora lengthy, involved, and complete cleaningand parts replacement, in addition toa rejuvenation of the diffusion pump itself.The proper vent cycle includes acooldown for the diffusion pump oil beforeventing (30–60 min). A warm-uptime is also required during the pumpdowncycle for the system to reach itsfull pumping speed. The ability ofturbomolecular-pumped systems to cyclefrom atmosphere to working vacuum in ashorter time is an advantage in some situations,but the initial cost of a diffusionpumpedsystem is lower. Costs of electricityand cooling water required fordiffusion pump operation also have to befigured into a cost-based decision betweenthe two approaches to achievinghigh vacuum.The turbomolecular pump must alsoimpose a net directional motion to theresidual gas molecules in the vacuum system.The turbomolecular pump does soby repeated collisions of the residual gasmolecules with rotating blades of a turbinerotor (Figure 2). The blades of therotor spin at speeds of 20,000–50,000rpm; the edge speed of the rotors approachesthe velocities of the residual gasmolecules themselves. When a collisionoccurs, a direction is imparted to the motionof the residual gas molecules towarda region of higher pressure and towardthe pump exhaust (again, a rough pumpoperating on the foreline).There are several stages of spinningrotors, with the angles of the rotor bladeschanging along the profile of the pumpfrom inlet to exhaust, as shown in Figure2. Residual gas-phase molecules aremoved toward the exhaust and then areremoved from the foreline by the roughpump. Higher molecular mass gas moleculesare pumped more efficiently by theturbomolecular pump than lower molecularmass molecules; therefore, there isvery little backstreaming of lubrication oilmolecules into the main vacuum chamber.The turbomolecular pump has thecharacteristic of being a clean pumpingsystem with no backstreaming.Clearly there is no heater to warm upor cool down in a turbomolecular pump.A system so pumped can be vented to atmosphereand then pumped down morequickly than can a diffusion-pumped system(minus the use of isolation gatevalves in the latter, which decrease the effectivepumping speed). Diffusion pumpsmust be mounted at the bottom of a vacuumchamber, and in a vertical orientationso that gravity can return the condensedworking fluid to the boiler. Aturbomolecular pump can be mounted inany orientation relative to the vacuumchamber. There is residual vibration froma turbomolecular pump, so the mountingsystem often includes an integral vibrationisolation device. Finally, although thepumping speed is not as high as with diffusionpumps, the smaller size of the turbomolecularpump is often an advantagein today’s smaller mass spectrometers,and there is no need for a cooling waterconnection.Mass spectrometrists seem to pay lessattention to vacuum-related issues todaythan in the past. High vacuum has becomeeasier to attain, pumps are more reliable,and pumping systems are moreautomated. In vacuum, as with so muchelse, the modern mass spectrometrist exhibitsthe habit of deferring or ignoringmaintenance. And, as with much else,there is an ultimate cost to be paid for sodoing. In a mass spectrometer, the cost isdecreased performance in terms of lowersensitivity or lower resolution; mass spectrathat deviate from those measured andarchived under more ideal conditions;and shorter lifetimes for replaceableitems such as electron multipliers, ionizationgauges, and source filaments. Therough pump has the final task of exhaust-16 SPECTROSCOPY 16(5) MAY 2001 www.spectroscopyonline.com

............................. MASS SPECTROMETRY FORUM .............................ing the gas molecules to atmosphere, andmaintenance of the rough pump is probablymost important.Rough pump maintenance involveschanging the rough-pump oil on schedule,checking and replacing the sealswhen necessary, managing the heat generatedby these pumps, and regularly inspectingvacuum hoses and clamps. Buthigh-vacuum diffusion and turbomolecularpumps also require maintenance. Internalmaintenance requirements are describedin the operators’ manuals andinstructions; these requirements may involvechecking the pump oil properties ina diffusion pump, or seals and rotor maintenanceprocedures in a turbomolecularpump. Diffusion pumps generate a greatdeal of heat; the pump may be air cooled,in which case the fan and ductwork mustbe kept clean. If the diffusion pump iswater- or fluid cooled, there is more to bedone. A closed-loop chiller will engenderits own specific maintenance requirements,including regular checks of thethermal trips that shut down the systemwhen there is a loss of chilling capacity. Atap water–cooled diffusion pump systemwill include an in-line filter that requiresregular replacement. If you drink tap water,don’t look too closely at the filter orthink too deeply about why monthly filterreplacement is needed. Check the integrityof the water connections on a regularbasis. A flooded laboratory is not apleasant way to begin one’s morning,mop in hand, with abject apologies to colleagueson the floor beneath yet to bemade.REFERENCES(1) G. Lewin, An Elementary Introduction toVacuum Technique (American VacuumSociety, New York, 1987).(2) N. Harris, Modern Vacuum Practice (Mc-Graw-Hill, New York, 1989).(3) M. Hablanian, Diffusion Pump Performanceand Operation (American VacuumSociety, New York, 1983).(4) Web site: http://web2.thesphere.com/SAS/SciAm/1996/Vacuum/Vacuum3.html.(5) Web site: http://www.ee.ualberta.ca/~schmaus/vacf/vacsys.html.Kenneth L. Busch is associate dean of SponsoredPrograms and a professor of chemistryat Kennesaw State University (Kennesaw,Georgia). He can be reached bye-mail at kbusch@kennesaw.edu. He hasspent a few mornings in the laboratorywith mop in hand, reflecting on why waterconnections on diffusion pumps alwaysseem to fail at 3:00 A.M. local time, regardlessof the actual geophysical location of thelab. Retentive in nature, he has kept a 25-year observation log of diffusion-pumpcooling-water connections, noting that approximately15% of home-built systems areplumbed in reverse. The coldest water directlyfrom the tap should enter at the highvacuumtop of the diffusion pump. Late atnight, he wonders who first came up withthe term “guppy pump” and whether theworld would be a better place if mass spectrometristsheld all positions of power andinfluence. ◆Circle 1818 SPECTROSCOPY 16(5) MAY 2001 www.spectroscopyonline.com