outgas with its maximum saturation reached in two to three days (dependent on atmospheric humidity and temperature). Aside from water vapor that has condensed (adsorbed) on the surface of the vacuum system, there is also water that has absorbed into the material of the vacuum system itself. With a glass vacuum system, the depth of absorbed water can be as deep as 50 molecules. Because it is almost impossible to remove absorbed water without baking the system, this water (in a glass system) can be considered nonremovable.

O-rings and other flexible seals also require an outgassing period. Neoprene or perbunan rubber O-rings can be used to pressures of 5 x 10"7 torr, but cannot be heated beyond 100°C. These materials give off small quantities of water vapor and CO. Viton O-rings, which show very little propensity for outgassing, can be heated to 250°C and can be used in pressures as low as 10~9 torr.66

If your system contains materials with high vapor pressures that cannot be baked out, you cannot achieve a vacuum greater than those vapor pressures. For example, a piece of paper placed in a vacuum system will limit the system so that it can never achieve better than a 10"5 torr vacuum.67

A dirty system takes longer to outgas (and may never outgas*) than a clean system, and a rough surface takes longer to outgas than a smooth surface—the former because of extra materials, and the latter because of increased surface area. Rough surfaces tend to hold water tighter than smooth surfaces.

The first and foremost step to solving any vacuum problem is to know your system. The more familiar you are with your system, the less time you will waste on wild goose "leak hunts." One technique that will help you know your system is to maintain a log book. Kept within this log could be information such as the date the system was used, how long it was used, what vacuum was achieved, what type of work was done, and when pump oils were changed. These data can provide invaluable aids for determining if any changes are gradual or sudden, if there is any relation to using new pump oils, and any other indications of change. A log book should be mandatory any time more than one person uses a vacuum system.

7.6.4 Real Leaks

There are two ways to demonstrate that a leak in a vacuum system really exists: Either you are unable to obtain a dynamict vacuum that previously could be obtained with no problem, or the system is not maintaining a static* vacuum that it previously could maintain.

'After reassembling a very powerful vacuum system that had been left open for a weekend, a friend of mine could not obtain a vacuum greater than 10"3 torr. Finally, in desperation he dismantled the system to find the remains of a small mouse whose outgassing rate set a limit for a system that nor-

mally was capable of significantly greater vacuum.

*A dynamic vacuum is when a pump is actively evacuating a system.

*A static vacuum is when the passageway from the pump to a system has been closed, and the quality of the vacuum within the system must be self-maintained.

Trapped gas

Fig. 7.51 An O-ring caused virtual leak.

Verifying that a real leak exists only tells you part of the story because there are four types of leaks. You must eliminate them by analysis of the symptoms, experience, the history of the vacuum system, and/or trial and error. The following are the four types of real leaks:

1.Virtual leaks

2.Leaks at demountable seals*

3.Real leaks through the walls of the vacuum system

4.Backstreaming from pumps

Virtual Leaks. A virtual leak is an honest-to-goodness leak, but because it is inside the vacuum system, it cannot be detected from the outside. A formal definition of a virtual leak is "a self-contained gas supply within a vacuum system." There are two types of virtual leaks:

1.Those leaks in which a gas is physically trapped (such as a gas which is poorly, or improperly, frozen or has had insufficient outgassing)

2.Those leaks where the gas is mechanically trapped as shown in Fig. 7.51 and Fig. 7.52

If frozen gases within a cold trap are too close to the top, they are likely to evaporate as the liquid nitrogen level in the trap drops. This situation will cause the pressure in the system to rise and/or fluctuate (see Sec. 7.4.3). Different problems are caused when the wrong coolant is used for trapping such as when dry ice, rather than liquid nitrogen, is used as the coolant for a cold trap. If water vapor is in such a trap, the lowest possible achievable pressure would be 10"3 torr. This level is equal to the vapor pressure of water at the freezing temperature of carbon dioxide. If liquid nitrogen is used as the coolant, the system could potentially achieve 10~15 torr,69 the vapor pressure of water at liquid nitrogen temperatures.

Demountable seals are sections of a system that can be removed and replaced without damage to the system, yet maintain integrity of the system.

Fig. 7.52 Improper and proper welds for a metal vacuum system.

Although outgassing is one type of virtual leak, not all virtual leaks are results of outgassing. One example of a mechanicallycaused virtual leak is the space trapped in the channels of an O-ring joint by a compressed O-ring. As you can see on the left half of Fig. 7.51(a), an O-ring is compressed in the O-ring groove. In the blow-up seen at the left of Fig. 7.51(a), you can clearly see the areas of trapped gas in the corners of the groove. Because these areas are at a higher pressure than the system, they will "leak" into the vacuum. However, because they are contained within the system, it is impossible to find the leak from outside the system. To prevent this type of virtual leak, the groove that supports an O-ring should be made without sharp corners. In addition, radial "pie cuts" can be made which allow gas passage. These cuts can be seen on the right half of Fig. 7.51(a) and on the top view of an O-ring joint seen in Fig. 7.5l(b).

A second example of a virtual leak occurs in metal vacuum systems by a double weld being made instead of a single weld (see Fig. 7.52). Because only one weld should be required, a second weld can only lead to possible problems. The statement "If one is good, two is better" does not apply to this case. Incidentally, all welds should be made on the vacuum side of any given chamber. When a weld is on the outside, channels are created on the vacuum side. These channels can collect contamination that may be difficult or impossible to remove.

Demountable seals are sections of a vacuum system that are not permanently mounted. These seals include sections or pieces that can be moved or rotated such as traps, stopcocks, or valves. They differ from permanent seals, which are pieces of a vacuum system that are welded, or fused, together and cannot be separated or rotated without damaging the pieces or the system as a whole. Demountable seals require some material between the sections to be compressed, such as O-rings, grease, or gaskets that can prevent gases from passing. If there is inadequate or improperly aligned compression or cutting, or inadequate or improperly placed (or distorted) material to be compressed, or cut into, there will be a leak.

Leaks at Demountable Seals. These leaks are caused by wear, old stopcock grease, poorly applied grease, poor alignment, twisted or worn O-rings, inadequate stopcocks or joints, mismatched plugs and barrels, or dirt between pieces. This list is by no means complete, but it gives you an idea of the range of problems.

438 Vacuum Systems

Among the items that may be initially checked on a glass vacuum system are to ensure that all plug numbers of glass vacuum stopcock plugs match their respective barrel numbers. As mentioned on page 193, these numbers must match.* Another factor to consider with vacuum stopcocks is whether they are old, worn, or just defective. Additionally, you should examine that the vacuum stopcocks have been properly and recently greased.f

One specific example of how leaks can originate in a demountable seal is the example of a worn standard taper joint being mated with a new joint. This situation can occur when an old cold trap bottom had been broken. The new replacement does not fit into the worn areas of the old, original member. If the worn areas (caused by the original members rubbing against each other for years) form a channel, a leak is created. In addition, standard taper joints and glass stopcocks can also be damaged by being left in HF, a base solution, or a base bath for an extended period of time. Unfortunately, the application of excess stopcock grease to cover up these imperfections is likely to create virtual leaks.

The advantage of O-ring joints is their easy demountability. However, if they are mounted unevenly or torqued unevenly, a leak can be easily created. Helium is readily absorbed into O-rings and result in long outgassing times. If previous leak detection was done on a vacuum line with O-rings, be prepared for such an occurrence. If a solvent is left in contact with an O-ring for an extended period of time, anything from swelling to destruction of the O-ring can result. This situation, of course, is dependent on the type of solvent and type of O-ring. One way to find out if there will be a problem with an O-ring and particular solvent is to pour some of the solvent in question into a small jar and drop in the type of O-ring you plan to use. Let it sit for several hours. If there are any effects, they should be obvious. A trick that can be used to provide limited protection from solvents is to apply a small dab of stopcock grease (select one not affected by that solvent) on an O-ring before assembly. Place a small amount of the protective grease on a Kimwipe and, while holding the O-ring in another Kimwipe, smear a thin film of grease on the O-ring.* The grease will provide a limited protective barrier for the O-ring.

The author has observed a system with two plugs and barrels mismatched that could not achieve a 1-torr vacuum. Once the plugs and barrels were properly matched, the system went to 10"3 torr in five minutes.

+ The author has also observed systems where the stopcocks looked perfectly good, but they had not been used in over a year. The user could not attain a vacuum better than 102 torr. Simply by regreasing all the stopcocks, the system went down to 10'5 torr within half an hour of restarting. In this situation, the poorly greased stopcocks could not provide adequate seals for a vacuum, but were able to prevent atmospheric moisture from contaminating the walls of the system. Thus, the system could achieve a reasonable vacuum in a short time.

*The Kimwipe is not used to remove dirt from your fingers, although that benefit will also be realized. Rather, it is to prevent your skin oils from contacting the vacuum equipment. Skin oils can create virtual leaks from their outgassing.

Fig. 7.53 Pattern for torquing bolts on metal flanges.

Ball-and-socket joints should not be used for high-vacuum systems because they are not intended for vacuum work. It is possible to obtain specially made ball joints with O-rings that are acceptable for some vacuum work. Some manufacturers supply sockets that have not been ground, and when used in tandem with ball joints with O-rings, they can achieve satisfactory vacuum performance.

Metal flanges require a gasket between their two sections. The bolts of metal flat flanges must be equally torqued by first tightening all bolts finger-tight. Then, using a wrench, you must tighten the bolts with less than a quarter turn each following the pattern shown in Fig. 7.53. When you arrive at the sixth bolt, repeat the same star pattern, but go to bolts 2, 5, 1, 4, 6, and 3 in turn.

Some flanges will have eight or more bolts. Use the same alternating pattern for making a proper seal. Continue with this rotational pattern until the bolts are tight. If a metal flange has a knife-edge that cuts into a gasket, the gasket cannot be reused. No matter how well you try to line up the cuts with the knife-edge, it will not work, and it will leak. If you are desperate, you might try to have a machine shop remove the top and bottom surface of the gasket. However, experience has shown that this attempt is less than 50% successful for being leak-tight.70

Avoid using pipe threads whenever possible because they create long leak paths. Long, thin leak paths require long time periods for leak verifications. If pipe threads must be used, do not use Teflon tape because it voraciously absorbs helium and will later thwart the use of a helium leak detector. Rather, use a small amount of thread sealant paste that contains Teflon (see Fig. 7.54). Glyptol, a common temporary leak sealant, is sometimes used as a pipe sealant. This sealant is not recommended because it ages and, after some six months or so, can develop a leak.

Thermocouples often come with pipe threads for attaching onto vacuum systems. Avoid using these pipe joints entirely by attaching a thermocouple to a glass system by silver soldering the thermocouple to a premanufactured glass-to-metal seal.

Use a cadmium-free silver solder and a suitable silver-solder flux. You will need a gas-oxygen torch to make this seal (see Chapter 8 for information on gas-oxygen torches). Do not use a "hissy" torch flame because that indicates a flame too rich in oxygen. A soft blue flame is reducing and will prevent oxidation of the metal pieces, which prevents a good seal.

leak is between 10~5 and 10"3 std cm3/sec. The next most likely type leak is one in the transitional flow range of gas movement. Here, the molecules of gas are about equally likely to hit a wall as another molecule. This size leak is between 10"6 and 10"5 std cm3/sec. The least likely type of leak found is when the size of leak is so small that a sufficient vacuum is achieved to cause molecular flow in the region of the leak. This size leak is below 10"7 std cm3/sec.71

Leaks up to the size of 10"5 std cm3/sec are not difficult to find in a laboratory vacuum system achieving a 10"6 torr vacuum. Below that range, baking the system may be required to fully open the leak from water vapor, fingerprints, or a host of other possible contaminants. Leaks smaller than 10"5 std cm3/sec may not affect your system, but the amount of time spent in looking for these small leaks needs to be justified.

In vacuum systems, because measurements are usually made in liters/sec and pressures are stated in torr or Pascal, leak rates are known as torr-liters/sec, or Paliters/sec. These expressions are not equivalent leak rates, but they are proportional in the following ratios:

1 std cm3/sec = 0.76 torr-liter/sec = 101.13 Pa liters/sec 1 torr-liter/sec = 1.3 std cmVsec = 133.3 Pa-liter/sec

1 Pa-liter/sec = 0.0075 torr-liter/sec = 0.0098 std cm3/sec Backstreaming. There are two types of backstreaming; both are not desirable,

but one can be quite deleterious. Once the vacuum system achieves its ultimate vacuum and is in molecular flow, gases from the system can drift back into the system. This is simply the results of equilibrium. On the other hand, if oils or vapors from the diffusion pump drift into the system, or oils from the mechanical pump drift into the diffusion pump, the effects are more profound.

Although a system typically will not act as if it has a leak until it is in the ultravacuum range (>10~8 torr), at lower vacuum ranges (<106 torr) other complications from backstreaming can occur such as:

1.Silicon diffusion pump can coat and damage the filament of a mass spectrometer.

2.Mechanical pump oils (with their high vapor pressure) that drift into the diffusion pump can decrease its pumping efficiency.

3.Oils in a vacuum system can negatively interact with vacuum gauges (mercury can destroy thermocouple gauges and hydrocarbons on thoriated iridium filaments of hot-ion gauges require constant recalibration).

What can be limited is the backstreaming of condensable vapors. This can be achieved through the effective use of baffles, traps, and using good vacuum system design. The simple resolution for backstreaming of condensable vapors from standard pumps is proper system design and proper maintenance of the traps.

If there are any pressure blips in the system and the mechanical pump cannot remove them fast enough, a diffusion pump will be swamped. That is, there is not enough backing pressure to maintain the gradient pressures within the diffusion pump and the gases contained within backstream into the system. The only pre-