ASPECTS OF VACUUM TECHNOLOGY In this exercise we wish (a) To see something of the operation of vacuum pumps; (b) To look at various vacuum gauges; Then, we will use the vacuum system for one of two experiments: (a) To study the meaning of pumping speed, and to make a measurement of it; or (b) To measure the outgassing in the vacuum system chamber when it is heated. This option may require you to do the first part of the computer data acquisition experiment if you haven t already done it. NOTE: There are two preliminary questions that you will need to answer and write up in your laboratory notebook before you start this experiment see below. 1 DISCUSSION: In the history of science it has frequently happened that some advance in technology has opened the door to fresh discoveries in basic knowledge; many important experiments became possible only when developments in apparatus had reached suitable levels. Vacuum technology is one of the best examples of this sort of thing. Many experiments which depended for their success on the attainment of low gas pressure were suggested, but were frustrated because sufficiently good vacuum pumps did not exist; and conversely, when good vacuums were produced there were immediate benefits to physics. Of course vacuum technology is extremely important today, as an adjunct to research in physics, and every physicist needs to have some acquaintance with vacuum techniques. If you don t mind a little pedantry, we shall begin by classifying vacuum pumps into two categories: mechanical (depending on gas streams, chemical effects, and so on). This distinction is useful and important even if it is obvious. 1
Mechanical pumps are made in various forms, some of which are capable of very advanced performance; we shall limit ourselves to the common laboratory type, and you can read further if you are interested. The usual laboratory mechanical pump (often called a forepump ) is essentially a device that opens up a certain volume and then squeezes it out again, repeatedly.the action is illustrated schematically but not with literal accuracy in Fig. 1. With the piston all the way up, the pump volume is essentially zero. Now valve 1 is opened and valve 2 is closed. The piston is drawn down, and the pump volume fills with gas from the system being evacuated. Then valve 1 is closed and valve 2 is opened, and the piston is pushed up, squeezing the gas out into the room. This action is repeated indefinitely. We emphasize that Fig. 1 is designed to show the essential principle, not the actual construction details; a typical pump does not have a piston and cylinder, but rather uses a rotating eccentric. The pump is immersed in oil to prevent gas leakage through its various mechanical seals. Since the pump opens up a certain volume in its cylinder every revolution, and since it makes a certain number of revolutions per second, we can say that it opens a certain volume, which we call S, per second. In time dt, the pump opens a volume Sdt; and, working at pressure P, pumps out P Sdt/kT molecules (this result is from the ideal gas law). Now every system leaks some the pump leaks back through itself, and there may be some little cracks or holes in the walls of the apparatus being evacuated. In addition, every surface evolves gas in a vacuum even solid walls absorb gas, which is given off again when the pressure is reduced. Also every substance has a tendency to vaporize; obviously this is true of such things as oil or water if we are so unfortunate as to have them in our apparatus, but it is also true (clearly to a lesser extent) of solids like brass or glass. So, as we say, the system exhibits leaks and it exhibits outgassing. For all these reasons, the system has a certain rate (call it M molecules/sec) of gas coming in. Thus in time dt the net change in the number of molecules in the system is 2
P Sdt/kT + Mdt. Again using the ideal gas law, we can write dp = kt V where V is the volume of the system. [ Mdt PS ] kt dt (1) If the pumping-out rate exceeds the input rate, of course the pressure falls. But since the pumping-out rate gets smaller as the pressure goes down, while the gas input rate M (on this theory) stays the same, it is not hard to see that the two rates eventually come into balance, after which the pressure stays steady. Then we say that the system has reached its ultimate pressure P s. Since then we have dp/dt = 0, it is easy to see from Equation (1) that P s = ktm S (2) Thus the ultimate pressure is high if we have outgassing and leaks (M) and low if we have a fast pump (S). Nobody should be surprised by those conclusions! A typical good forepump on a reasonably tight system may give an ultimate pressure in the neighborhood of.001 torr. (One torr is one millimeter of mercury; recall that normal atmospheric pressure is 760 mmhg.) It is not difficult to integrate Equation (1). Doing this, with the boundary condition that the pressure is some initial value P 0 at t = 0, we find P = P s +(P 0 P s )e S V t (3) and the general shape of the curve of pressure versus time is shown in Fig. 2. One part of our exercise is to check this out by starting up a forepump on a system from scratch, that is with initial pressure equal to atmospheric, and following the pressure as a function of time. It is not quite accurate to say that the gas input rate M is a constant. For one thing, leaks depend on the difference between internal and external pressure, so that they have more effect as the system pressure goes down; but a leak tends to a constant rate as the system pressure drops. Outgassing tends to decrease with 3
time, as the system cleans up. Mostly for the latter reason, the system pressure may continue to fall very gradually for days. But this occurs so gradually that we can do a good experiment along the above lines over a period of an hour or so. As we have defined the pumping speed S, it is a mechanical constant depending only on the geometry and rotational speed of the pump. If a manufacturer gave only this number for his pump, he would be telling only half the story, because we are also interested in how much the pump leaks back through itself, as every pump does; this backstreaming rate determines how low an ultimate pressure the pump will produce on a tight system with no outgassing, and is a measure of the quality of the pump. To take this into account in the pump specifications, we can define the effective pumping speed S eff by writing S eff = kt P [ PS kt M ] = S ktm P in which case Equation (1) becomes (4) dp dt = PS eff V (5) and we see that S eff goes to zero at the ultimate pressure specified by dp/dt =0. A typical forepump manufacturer s curve of S eff versus pressure is illustrated in Fig. 3; the graph is constructed from measurements on a forepump connected to a tight system with no appreciable outgassing. 4
Non-mechanical vacuum pumps are made in various forms. We shall limit out attention to one of the most important types, namely the oil diffusion pump. This pump has no mechanical moving parts, but depends on a jet of oil vapor. Oil is heated in a pool at the bottom of the pump; the vapor rises through a central chimney and is deflected downwards again along the wall of the pump. (See Fig. 4.) Gas molecules which wander into the jet region are kicked downwards by collisions with oil molecules, and are pumped out at the exhaust by a forepump connected to that point. The oil vapor on striking the wall of the pump is condensed, since the wall is kept cool, and drips into the oil pool. So the oil is continually recirculated. A diffusion pump MUST NOT BE OPERATED WITHOUT ITS WATER COOL- ING. Also it CANNOT OPERATE AT ATMOSPHERIC PRESSURE, but requires to be brought down first with a forepump. If either of these precautions is ignored, very serious damage to the system will result. Typically, the oil oxidizes into a black sludge that is very hard to remove. Vacuum systems are often provided with automatic safety devices to prevent accidents but it is best not to test them! A good oil diffusion pump operating on a tight system with low outgassing is capable of producing an ultimate pressure in the neighborhood of 0.1 microtorr (1 10 7 mmhg) or better. The limit is determined mostly by leakage back through the pump, or backstreaming. A representative manufacturer s curve, for a certain small diffusion pump, is illustrated in Fig. 4-B. The ideal limit to the speed of the pump is determined by the area of its input throat, so that pumps of different sizes have different speeds; some large diffusion pumps have speeds of many hundreds or even thousands of liters per second. 5
A liquid nitrogen cold trap also improves the quality of our vacuum. First, it reduces backstreaming of diffusion pump oil up into the chamber. Second, it freezes out any readily condensed gas that may be present. Surprisingly, much of the gas that is pumped out at low pressures is water vapor that outgasses from the chamber walls! How are these low pressures measured? A mechanical gauge, such as a Bourdon gauge or a liquid column, is satisfactory between atmospheric pressure and (say) one mmhg, but at lower pressures some other scheme must be used. Many different types of pressure gauges have been employed, but the two most frequently used are the thermocouple gauge and the ionization gauge, both of which we shall use in our exercise. The thermocouple gauge depends on the fact that conduction of heat through a gas is a function of pressure. Perhaps you learned in kinetic theory that the thermal conductivity of a gas does not depend on pressure; but that is true only at high pressures. When the pressure is reduced so that the mean path is comparable to the container size, than the thermal conductivity of the gas begins to drop; and the conductivity goes essentially to zero when the mean free path becomes large. To use this effect, a thermocouple is made, with one junction in the external air and one junction in the vacuum system. The internal junction is heated electrically at a constant power. As the pressure goes down, the temperature of this junction tends to rise, because its ability to lose heat through the gas decreases. So the thermocouple output current is a measure of the pressure if the gauge is calibrated. The effect occurs roughly between 0.1 torr and.001 torr, thus limiting the range of usefulness of the thermocouple gauge. But that is the range appropriate to a forepump. Since the thermocouple gauge is simple, rugged, and dependable, it is very extensively used to measure forepressures. The ionization gauge is useful at pressures from roughly 10 torr to 10 8 torr, so it is appropriate for measuring the pressures produced by diffusion pumps. Schematically it works as shown in Fig. 6. Electrons from a filament are accelerated to a 6
positive grid, and their number is measured by reading the emission current. Some electrons penetrate through the holes in the grid, into the space between the grid and a negatively-charged collector. Such an electron may collide with a gas molecule and ionize it. If that happens, the positive gas ion drifts to the collector; the electron goes back to the grid. The ionization current is a measure of how many positive gas ions are produced. Clearly the number of positive ions produced is proportional to the number of incident electrons and to the number of gas molecules available for collisions. But the number of gas molecules per unit volume is proportional to the pressure; hence the ionization current is proportional to the pressure, if the emission current is constant. The calibration of the gauge is of course proportional to the emission. Since the ionization gauge becomes rather warm in operation, its surface outgas quite a lot when it is first turned on. The gauge gives reliable readings only after this outgassing process is essentially complete, which may require on the order of ten or fifteen minutes. 7
2 PUMPING SPEED EXPERIMENT Experimenters are often interested in measuring pumping speeds, and we shall perform such a measurement on the oil diffusion pump. For this purpose we connect a little manometer, illustrated on Fig. 7, to the input of the diffusion pump. Initially both sides of the manometer are at atmospheric pressure. Now the leak valve is opened a bit, and gas flows into the pump; if the manometer refill is now closed, the manometer acquires a pressure differential and the liquid level shifts, over a period of time, so that we can see how much gas was pumped. Quantitatively we argue as follows: Let the leak valve be closed, and let the pump be working at its ultimate pressure, so that we can use Equation (2). Now let the leak be opened so that φ molecules per second flow through it; the pump moves to a new equilibrium pressure P, and (since the total gas flow through the pump is M + φ molecules/sec, we can write in analogy with (2). kt(m + φ) P = S = P s + ktφ S (6) Now the gas in the manometer starts at a pressure P atm and some initial volume V 1 ; after some time it is reduced to some pressure P 2 and and volume V 2. From the ideal gas law, the number of gas molecules removed is P atm V 1 P 2 V 2 kt Let V 2 = V 1 V, where we can get V from the measured movement of the oil column h (Fig. 7) and the cross section of the manometer tube. Also let P 2 = P atm P, where we can get P from the height 2 h and the density of the manometer fluid. Then we can write remembering that φ molecules are pumped through the controlled leak in time t φt = 1 kt (P atmv 1 P 2 V 2 ) 8
= 1 kt (P atm V + V 1 P P V ) (7) By combining Equations (6) and (7) and performing the indicated measurements, we obtain the pumping speed S. Remember that we are dealing here with pressures in two different places; in (6) the pressures are those in the vacuum system, while in (7) the pressures are those in the manometer. Perhaps there is no real danger of confusion on this point. 3 OUTGASSING EXPERIMENT When the glass or metal surfaces that make up most vacuum systems are heated, the rate of outgassing increases. Consequently, baking a vacuum system and then letting it cool off is a good way of boiling off some of the gas that is adsorbed on the surfaces, and hence improving the ultimate vacuum. The experiment is performed by wrapping a heater tape around the vacuum chamber and controlling its temperature by plugging it into a Variac a transformer that controls the voltage to the heater tape. We measure the temperature of the chamber surface by attaching a solid state temperature sensor to the chamber. The current from the sensor is measured on a digital multimeter; the sensor is calibrated so that the current in µa is approximately equal to the absolute temperature. NOTE: The temperature of the surface must be kept well below 200 C, the temperature at which the rubber O-rings in the chamber start to burn up! To get a reasonable amount of data, the chamber should be allowed to heat up and outgas for at least several hours, and then allowed to cool for the same amount of time. Hence this experiment will benefit from computer automation. If you have not done the computer data acquisition experiment, you should do the first part of it as part of this experiment. 4 PROCEDURE: NOTE: You should write a detailed description of the actual procedure you follow in your laboratory notebook. This description should be written while you are in the lab doing the experiment. Do NOT rely on your memory or on rough notes taken on scrap paper!! Do NOT simply copy the suggested procedure in this writeup write down what 9
you actually do!!! Inspect the vacuum system carefully and identify all its components and controls. Make a rough diagrammatic sketch of the system, showing every component and control, with labels. Take a few minutes to work out your procedure for starting up the system. Make sure that the valve between the forepump and the diffusion pump is OPEN, and that the gate valve above the diffusion pump is CLOSED. Also make sure that the manometer leak valve is CLOSED. Now turn on thermocouple gauge No. 1 (the one of the high-vacuum side of the diffusion pump). 0.64 amp AC. Pump the system down from scratch with the forepump, recording the output reading of thermocouple No. 1 as a function of time, to see how Equation (3) and Fig. 1 work out in practice. Since we do not know S or V, we cannot check (3) quantitatively; the most we can do is obtain a curve like Fig. 2 and see generally what it looks like. When ready for the next phase of the exercise, turn on the diffusion pump water supply and the diffusion pump. Be sure the gate valve to the vacuum chamber is closed. Naturally the forepump must be left on! The gate valve should be closed, and the vacuum chamber roughed out to about.01 torr. Shortly after the diffusion pump is turned on, thermocouple gauge No. 2 may indicate a substantial rise in pressure; this effect is puzzling until one realizes that the oil in the diffusion pump outgasses as it is heated. After a time, say 25 minutes, the thermocouple gauge iwll indicate a lower pressure than was obtainable with the forepump alone. This lower pressure is evidence that the diffusion pump has begun to operate. Check thermocouple gauge No. 1, and be sure the pressure is on the order of 0.1 torr. At this point, open the gate valve; the pressure should drop rapidly, and within a few minutes it should be possible to turn on the ionization gauge. When the system has reached a pressure of about 10 5 torr, we can add liquid nitrogen to the cold trap. Then we wait for the system to pump down to its ultimate pressure, a process that will take on the order of hours. The ultimate pressure should be of the order of a few microtorr or better. Then perform the pumping speed measurement as outlined in the Discussion section. Make this measurement several times at a wide variety of pressures. CAUTION: the manometer leak valve needs to be opened ONLY a LITTLE BIT open it slowly and carefully. CAUTION: watch the manometer to be sure that its fluid does not get sucked over into the vacuum system. 10
To shut down the system: turn off the ionization gauge; turn off the DIFFUSION pump. LEAVE THE WATER ON. LEAVE THE FOREPUMP ON. Turn off the thermocouple gauge. RESULTS DESIRED: 1. Display your sketch of the system, showing all components and controls. 2. Give a careful explanation of how the pumps and gauges work. 3. Display your curve of pressure versus time taken with the forepump. Compare its appearance with Fig. 1. 4. If you did the pumping speed experiment, compute the pumping speed (in liters/sec) obtained in each measurement, and put an error bar on each determination. Does the pumping speed come out about the same, within uncertainty, regardless of the pressure at which it was measured? 5. If you did the outgassing experiment, make careful graphs of your data and discuss them. Can they be interpreted in terms of adsorbed gases being boiled off the chamber walls? In any case, give a careful discussion of your results. 6. Calculate the number of molecules per cm 3 at the lowest pressure you reached. In view of your result, remark on why we call this system a vacuum? (Hint: This question is not entirely obvious, and will require a little reading and a little thought. Begin by reviewing mean free paths in any text and calculate the mean free paths at atmospheric pressure and your lowest pressure. Think about how you can associate the properties you associate with a vacuum with mean free paths.) PRELIMINARY QUESTIONS 1. Suppose we have a forepump with a speed S of 0.1 liter/sec working on a volume of 20 liters. If the ultimate pressure P s is.0001 torr, how long will it take the pump to work down from atmospheric pressure to a pressure of 0.1 torr? See Equation (3). 2. A diffusion pump will handle 100 liters/sec at a throat pressure of 1 microtorr (10 6 mmhg). Its exhaust is a forepump with a speed of 0.1 liter/sec. The forepump must maintain a pressure of 100 microtorr (10 2 mmhg) at the diffusion pump exhaust port if the diffusion pump is to operate properly. Is the forepump fast enough, or should we get a faster one? 11