UNITS: SI including Pascal, 1 Pa = 0.01mbar. (1 Torr  = 133.322  Pa)

1. Introduction

UHV has long been an important tool in research and development laboratories for the physical sciences particularly those concerned with the preparation and analysis of materials often in the form of thin films. The need for rapidly and reliably obtained UHV is also a developing need in certain production industries and perhaps particularly those concerned with semiconductor manufacture. Stainless steel has been traditionally used for such systems and remains the most popular material for ease of fabrication and economic reasons, although in recent years there has been some move towards the use of aluminium (Ishimaru et al., 1992).  UHV  nowadays  is standard technology and can be achieved rapidly and reliably providing cost is no object. In this tutorial we will consider the practical realities of the research laboratory which often involve working with existing systems, complex internal fittings for experimental purposes and sometimes the enforced use of non-UHV compatible materials. The aim is to be able to make estimates of what can be expected in less than ideal situations where a compromise may have to be reached using the available equipment.

2. Basic principles

UHV may be roughly defined as vacuum pressures below E-8 mbar. For fusion or outer-space simulation applications, pressures below E-10 mbar with low hydrogen content would probably be required, whereas some high rate deposition processes for example, although sensitive to impurities, would tolerate E-8 to E-9 mbar and hydrogen may or may not be a problem. Hydrogen is highlighted for good reason; once surface adsorbed water, carbon oxides and hydrocarbons are removed by baking and assuming no leaks or build-up of poorly pumped gases, then hydrogen diffusing from the walls of stainless steel chambers and from internal fittings is the main source of residual gas and will be directly related to the wall thickness of the vacuum vessel. It can be shown experimentally and theoretically from Fick's Law (Jost, 1960), that long-term high-temperature bakeouts are necessary for thick-walled vessels if these are to be used at elevated (300C) operating temperatures (Santeler, 1991).
Before examining the significant factors in the attainment of UHV it is worthwhile to take as a simple example a typical vacuum chamber of say 450 mm diameter and 750 mm length, an area,A, of about E+4 cm2. The pressure,p,  achieved in the chamber is given by the relation,

p  =  QA/S  mbar        Equ. 1

where Q is the outgassing rate from the walls of the chamber in mbar l/ s cm2, and S is the pumping speed in l/s. Outgassing rates from cleaned and /or low temperature-baked, (<150C) stainless steel have been quoted in the literature over a wide range from E-8 to E-11 mbar l/s cm2, but using a commonly quoted (but fairly conservative), value of E-10, (Young, 1969; Sullivan, 1992; Dylla et al.,1993) gives a pressure,p, of E-8 mbar for a pumping speed of 100 l/s (a small ion-pump or turbo); or E-9 mbar for a speed of 1000 l/s (a much more expensive pump). Hence to achieve low pressures  in systems containing fittings and fixtures which may involve a considerable amount of metal area and gas traps (if for example mechanical parts or manipulators are involved), it is necessary to reduce the outgassing rate by orders of magnitude if reasonable size pumps are to used. If costs are no object,  pump size may be increased with advantage but not without limit, since the pump itself adds a considerable area of metal in its connecting tubulation; the outgassing from this area must be reduced to the same level as the rest of the system if advantage is to be taken from the increased pumping capacity. One should note also that in a laboratory where old equipment may have been pressed into service, the outgassing rate may well be considerable higher than in the above example even with rigorous cleaning techniques.
The principal method of reducing outgassing is baking at elevated temperature although other factors as discussed below can also be significant.

3. Factors involved in the achievement of UHV

3.1 Materials for construction and hardware
Stainless steel is the most frequently used material for fabricating vacuum chambers and for internal hardware but significant variations of outgassing rate have been found between different grades as reviewed for example by Adams (1983). Early systems and components were often made with 304 and later with 316 but considerations of outgassing, fabrication and quality control have led to the almost standardized use of 316L. Considerable variations in outgassing performance have been reported  even for this one material leading to the conclusion that the quantity of gas held trapped in the metal is in many cases uncontrolled in the manufacturing process. Thus for the highest performance as required for example in fusion devices, space hardware and particle accelerators, or experiments involving the elimination of monolayers on surfaces, vacuum-melted 316L is the material of choice. The outgassing rate for cleaned but unbaked 316L is quoted as E-9 to E-10 mbar l/s cm2,  whereas vacuum induction melted/vacuum arc remelted (a double vacuum-melting process) has an unbaked outgassing rate of E-11 mbar l/s cm2, (Sullivan et al., 1992). The use of such a material is a major advantage and can yield outgassing rates a further two or three orders of magnitude lower when surface passivation and baking techniques are subsequently used.

3.2  Surface condition
Desorption of surface gas is a major component of outgassing load and it is therefore to be expected that the condition and chemistry of internal surface layers would be of importance in determining achievable vacuum. Significant factors include: chemical composition, structure and thickness of oxide layers since these affect adsorption and desorption from the surface as well as diffusion from underlying metal (Ishikawa, 1995); surface passivation to produce Cr oxide or fluoride rich or boron nitride layers (Tomari, 1991; Yoshimura et al., 1991; Tohyama et al., 1991); effect of metal forming and subsequent machining or surface finishing techniques employed (Young, 1969); polishing, which may be chemical, electro-chemical or mechanical; and cleaning where a wide range of possible techniques is available.

3.2.1 Surface chemistry
Early work (Huffine and Williams, 1960), suggested that oxidation of the surface of stainless steel could reduce hydrogen outgassing by a factor of ten or more. Other workers  have since found that outgassing can be reduced by oxidation. More recently Ishikawa (1995),  examined the effect of a surface oxide layer on outgassing from stainless steel and found that oxidation in air produced a surface layer of iron oxide whose effect was to form a diffusion barrier for hydrogen diffusing from the bulk metal. The improvement is rather less than an order of magnitude but evidence is given that carbon contaminants also are reduced and hence carbon containing outgassing species. Young (1969),  found evidence that oxidation of 304 stainless steel in air at 450C  reduced outgassing by depleting hydrogen from the steel rather than any diffusion barrier effect.  According to Sullivan (1992), the important chemical processes that cause the formation of volatile gases, occur in the near surface layer which extends through the native oxide and to some extent into the bulk. These authors conclude that a thin dense and chemically clean surface oxide layer reduces the probability of chemical surface processes contributing to outgassing rates.

3.2.2 Surface structure
Some early workers tended to assume that outgassing would be higher from rough rather than smooth surfaces on the assumption that a rougher surface would contain more gas traps and have greater surface area. This view was questioned by Young (1969), who produced evidence with 304 stainless steel to show that outgassing from smooth electropolished surfaces was no different from glass bead shot-blasted surfaces, both surfaces having subsequently been baked at 250C. His results interestingly, agreed quantitatively with Calder and Lewin (1967) who used U15C material and who did not state any particular surface treatment. Dylla et al. (1993), examined the effect of surface roughness on outgassing performance and showed that there was no effect when the surface roughness was varied by over two orders of magnitude. It may be stated that there is little if any real evidence to show that electropolishing together with other forms of fine surface machining reduces outgassing and while such processes are commonly used as finishing techniques by manufacturers, their main function  would seem to be cosmetic.

3.2.3 Surface Cleaning
It is obvious that rigorous cleaning should be undertaken before exposing surfaces to vacuum and bakeout. ESCA studies (Sullivan et al., 1992), have shown that electropolishing alone or in combination with abrasive polishing does not result in noticeably lower surface carbon content than just trichloroethane vapor degreasing, and this conclusion applies to most other impurities examined. The authors criticise the use of electropolishing in manufacturing practice for quality control reasons, since commercial electropolish suppliers may not change solutions frequently enough to remove impurities which can lead to outgassing products as well as particles and corrosion centers. They conclude that for vacuum component manufacture, an environmentally safe cleaning process using an alkaline-based detergent, hot deionized water rinse and drying is an effective technique resulting in surfaces that desorb water efficiently and also have extremely low particulate generation.
A form of cleaning which has been used successfully in the fusion energy field is glow discharge cleaning (Santeler, 1991). In this process the higher energies (compared to thermal motions), of the plasma particles removes tightly bound carbon oxides and and hydrocarbons which are removed by the pumps. This is of vital importance when a low surface contamination is necessary (for example when surfaces are to be subsequently bombarded during a processing schedule). Plasma cleaning may be combined with simultaneous baking for maximum effectiveness.

3.3 Baking
Early workers baked systems only to about 200C which was effective in removing weakly bound surface water and hydrocarbon molecules. It was not long before it was discovered that baking to higher temperatures was required to remove more tightly bound species and the hydrogen diffusing from the bulk, both of which had the effect of limiting the ultimate system vacuum. The effectiveness of high temperature bakes is amply demonstrated  for example by Calder and Lewin (1967), who showed that outgassing could be reduced to about E-16 mbar l/s cm2 by baking for 11 days at 300C or 1 hr only at 635C. Barton and Govier (1968),  showed that baking new 18/18/1 stainless steel components at 450C in vacuo, with the exception of traces of hydrogen, successfully removes all gases resulting from the previous history . Interestingly these workers found that gas adsorbed at the surface of stainless steel on re-exposure to atmosphere could be remove by baking at this temperature for 2 to 3 hrs. If pressures only of the order of E-9 mbar were required then sufficient gas could be removed by a 12 hr bake at 200 to 300C;  for some applications it is thus worthwhile to use a separate vacuum oven for baking new components at 450C thus avoiding the need for the main system to withstand high temperatures.
Santeler (1991), quotes early work by Aero Vac Corporation which gives valuable data on the effect of baking to different temperatures which is reproduced here below:

Outgassing rates in Torr l/s cm2
                                                        BAKING TIME
         20 hrs
40 hrs
100 hrs

The data above bring home clearly the benefits of high baking temperatures and the diminishing returns from increasing baking time much beyond 20 hrs for baking temperatures up to 400C; at 500C  the table shows that it is worthwhile to bake for as long as 200 hrs although one might legitimately question the unusually low value of 8.0E-19 torr l/s cm2 that is quoted. Recent work (Ishikawa, 1995; Ishikawa et al 1991; and Ishikawa and Odaka1990), confirms the values above for lower temperatures but indicates that surface treatments and raw material quality can improve these figures substantially, (section 3.2 above).

3.4 Pumping systems
The selection of pumps, whether for research or production, is not always straightforward and invariably involves cost considerations. All four main types of high vacuum pump:  turbo, cryo, diffusion and ion, are capable of producing pressures in the UHV region, often with the addition of a titanium sublimation pump (TSP), to achieve the lowest pressures. The choice of pump will often be determined by the application. However, there are characteristics of pumps which limit their ultimate UHV performance and capability of removing products of outgassing and these are now considered.

3.4.1 Turbomolecular pumps (TMPs)
These are compressors and are limited inherently by the low compression ratio for hydrogen which is the main residual gas in a UHV system. To achieve pressures approaching E-10 mbar or less it is advisable to combine this type of pump with a titanium sublimation pump for the above reason. TMPs can usually be baked to a maximum of about 120C and for some models less, hence some thought is necessary when designing connecting manifolds to avoid overheating the pump and at the same time minimising the area of metal that is heated only to low temperature during bakeout. Lower cost TMPs have hydrocarbon oil or greased bearings which is a potential source of contamination but providing the pump is operated correctly hydrocarbon molecules will not migrate into the vacuum vessel. Correct venting is particularly important and the pump must not be stopped under vacuum. Higher cost TMPs are available with air or magnetic bearings which avoid these possible problems.

3.4.2 Cryopumps(CPs)
These pumps are limited by their capacity to cryosorb hydrogen onto cold charcoal.  An ultimate system pressure somewhat below E-9 mbar is typical which can be extended to below 5E-10 mbar with a TSP. The capacity to adsorb hydrogen will decrease with time once the hydrogen cryosites are used up and this can result in an increase in system pressure. CRPs cannot be baked and moreover must be shielded from radiated or conducted heat caused by baking or processing schedules of the main system. The area of unbaked shields and tubulation must be minimal if UHV pressures are to be achieved. Some manufacturers have incorporated additional radiation cold shields and baffles into their pump designs to reduce the scale of this problem. Cryopumps have very high pumping speeds for water-vapour, but in a UHV system water-vapour is removed by baking so that this may not be such a great advantage. CRPs require periodic regeneration and a high vacuum isolation valve; for the lowest pressures an all metal bakeable design is necessary, but many UHV systems operate satisfactorily with a viton plate seal and copper gasket or wire seals for flanges with a bellows sealed drive. The viton seal can be baked safely to 120C closed and 150C in the open position. As stated above the area of metal not heated to the full baking temperature must be minimised.

3.4.3 Diffusion pumps (DPs)
DPs will pump to about 5E-10 mbar which can be extended to well  below E-10 mbar with a TSP. Current practice is to use low vapour pressure hydrocarbon based fluids with a liquid nitrogen and/or water or cryo-cooled baffle. Thorough tests have shown, (for example, Gay et al., 1994;  Dennis et al., 1982) have shown that even with only a water-cooled baffle hydrocarbon contamination from the pump is at the limits of measurement and effectively close to zero for all practical purposes. The work by Gay et al., is particularly illustrative; the experiments were made to confirm contamination free conditions for a GaAs source which could not tolerate hydrocarbons or CO and revealed the importance of correct operation not only of the diffusion pump, but the foreline where trapping and purging procedures were found to be essential to ensure isolation from rotary pump backstreaming. Dennis et al., showed that integral diffusion pump/baffle/valve systems could be cleaner than hydrocarbon lubricated turbopumps.
DPs do carry the risk of gross contamination from the consequences of accidental air-dumping or partial contamination from faulty operation. Diffusion pumps, like turbopumps, act by compression and hence their pumping action in a UHV system is also finally limited by the compression ratio for hydrogen, but the problem is less severe with a DP compared to a TMP. A high vacuum valve is normally used to isolate the pump when the system is vented, but some designs of UHV system would eliminate this to achieve higher pumping speeds and remove a source of outgassing.

3.4.4 Ion pumps (IPs)
IPs will pump to 5E-10 mbar or less with extension to below E-10 mbar with a TSP. In many respects the ion pump which has no moving parts or working fluids is ideal for UHV in cases where there is no subsequent process gas load. The IP does have certain characteristics which the user needs to be aware of and may cause problems in some circumstances: (i) Selective pumping. Pumping is by gettering for chemically active gases and  burial for inert gases, which occurs at a much lower pumping speed. Inert gases will therefore tend to build up as a residual gas background, the lighter inerts (He and Ne), are particularly prone since these may diffuse out again from beneath the surface. Hydrogen readily diffuses into titanium as a solid solution, but will be released again on heating. The presence of argon in the residual enhances sputtering and hence the chemical removal of hydrogen as titanium hydride. Methane is normally present in the residual since this is generated at titanium surfaces although also removed by dissociation in the pump discharge. (ii) Memory effects. These are characteristic of IPs and will affect the residual gas composition, gas ions buried in the cathodes earlier in the life-cycle are released by subsequent sputtering. The effect can be particularly bad for inert gases and leads to the phenomenon of argon instability where significant amounts of released argon lead to periodic increases of pressure. This has been designed out of some models of pump by slotted cathodes, triode structures or Ti/Ta cathodes. (iii) No tolerance to gas load. IPs are not suitable for even small loads and particularly inert or hydrocarbon gases; once the pressure inside the pump rises above E-6 to E-5 mbar, the pumping speed falls and pumping may become unstable. This can eventually lead to heating inside the pump and the additional release of hydrogen, helium and neon. Undetected leaks during starting can be a particular problem. (iv) IPs have magnetic fields which can interfere with processes unless screened which is not always possible.
The ion pump may be operated with sorption pumps (some workers use a turbopump) for starting, to produce a completely contamination free system. The whole system including the pump can be baked, the baking temperature depending on whether the ion-pump magnets are removed.  A high vacuum valve, which increases outgassing load and reduces pumping speed,  is not necessary although this can be useful for isolation for faster starting during pump-down.

4. Example from a silicon processing research laboratory
A small stainless steel chamber 30 cm diameter by 50 cm long and used for SIMS analysis is attached through a Viton-sealed gate-valve to the central distribution chamber of a cluster-tool which processes 8 inch diameter silicon wafers. The gate-valve is of 'letter-box' design allowing the wafer to pass back and forth through a slot between the two chambers. The SIMS chamber is pumped directly by a 150 l/s ion pump with no high vacuum valve, the chamber and ion-pump being initially roughed from atmosphere through a small Viton-sealed valve connected via a foreline to a turbopump. A titanium sublimation pump is available and an additional 50 l/s turbopump is used to evacuate the SIMS ion-gun, but this pump is ineffective in pumping the chamber because of the constriction to gas-flow by the ion-optics.
After loading the wafer and closing the gate-valve the system is required to be pumped to less than E-9 mbar and pressures approaching E-10 mbar would be preferred. A fast cycle time is important to prevent a bottle-neck as processed wafers are passed from the other process chambers. The budget on this project is exhausted and only minor purchases can be considered.
What baking times and temperatures are required and are they consistent with all the requirements? Would the titanium sublimation pump overcome any limitations of the baking schedules?

The total area of the chamber with flanges and tubulation is about 6000 cm2. It is difficult to calculate the area of the internal parts which contain many gas traps in the form of bearings and screw threads of manipulators, several sets of bellows and in addition the ion optics and detector components of the surface analysis system. For the sake of the example we will assume a total area of three times the above, that is 18000 cm2. For a preliminary estimate, assume a 20 hr bake at 150C which creates no problems for the Viton seals provided the valves are open (this assumes the distribution chamber is open during the bake which may not be desirable). Using Equation 1, and Santeler's values above for Q, then:    

p  =  6.3 * E-11 * 18000/150  Torr  =  E-8 mbar

This is clearly inadequate for the application and too high to make worthwhile use of the titanium sublimation pump. To achieve a pressure just an order of magnitude lower would require baking for 20 hrs at 250C, so that for the required lower pressures, baking at 300C would probably suffice. The difficulty is protecting the Viton seals. The options available are: (i) Cool the seals. This is difficult and may be impossible to engineer with the geometries and existing construction of the gate-valve although the small roughing valve could cooled without difficulty. (ii) Replace the valves with metal sealed versions. In this example the budget does not allow new capital purchases, moreover, a metal sealed UHV gate-valve which would pass an 8 inch diameter wafer would be of high cost although the small roughing valve could possibly be replaced with a small conflat sealed valve. (iii) Use larger pumps. The example shows that a pump of considerable more than 1500 l/s would be required and again this is beyond budget. (iv) Bake for longer times. This is not acceptable, even 20 hrs is too long for process turn-round requirements. The table above shows that in any case the gains from longer bakes are only minimal.
One is left with a compromise situation; unless residual vacuum requirements are relaxed, one must try and bake the chamber at the highest temperature possible while relying on the large mass of the distribution chamber to keep the gate-valve at an acceptable temperature to protect the seals. It is worthwhile to estimate the permitted fraction,x, of the area of the chamber metal baked at lower temperature in order to achieve required pressures. The final pressure will then be the sum of two components from metal outgassed at lower and at higher temperature. For the estimate we will ignore the temperature gradient and assume average values.
If x not too large then the contribution from baking the main chamber at 300C for 20 hrs assuming an outgassing rate of E-12 mbar l/s cm2,  is given to an approximation by:

p(300)  =  E-12 *(18000)/150  mbar                     Equ. 2

The contribution from baking a smaller area,  6000x cm2, at 150C for 20 hrs with an outgassing rate of 8*E-11 mbar l/s cm2 is given by:

p(150)  =  8*E-11 * 6000x/150  mbar                     Equ. 3

Adding  Equs. 2 and 3  yields:

p(300)  +  p(150)  =  1.2 E-10  +  3.2 E-9x             Equ. 4

The second term in Equ. 4 equals the first when x  is about 4%, yielding a pressure of 2.4E-10 mbar, if x is allowed to increased to 10%, the total pressure increases only to 4.4E-10 mbar, which is perfectly acceptable for the application. This approach looks feasible, the example illustrates the initially assumed fact that outgassing is dominated by the internal parts and it is clearly important to ensure thorough degassing of the internal mechanisms when the gate-valve is maintained at lower temperature. Once the pressure is below E-9 mbar it is worthwhile to use a titanium sublimation pump for attaining lowest pressures.

5. References
Adams R O, "A review of the stainless steel surface,"  J Vac Sci Tech. A1 12 1983 pp. 12-18
Calder R and Lewin G, " Reduction of stainless-steel outgassing in ultra-high vacuum," Brit J Appl Phys.  (1967) pp. 1459-72
Dennis N T M, Colwell B H,  Laurenson L and Newton J R H, "Further developments with single structure vapour pumping groups,"  Vacuum 28 1978 pp. 551-8
Dylla H F,  Manos D M, and LaMarche P H, "Correlation of outgassing of stainless steel and aluminum with various surface treatments,"  J Vac Sci Tech. A 11(5) 1993 pp. 2623-36
Gay T J,  Brand J A , Fritts M C, Furst J E, Khakoo M A, Mell E R, Sieger M T, and Wijayaratna W M K P, "Clean ultrahigh vacuum system with single-structure diffusion pumps," J Vac Sci Tech A 12(5) 1994              pp. 2903-10
Huffine C L and Williams J M, Corrosion 16 1960 p. 430
Ishikawa Y, Koguchi Y and Odaka K, "Outgassing rate of some austenitic stainless steels," J Vac Sci Tech. A 9 1991 pp. 250-3
Ishikawa Y and Odaka K, "Reduction of outgassing from stainless surfaces by surface oxidation," Vacuum 41 1990  pp.1995-7
Ishikawa Y and Yoshimura T, "Importance of the surface oxide layer in reducing outgassing," J Vac Sci Tech. A 13(4) 1995 pp.1847-52
Ishimaru H, Itoh K, Ishigaki T and Furutate M, "Fast pump-down aluminum ultrahigh vacuum system," J Vac Sci Tech. A 10(3) 1992 pp. 547-52
Jost W, Diffusion in Solids, Liquids, Gases  (Academic, New York, 1960)
Santeler Donald J, "Estimating the gas partial pressure due to diffusive outgassing," J Vac Sci Tech A 10(4) 1992 pp.1879-83
Sullivan J, Schaffer S and King S, "Contamination control in the design and manufacture of gas flow components," J Vac Sci Tech A 10(4) 1992 pp. 1869-74
Tohyama et al., J Jpn Inst Met. 54  1990 p. 247
Tomari H, Hamada H, Nakahara Y, Sugiyama K and Ohmi T, "Metal surface treatment for semiconductor equipment: oxygen passivation, Solid State Tech. 34  Feb 1991 pp.S1-S5
Yoshimura N, Hirano H, Sato T, Ando I, and Adachi S, "Outgassing characteristics and microstructure of a 'vacuum fired' (1050C) stainless steel surface,"  J Vac Sci Tech A9 1991 pp. 2326-30
Young J R, "Outgassing characteristics of stainless steel and aluminum with different surface treatments," J Vac Sci Tech 6(3) 1969 pp. 398-400