Hoffman D.M., Singh B., Thomas J.H. (Eds). Handbook of Vacuum Science and Technology

Подождите немного. Документ загружается.

2.5.5 Cryogenic Pumps Specifically for Water Vapor 177

pump during regeneration can be irritating or toxic. Under certain process condi-

tions that are not well defined or limited, more efficient ozone generation takes

place. The likelihood for a spontaneous ignition or explosion during regeneration

increases due to the catalytic reaction of fine dust from the sputtered reactive

metal and liquid ozone. Cryopumps that accumulate oxygen or other reactive and

flammable gases should be regenerated as often as possible.

2.5.4.4 Ion Implantation

As previously discussed, ion implanters produce heavy hydrogen loads through

the decomposition of photoresist on the silicon wafers by the energetic ion beam.

Some additional hydrogen load may be added from the ion source

itself,

although

this is usually small by comparison. On an implanter with a high-vacuum pump

on the load lock, the regeneration schedule of the cryopump will be determined

solely by the hydrogen load. Other things being equal, the hydrogen generated is

proportional to the beam current or total dose of the implant. The hydrogen load

also goes up as the beam energy (voltage) increases. The impinging ion beam also

dissociates the photoresist into heavy hydrocarbons and other gummy by-products

that gradually migrate into the cryopump. Implanters without high-vacuum pumps

on the load locks are subject to higher water vapor loads in the end station.

For these reasons, special cryopumps have been developed for implanters.

First- and second-stage arrays have been optimized for higher hydrogen speeds

and capacities. Unique regeneration routines for purging out all the hydrogen and

driving out water and photoresist by-products are used. Implanter pumps pre-

sent special challenges during overhaul and maintenance procedures because the

accumulated deposits in the pumps may contain hazardous amounts of arsenic,

boron, or phosphorous. Pumps returned for service should be carefully packaged

and clearly marked to indicate the presence of hazardous materials. Generally, ar-

ray parts from these pumps should be discarded under strict hazardous material

controls, and cleaning fluids (water and solvents) must also be treated.

2.5.5

CRYOGENIC PUMPS SPECIFICALLY FOR WATER VAPOR

A cold surface held in the range of 105-135 K has very high efficiency for con-

densing water vapor. At these temperatures, it will condense no other normal

gases,

except heavy hydrocarbon vapors. Using only a single-stage refrigerator

operating at about 107 K, several alternate designs are available to provide eco-

178

Chapter 2.5: Cryogenic Pumps

Rg.l2.

ao4

An in-line water

pump

configuration with annular condensing ring. (Line drawing of On-Board

waterpump courtesy of CTI-Cryogenics.)

nomic, high-speed auxiliary pumping of water vapor in vacuum applications. In

some semiconductor processes, such as etching, reduction in water vapor has re-

sulted in lower particle counts on wafer surfaces.

For process chambers on load-locked systems, the refrigerator may be mounted

on the outside wall of the chamber with the cold tip of the cylinder passing

through a small-diameter hole in the chamber wall. A copper or aluminum plate

within the chamber is attached to the refrigerator heat station. With a net water

pumping speed of 10-20 liters/second per square centimeter of plate area, the

size of the plate is matched to the heat load it will generate on the refrigerator. A

thick plate has lower temperature gradients and can therefore pump effectively at

higher average temperatures, although a thinner plate will cool down faster.

The in-line water pump configuration, as shown in Figure 12, uses a cold an-

nular ring in front of a turbomolecu.ar pump. Since turbopumps are relatively

expensive for the pumping speed they provide, it becomes economical to add a

single-stage cryopump for water. Then high water vapor speeds are obtained for

process cleanliness, and adequate speeds for process gas still result. The water

pump array configuration produces little loss in oxygen, helium, or hydrogen

speed of the turbopump while collecting 95% of the incident water vapor. A

turbopump/waterpump combination is useful for applications where the process

gases may not be suitable for a standard cryopump.

2.S.6 Comparison of Cryopumps to Other Types of Pumps 179

2.5.6

COMPARISON OF CRYOPUMPS

TO OTHER TYPES OF PUMPS

2.5.6.1 Benefits of Cryopumps

There are a number of benefits from using cryopumps.

SPEED

Cryopumps have the highest speeds for any given inlet diameter. For water, cryo-

pumps have essentially 100% of the maximum theoretical pumping speed. For ar-

gon, oxygen, nitrogen, and other Type II gases, they are about 40% efficient. Hy-

drogen efficiencies of cryopumps are about 18-25% of theoretical maximum.

This compares well to the 8-13% efficiencies of turbopumps for all gases and the

20-25%

efficiencies of LN2 trapped diffusion pumps for all gases except water

(—100% water efficiency of the trap). Ion pumps achieve about 20% for active

gases but only 5% for argon and other neutral gases.

COST

They have the lowest cost in liters/second per dollar invested. Although a basic

diffusion pump is cheap, the addition of a trap and control systems raises the cost

substantially, and operating costs are high for power and cooling water. Turbo-

pumps are expensive to buy for the equivalent pumping speed, and the required

inlet diameter is larger. To achieve complete freedom from oil in a turbopump, a

magnetic levitation system is required, doubling the cost of the turbopump. The

turbopump also requires a backing pump. Ion pumps are not practical in size or

cost for high pumping speeds; they cannot be used in high-throughput situations.

CLEANLINESS

In the event of worst-case power failure, cryopumps can only put back into the

vacuum system what they took out initially. Cryopumps do not backstream oil,

and they cannot provide a path for oil from a backing pump.

RELIABILITY

Cryopumps are highly reliable. The relatively simple mechanical parts of a pump

run at very slow speeds, so they do not wear out rapidly. Most pumps will run two

180 Chapter 2.5: Cryogenic Pumps

to five years without maintenance. Overhaul costs are a small fraction of the ini-

tial purchase cost. In the event of a mechanical failure, none of the parts are in-

jected into the vacuum system.

ORIENTATION

Cryopumps can run in any orientation. Cryopumps operate the same whether they

are flange side up, flange side down, or sideways. There is no problem with ori-

entation as in oil-lubricated or magnetically levitated turbopumps, or as in diffu-

sion pumps. You may, however, want to think about where liquids go when the

cryopump regenerates.

2.5.6.2 Limitations of Cryopumps

As good as cryopumps are, they do have some limits.

REGENERATION

Cryopumps do need to be regenerated. The capacities for argon and nitrogen are

high enough to get through a week or so of heavy process gas flows, but then one

to three hours must be devoted to regeneration. The same is true for ion implanters.

Some applications have such high gas flows that daily regeneration is required.

HEAT CAPACITY

They have limited heat capacity. Processes with elevated temperatures in the pro-

cess chamber may overheat the cryopump. Thermal shielding may reduce the net

pumping speed to where another type of pump is more practical.

PROCESS GASES

Cryopumps cannot pump all processes. Systems that produce high gas loads of

hydrogen or helium are poor candidates for cryopumps. Processes that use highly

toxic,

flammable, or other dangerous gases in large quantities should not use cryo-

pumps because of concerns for operator safety. This includes arsine, phosphine,

silane, and methane, and ozone, among others. Vacuum ovens and furnaces that

load oily parts should consider another type of pump.

2.5.7 Future Developments

181

2.5.7

FUTURE DEVELOPMENTS

The last

or 10 years have brought significant improvements in cryopump speeds,

capacities, and reductions in regeneration time. Automatic control and com-

munication functions have been added to improve integration of the cryopump

with the rest of the vacuum system. Continued advances in all these areas can be

expected.

Control of the speed of the drive system of the cryopump promises to bring a

number of benefits. At more rapid stroking speeds, the refrigerator gains effi-

ciency. Significantly shorter cooldown and refrigeration times result as the stroke

rate is doubled or tripled, if somewhat higher vibration levels can be tolerated

during this period. Because this speedup occurs only for a short interval, little re-

duction in overall life occurs. At slower than standard speeds, gas utilization from

the compressor is improved and vibration levels are reduced.

Development of microprocessor controls incorporated into the cryopump, and

also distributed in cryopump network controllers, has already enabled pumps to

perform regeneration processes that are not practical with manual controls. Fur-

ther integration of these capabilities can allow the efficient brokering of system

resources, such as rough pumps, between load-lock rough-pumping and cryo-

pump regeneration. The ability to sense and control more parameters of pump op-

eration allows analytic and diagnostic functions to be built into the pump. Needs

for maintenance can be predicted and communicated to the system operator, ei-

ther locally or remotely.

It is also becoming more conmion for cryopumps to be purpose-built for spe-

cific processes and even for specific vacuum systems. Various performance para-

meters can be traded off to obtain the best pump characteristics for a particular

process. For these reasons, system designers and pump specifiers need to under-

stand the operating characteristics of cryopump systems to obtain the best level of

performance at the right cost.

REFERENCES

J. Dewar, Proc. Roy. Inst., 18 (1907) 747-756.

W. E. Gifford and H. O. McMahon, Prog. Refrig. Sci. Technol, 1, eds. M. Jul and A. Jul (Perga-

mon, Oxford, 1960), p. 105.

J. F. O'Hanlon, "A User's Guide to Vacuum Technology" (2nd ed.) (Wiley-Interscience, New

York, 1989), p. 410.

M. A. Paul, Physical Chemistry (D.C. Heath, Boston, 1962), p. 205.

R. E. Honig and H. O. Hook, RCA Review

21,

(1960) 360.

182 Chapter 2.5: Cryogenic Pumps

6. P. A- Lessard,

Vac.

Sci.

TechnoL,

A7 (3), (1989) 2373.

7. R. C. Weast, ed., "CRC Handbook of Chemistry and Physics —

66th

edition" (CRC Press, Boca

Raton, Florida, 1985).

8. R. B. Liebert, L, Ficara, R. Eddy, W. Ghen, and P. A. Lessard, Ion Implant Technology—94, ed.

S. Coffa, G. Ferla, F. Priolo, and E. Rimini (Elsevier Science B.V., Amsterdam, 1995), p. 548.

9. D. H. Holkeboer, D. W. Jones, F. Pagano, and D. J. Santeler, "Vacuum Technology and Space

Simulation" (American Institute of Physics, New York, 1993), p. 184.

10.

M. J. Drinkwine and D. Lichtman, Partial Pressure Analyzers

and Analysis

(monograph) (Amer-

ican Vacuum Society, New York, 1979).

11.

G. S. Ash, 37th

Ann.

Technical

Proceedings, Soc. of

Vac.

Coaters (Albuquerque, 1994), p. 1(X).

CHAPTER

2.6

Turbomolecular Pumps

Hinrich Henning

Leybold Vakuum GmbH

2.6.1

TURBOMOLECULAR PUMPS (TMP)

The turbomolecular pump (TMP) invented by Becker [1] in 1957 became com-

mercially available in 1958. Since then it has become very popular in every field

of high- and ultra-high-vacuum technique, due to the clean, consistent, and pre-

dictable vacuum created, the easy operation, and the advanced degree of operat-

ing reliability. The TMP is the only mechanical vacuum pump that together with

a roughing pump can attain ultimate pressures in the range below

10 "^^

torr.

2.6.U General Information

PRINCIPLE OF OPERATION

The turbomolecular pump is a bladed turbine that compresses gases by momen-

tum transfer from the rapidly rotating blades of the rotor disks to the gas mole-

cules.

The rotor impulse is transmitted to the particles by the superposition of the

thermal velocity of colliding particles with the velocity component of the mov-

ing rotor surface. The nondirected motion of the particles is changed to a directed

motion, creating the pumping process. When the mean free path of the particles

is larger than the spacing between rotor and stator blades (molecular flow range,

*^^-^ All rights of reproduction in any form reserved.

183

184 Chapter 2.6: Turbomolecular Pumps

typically <10~-^ torr), particles collide primarily with the rotor, resulting in an

efficient pumping process, and there is no interacting influence of the different

gases.

In the laminar flow range (typically >10~^ torr), the action of the rotor is di-

minished by the frequent collisions between the particles. Therefore, normally, a

TMP is not capable of pumping gases against atmospheric pressure and must be

backed by an adequate roughing pump.

HISTORY OF TMP

Historically, the development of the TMP goes back to the year 1913, when

Gaede introduced his "molecular drag pump". The "early" molecular pumps

(Gaede [2] "molecular drag pump", 1913; Holweck [3] "dual-flow molecular

pump", 1923; Siegbahn [4] "disk-type molecular pump", 1940) were never really

successful because of their relatively low pumping speed and their questionable

reliability, and furthermore, in those days there was no real demand for these

pumps. To attain low ultimate pressures with these pumps, the clearances be-

tween rotating and stationary parts were made a few hundredths of a millimeter.

Therefore, any change in temperature or intruding solid particles could result in a

failure of the pump, caused by a seized rotor.

Recently, however, the basic ideas of Gaede (disk) and Holweck (drum) were

used successfully in the design of modern pumps (molecular drag pumps; combi-

nation pumps), in order to attain extremely low pressures and/or to make use of

simple dry roughing pumps.

The Becker design ("turbomolecular pump") avoided these disadvantages: The

pump is composed of a series of disks with a row of blades, alternately fixed and

moving. In each disk the blades are inclined with respect to the disk plane, in one

direction for the blades of the moving disks and in the other for the blades of the

fixed disk. The moving disks have a high rotational speed, so that the peripheral

speed of the blades (up to 500 m/s) is of the same order of magnitude as the speed

of the molecules of the pumped gas.

The distances between these disks are in the range from several tenths to a few

millimeters. The channels between the inclined blades of the disks act like ele-

mentary molecular pumps, similar to the Gaede molecular drag pump. All chan-

nels on one disk, «20 to 50, are connected in parallel and together can yield a

high pumping speed up to several thousand L/s.

2.6.1.2 Theoretical Considerations: Performance Data

Fundamental theoretical work was performed by A. Shapiro [5].

The main performance data of

TMP, the compression and the pumping speed,

can be calculated from the data of single rotor and stator disks. A rotating disk

2.6.1 Turbomolecular Pumps [TMP]

185

with blades pumps molecules from one side to the other, and some particles move

opposite to the primary flow direction through the blades.

The structure of most common TMPs is shown in Figure 1 [6]. Distribution

for the probability of particle direction from the rotor coordinate system is gov-

erned by the cosine law. As seen from the moving rotor this distribution will

be deformed to a nearly elliptic shape by subtracting the constant rotor speed

Vb = — Vf. Figure 1(a) shows clearly that particles coming from the high-vacuum

side will preferentially pass the rotor structure without collision with the blades.

Particles coming from the discharge side will preferentially strike the "lower"

side of the blades, where they are desorbed again but now with an undeformed

cosine distribution, i.e., less than 50% of them to the high-vacuum side. The ratio

Fig.l.

Pumping process.

(a)

(a) Probability of molecule's flight direction, as seen from the mov-

ing rotor coordinate system. The cosine distribution is deformed by

the constant speed V, of the gas.

HV low-pressure side V,„ average molecule velocity

VV high-pressure side

— Kb

relative gas velocity, as seen

blade velocity from the rotor coordinate system

(b) The transmission of molecules desorbing from the front side of a

rotor blade [6] (courtesy of SFV):

Molecule C coming from the low-pressure side

Cj,

C2,

angle for transmission, sorption on rotor surface, rejection

Molecule D coming from the high-pressure side

d|, d2, d3 angle for transmission, sorption on rotor surface, rejection

186 Chapter 2.6: Turbomolecular Pumps

of particles coming from the high-vacuum side to the discharge side will be

clearly higher than in the opposite direction. As for particles striking the blade

surfaces. Figure 1(b) shows

A molecule coming from space 1 collides with the blade of the pump rotor;

the probability of returning to space 1 is proportional to the angle c^ and is

smaller than the probability of rebounding on another blade c

or to be trans-

mitted into space 2

(C3

probability).

The situation is quite different for a molecule coming from space 2, which

has a lower transition probability of crossing the blade structure than a mole-

cule coming from 1. The pumping speed of the pump is the net difference be-

tween the molecules 1-^2 and the molecules 2 -> 1.

To calculate the performance of a single disk, we must know two extreme val-

ues,

the maximum compression

K,„ax

(^^ zero flow) and the maximum pumping

speed

S„j„^

(for equal pressure on both sides of the disk). These values depend on

the transmission probabilities of the particles moving in and opposite the pump-

ing direction. For the molecular flow, these probabilities depend on the ratio v/u

of the blade speed v and the mean thermal velocity u of the gas particle, the blade

angle a, but not on the pressure.

K^^ax

of a TMP is exponentially dependent on

the blade speed v, a pump-specific factor g (rotor and stator geometry), and the

square root of the molecular weight M of the particles pumped:

K,„,,'-oxpiM'^^vg) (1)

Sf,jax

is proportional to the product of a pump-specific factor G (rotor and stator

geometry) and the blade speed v. It does not depend on the pressure and the

species of gas pumped:

S,„ax

~VG (2)

A disk in operation works between these two extremes. For real pumping con-

ditions (gas throughput Q = S

- Piniet)

and different pressures on both sides of the

disk {K =

PoutieflPiniet)^

^^ foUowiug relation exists between compression K and

real pumping speed 5 of a single disk:

„ _ „ ^max ~ ^ _ ^ ^max " ^^^v ^ ,^.

^ " ^max j^ _ ^ ~" ^max jr _ ^ ^ w/

'^niax ^ ^max ^

More exact calculation takes into account the gap between the rotating disk's

outer diameter and the inner diameter of the stator housing and the variation of

the blade geometry with the radius. However, the basic relation shown in Equa-

tion (3) remains valid.

A TMP is assembled from several disks in series with different blade geome-

tries.

Each disk can be regarded as a separate pump. If the pumping speed S,, of

the next disk downstream is known and the throughput is constant, K in Equa-

tion (3) can be replaced by S/S^, and beginning with the pumping speed S^ of