Académique Documents
Professionnel Documents
Culture Documents
C. M.
VAN ATTA
>
Low Pressure;
0)
Vacuum Measurements;
Design and Operating Features of
Q m z o m > z
Jouge""
'"i
^"^^
^"^ ^^^P
Gate valve
m z o z m m
^z
>^
Liquid -nitrogen
trop
3
12
Iritefcorrectmg valve
Forevacuum
Roughing
mechanical
pumptrops
Thermocouple
o
55
gouge
Diffusion pLtrnp^^,
.--ll
621.
X-
VAN
McGraw-Hi(l_
Diffusion-pump-''
torelme trap
connpressof
-Baching pump
'-
Roughing pump
66854
C.
M.
VAN ATTA
Vacuum Technology
Consultant on
London
Sydney
FOREWORD
b6->lSi*i2**
,,._.
*^*---.--.^.<.^
HARRIS
'X
industry with The Design for High Vacuum Systems in 1945 and a The presence of many revised edition under the same title in 1958.
of these worn, dog-eared manuals on engineers' and scientists' desks, as well as the great demand for new and replacement copies, stands as testimony to Dr. Van Atta's success in meeting the needs of practitioners in the field.
6 '3-^
'>2r
jifi^
c^
797%> - --
^ '^
By comparison,
is
far superior to
prior book.
We
Vacuum Science and Engineering and is certainly much more comprehensive than the believe that Dr. Van Atta has achieved in this new
this
new
effort
and up-to-date coverage of his subject, which should again meet the needs of the industry and become a standard text and reference for all those who wish to study or practice in the
writing a complete
field
of
vacuum
science
and engineering.
been a privilege for the Kinney Vacuum Division to encourage and support this work. It is with admiration and respect for the eminently qualified author that we submit this book for your use.
It has
J. E. Chappell, General
Manager
Number 65-17497
66854
23456789-MP-9876
PREFACE
Over the past forty years vacuum technology has evolved from an
incidental but essential tool of scientific research to a rapidly growing
branch of engineering. In the 1930s the principal engineering application of vacuum technology was in the manufacture of light bulbs and radio tubes, for which processes ingenious equipment was developed largely by an empirical approach to the problems of evacuating and surface conditioning. The transition from a research tool
to
engineering application was greatly accelerated during World II, particularly by the multilateral attack on the release of atomic energy by the Manhattan Project. Many divisions of that
War
and greater
capabilities
previously.
Subsequently the
and
metal degassing made further engineering applications of the vacuum technology developed during the war. More recently the requirements of controlled thermonuclear research and space simulation have converted to an engineering scale the techniques of ultrahigh vacuum which had previously been applied only to small scale laboratory experiments.
The process of evolution and growth of vacuum engineering is by no means complete. Requirements in many fields of research and
materials processing are even now inadequately met, either because the desired vacuum conditions cannot be reliably attained, or because the cost of doing so is excessive. Improved methods of vacuum
pumping, surface degassing, and the measurement of low pressures are needed to meet these present requirements. The role of sorption (adsorption, absorption, and chemisorption) on surfaces is imperfectly understood, so that significant further progress will depend upon a concerted experimental and theoretical effort to understand the basic phenomena involved in the interaction between gases and surfaces at low pressure.
Vlll
PREFACE
In Vactmm Science and Engineering the objectives are to give in form the scientific basis of vacuum technology, to describe in some detail the performance characteristics and limitations of vacuum pumps, gauges for measuring gas pressure, and other components of vacuum systems, and finally to provide design criteria in sufficiently general form to be useful in designing vacuum systems for a wide range of applications. Throughout the text an effort has been made to describe in some detail the physical processes which determine the operating features of the various devices which are discussed. The object in doing so has been to give the reader not merely a catalogue of typical vacuum components and perform^ance data, but in addition a basis for judging the importance of various phenomena which occur in vacuum systems. It is my belief that only by this approach can one provide guidance for the optimization of the design of vacuum systems for a variety of uses. Aside from my own experience in large scale experimental research and industrial vacuum development, I have drawn heavily upon the expanding technical literature dealing with vacuum technology. The emergence of the published proceedings of the American Vacuum Society and its predecessor organizations, and those of the International Organization for Vacuum Science and Technology, as well as such journals as Vacuum (Pergamon Press, London), Le Vide (la Societe Francaise des Ingenieurs et Techniciens du Vide, Nogentsur-Marne (Seine) France), and Vakuum-Technik Springer- Verlag OHG, Berlin) has greatly eased the task of locating literature on new developments in vacuum technology. With the kind cooperation of Dr. J. H. Leek, I have found his excellent book. Pressure Measurement in Vacuum Systems (published for the Institute of Physics and the Physical Society by Chapman and Hall, Ltd., London), most helpful in writing Chapter 3 of the text. It is with deep appreciation that I acknowledge the incentive and support provided by the New York Air Brake Company for undertaking the task of writing a book of this character. The critical comments of R. R. Cyr and Z. C. Dobrowolski of the Kinney Vacuum Division of the company contributed significantly to the final version, particularly of Chapters 5, 6, 7, and 8 of the text. I am deeply indebted to Miss Margaret R. Thomas, who not only typed the manuscript with its many revisions, but also maintained order in the growing lists of references, permissions, and credits. I also wish to thank the many authors, publishers, and vacuum equipment manufacturers who have responded so generously to requests to use illustrative material and who, in many cases, have provided the glossy prints necessary to reproduce photographic illustrations. C. M. Van Atta
fairly classical
(
CONTENTS
Foreword
Preface
v
vii
Chapter
1-1. 1-2. 1-3.
1.
......
.
of
Gases
1
Molecular Constitution and Kinetic Theory of Gases Pressure Related to the Average Molecular Kinetic Energy
4
5 8 11
1-4.
1-5. 1-6. 1-7. 1-8. 1-9.
The Maxwell-Boltzmann Distribution Law Velocity of Sound in a Gas Flow of Molecules through a Hole Molecular Mean Free Path
12
13
Van
15
18
BEFEBENCES
......
Chapter
2.
.
22
Gas Flow
23
23
2-1.
2-2. 2-3.
2-4. 2-5. 2-6.
Gas Flow
in
Vacuum Systems
Pumping Speed and Conductance Viscous Flow Poiseuille's Law Pressure Drop Formula Turbulent Flow in Vacuum Systems Correction to Poiseuille's Law Due to Surface Slip Gas Flow in the Transition Pressure Range Gas Flow at Low Pressure Conductance of a Long Tube at Low Pressure
26
30
31
34 36
....
43 44
47
2-10.
2-11.
2-12. Clausing
2-13.
and Monte Carlo Corrections to the Knudsen Conductance Formulas Summary of Gas-flow and Conductance Formulas
RErEBENCES
t^*
Chapter
3.
Pressure Measurement in
3-1. 3-2.
for
End Effect
49
51
57
62
Vacuum Systems
63 65
X
3-3. 3-4.
CONTENTS
The Dubrovin Gauge The McLeod Gauge Thermal Conductivity Gauges Hot-cathode Ionization Gauge The Bayard-Alpert Ionization Gauge Hot -cathode Magnetron Ionization Gauge
.
CONTENTS
6-5.
XI
Pumping Speed
of Diffusion
69
78
6-6.
3-5. 3-6.
90
103 107
Limiting Forepressure for Diffusion Pumps 6-7. Factors Contributing to the Ultimate Pressure of a Diffusion 6-8. Fractionation and Purging
6-9.
)C_3-7.
3-8. 3-9.
Beam
.
Ionization
Gauge
111
Gauges
3-11. The Alphatron Gauge ^3-12. The Knudsen Radiometer Gauge 3-13. Calibration of Vacuum Gauges 3-14. General Remarks on Ambiguities of Pressure Measurement in Vacuum Systems
....
Chapter
7.
KEFBBBNCES
.... ......
128
131
jfJ:"^-
Pumping Speed Measurement of Gas Flow Mechanical Pump Speed Measurements Measurement of the Pumping Speed of Diffusion Pumps RBFBBENCES
Alternative Definitions of
. . .
249 254
257
Pump
The Measurement
of
Pumping Speed
....
274 277
291
293 302
Chapter
Detectors
133
143
8.
The Design
Chapter
4-2. 4-3.
4.
Vessel
Seals
.........
of
Vacuum Systems
303 307
4-1. Magnetic-deflection
4-4.
4-5.
Mass Spectrometers The Omegatron Mass Spectrometer Linear High-frequency Mass Spectrometers Halogen Leak Detector
Leak-detection Techniques
313 318
Vacuum
Vapor
Valves
BErBBBNCES
Chapter
5-1. 5-2. 5-3.
5-4.
Functions of Mechanical Pumps General Features of Oil-sealed Mechanical Pumps Pumping Speed of Oil-sealed Mechanical Pumps The Effect of Condensable Vapor upon Mechanical
.
..... ......
.
328 341
348
8-8. Selection
5.
Mechanical
Vacuum Pumps
358
.
362
Chapter
9.
Ultrahigh
Vacuum
ance
5-5.
5-6.
Gas Ballast
......... .........
Pump
.
Perform
9-3
..... ....
.....
.
. .
363
Other Methods of Preventing Contamination by Condensables 5-7. Mechanical Booster Pumps 5-8. Analysis of Mechanical Booster-pump Performance 5-9. Computed Performance Curves for Mechanical Booster Pumps 5-10. Measured Performance Curves for Mechanical Booster Pumps 5-11. Overheating of Mechanical Booster-pump Rotors 5-12. Vapor Compressor Action of a Mechanical Booster Pump
5-13. Molecular-drag
......
9-6
9-7
Evaporative Deposition of Reactive Metals 9-9 Cryogenic Pumping 9-10 Ultrahigh-vacuum Systems
9-8
Pumps
RErBBENCBS
BEFEBENCES
APPENDIX APPENDIX
APPENDIX
I
Chapter
,
Vapor-jet
Vacuum Pumps
219 227
APPENDIX IV APPENDIX
Author Index
Suhj ect Index
6-1.
6-2. Diffusion
Pumps
6-3. Theoretical
6-4.
Pump
Working Fluids
Pumps
230 240
..
III
.
.
II
..... .....
. .
440
441
442 445
447
451
In some cases it has not been convenient to avoid the use of a symbol for more than one purpose. The most prevalent meaning of each commonly used symbol is defined in the following list. Exceptions are clearly indicated
in the text.
A B
c
area
C
Cj,
conductance
specific specific
C^
D
e
E
/
h,
R Ho
ij^
i_
electron current
electrical current
1
Ic
/v
conductance factor
length
mass of molecule
molecular weight
M
n
Wmoi
N
p
number of molecules per unit volume number of molecules in one mole total number of molecules present
probability of ionization
pressure
gas throughput,
PdVjdt
w,,,^]/^
R
Rg
Re
s
Reynolds number
sensitivity
pumping speed
displacement speed of a mechanical
pumping speed
t
at the inlet of a
pump pump
time
T
u
temperature
drift velocity of
a gas
U
V
velocity
velocity
F w
W
z
Z
a
ion
y
e
r]
GJC^
viscosity
mean
free
path
heat conductivity
A
V
free molecular
number
one second
I
molecular diameter
density,
p a
T
period
CHAPTER
1-1.
of permanent gases
Our understanding of the behavior based upon the experiments of Boyle, Charles,
to the general gas law. Experiments by Boyle resulted in the conclusion that the volume of a body of gas at constant temperature is inversely proportional to the pressure, which
is
PV =
const
(1-1)
where the pressure is defined as the force per unit area exerted by the gas on the walls of the containing vessel. Charles and Gay-Lussac observed that if the volume of a body of gas is kept constant and its
temperature varied, the pressure increases linearly with the temperature, so that
^1
in
-Po(l
aT)
(i_2)
which T is the temperature on any chosen scale, such as centigrade F^is the pressure of the body of gas at zero on the same temperature scale, and is a constant. If Eq. (1-2) is multiplied by V^, the initial standard volume of the gas sample,
PoFo(l
ocT)
(1.3)
PV =
which can be written as
PF(1
aT)
(1.4)
PV = P.V^T +
The experimental
fact
is
1/a)
(1.5)
the temperature is measured on the 273.I6C, that is, the volume of a body of
if
that
^^^ changes by an amount equal to 1/273.16 of its value T^^o^''?* at C for each degree change in temperature. This constant is essentially the same for a large number of gases (hydrogen, helium
VACUUM SCIENCE AND ENGINEERING and others) and therefore has very broad signifione chooses a new temperature scale such that 273.16 C zero, then one can write
If
nitrogen, oxygen,
cance.
is
PV = PoFoaT
where now the temperature
or Kelvin, scale.
is
(1-6)
The implication
at constant
of Eq. (1-6)
by a gas
0K.
volume approaches zero as the temperature approaches Although many common gases follow Eq. (1-6) over a wide
Table
1-1.
temperature and pressure the mass of a standard volume of gas is proportional to its chemical combining (or molecular) weight. Consistent with Avogadro's law, precise experiments have shown that under standard conditions of temperature (0C or 273.16K) and pressure (normal atmospheric pressure defined as 760 torr) 1 gram molecular weight of any gas occupies a volume of 22,415 cm^. This is the volume occupied by 32.00 g of oxygen (0^) at STP standard temperature and pressure, which is the arbitrary standard on the chemical scale of molecular weights. A partial list of molecular weights of
given in Table 1-1. A more comprehensive table of molecular weights of gases is given in Appendix I. By referring to Eq. (1-7) we can now write the general gas law in terms of the molecular weight of the gas, as follows:
gases
is
common
some
Chemical formula
Ha He Xe
^'2
^^=|^oT
where
(1-8)
Air (mean)
O2 Ar
CO2
CL
* See, for
70.91
molecular weights) of the gas present. The numerical value of E^ depends upon the units of mass, pressure, volume, and temperature used. If the pressure is measured in torr, the volume in liters (1 hter = 1,000.027 cm^), and the temperature in degrees Kelvin, then for 1 mole (W/M 1) of gas
= universal gas constant per mole W/M is the number of moles (gram
from this relationship at sufficiently large values of the pressure and low values of the temperature. Thus only for an ideal gas would the pressure actually approach zero as the temperature approaches absolute zero.
all
PV = R,T
Under standard conditions
(i-sa)
Returning to Eq. (1-6), since at constant temperature the product PV for a given body of gas is constant for a given mass of gas, and
since twice the
volume
not only to the absolute temperature but also to the mass of the body of gas W, and we may
PF is PF in
same pressure contains twice the mass, proportional to the mass of the body of gas. Thus
at the
so that
E(,
P= F = T = = PV
T
760 torr
22,415/1,000.027
22.415
liters
273.16K
0C
760 X 22.415
273.16
Eq.
(1-6) is proportional
PV =
P,V^T
WRT
(1-7)
where
i? is a constant of proportionality. Equa,tion (1-7) is one form of the general gas law which describes the behavior of an ideal gas
the numerical value of E^ is given in the mass of a body of gas is of no concern, but the changes in pressure, volume, and temperature are of interest. In this case a convenient form of the general gas law is
For the
Table
1-2.
PiFi
which follows directly from Eq.
P2F,
(1-9)
(1-8), since for
and
is
many common
range of practical conditions. Further understanding of the nature of gases was contributed by Avogadro, who demonstrated experimentally that at the same
PVjT
a given
body
of gas
is
a,
constant.
Table
KuMERicAL Values of Bg Gas Constant peb Mole fob Vabious Systems of Units* f
p
dynes/cm^ newtons/m^
torr torr
V
cm^
T
"K K K K "K
"R
i?
molecules present in 22.415 liters of any gas under normal conditions. The number of molecules in a unit of volume of gas under normal conditions is therefore
ti,
mS
cm^
liters
6.023 X 1023
^^^Y5
It
^-^^^ ^
^^" molecuIes/cm3
atm
psi
cm'
ftS
torr
cm'/K
torr liters/K
atm cm'/K
lb ft/R
is worthwhile pausing to note the magnitude of this number. Its meaning can perhaps be visualized best by noting that if the molecules in a cubic centimeter of gas under standard conditions were arranged
number of such
10i8)'/iS
cells in
a centi-
mole of gas occupies 359 ft' at 32F and atmospheric pressure (14.67 psi). The Bankine absolute temperature scale is based upon the Fahrenheit scale for which absolute zero temperature is 459.69F. Thus T R = T F + 459.69 just as T K = T "C + 273.16. Physical Tables (Smithsonian Instit Sources: W. E. Forsythe, Smithsonian tution, Washington, D.C., 1954, 9th rev. ed.; T. Baumeister (ed.), Marks' Mechanical Engineers' Handbook (McGraw-Hill Book Company, New York,
*
In engineering units,
lb
lO^")!^ <^
(27
10"
sothatthedistancebetweenmolecules would be about 3.3 x 10"' cm, which is approximately the average distance between molecules for a gas under standard conditions.
Actually, in a gas the molecules are not arranged in a simple pattern but are moving randomly relative to one another. During their motion the molecules suffer collisions between themselves and bounce off one another. The average distance which the molecules move
1-2.
From the time of the Greek philosophers is made up of tiny indivisible particles
.
sporadically put forward to explain one or another of the observed On the basi s of the experimental results reported properties of matter by a number of independent investigators, Avogadro concluded that
between collisions is called the mean free path (A). The molecules impinge on the walls of the confining vessel and bounce back (or are reemitted) from the wall. The momentum transfer from the
also
equal volumes of all gases under the same conditions of temperature and pressure contain equal numbers of molecules. We have already seen that under standard conditions (760 torr pressure and 0C) a gram molecular weight of any gas fills a volume of 22.415 hters. The number of molecules contained in this standard sample of gas is obtained from the precise measurement of the faraday,
molecules to the walls of the vessel produces the outward force exerted by the gas upon the walls of the vessel. This force per unit area is precisely the pressure which appears in the general gas law
discussed in the previous section. The molecules of an ideal gas may be considered as a mation to be elastic spheres about 10"* cm in diameter.
first
approxi-
The random
F =
96,488 coulomb
the electrical charge necessary to deposit a gram equivalent of a substance in electrolysis, and the charge on an electron,
e
1.602
X 10-19 coulombs
The
96.488
1.602
10-1-
6.023 X 1023
is
known as Avogadro's number; it is the number of molecules in a gram molecular weight of a substance, and is therefore the number of
motion of the molecules consists of motion in all possible directions with various individual molecular velocities. As will be shown, the temperature of the gas determines the mean square velocity (i.e., the average value of the square of the velocity) for any particular type of molecule and the average kinetic energy for molecules of all types in a mixture of gases. The kinetic theory of gases consists of the statistical mechanical treatment of the microscopic molecular motions and leads to a basic understanding of the gross behavior of a gas in terms of molecular motions. 1-3. Pressure Related to the Average Molecular Kinetic Energy. Consider a box of rectangular cross section filled with a pure gas such that the density in the box is n molecules/cm', each molecule having a mass of m grams. The two walls of the box perpendicular to the x axis suffer collisions by molecules by reason of
their x
If the molecules in the box are moving in a random manner, half will be moving with positive v^ and half with negative v^. Those moving in the positive x direction
with a specific value of v^ will strike the wall in a time ht if they are contained in a sheet of thickness v^ dt adjacent to the wall. The number striking a unit area of the surface in the time dt is thus n^v^ ?)tj2, so that the number striking a unit area in a unit of time is
If each molecule on striking the surface of the contamer experiences an elastic collision, the magnitude of v^ does not change on collision with the wall but the direction reverses. The other components of velocity, f and v^, are not changed in the collision since the surface
proportional to the kinetic energy of the molecules contained in a volume nmv^j2. If the mass of the molecules is measured in grams and their velocities in centimeters per second, the pressure is measured in dynes per square centimeter. Comparing the expression for the pressure given in Eq. (1-11) with that of the general gas law given in Eq. (1-8), one concludes that
P =
If this expression
is
W R.T VM
2
^/snmv^
(1-12)
is
y^mv'
since
under consideration is parallel with the yz plane. The change in momentum of a molecule on striking the wall is thus 2mv^. If in the above discussion n^ is the number of atoms with x components of
then the number the range dv^ is in velocity the with striking a unit area each second for this group momentum of change of Therefore the rate n^v^ dvj2.
velocity in the range between v^
VM
=
n
K
(1-13)
2 Wmoi
Wmol
W
w
:
and
v^
dv^,
Introducing
- =
i?n
1.3805 X 10-ierg/K
(1-14)
known
of molecules
is
is
nv^ dv^
[2mv^
n.mvj
dv^
i.e.,
E =
Yzniv^
= %kT
(1-15)
For the entire distribution in v^, the total rate of transfer of momentum, and therefore the force per unit of area, is
f*CC
dv^
nmv^^
(1-10)
the average kinetic energy of a molecule is proportional to the temperature, and the energy associated with each coordinate direction (x, y, z) in its motion is therefore A;T/2. Combining (1-15) with (1-12) results in the important fundamental expression
Jo
P =
Similarly
%n(}4mv^)
34n(%kT)
= nkT
dynes/cm^
(/^bar)
where
v^^ is
the
mean
(1-16)
nmVy^
and
P,
nmv.
Since the motion is random there is no difference in the average motion in the various coordinate directions, so that
,/
The derivation of the preceding relationships from elementary kinetic theory does not depend upon the molecular mass of the gas involved. Remembering that from Avogadro's law equal volumes of gas under the same conditions of temperature and pressure contain equal numbers of molecules, it follows that for two different gases
y2miV^^
If
=
V^''
V,
=
V,
J^ma^a^
/4kT
(1-17)
we now
define
Also since
V^
=
in
VJ'
then
Vj
+ =
V,
(;2)'^
(j.jg)
}/3V^
any direction
P =
From
P^
P^
= P,=
y^nmv^
(1-11)
^=
[m-J
(1-19)
by a gas
is
inversely proportional different gases at the same temperatures are of their gram molecular to the square roots of their molecular masses, or
weights.
From Eq.
(1-11)
P =
exponential term is e = 1. It also goes to zero as v approaches infinity because of the dominance of the exponential term. For all values of v between zero and infinity /^ is positive and therefore must have a maximum value, which is the most probable value of the
velocity.
(1-20)
\nml
where p
=nm\s the
is
density of the gas in grams per cubic centimeter microbars in dynes per square centimeter or
1-4.
In
the
to inquire into discussion of the previous section it was not necessary might assume gas of a molecules the which the distribution in velocity the molecules if even that clear is It as a result of mutual collisions. the subvelocities directed randomly but of a gas were to have equal would quickly introduce a wide distribution
08
V
1.0
in
1.2
14 16
1.8
2.0
22 24 26
2.8
30
units of vp
in velocity.
and
after each
of the Since in an elastic collision the total kinetic energy is the same before m{v^^ quantity v^)l2 the preserved, molecular collision, even if v^ and v^ must change to
Fig. 1-1.
momentum.
By
of a large
number
^ ' remains
n
constant
to zero, the
and setting the result equal most probable value of the velocity can be determined.
4
/
of even though the individual molecular velocities change because also therefore and mutual collisions. The root-mean-square velocity the velocity the pressure exerted by the gas are independent of features Some collisions. distribution which may result from mutual equilibrium the of form of gas behavior, however, do depend upon the so that a distribution in velocities which gas molecules assume, complete a to essential knowledge of the velocity distribution is understanding of the properties of a gas.
Y"
dv
so that
^\2fcy/
kT ^
'
_ l2kTY
*
\
(1-22)
TO /
The
deduced by
gas was actual distribution in velocity of the molecules of a scope the beyond Maxwell and Boltzmann by arguments
The meaning of the most probable value of the velocity is that more molecules have this value than any other value of the velocity. The most probable velocity is not, however, the arithmetic average value of the velocity, which is calculated as follows
/oo
in of this text. The Maxwell-Boltzmann distribution law expressed is terms of the velocity in a randomly chosen direction
^
dv
l2kT\^
1.128t;.
(1-23)
dn
l.-mv'ltkT
(1-21)
\2kTj
Finally, the
mean square
velocity
is
of molecules in dv is the fracdnjn the velocity range between v and v + dv, bo that of velocity unit per tional number of molecules in this velocity range
is
number
range.
f^
plot of /^
is
goes to zero at v
shown in Fig. 1-1. The distribution function the because of the w^ term, since at
=:
Joo
= 3 kT
TO
(1-24)
IQ
11
without considerconsistent with Eq. (1-15), which was arrived at from Eq. (1-25) ation of the details of the distribution law. Also we have for the root-mean-square velocity
and
V,
= 1.128 X 4.09 X 10* = 4.61 x 10* cm/sec = 1.225 x 4.09 x 10* = 5.01 x 10* cm/sec
In general, the above basic factor may be written in terms of the mass of a molecule of unit molecular weight, m^ = 1.66 x 10-^* g
/2kTY_ /2kY( TY
average Which of the above velocities is of interest as representing the consideration. behavior of a gas depends upon the process under representative Table 1-3 provides a convenient summary of these to another. one from velocities and the numerical factors for converting
2
1.38
vA/ y X io-i\^/y Y^
1.66
10-2*
(
=
when T
is
T\A
cm/sec
(1-27)
1.29
X 10*177)
Table
1-3.
expressed in degrees Kelvin and in grams per mole. Sound in a Gas. A velocity closely related to those discussed above is the velocity of sound in a gas, which according to elementary texts on sound is given by
1-5. Velocity of
0.816
(t7
^av
(1-28)
2 /2J;TV^
1.128
0.921
in
1.225
which
(1-29)
1.086
many factors in Table 1-3 are so near unity that for between distinguish to approximate calculations, it is not necessary
The conversion
velocities. the most probable, the average, and the root-mean-square factors numerical with The basic factor (2A;T/m) appears in all three weight molecular average which are not very different from one. The in a gram molecules of of air (28.98 g/mole) divided by Wmoi, the number molecule air an molecular weight, gives for the average mass of
is the specific heat of a gas at constant pressure and G^ is the specific heat at constant volume. According to the kinetic theory of gases the value of y is simply related to the number of degrees of
where C^
freedom of the molecule in terms of the independent coordinates of motion (three represented by x, y, and 2), the axes of rotation, and the
modes of vibration. The result is that for the common gases, the molecules of which are monatomic (one atom per molecule), diatomic (two atoms per molecule), or polyatomic (more than two atoms per
molecule), the predicted values of y are respectively 1.66, 1.40, and 1.33. The observed values of y for a number of common gases are given in Table 1-4. The velocity of sound is
%=
%=
%=
28.98
4.81
6.023
10^3
X 10-" g
Using
(1-14),
mass and the value of h given in Eq. the most probable velocity of an air molecule at 68r = 20C
(
293K)
is
_ lykTf \ m I - (ir(1-30)
2
( \
1.371
^^^ ^
^^,
J2
13
For
68F
v^
20C, therefore,
10*
cm/sec
by reference to Eq. (1-27) when n is in molecules per cubic centimeter and A is in square centimeters. The volume of gas at the pressure in the vessel escaping each second is obtained by dividing the molecular flow rate q by the density n, so that
in a gas is of the order of threegas molecules. It is evident the of fourths the average speed (vav) of the density (n) independent is speed sound from Eq. (1-30) that the the temperature of the gas, but is proportional to the square root of velocity. as is the average molecular
and
dV
dt
Av
n
= 3.64
10^
JT\^ A \M
cm^/see
(1-33)
which for
air at
0C would be
dV_
Table
1-4.
11
1-401
273
\^
3.64
10^
\28.98/
Monatomic
11.2^
1-408 1-402
1-400 1.401
liters/sec
(1-34)
Air
1.667
1-66
. .
Hydrogen
Nitrogen
1.66
Oxygen
Carbon monoxide ...
Polyatomic
.
For the flow rate through a hole to agree with this value the thickness of the wall at the point where the hole perforates the wall and the diameter of the hole must be small as compared with the mean free
path for collisions between the molecules, a dimension to be defined and evaluated in the next section.
1-7. Molecular Mean Free Path. The concept of the mean free path of the molecules in a gas is of considerable practical importance in defining the behavior of a gas in a vacuum system. As was previously mentioned, the molecules of a gas may be considered to act in a first approximation as if they were elastic spheres of diameter of the order of 10-^ cm. Let us designate the molecular diameter of a particular gas as | and attempt to calculate the distance a molecule will travel on the average before undergoing a collision with another
Ozone
1-29
Water
(vapor) ....
.
Ammonia
1-6.
Flow of Molecules through a Hole. A quantity of some molecules which interest in vacuum technology is the number of
strike a surface in a unit of time.
It
of the walls of gas molecules which impinge on each unit of surface area of a containing vessel during a second is given by^-*
molecule.
yinvav
~ 2^1 m
molecules/cm^ sec
/
(1-31)
The above in which the value of v^y given in (1-23) is substituted. molecules of flow of result can be applied to the calculation of the rate is area, of hole through a tiny hole in a very thin plate. If a tiny
A molecule having a velocity v moves a distance v dt in the time dt. The molecule under consideration suffers a collision with another molecule if anywhere its center is within the distance f the molecular diameter, of the center of another molecule and therefore sweeps out the volume
,
dV=TTPvdt
Since there are
(1-35)
cut in the thin wall of the vessel beyond which the gas density zero, the rate at which molecules of gas leave the vessel is
is
nA /2kT\
^
= ''^=2^1"-^) 2^\~m"j =
3.64
n molecules/cm', the volume associated with one molecule on the average is l/n cm^. Thus when the volume in Eq. (1-35) becomes equal to l/n, it must contain on the average one other molecule and a collision has occurred. That is, for a collision to occur on the average
X W^iTIMY'^nA
molecules/sec
(1-32)
when
dt
TTi^VT,
(1-36)
t,
14
path is the distance traveled, that corresponding to Eq. (1-36) and is therefore
The mean
free
is,
15
=
TTWl*
(1-37)
Although the above derivation gives the correct general dependence of the mean free path on the density and molecular diameter, it is oversimplified by not taking into consideration the fact that the molecules have a distribution in velocity and that not only the reference molecule but also all others are in motion. The exact calcu25^ in the lation, ^ which will not be given here, introduces a factor of path is free mean denominator so that the corrected value for the
as the dimensions of the tube, both types of important and the flow is of intermediate character. The range of pressure over which the flow is not strictly viscous or molecular is referred to as the transition region. Flow rates for this pressure regime will be discussed in a later section.
path
is
of the
same order
collisions are
of State. From the general gas one concludes that at T = 0K and finite pressure the volume of the gas becomes zero. This conclusion is in contradiction to the concept of a definite molecular diameter, i, Clearly the gas cannot be compressed discussed in the previous section. to a volume smaller than that occupied by the tightly packed molecules without destroying the __
1-8.
kT
since
Even
be-
from
(1-16)
P =nkT
The mean free path is thus inversely proportional to the molecular density and to the square of the molecular diameter. Alternatively, it is proportional to the temperature and inversely proportional to the pressure. As the pressure in a vacuum system decreases during pumpdown, the mean free path changes from the order of 7 x 10"* cm
10"* torr. at atmospheric pressure to about 5,000 cm at a pressure of Values of the mean free path for various gases, determined from Eq.
different types of
(1-38) using values of the molecular diameter f deduced from two measurement, are given in Appendixes II and III.
The mean free path is a parameter which enters into the determinaIf tion of the flow of gases under the influence of pressure difference. the mean free path is short as compared with the diameter of a tube
through which the gas is flowing, collisions between molecules will predominate and collisions with the walls of the tube will occur only The flow dominated for those molecules which are close to the walls. by intermolecular collision is referred to as viscous flow because the However, if the mean free flow pattern is that of a viscous fluid. path is long as compared with the diameter of the tube, collisions of the molecules with the walls predominate and intermolecular collisions are not important. In this latter regime the molecules move independently of one another and nothing like viscous flow occurs. The molecules move randomly in straight lines, colliding with the On the walls of the tube and proceeding in a chaotic manner. average, more molecules move from a region of higher density (or
pressure) to a region of lower density for purely statistical reasons.
reached the assumption of entirely valid because of the restrictive effect of collisions between closely spaced molecules. Moreover, the finite size of each molecule eliminates a certain amount of Fig. 1-2. Exclusion volume which cannot be occupied by another volume in molecular collisions. colliding molecule, no matter how low the molecular density may be. Consider a molecule A which undergoes a collision by molecule B moving in a given direction, as shown in Fig. 1-2. The center of the molecule B can lie anywhere outside the hemisphere of radius |, the molecular diameter, but cannot lie inside this hemisphere because of the collision process. The center of molecule B is therefore excluded from the volume
molecules
is
V^
%7Ti^
(1-39)
Such an exclusion volume is associated with each molecule so that the volume available for random motion of the molecules is not V, the volume of the vessel, but is V - b where
b
VzttNP
(1-40)
which
This process
is
When
the
mean
free
molecules during impact is an between two colliding billiard balls. However, it is known, for example, from the behavior of condensed gases that definite attractive forces act between molecules. Surface tension IS an evidence of the existence of such forces.
elastic force similar to that
no forces acting between the molecules of a gas until a and that the force acting between the
is a volume four times that of the N molecules present. The general gas law is also based upon the assumption that there
are
collision occurs
The
origin of the
16
attractive force
is
17
coefficients of
the mutual electrical polarization of one molecule attraction between the resulting electrical dipoles. the and another by This attractive force is relatively long-range in terms of the molecular separation distances over which it remains effective and is referred to as the
be equal.
see that
Comparing the
we
3F.
RT.
6
Po
3F.2 "
is
F3:
P,
'
Ab
P,
empirical basis,
had proposed an
critical
point
F
b)
A
36
2762
b,
7'.=
8t7
276^
(1-45)
{^-v)^^
= RT
(1-41)
and
R in terms of the
-R
"triple point"
which differs from the general gas law for perfect gases in the correction term AjV^ added to the pressure due to the attractive forces between the molecules and the subtraction of b from the volume to allow for the region in space from which the molecules are excluded by one another
because of their finite dimensions. By multiplying out equation and rearranging terms one has
^=
3F,''P,
'
6=-!-^
3
= ^7^
3T,
8P F
(1-46)
PV
A\ A\
1_
(1-42)
V
plus terms of higher order such as b^jV, etc., which are negligible. It has been shown* that Van der Waals' empirical equation (1-41) is consistent with the equation of state for an imperfect gas calculated from basic assumptions about the intermolecular forces to terms of as
Returning to the form of Van der Waals' equation given in (1-42), evident that for large values of the temperature the term b in the coefficient (6 AjRT) dominates and the pressure is higher than that of a perfect gas. Alternatively, when the temperature is very low, the term AjRT dominates and the pressure is less than that for a perit is
fect gas.
In Fig. 1-3 the behavior of the isothermal curves (dependence of pressure on volume for a constant value of the temperature) above
is
high an order as we normally need to consider. For our present purposes it is instructive to write equation (1-41) in descending powers of F as follows
Van
der Waals'
and below the critical temperature shown. Curves for temperatures below the critical tempera-
At
F3
V^
+ jV
Ab
(1-43)
This is a cubic in V for any given values of T and P. In Fig. 1-3 are plotted typical isothermal curves (pressure versus volume) for various values of temperature. For sufficiently small values of the pressure and temperature there are three values of the volume, corresponding to Eq. (1-43), for a given value of the pressure as represented by the horizontal dotted line in Fig. 1-3 passing through the points i,j, and k. However, when the temperature is sufficiently high the isothermal curve
From
on the
the
iso-
point
to the point k
has only one root. The critical point is that indicated at c where for specific values of temperature, pressure, and volume all three roots of the equation are equal. At this point
thermal curve the actual behavior of the substance does not follow the Van der Waals equation with a minimum and a maximum in the pressure. In the mixture domain ,, the pressure remams constant as
the volume
is
(7
7^)3
F*
37^7,
3FF,2
_ F/ =
increased until
all
j it o rr, nr i 1-3. Typical Van der Waals isoFig. i values of various thermal curves for the temperature.
(1-44)
18
19
converted into vapor. This is illustrated by the horizontal After the liquid is all evaporated, the pressure straight line from i to k. decreases as the volume is further increased in a manner somewhat In this third, or vapor, similar to the behavior of a permanent gas. domain the behavior differs from that of a permanent gas in that it does not quite follow the general gas law since the Aj V^ and b terms of Eq. (1-41) do not become negligible until the volume has become quite large and the pressure small as compared with their values at the
the liquid
point
k.
with the lower plate is zero, whereas that in contact the gas in contact The drift m, the velocity of the upper plate. with the upper plate is to the proportional point is some intermediate velocity of the gas u at that plate, so distance y from the stationary
or
(1-47)
The
drift velocity
temperature the isothermal curve passes through the triple point where the region of liquid-vapor mixture has narrowed down to a zero range in pressure. At all values of the temperature above the critical temperature there is no condensation of the substance into a liquid so that for all values of the pressure, no matter how large, no condensation occurs. This is the condition in which the substance The higher the temperature above the is called a permanent gas. critical temperature the more nearly does the P V curve approach that of a perfect gas as represented by the general law, Eq. (1-8). All substances conform approximately to the behavior described above, the difference being in the values of the parameters given in
At the
critical
u is superimposed upon the random velocity of the most physical situations is much less than the average
direct proportionality for the above equation is Newton and has been confirmed experimentally of that consistent with of presrange wide for gases over a case the In liquids. ' sure and for
The assumption of
of a gas the uniform velocity gradient can be understood in terms of the transfer of momentum from one layer of gas to the adjacent
W/^^X^^W^^^/^^X^/^/Z/W//^
1
u.
^y
In Appendix II values of the critical temperature, the Van der Waals parameters A and b, the molecular diameter | calculated by Eq. (1-40), and the mean free path calculated by Eq. (1-38) are given for several
gases and vapors. 1-9. Dependence
In order to approximate layer. the momentum transfer we note that molecules on the average move up or down because of their random motions a distance A, the
j%mMm^^^%^^%^^^%;^^,
Fig.
1-4. Drift
11/
velocity
distribution
due to viscous
forces.
mean free path. Molecules from a distance A above move the layer under consideration with a momentum given by
(mu).
down
into
of
Viscosity on
Molecular Diameter.
Because the molecules of a gas have an appreciable cross section for collision determined by the molecular diameter discussed in Sec. 1-7, the flow of a gas through a tube or hole due to a pressure difference is characterized by a viscous drag. The viscosity of a gas can be derived in terms of the molecular diameter, which also determines the mean free path and the correction to the volume term in Van der Waals' equation of state. Viscous forces appear when there is a variation in the drift or flow velocity of a gas from point to point. Such variations will occur whenever gas flows through a tube because of the maintenance of a pressure difference, the gas flowing from a region of high pressure to one of low pressure. Our quantitative understanding of viscosity originated in the work of Newton, who assumed that the internal viscous
forces are directly proportional to the velocity gradient in the fluid.
= my +
Wy
(1-48)
However, molecules from the layer at the height y the higher layer at y with a momentum
(mu)_
^ also
move
into
=m
(1-49)
In the random motion one-third of the molecules may be regarded as in the vertical direction and one-third each in the directions parallel and perpendicular to the motion. Only the one-third moving the vertical direction contribute to a transfer of momentum between
moving
Consider a body of gas, as shown in Fig. 1-4, between two parallel by a distance w, with the lower plate at rest and the upper plate pulled to the right with a force F. The drift velocity of
plates separated
Of these one-half are moving upward and one-half downis the average random velocity of the molecules, the number crossing an area A is the number contained in the volume Ava.yt and moving in the required If the total number of direction. molecules per cubic centimeter is n, then the number crossing from
ward.
If
Va,v
the layers.
20
21
KnAv^v
The above
HnAva,v('mu)+ A<
= Hnm
ndom
Ave.yU A<
(1-50)
calculation of the viscosity is only approximate since the motion of the molecules is treated inexactly by assuming
all
move
in their
assumed to travel a
collisions.
mean
free
path between
When
the
moving upward
A2)_
is
}4nAva,y(mu)_
M = Hnm
Avs.yU
(1-51)
and the distribution in free paths are distribution in becomes much more complicated and problem the taken into account, approximations.*'^ only successive at by arrived, be the solution can
random
velocities
of the gas below the boundary the difference between these two quantities, so that
The
momentum
The
is
is
^p = ^p+ -
AiJ_
M
nmXAva,yU,
is
,
(1-52)
0.499
,, ,, (1-58)
Since the rate of change in momentum below the boundary by that above.
0.998 /mA;T\^
F=
-r-
}/inmVa,\X
A<
(1-59)
-
(1-53)
By definition the coefficient of viscosity is tangential force per unit area divided by the resultant velocity gradient dujdy of the shear motion in the fluid. In this case
?
the ratio
between the
Equation (1-59) provides another basis for measurement of the molecular diameters of various gases. Solving for the molecular diameter,
0.999
/mkTY'
i\i4
(1-60)
=
rj
FjA
dujdy
FjA
ujw
(1-54)
collisions
which according to
(1-53) is
HnmVavX
free
(1-55)
(1-38),
a=
for
0.998 /mA;T\'^
77|2
r]
-)
\
TT
(1-61)
mean
path A given in
we
find
Substituting numerical values for k and tt and the molecular weight 6.023 x lO^* ^ ^^ ^j^g following numerical results are Wmoim
'
3(2)'^77|2
obtained
in
which f
is
5.22
10-11
iv)
cm
cm'
(1-62)
2 /2fcT\^
=-u m
tt'-^V
and
8.52
10-21
(MT) H
(1-63)
so that
n
(1-56) it
37Tf
'\)
^1^
I)
Appendix III together with computed values of the molecular diameter f from Eq. (1-62) and of the mean free path A from Eq. (1-38). These
values of f
From Eq.
1
can readily be seen that the dimensions of viscosity are mass/length x time. In the cgs system the unit of viscosity is
poise
and A based upon viscosity measurements should be compared with those given in Appendix II based upon the experimental
values of the
g/cm
sec
dyne sec/cm^.
Van
b.
22
The viscosity result is rather surprising in two respects. to increase and density the of independent predicted to be is a gas of common of case the in whereas temperature, root of the the square as liquids the viscosity is known to decrease as the temperature is Also for two different gases the viscosities at some standard increased. temperature should be proportional to the square roots of their molecAs surprising as these results may be, experimental ular weights. measurements of the viscosity of gases confirm them for a wide range
The above
CHAPTER 2
GAS FLOW
of temperatures and pressures. However, at extreme values of the pressure, both high and low, the viscosity of a gas departs from this prediction. At very high pressure the average distance between the
2-1.
Gas Flow
in
Vacuum Systems. An
understanding of gas
molecules is so small that the intermolecular forces become important and the momentum transfer differs markedly from that assumed above. At values of the pressure which are so low that the mean free path exceeds the distance between the walls, colUsions between molecules only rarely occur so that collisions with the walls predominate. In this case there is no transfer of momentum directly from one layer of gas to another, but only between the moving gas and the walls of the tube or Thus, as mentioned in Sec. 1-7, the mean free path is a vessel. characteristic dimension which determines the behavior of a gas and
in particular determines
flow over a very wide range of pressure is essential to an intelligent approach to vacuum-system design. A system is generally at atmospheric pressure initially, is then "roughed out" by mechanical vacuum pumps, and is flnally evacuated to the desired limiting pressure by diffusion pumps with an appropriate arrangement of water-cooled The mean free path of the gas in the system is baffles or cold traps.
initially
or of molecular flow. The flow of gas due to a pressure difference will be discussed in the next chapter. In that discussion the relative magnitude of the mean free path as compared with the dimensions of the tube or passage
will
be of basic importance.
REFERENCES
1.
very smaU (about 7 x 10-" cm), as discussed in Sec. 1-7, so that the dimensions of the pipes and manifolds are many times the mean free path and the flow of gas is limited by viscosity. When very low pressures are finally attained, the mean free path may be large (perhaps hundreds or thousands of centimeters, depending upon the pressure attained), so that the dimensions of the pipes and manifolds are then a very small fraction of the mean free path of the molecules. Viscosity then no longer characterizes the gas flow, which is then referred to as molecular flow. At some value of the pressure, depending upon the cross-sectional dimensions of the particular part of the system under consideration, the molecular mean free path is about equal to those dimensions and the flow is neither purely viscous nor purely molecular in character. This is the transition region for which the
equations for gas flow are rather complex.
2. Ibid.,
pp. 86-88.
3. J. C. Slater
4.
and N. H. Frank, Introduction to Theoretical Physics (McGrawHill Book Company, New York, 1933), pp. 462-465. S. Chapman, Phil. Trans. Roy. Soc. London 211 A, 433 (1912); 216A, 279 (1916); 217A, 115 (1918).
D. Enskog, Kinetische Theorie der Vorgange in massig verdilnnten Gasen, Dissertation, Uppsala, 1917.
The characteristics of viscous and molecular flow will be developed some detail in this chapter, and a number of formulas of practical interest given. The transition region will be discussed qualitatively
in
5.
to illustrate the approximate behavior of gas flow in this range of pressure. In addition, the conductance formulas so essential for the design of vacuum systems to meet specifications will be developed.
Applications to typicd vacuum-system situations will be made in Chap. 8, dealing with vacuum-system design. 2-2. Pumping Speed and Conductance. Before proceeding
With a detailed discussion of the characteristics of gas flow it will be useful to define the terms pumping speed and conductance as used in "Vacuum-system design.
23
24
GAS FLOW
25
The prime movers in vacuum systems are the mechanical vacuum pumps, steam or oil vapor ejectors, diffusion pumps, and a variety of These specialized pumping devices, such as ion and getter pumps. which at a rate devices all remove gas from the system to be evacuated is measured by the pumping speed Sj,, which is defined as the volume of gas per unit of time dVldt which the pumping device removes from the system at the pressure existing at the inlet to the pumping device. The common units of pumping speed are liters per second, cubic feet per minute (cfm), and cubic meters per hour; several other combinations of volume and time are used occasionally. A convenient table of conversion factors for these units is given in Appendix IV. The gas flow into a pumping device, called the throughput, is defined as the product of the pumping speed and the inlet pressure,
I.e.
The pumping speed of a vacuum pump according to (2-1) is 8^, = QlPm- By analogy with this expression it is convenient to define the pumping speed at any point in a vacuum system as
S
where
=Q
(2-5)
is the gas flow in the system and P is the gas pressure at the the pumping speed is defined. In the case of an opening which point at which the quantity of gas Q is flowing from a region at through or pipe region at Pj, the pumping speeds at the two points in a to pressure Pi given by are the system
and
Sn
T^
Q =
for
P,A =
dV
^i
(2-1)
so that
-^
and
Q P =
into (2-4), the result
is
(2-6)
~di
liters per second, torr cubic feet = micron = 10"^ torr), listed with second per liters minute, per // (n conversion factors in Appendix IV. The throughput is proportional to the mass flow of gas since under conditions of steady flow such that the
and Pa
pressure
is
constant
d Q = ^'7 dt "dt
rl
dV
iW)^
is
/W
R.TdW BoTdW
Dividing through by
(2-2)
Q and
dt
by
reference to
(1- 8).
8,
(2-7)
dW
(2-3)
dt
Thus the pumping speed at any point in the system can be obtained from the known pumping speed at some other point and the conductance of the portion of the system (pipes, holes, or passages) in between. In
particular, for the combination of a
generally passes through a series of pipes or conduits which present a resistance to flow, so that between any two points along the flow path
pump
pipe of conductance
the combined
1
between the ends of a pipe leading to the pump) a pressure In fact, a net flow will occur only if such a pressure difference does exist. By analogy with an electrical circuit, the conductance between two points along the flow path is so defined that the quantity of gas flowing through the system is the product of the conductance and the pressure diff'erence, i.e.,
(e.g.,
^ =
8
or
8,
+C
(2-8)
Q = (P,- P,)G
(2-4)
By reasoning similar to that given above, it can easily be shown that lor several openings, pipes, or conduits in series, each with individual
conductances G^, G^, C3,
etc.,
Since Q is the quantity of gas per unit of time entering the pipe or conduit at the pressure Pj, then if no additional gas leaks into or is removed from the pipe between the points of interest this same quantity of gas comes out the pipe at pressure P^.
is
given by
(2-9)
(series)
= + G\
Cj
C,
26
GAS FLOW
27
Also for several openings or pipes in parallel, so that the gas flow divides between them, the combined conductance is
(parallel)
Ci
C^
C3
(2-10)
center moves faster, and that farther from the center, and thus nearer the wall, moves more slowly than that in our sample cylinder. The component of force due to the faster-moving gas just inside the cylinder under consideration can be written by reference to the definition of viscosity given in Eq. (1-54) as
The analogy with the conductance in an electrical circuit is obvious. The performance of a vacuum system is normally calculated by first
determining or assuming the total gas flow Q to be expected, then choosing a combination of pumping speed and conductance such that the desired working pressure can be maintained. The individual conductances of components {G^, C^, C^, etc.) are first determined, then the combined conductance is calculated by some combination of Eqs.
(2-8). (2-9) and (2-10), and finally the pumping speed is calculated by This resulting pumping speed 8 for the system must then satisfy (2-5) when Q is the gas flow expected from the vacuum process concerned and P is the specified pressure to be maintained. The definition of pumping speed under more complex conditions involving the outgassing of surfaces at reduced pressure, residual leaks into the system,
dv
P,
= coefficient of viscosity of the gas = surface area of the cylinder dvjdr = velocity gradient S = 277r Sx is the area of interest, and the
where
rj
;Sf
inside gas
is
thus
-27T7]r
dx
dv
(2-12)
and variation of the pressure with time is discussed in Chap. 7. Poiseuille's Law. The characteristics of 2-3. Viscous Flow flow of highly viscous liquids, such as "heavy" oil or molasses, have been observed qualitatively by nearly everyone. The flow rate of such
force
27Tw(r
a hquid through a tube is proportional to the pressure difference causing the flow and to a high power of the diameter of the tube. This same tj^e of behavior occurs for gas flow at relatively high density. The formulas governing gas flow under these conditions will now be
discussed.
cross section (see Fig. 2-1).
8r)
8x(^]
\dr/r+ar
is
i;
However,
a,t
dv\ v\
^^
dr
Consider the flow of gas through a long tube of uniform circular Between the ends of a segment of this tube a pressure difference Pj P^ exists, and gas flow occurs along the tube from the region of higher to that of lower pressure, i.e., from Pj to Pj. The gas contained within a thin-walled cylinder of radius
r
Tr) r/r+ir
dr\
and
P,
27Tri{r
dr)
dx\v+^dr
dr\
dr
(2-13)
dr,
and within a
Equilibrium will occur when the force due to the pressure difference given in Eq. (2-11) is just balanced by the sum of the viscous forces
experiences a force in the direction of flow given by the cross-sectional area 2Trr dr multiplied by the pressure difference 8P, so that
(2-13), so that
^F = 2^r8PJ^r
(2-11)
dP
dr
dv
-27Tri dx
dv
'dr-^"^^-^drV^dr^n.
I
,
The flow quickly reaches a steady state in which an equilibrium exists between this force and the
viscous forces from the gas, both at smaller and larger values of the radius. The gas nearer the
ZTTf)
which yields
dx\dr\
dv
dr
d^v
1-
y r dr
^
dr^
\
d^v
dr^
dr^
L
Fig. 2-1. Viscous flow through a tube,
rdP
(dv
d^v
d^v
28
If
GAS FLOW
Since Pj^
29
we drop
P^^
(Pj
P2)(Pi
Pa),
and
since
we
define Pay
(Pj
P2)/2, Poiseuille's
Tra*
-*
dP
dx
= dH
dr^
\dv
1
.-
(2-14)
r]
r dr
is
8rjL
av(Pi "''
'
Pa "
128;L
Pav(Pi
(2-20)
It
is
of the form v
=A +
Br^
is
A in
dP
.
(2-15)
2a is the diameter of the tube. As this equation is written the pressure is measured in dynes per square centimeter and the radius The quantity Q (or diameter) and length of the tube in centimeters. is thus measured in (dynes/cm^)(cm^/sec) = microbar cm^/sec, since
where
D=
4jy
ox
10* djrnes/cm^
by
definition
is
bar.
which
is
a constant determined by the boundary conditions. In the event that there is no motion of the gas
in contact with the wall of the tube then ?; a, and in this case at r
Referring to the definition of conductance given in Eq. (2-4), it follows from (2-20) that the conductance of a tube of circular cross section at high gas pressure is
C(cm3/sec)=^-^^=^Pa.
in
(2-21)
(2-16)
4:r)
ox
According to (2-16) the gas velocity as a function of the radius is parabolic with maxiFig. 2-2. Velocity as a function of the radius in viscous flow through a tube.
mum
axis
(r
velocity
(r
Vraa.y.
which the quantities are all in basic (cgs) units, i.e., the dimensions and L are in centimeters, -q is in poises, and Pav is measured in dynes per square centimeter. Since 10* dynes/cm^ = 750.06 torr (mm Hg), the expression for the conductance in (2-21) becomes
{\r)){dPldx)a^
on the
0)
at the wall
^
when the
conductance
(^^^/^^^)
m^
ttD*
10*
D*
^-
7^6
^''
^ ^ Pav
riL
^^-''^^
The volume of gas flowing through the is obtained by integrating Eq. section of the tube, so that the flow in volume F
is
pressure Pav is measured in torr. more usual unit of is liters per second, for which (2-21) becomes
(liters/sec)
= lO-V
(cm^/sec)
3.27
x IO-2
(2-226)
dt
=
Jr=0
2iTrvdr
=-]
^n dxJr=0
(a^
r^)rdr=^'-^
8?^
(2-17)
ox
in
for
define the gas flow as in (2-2), then from Eq. (2-17) the gas flow through the cross section of the tube
If
we
we have
centimeters,
which the pressure is again in torr, the dimensions L and and the viscosity is in poises. Finally, since 28.3 liters, 1 in. = 2.54 cm, and 12 in. = 1 ft,
D
1
are in
ft^
^
If
(2-18)
dt
8r]
(cfm)
3.27
X 10-^
Ift^
dx
^ Pav X ^^ min
rjL
1
2.54
cmV
lin.
we now extend
1ft
12 in.
L =
j dx,
then
28.3 liters
na^j^ri
P(dPjdx) dx
dx
=
na*
(Px'
l6r]L
9.46
X 10-2
Pav
=
(2-22c)
Q =
P^')
(2-19)
f
Jo
This is one form of Poiseuille's law describing the flow of a viscous, compressible fluid through a tube of circular cross section.
is measured in inches, the length of the tube in the viscosity in poises, and the pressure in torr. The viscosities of various gases at a temperature of 20C are given in Appendix III. For the particular case of air at 20C 68F the
30
viscosity
is
GAS FLOW
31
X lO"*
becomes
(2-23a)
2-5.
Turbulent Flow in
C
in
(liters/sec)(air at
20C)
179
Pav
through a tube
may
Corre-
where p
(2-236)
C
in
(cfm)(air at 20C)
is
Pav
Ij
= = D= =
V
rj
in torr,
D in inches,
From
and
in feet.
Eqs. (2-23a) and (2-236) a convenient "pressure drop" formula can be derived which provides guidance in the selection of a pipe size to be used for pumping air with
This is known as the Reynolds number. As the fiow velocity increases, the Reynolds number increases and the pressure difiference between two points along the tube increases consistently with Eq. (2-20) (where
(2-1)
and
(2-4)
Q = ttD^PvI4:) until the Reynolds number exceeds a critical value. When Re > 2,000 (approximately), then the character of the flow
Instead of flowing in a smooth pattern of continuous flow by viscous flow, the fluid becomes turbulent and erratic with the appearance of eddies and oscillations This is turbulent flow, the onset of which can be predicted for any fluid by evaluating the Rejmolds number. The transition from viscous to turbulent flow will occur for any fluid approximately in the range 1,000 > Re < 2,000.
changes.
lines characterized
.
-P,^^ = -9Il.
If a pipe of diameter
(2-24)
is
connected to a mechanical vacuum pump of pumping speed Sj,, then, since the pressure at the pump inlet is the same as P^, the gas flow into the pump is Q = PiSj,, so that
pump capacity the rule is adopted that the pressure drop in the pipe must not be greater than one-fifth the pump inlet pressure, since Pav = (Pi + P2)/2, then the quantity
If for the efficient utilization of the
nearly always viscous in character and 2,000 the flow is nearly always turbulent in character, in which case the pressiire difference between two points along the tube no longer is given by Poiseuille's law (2-20), but becomes erratic in value and greater than that corresponding to viscous flow. At what exact value of the Reynolds number the flow
is
For Re
<
When Re >
2P,
0.9 to 1.0 (2-26)
Pav
is
Pi
"T
P2
becomes turbulent depends upon the roughness of the surface of the tube and other factors, but in any case is in the range 1,000 to 2,000. The throughput may be expressed in terms of the Reynolds number
as follows
approximately unity, as long as the rule stated for selection of pipe Thus (2-25) yields for the approximate value of sizes is followed.
Pressure drop (torr)
tD^
Q
But
since
Pv
ttD^
fjHe
rjl
pD
ReZ)
S^
D*
5.6X10-"%^
centimeters
(2-27a)
from
(1-8) p
= WjV
PMjR^T
77,?PT
(2-29)
179
Similarly
(2-276)
1.9
517 D*
X 10-
For
air at 20C,
r]
T =
293.16K, and
M = 28.98 so that
=
(1-829
"L
1.829
x lO^*
poise,
R^
These cubic feet per minute, in inches, and L in feet. expressions provide a means of obtaining a rough check on the selection of the proper size for a given operating pressure range. If the pressure
drop so calculated is not small as compared with the desired operating pressure, a larger pipe size must be selected.
Q
when
(air
20C)
x 10-^)(62.364)(293.16)
.28.98
^^
^
(2-30)
= D is
measured
9.06
X 10-2 Re
torr liters/sec
in centimeters.
32
If
GAS FLOW
33
sets in at
Re
be turbulent
liters/sec
ISlDtorr
(2-31)
vacuum practice in which turbulence occurs. Consider the case in which a vacuum tank has been evacuated to a very low pressure, and then air is admitted by opening a valve to atmospheric To be specific, assume that the valve is connected to the pressure. vacuum tank by a tube 1 cm in bore diameter and 10 cm in length.
Situations do develop in
pump of 110 cfm or about 50 liters /sec displacement speed is to be used and the connecting pipe length is 2 m. The pumping speed of such a mechanical pump at a pressure of 0.05 torr is about 38 liters/sec. (See Chap. 5 for mechanical vacuum-pump characteristics.) Then according to (2-27a)
assume that a mechanical
0.01
38 X 200
179
D*
is
According to
(2.19)
and
179 Z)*
D=
(P,^ -
8.1
cm
3.3 in.
Q
Initially P^*
p^)
< Pi'' so that according to Poiseuille's law Q = i7^Ho(760)2 = 5.17 X 10' torr liters/sec
whereas according to (2-31) turbulence occurs for any value of throughput greater than Q which for our chosen example is just 181 torr liters/ Thus initially, when the pressure Pj in the tank is low, the insec.
rushing air will experience turbulent flow. The flow will continue to be turbulent until the internal pressure has reached a value such that
assuming that the connecting pipe has an inside diameter of length of 200 cm, the question is whether pumping down the system from atmospheric pressure will result in turbulent flow for some portion of the pumpdown cycle. Since at high pressure the pumping speed will be nearly equal to the theoretical displacement speed of the pump, the pumping speed S^ s 50 liters/sec. From Eq.
8.1
Now
cm and
(2-23a)
= S,P = =
C(P,
Pi)
179
P,y(P, Z)<
P,)
Oo or
1^^0(7602
Pa^)
181
179
D*
Ij
P(O.Ol)
759.5 torr, so that turbulent flow occurs essentially that is, until Pj for the entire period of admitting air to the vacuum tank since the throughput does not drop below the critical value for turbulent flow until the pressure in the tank is within about 0.5 torr of atmospheric
pressure.
From
179
throughput
Q,
1812)
-
jj
Z)<
P,(0.01)
However, when pumping situations are considered, the occurrence of throughput in excess of the critical value for turbulence is not usual and in any case does not lead to the need for design alterations. At low pressure, the throughput is automatically low since in the region of viscous flow the throughput depends upon the square of the pressure.
pumpdown
until the
P^=
is
181
^-
102
38.5 torr
179 Z)3
reached, below which the flow will be viscous. At this value of the pressure the pipe conductance, according to (2-23a), is
flow during the roughing-down period of a system is turbulent during the early (high-pressure) portion of the pumpdown cycle when a reasonable combination of mechanical pump
displacement and connecting pipe is used. According to the criterion suggested in Sec, 2-4, the pressure drop in the pipeline should not exceed one-fifth of the pump inlet pressure when the system is at the operating level. Consider a system for which the mechanical pump must maintain a base pressure of 0.05 torr, so that the pressure drop in the pipeline should not exceed 0.01 torr. To be ouite specific, let us
(at
P=
38.5 torr)
= = =
179
D* -
(38.5)
(8.1)*
' 1'79^177T7^(38.5)
200
1.46
10^ liters/sec
34
GAS FLOW
35
is The conductance of the connecting pipe at this value of the pressure of speed the that pump of the so much larger than the pumping speed torr) evacuation is not affected by the very small pressure drop (0.01 difference pressure a creates Even though turbulence in the pipe. from 760 down significantly greater than 0.01 torr at higher pressures,
Molecules experiencing absorption and reemission represent a layer of gas which is at rest next to the wall and provide the viscous drag discussed above. It can be shown that the coefiicient e, which defines
this effect,
is
given
by
measurably decreased to 38.5 torr, the actual speed of evacuation is not the criterion in Sec. as long as by the presence of turbulence. Thus, the occurrence of pipe, connecting 2-4 is followed in the design of the need for altering any present not turbulent flow during pumpdown does be expected to may flow turbulent the choice of pipe size. Although in almost any cycle pumpdown occur during the early phase of the flow in the viscous upon based the design parameters
U R^Tl
-/
(2-33)
where / is the fraction of molecules which are absorbed and reemitted, and 1 / is the fraction which are specularly reflected. By substituting this expression into Eq. (2-32) one obtains for the gas flow through a tube of circular cross section
vacuum system,
dependent upon low-pressure range, where the performance is critically altered by the be not need pipe, the dimensions of the connecting of the pumpportion high-pressure the occurrence of turbulent flow in systems, vacuum in occur does therefore flow down cycle. Turbulent requirement.^ design additional any imposes but not in a way which Slip. 2-6. Correction to Poisseuille's Law Due to Surface confirmed been has (2-20) Eq. given in as flow Poiseuille's law for viscous experimentally over a wide range of gas pressure and tube diameter.
Q
-128j
16\2
c,D^)
-f
I
D^
Pi
f
(2-34)
CiPavi>*
Pl-P2
where
c,
=
128>;
(2-35)
and
However, as the pressure is decreased for a given diameter of tube, the Poiseuille's flow rate eventually begins to deviate from that predicted by the gas that assumed it is law Poiseuille's of In the derivation law.
If this is not the case, but velocity drops to zero at the tube wall. velocity, referred appreciable an wall has the instead if the gas next to of the form expression an given by is the flow to as surface slip, then
'^^
16\2 lW~/
(2-36)
When the pressure Pav is sufficiently high, the term c^P^yD^ dominates the term c^D^ and the flow follows Poiseuille's law as given in Eq. (2-20). When the value of the pressure is such that in (2-34) the term CiPav-D* is equal to the term c^D^, the character of the flow departs
significantly
from that of
Poiseuille's law.
The pressure
for
which
Q = -^Pav(P.-Pi)(l +
128yL
and
(2-32)
is
given by
Pi)(
D
16^
c,P,D*
or
c,D3
(2-37)
L
in
\l28rj
the coefficient which determines the velocity of the gas at the inner surface of the tube. The interaction of the gas with the walls can be analyzed in terms of two distinct processes. Some gas molecules in striking the wall experience specular reflection and thus retain the same component of If all molecules velocity in the direction of flow as before the impact. there reflection specular striking the tube wall were to experience
P.=
which
e is
c^D
At pressures
significantly below P^ the viscous flow term c^Ps^vD* is of decreasing importance and the nonviscous term CjD^ dominates. The correction to Poiseuille's law due to slip is therefore negligible at values of the pressure which are large as compared with the transition
pressure, but
would be no "drag"
uniform over the cross section of the tube. Other molecules strike microscopic irregularities in the wall and bounce several times. Under these conditions a molecule may be absorbed by the wall and then reemitted later with a random distribution in angle and velocity.
becomes so important below the transition pressure that the character of the flow is completely altered. From (2-34) and the definition for conductance
C=
P.
C^P^yD*
CaD^
(2-38)
-P.
36
GAS FLOW
so that
37
where
From the disare given by Eqs. (2-35) and (2-36). and the form of Eq. (2-38), the character of the conductance of a tube can be seen to change radically, depending upon the
and
c^
cussion above
pressure.
K=
3.81
X 103
is 1.
If the pressure
is
large
compared with
If the empirical expression for c^ at very low pressure given in (2-39) compared with the theoretical expression given in (2-36), one finds
is
given by Eq. (2-21) and is (a) proportional to D*, (6) proportional to the pressure, and (c) inversely proportional to the viscosity. is 2. If the pressure is small compared with P the conductance inde(c) and (a) proportional to D^, (6) independent of the pressure, pendent of the viscosity.
that the value of / (the fraction of molecules which are absorbed and reemitted randomly when they hit the wall) is approximately 0.74, and therefore the fraction specularly reflected is about 0.26. If the pressure is sufficiently, high the terms k^Pa,^ and k^P&Y are very large
An exact 2-7. Gas Flow in the Transition Pressure Range. treatment of gas flow in the pressure range in which both viscous and molecular flow are important is difficult and unsatisfactory because the coefficient of slip, the e which appears in (2-32) and is defined in (2-33),. An empirical approach to this is not calculable from first principles. problem was offered by Knudsen^* based upon a series of carefully controlled experiments on gas flow. Knudsen found experimentally that the coefficient Cj given in (2-36), which determines the magnitude of the nonviscous term in the corrected form of Poiseuille's law (2-34),
can be expressed in the form
x lO^TIM)'^. The Thus Knudsen's results imply that the fraction of molecules absorbed and reemitted, as contrasted with those which are specularly reflected, changes slowly in the transitional pressure region. Including Knudsen's results in the
3.07
compared with unity and c^ Kikjk^) = corresponding value of/ is then about 0.85.
C
in
(cm3/sec)
Pav
128??
^+
3.81
M^f ^-^1^5^ ^
k^Pav
L'
(2-43)
which the values for k^ and fcj are those given in (2-41) and (2-42), the viscosity is measured in poises, the dimensions in centimeters, the
pressure in dynes per square centimeter {fi bars), the temperature in degrees Kelvin, and the mass in grams, so that the conductance is measured in cubic centimeters per second. Converting units to torr
for pressure
=K
in
-f fciPav
(2-39)
kari 2^ av
which
and to
liters for
A-=-.,
by
^x4(5^f. 12 n^X m
J
3.81X10.(1)" \M]
(2-40)
(liter/sec)
3.269 X
10-2-^
rj
substituting the value of Vav given in Eq. (1-23) value for {2kTlm)''^ given in (1-27). The values
Z)*
3.81
(2-41)
1. 1.333 1. 333
X X
103A;iPav
lO^jfcoPav
D^
(2-44)
i"
and
k.
1.24
5(i!L)"=1.38xlO-.:?(|| \ TJ n \kT/
rj
(2-42)
C-
=3.269
10-
PavJ>
3.81
/TWl +
\m)
1
0.1 .liliMITf^iP^^DIr])
+0.1 81(MITf^{P^^DIrj)
(2-45)
were obtained by
fitting the experimental results of many measurements of gas flow and pressure difference. At sufficiently small values of the pressure the terms JfcjPav and AjaPav are both negligible compared
+ +
The quantity on the left side of (2-45), the conductance multiplied by the length divided by the cube of the diameter of the tube, is therefore a simple function of the variable Pav-D/? as shown graphically in Fig.
2-3.
fciPa
1
fc^Pa
Equation (2-45)
is
of the form
References indicated
by
chapter.
ex
ax
-\-
b
I
dx
(2-46)
38
GAS FLOW
39
where
y=^CX
L
/MY'
the conductance, reaches a minimum. For pressures less than Pmin the conductance increases asymptotically toward the low-pressure value to be discussed in the next section, and at pressures greater than Pmin
=
n
a
Differentiating
3.269 X 10-2
0.181
(ff
the conductance increases with increasing pressure and eventually becomes proportional to the pressure as given by Poiseuille's law (2-21). In (2-37) the transition pressure P, is defined as that value of the pressure for which the viscous term CjD^Pav in the expression for the
C2D^.
with respect to x and setting the resultant has a derivative equal to zero determines the value of x at which y
(2-46)
conductance of a tube given in (2-38) is equal to the nonviscous term When (2-38) is compared with (2-45) and (2-46) it is evident
when Pav
=
ex
P(,
minimum
value.
The
result
is
3^Tnin
'b(d
,.
14
ax
(2-47)
dx
Solving this expression for x yields
{(be
2ad
-a)
[(be
ay
iabdf"'}
(2-53)
(E^) =5.47(^f
\
ri
(2-48)
/mm
\M/
the value of Pav>/? at which CLjD^ has a According to (1-58) the viscosity is given by
This
is
minimum
value.
in which the negative sign leads to a meaningless negative value for x. Using the positive sign, substituting the values for x, a, b, e, and d given above into (2-53), and noting that Pav is now the transition pressure Pj, we have
T]
0.499 wmUavA
1.497
XP
Vav
(2-49)
^
By
Thus
P.D
= ''-Am)
T\^
^'-''^
by
Thus from
(2-49), (1-27),
and Table
1-3
1 ^ ^ 1.497 av A
1.145
10^
(2-50a)
\m)
'l
when the
pressure
is
measured
in
(/xbars)
750.06 torr,
8.59
96?7
D
(2-56)
11.14
1.145
10*
X 750 X 10"'l]^l \
(2-506)
That
''Am)
when the
pressure
is
\
(2-49) it
is, at the transition pressure the diameter of the tube is about 11 times the molecular mean free path. Since for air at 20C the viscosity r] 1.829 x 10"* poise, the mean
free
measured
is
1
293
8.59
829 X 10-*
x 10-3
-^
so that
,n
\28.98/
(2-51)
1.57i)
(2-52)
where A is given in centimeters and in torr. Thus for air at room temperature the minimum value of the conductance of a tube occurs at a pressure
-i
P P
(2-57)
That
path of the molecules is 1.57 times the diameter of the tube, the parameter GLjD^, and therefore also
is,
free
min
X 10-3
X 10-3
1.57Z>
3.18
10-3
40
in
GAS FLOW
41
meters.
so that the transition pressure range within which the character of the lOP, to P, O.IP,. flow changes ranges essentially from P 28.98 and 293K, rj 1.829 x 10^* poise, For air at 20C g,
M=
Ft-
10-3
X 10-3 X 11.14
5.57
10-2
(2-58)
is
D =
P,
(air
20C)
293
95.7
-
V-^
1.829
X 10-4
1 cm is a tube of i) so that, for example, the transition pressure for 5.57 X 10-2 torr or 55.7 fi. . r dependence ot At the transition pressure as defined above, the The pressure conductance on pressure is inconveniently complex.
28.98,
D
(2-65)
5.57
X 10-2
D
0.557
so that
predominantly viscous region over which the conductance changes from defined in terms of be can character to predominantly molecular in limit P and a lower upper an between bracketed
the second limit Pi. convenience For character. in viscous term, the flow is predominantly Then greater by a factor of 10. let us require that the first term be
is
lOPt
5.57
D
X 10-3
(2-66)
(2-45)
and
(2-46)
and
O.IP,
(2-67)
ax
106
I
+ +
ex
(2-59)
dx
is given in torr and the pipe diameter in centimeters. given the range of pressure for the transition region for various pipe sizes in the case of air at 20C.
when the
In Table
pressure
2- 1 is
Table
[(106c
2-1.
-{(106c -a)
2ad
a)'
40abdf^}
Transition Pressure Ranges for Various Pipe Sizes fob Air AT 20C
Transition pressure range, torr
Pipe diameter
Centimeters
0.254 0.635
1.27
meaningless negative in which, as before, the negative sign leads to a for a, 6, c, and d, value of X. Using the positive sign and the values
the result
is
Inches
0.1
Pi
2.2 X 10-2 8.8 X 10-3 8.8
Pt
0.22 X 10-2
Pu
2.20 0.88
^
P
if
= 948(^f
(2-60)
0.25
0.5
1.0
4.4 X 10-3
2.2
1.1
2.54
so that,
by comparison with
(2-55),
9.91P,
(2-61)
5.08 10.16
20.3
X 10-3 X 10-3
0.44 0.22
0.11
5.5 2.8
1.4
X 10-2
5.5 X 10-4
2.8 X 10-4
1.4
X 10-3
Correspondingly,
times the
first,
10 the second term on the right side of (2-46) is character. in then the flow is predominantly molecular
1
40.6
16.0
X 10-4
1.4 X 10-3
That
is,
lOax
Solving for the value of x yields
=b
+ cx + dx
rfsM
^2-62)
pressure range
Also for air at 20C the conductance of a long tube in the transition is obtained bj^ substituting the above constants into (2-45), with the result
C
(2-63.)
(liters/sec)
1)4
252.lPavi)-D3
12.12
(2-68)
178.7Pa
=
so that,
10.99
311.7Pavi)T
Pi
0.114P,
(2-64)
Note that the first term on the right side of (2-68) is and is therefore the viscous-flow term. Equation
identical to (2-23a)
(2-68)
may
also be
42
GAS FLOW
43
(liters/sec)
=
=
12.12
^|l4.74Pav-D
252.lPav-D
311.7Pav>
(2-69)
pass through the transition region rapidly as the system is pumped down. In such cases the laborious calculation of conductance in the
12.12
1
1
where
G = 14.74Pav^ +
+ +
252.lPav-P 311.7Pav-D
(2-70)
not justified and the operation of the by using the viscous-flow or Poiseuille form of conductance given in (2-23a to c) from atmospheric pressure down to the transition pressure P^ and the molecular-flow form of conductance to be discussed in the next section for values of the pressure below P^.
transition pressure range
is
system
is
At
+ +
311.7Pavi>
r"
r rr
in
0.808
r
[
311.7
t%\Z-Zt^]
first term which is then negligible as compared with 14.74Pavi>. The viscous for to (2-23a) in the bracket then dominates, and (2-69) reduces becomes However, when P^^D is very small, the first term flow. approaches unity for negligible as compared with the second, which The value of G is then 1 and sufficiently small values of the parameter.
/
1 1 1
[ill
1
/ /
180
1
160
1
low-pressure the conductance is given simply by 12.12D3/L, the next section. In the in discussed be to conductance the of value parameter Table 2-2 are given values of G from (2-70) as a function of the
limiting
/
f f J
140
120
/
/ ;
00
rrJ
Table
2-2.
ll
/
/;
100
80 60
/
,
y
12.12
; ;
rt:
i
^
Pe^yD,
torr
yj
K**
2 3 5 7
1
40
20
cm
1
'11
'.-'
^ji
2 3
L
5 7 ]q-I
10-5
io- 10-3
10-2
10-1
1
IQ-5
7jQ-4
7|q-3
7 jq-2
2.289
15.55
148.2
10
102 103
1,475 14,740
2-8. Gas Flow at Low Pressure. At low pressure, i.e., at values of the pressure at which the mean free path for collisions between molecules is long as compared with the dimensions of the tube or
These values may be obtained from Fig. 2-3 by dividing the Pav-D. values of the term CL/D plotted in the graph by the factor 12.12. In designing a vacuum system in which prolonged operation is expected to occur in the transition pressure range, calculation of conductances by (2-45) or (2-68) may be justified. However, in most practical systems the pressure in the forevacuum portion of the system remains in the viscous-flow regime during the crucial period of operation,
conduit through which gas flows, the mechanics of flow are entirely from those at high pressure. The gas molecules move in random directions with a velocity distribution characteristic of the temperature as given by the Maxwell-Boltzmann distribution law
diff"erent
pump
(1-21), and by pure chance individually progress from one point in the system to another. Collisions between molecules are very rare events, whereas collisions with the walls of the system dominate so that the molecules, instead of jostling each other by collision processes, move independently of one another. Pressure is not transmitted from one
44
GAS FLOW
thus
45
momentum from region in the system to another by direct transfer of region of lower molecule to molecule, thus producing a flow toward the the molecules pressure; instead, the transfer of momentum is between character of and the walls of the system. In spite of the independent occur motion of the individual molecules, net flow does nevertheless lower of from the region of higher density (or pressure) to the region a leaving molecules density. Net flow results because the number of the to proportional unit volume of any given region in the system is in the unit of arriving number the whereas region, that density in those other volume from elsewhere is proportional to the density in By a purely statistical effect, therefore, net flow is always in
regions.
dp
'di
dn mn (2kT\^ mu- = ^m su
dt
2tt^\
Az
(2-72)
Since the rate of change of momentum represents a force exerted by the gas molecules on the tube A/ = dpjdt, the tube reacts with a retarding force on the flow of gas of this same magnitude. This retarding force acts over the cross section of the tube so that the change
in pressure
is
1
dp
(2-73)
dt
A
where
(2-72)
and thus from the direction tending to equalize the density everywhere In the pressure regions of higher to regions of lower density or pressure. process is called regime for which the above conditions hold the flow molecular-flow rates molecular flow. Approximate formulas for princithrough tubes and apertures of various shapes were developed
pally 2-9.
is
we have
Combining
(2-73) with
mnl2hT\^s
277-^ \
(2-74)
by Knudsen.^ Conductance
of a
Long Tube
at
Low
Pressure.
In a
is
(see Fig. 2-4) tube through which gas is flowing at very low pressure wall at the the the molecules move in random straight lines, striking Maxwell-Boltzmann end of each free flight. If the molecules have the
AP
"a7
(2-75)
distribution
(1-21)],
of
velocities
[Eq.
is
Q = PAu
from which
/jh&T cm*/sec
(2-76)
molecules impinging on a square centimeter of surface area each second is given by Eq. (1-31) as
Fig. 2-4. Motion of molecules at low
pressure.
V
Pu =
QjA.
WWav
/2)fcT\^
2n^
Az of the
AP/Az
in
(Pi
Pa)/^
dn
dt
n /2kTf
Stt'-^V
which Pj and Pj are the values of the pressure at the ends of a section of tube of length L. Substituting this value of AP/Az into
(2-77)
and solving
for
cross section
in
the periphery of the cross section of the tube, which might be circular or have any other shape. If each molecule is completely stopped by the impact at the surface and then reemitted randomly, there is a net momentum transferred to
which
s is
(2-78)
I
the wall of the tube, provided that there is a mean drift velocity u in the direction of flow. The momentum transferred to the increment of length Az of the tube will therefore be mu on the average for each molecule hitting the wall in the segment Az. This momentum transfer
The above derivation contains the implicit assumption that a uniform u is superimposed upon the random Maxwell-Boltzmann distribution of the molecules. Knudsen has shown that one should more reasonably assume that the superimposed drift velocity of a molecule is proportional to its thermal or random velocity. On this
drift velocity
46
GAS FLOW
47
numerical factor in modified assumption Knudsen found that the along a tube of flow the (2-78) must be multiphed by S/Stt, so that uniform cross section is given correctly by
when
D and L are D
measured
Cair
in centimeters,
and
(2-82a)
13.82-
cfm
(2-79)
when
is
in inches
and
in feet.
of which proves to agree with experimental results. The conductance therefore is section cross uniform a segment of a long, straight tube of
Q
'P^^Hp^
/2fcr\^^^
J
~ 3^\ m
of a long, straight tube well removed from the ends. They also apply to the case of a tube for which the length is very large as compared with the diameter so that the end effect is small. If the tube is short, however, an end correction is required
to obtain results which are even
a segment of length
sL
= ^,(1.29xl0^)y
3.44
^
cm^/sec
(2-80)
and
finite
length
L connect-
X 10V?'\'^^'
_34^/T\^^
liters/:
of the tube
is
decreased to zero,
crossFig. 2-5.
the result
is
an aperture of
Tube end
effects.
by
sectional
area
A =
ttD^I^.
The
which A^js
ttD^JW,
is
therefore
C
3.81
103
[mJ
IP
T
and
cfm
cm^/sec
formula for the conductance of the tube must become equal to that of the aperture as the length of the tube shrinks to zero. In order to complete the derivation of the conductance of a short tube, including end effects, it is necessary first to derive the conductance of the aperture. 2-10. Conductance of an Aperture. In accordance with (1-32) the number of molecules which pass through a circular aperture from the region at the left is
(2-81)
1i
liters/sec
when
D and L
% M =im\) m
277^^
(2kTY^
-^
ttD'^
i2kT^
C
when
4.34
(2-81a)
air at
D^
molecules/sec
(2-83)
B is
measured
in inches
and
in feet.
For
room temper-
Similarly, the number of those which pass through the aperture from the region at pressure P^ on the right is given by
ature (20C)
,
q^=v^A=-i--\
.if,
n^7T''^(2kTY
Z)2
molecules/sec
is
(2-84)
Ml
(7air
28.98/
Z)3
so that
at
Pj to the region at Pg
then
(3,810)(3.181)
g-i
277
^2
=
D"^
i)3
8 \
-:-
12.12
X 10^^^
cm^/sec
Y ^i \mkTI
(2-82)
(Pj
(wi
n^)D^
molecules/sec
Pj)
(2-85)
12.12-
Li
liters/sec
= PjkT
48
GAS FLOW
of area
49
put in terms of the volume of gas leaving the flow occurs, then region of higher pressure Pj from which the net
q (molec ules/sec)
at low pressure
is
^JcT
_L-I
J(Pi
Pa)
molecules/sec
(2-91)
^^.^^^^
(2-86)
n^ (molecules/cm^)
Pi
The flow in quantity of gas is defined as Q and in this case from (2-86) is given by
(dldt)PV(fihaT cm'/sec)
is
a special case.
= P,=qkT
n.
3.64
X lO'(^) '^(Pi
P2)
/^laar
cm^/sec
at
(2-92)
= !^(?^f Z).(P, _
8
\
P :)
(2-87)
low pressure
is
by
expression becomes
C=
3.64
cm^/sec
Q=!^(1.29xlO*)(|)V(Pi-P.)
=
(2-88)
3.641
TY A
1
liters/sec
(2-93)
2.86
10^ (-^)
D%Pi -
P2)
/^bar cm^/sec
when
is
measured
a circular aperture.
is
Also
rp\Vi
The conductivity of a
circular aperture
thus
C=
cm^/sec
49.
cfm
(2-93a)
P,-P
/
2.86
X lOM-
D^
when
(2-89)
2 86
\MI
TY D^
liters/sec
C =
when
11.6^
liters/sec
(2-94)
when
is
A A
is
C = C = 39.i/-|f Z)2
when
cfm
(TIM)'-^'
159 A
cfm
(2-94a)
(2-89a)
when
is
3.181,
(2-90)
2-11.
Conductance
when
D is measured in centimeters,
C =
and
cfm
(2-90a)
to the conductance of a tube of Joining circular cross section and limited length, as shown in Fig. 2-5. may be tube the two regions at pressures Pj and Pj respectively, separated and the are which considered as an aperture, the two sides of
End
Effect.
at
Low
125.31)2
when
D is measured in inches. Note that the above derivation could have been carried out for an aperture of any shape since the result given in (2-89) depends only on the cross-sectional area. In general, the net flow through an aperture
tube of length L connected between them. The result is a combination of two conductances in series, that of the tube C^ and that of the aperture Co, so that according to (2-9) the resultant conductance is
-=
or
\
C=
7;
(2-95)
50
in
GAS FLOW
51
which C^ is given by (2-81) and Cq is given by (2-89). Thus for a tube of limited length L the conductance at low pressure is
to (2-95), yield
a262
C C
2.86
(3.810)(2.86)
9.71
X
'
103
X 10<'(TIM)(D^IL)
-f 3.810
W)
(a
b)L
% ab
cm3/sec
X 103(T/Jf)'^Z)2
X \0^{T j M)'^^
cm^/sec
{D'' j L)
j'VA
9.71
3.810
lO''
(1
2)3
MI
(a
+b)L
+% ab
^
liters/sec
(2-100)
*AD
liters/sec
y\!^
3.810
(2-96)
(I)
measured
in centimeters,
b and in which the length L of In the case of a long slot in which a the slot in the direction of flow is not necessarily very large as compared with the width b of the slot,
when
and
are
and
C
Finally, for a long,
9.71
T ab^ Ml L + %b
slot
liters/sec
(2-101)
narrow
'^ b,
in
direction of flow
is
also large {L
> 6),
ab''
the conductance
when
"ls
liters/sec
(2-102)
C =
Z)3
12.12
109
L +
cm^/sec
by
In Eqs. (2-100) to (2-102) the conductance for air at 20C is obtained setting 9.7l(TjMy-^^ = 30.9, so that, as an example, the conductance of a slot from (2-102) is
Z>3
12.12
L +
%D
7)3
liters/sec
(2-97)
C =
ab^
30.9
-
Ij
(2-103)
when
and
are
measured
in centimeters,
and
cfm
(2-97a)
C =
13.82
O.llD
when D is measured in inches and L in feet. The above calculation was carried out for a tube of circular cross section. The conductance of a tube or conduit of uniform cross section of any shape can be derived by combining the conductances
given
in (2-80)
Another geometrical form of interest is the annular region between two coaxial tubes, of which the inside diameter of the outer wall is Dj, and the outside diameter of the inner wall is D2. Considering the periphery to be made up of the sum of the circumferences of the inner and outer walls and following a procedure similar to that for a tube as developed in (2-71) through (2-96), the conductance of an annulus is
found to be
(D,
Hi
liters/sec
(2-104)
D,)
and
which for
and the periphery s. Thus, for a channel of rectangular cross section with sides a and h and length L, the conductance at low pressure is made up of the two components
C,
9.71
air at
20C
is
C =
{D,'
12.12
D,^)(D,
D,)
liters/sec
(2-105)
103
TV Ml
lO^I
a^ft^
{a
+
ab
(2-98)'
b)L
(2-99)
2-12. Clausing and Monte Carlo Corrections to the Knudsen Conductance Formulas. The method of Knudsen, by which the
/rpVA rpVA
and
Co
3.64
above formulas for conductances at low gas pressure were derived, is only approximately correct, even for simple tubes and conduits of uniform cross section. Clausing3 has carried out a much more exact
52
GAS FLOW
53
calculation for tubes of circular cross section and has shown that Knudsen's formula (2-96) gives conductances which are too large for short tubes (i.e., tubes for which the length is not very much larger
3. The walls of the geometries are microscopically rough so that molecules are diffusely reflected according to the cosine law.
For more complicated configurations, such as occur in manifolds, baffles, and vapor traps, only very crude approximations can be made of the conductance using Knudsen's method. In of this case the procedure is to represent the system by a combination computed are conductances individual tubes, slots, and apertures. The
with
(2-9)
and
(2-10).
first
geometry
in series or parallel
by use of
(2-9)
or (2-10).
of which
is
The value
in
Recently, progress has been made in computing conductances at low pressure by applying Monte Carlo methods, i.e., by tracing individual molecules through the system analytically. The availability of large large electronic computing machines has made it possible to trace the determine thereby and molecules numbers of randomly selected
Table
2-3.
Knudsen and Clausing Conductance Factors and Ebbor Knudsen Factor for Tubes of Vaeious LjD Ratios
L/D
Kk
1.000 0.727 0.571
Kc
1.000 0.672 0.514 0.359 0.281 0.232 0.172 0.114 0.0613 0.0420 0.0319 0.0258
conductance from the net behavior of a large sample of molecules. Davis, Levenson, and Milleron*-^ have carried out a series of Monte Carlo calculations for the configurations shown in Figs. 2-6 through 2-13 and have compared the computed conductances with carefully
0.5
1
8.2
11.1
2 3
measured values.
In the computation a randomly selected entering by numerical computation. At each collision followed molecule is is assumed to be effectively absorbed and molecule wall the with the is then assigned random numbers molecule The promptly reemitted.
4
6
10 20 30
11.4
9.3 7.8 5.8
2.6
2.0
1.2
and direction after leaving the wall. The on the assumption of a Maxwell-Boltzmann based is velocity selection of the wall. The selection of direction temperature the distribution at emission, i.e., the molecules leaving a law of Lambert's is based upon
to specify the velocity / cos Q, unit area of the wall are distributed in angle according to Ig respect with angle the Q at leaving second per number where I^ is the leaving second per number the is and surface the / to the normal to
40 50
0.9
0.8
of Co in this case is the conductance of a circular aperture as given in (2-89). Thus the value of obtained from Knudsen's formulas for this case is
Kr
C
c:
3.810
2.86
Z3
normal to the surface. Davis, Levenson, and Milleron adopt as a reference the conductance geoCo of an aperture of area equal to that of the opening into the
metrical configuration (tube, elbow, baffle system, etc.) being investigated. The computed and measured conductance C is related to Co
D
Dl
3L +y3D
(2-107)
by a geometrical
factor
K such that
C = ZCo
(2-106)
in the calculations
Clausing's corrected values of this factor K^, arrived at by more elaborate methods, cannot be expressed analytically but can be
in
K^
arrangement are
The flow is steady-state with the molecular mean free path longas compared with the dimensions of the system. 2. The geometries under study connect effectively infinite volumes, i.e., volumes large enough so that diffuse flow is not inhibited.
computed by the shown on the graph. For this case it will be noted that the Clausing and Monte <-^arlo computed values are in excellent agreement and that the measured
are also
values approximate these results very closely, but are generally lower
54
GAS FLOW
55
than those obtained from the Knudsen formula by as much as 11 per cent. The good agreement obtained in this simple case between the Monte Carlo and Clausing calculated factors and the measured values lends confidence in the results obtained by Davis, Levenson, and Milleron in the more complicated geometries for which the Knudsen and Clausing methods are not so easily applied. The second geometry investigated by Davis, Levenson, and Milleron is a 90 elbow, the conductance factor of which, calculated by the
D,^)
(2-108)
:o
Knudsen Kk
Clausing Kc
h-A-H
The conductance for the annulus according to the Knudsen method The conductis given in (2-104).
ance factor in this case
is
then
Kk
C_
D,-D,
3L
Vz{D,
D,)
X
\
(2-109)
Measured values
and
is
denoted
Calculated byClousing tor tube
by
=
1
DJD2,
k
then
(2-109)
becomes
tor
elbow
L.
3 (A +
B)/R = L/R
{\-k)+
HiLjD,)
(2-110)
Fig. 2-6. Molecular-flow factors for a tube. [Taken with permission from L. L. Levenson, ^N". Milleron, and D. H. Davis, in 1960 Vacuum Symposium Transactions (Pergamon Press,
Fig. 2-7. Molecular-flow factors for 90 elbow. [Taken with permission from L. L. Levenson, N. Milleron, and D. H. Davis, in 1960 Vacuum, Symposium, Transactions (Perga-
L/Dz
London,
1961).]
mon
Press,
London,
1961).]
Monte Carlo method, is compared in Fig. 2-7 with that of a tube for which the length L is equal to the axial length of the elbow A + B. The computed values of K do not differ significantly from those of a straight tube such that LjD of the tube is equal to (A + B)ID of the elbow. (Note that R = Dj2 is used in plotting the figure.) The conductance measurements are in good agreement with this conclusion.
L/Dj a-re given in Table 2-4 and are shown graphically in Fig. 2-8, together with values of K computed by the Monte Carlo method.
Fig. 2-8. Molecular-flow factors for an annulus. [Taken with permission from L. L. Levenson, N. Milleron, and D. H. Davis, Le Vide 18, 42 (1963).]
It will be noted that deviations of the order of 10 per cent occur between the two calculations, but that the Knudsen formula for the conductivity of an annulus is surprisingly good.
The importance of this result is that in the low-pressure or molecularflow regime very little additional resistance to flow is introduced by the presence of bends in the pipeline. The conductance for a tube with bends is the same as that for an equivalent straight tube with length' equal to the axial length of the bent tube.
Fig. 2-8.
Results for the conductance of a cylindrical annulus are shown in In this case the conductance of the aperture calculated
In Figs. 2-9 and 2-10 the molecular-flow factors for various louver and chevron types of baffles are plotted as a function of the ratio of length to the width of the slots for various baffle angles. Baffles of the chevron type are frequently used in vapor-trap design, the main lunction of which is the trapping of condensable vapor in vacuum ^y^*nis as is discussed in Chap. 8. The experimental and theoretical (Monte Carlo) results of Davis, Levenson, and Milleron are given in the gures. The 45 chevron is most frequently used in vapor-trap design. he value of Kj^ = CjC^ for such a baffle can be computed according
56
GAS FLOW
57
Table
= DJD2
0.5
the width of the slot required for application of the slot formula The length of the flow path is iy = 2'-^ d, so that on substituting into (2-101) the conductance is
and
is
(2-101).
LID2
0.25 0.75
0.90
^ = ''Am)
The conductance
factor
is
2y^d
%m^) =
0.28
'''''''
(2-112)
1.000
1
1.000
0.571
2 3
4 6 10
thus computed to be
1.030
(2-113)
^K
It
3.64
to Knudsen's formula for the conductance of a long, narrow slot as The area oi given in (2-101), which is applicable if a rf in Fig. 2-10. the aperture is A = ad so that the conductance of the unobstructed aperture is, according to (2-93),
>
is interesting to note that this result is not drastically diff'erent from the value 0.25 calculated for a long 45 chevron slot by the Monte Carlo method and shown in the solid curve in Fig. 2-10. The measured values of conductances reported by Levenson and Milleron for the
45 chevron baffle are generally about 20 per cent below those computed by the Monte Carlo method. Furthermore, in a practical case cooling
3.64
(rplates
is
(2-111)
tubes for liquid nitrogen or other refrigerant will be attached to the baffles with the result that the effective area will be somewhat reduced.
Considering these factors, a realistic value of
dj2^^
chevron
of the
baffle is
about
0.20.
The
independent
I.O
0.4
Calculated points
-^
Calculoted points
Experimental points
d t~
an aperture of area A Since for such a baffle arrangment the length of the flow path decreases as the spacing between the plates decreases (and therefore
baffle plates
number of 45 chevron
used to
fill
Experimental points
"
A Aa // \
their
60 louver 45 louver
30 louver
0.6
\iV^ N
\
e=60'
is to be expected as long as the inappreciable as compared with their spacing. In a practical situation it is therefore important to choose a baffle thickness not more than 5 per cent of the perpendicular spacing between
number
0,2
04
the plates. In Figs. 2-11 through 2-13 the molecular-flow factors obtained by Davis, Levenson, and Milleron for a variety of useful geometrical shapes are shown. 2-13. Summary of Gas-flow and Conductance Formulas.
01
0.2
00
o/d
Cole.
00
Cole.
10 Exp.
o/d
10 Exp.
Fio. 2-9. Molecular-flow factors for louver geometries. [Taken with permission from L. L. Levenson, N. Milleron, and D. H. Davis, in 1960
Fig. 2-10. Molecular-flow factors for chevron baffle geometries. [Taken with permission from L. L. Levenson,Jf. Milleron, and D. H. Davis, in
The following summary of the gas-flow and conductance formulas derived and discussed in this chapter is provided for convenient reference. The gas flow in vacuum systems is usually maintained by a pump for which the pumping speed <Sj, is defined as the volume per unit of time which the pump removes from the system at the inlet pressure to the pump. Conversion factors between the various common units of pumping speed are given in Appendix IV. The gas flow, or
throughput, into the
pump
is
defined as
Vacuum
Symposium
Transactions
1961).]
1960
actions
1961).]
Vacuum Symposium
(Pergamon Press,
Q =
-The
PinS,
(2-1)
Trans-
London,
common
units
and conversion
Appendix IV.
58
GAS ELOW
is
59
It follows
in a
vacuum
defined such
(2-4)
g =
(Pi
P,)C
C
if
(cfm)
=
L
9.46
X 10~2 _, p^^
rjL
rj
(2-22c)
in which P^ and P^ are the values of the pressure at the two points in The pumping speed at any point in the system is defined as question.
(2-5)
is
measured in inches,
in feet,
torr.
'^
The
is
=p
For
that
air at
68F
1.829
x 10- poises so
where
is
P is the
If Si and respectively
pressure at the point where S is defined. ^2 are the pumping speeds at points where the pressure
C
and
(liters/sec)
=
517
179
- Pav
Pav
D*
(2-23a)
(cfm)
1j
(2-236)
(2-7)
The units for conductance are the same as those for pumping speed. The pumping speed of a system consisting of a pump of speed S^ with
an interconnecting conductance
is
(a pipe or conduit) of
conductance
for the pressure drop in air along a pipeline is applicable if the pressure drop is small (say not more than 20 per cent) as compared with the pressure o r (2-27a) Pressure drop (torr) = 5.6 x 10~^ -y-^
given by
11
s
>s
when
Sj, is
measured in
liters
c
(2-8)
centimeters
or
.SL
1.9
D*
(2-276)
when Sp
If several conductances are connected in series the resultant confeet.
is
measured
in inches,
and
in
is
given by
1 1
=++ Oj ^S
1
(^2
(2-9)
Flow in the Transition Pressure Region. When the gas pressure is such that the mean free path is of the same order as the cross-sectional dimensions of the pipe or conduit through which it flows, the conductance of a tube of circular cross section is then given by
ductance
is
C(parallel)
Ci
C2 -f C3
=
which
3.269 X 10-2
(2-10)
(^)--O^
1
T\^
0.
147(J//r)''^
P^yDIv
181(J//T)'^ Pavi>/?
(2-45)
High-pressure or Viscoiis Flow. The pressure region of viscous flow is that for which the molecular mean free path is short as compared with the diameter of the pipe or conduit. For these conditions the
in
G is measured
is
L and D are
r]
in centimeters,
is
in grams,
and
is
in poises.
For
air at
7)3/
C
if
(liters/sec)
3.27
10"
rjL
(2-226)
(liters/sec)
_j_
12.12 .^(l4.74
P..D
+
^
252
P D\
3,/,
^J
(2-69)
D and L are measured in centimeters, the viscosity rj is in poises, and the pressure is in torr, where Pav = (Pi + P2)/2, in which Pj and
For accurate calculation of the conductance of a long tube, these formulas should be applied over a range of pressure from lOP^ to 0.1 Pj,
60
gas flow
61
where P;
by
(2-55)
P,
C = G
a^b^
9.71
95.7
MI D
X 10-2
[mJ
1m, (a
+
+
h)L
+ %ab +
y^ab
(2-100)
which
for air at
20C
is
a%^
(air)
5.57
30.9(a
D
Calculation of conductances
(2-69) is
(2-58)
in
b)L
(2-lOOa)
which a and
b are the
by the
seldom
justified in practice.
is
the
viscous-flow value of the conductance as given in (2-22) and (2-23) from atmospheric pressure down to the transition pressure given in
1.0
and the molecular-flow conductance such as that given in (2-97) from the transition pressure on down. Molecular Flow at Low Pressure. At sufficiently low pressure, i.e., when the mean free path is large as compared with the cross-sectional
(2-58)
Calculated points
1
'
Expenmentol points
fnr(R/R/
dimension of the tube or conduit, the conductance is independent of the For most purposes the conductance formulas derived partly pressure. empirically by Knudsen are sufficiently accurate. In the following
formulas the linear dimensions are measured in centimeters, areas are in square centimeters, conductances are in liters per second, temperature The values for air are given is in degrees Kelvin, and mass is in grams. The less frequently needed equivalents for conductance in at 20C. cubic feet per minute and dimensions in inches and feet are given in the
text.
L/Ro
Circular Aperture
C
C(air)
in
Fig. 2-11. Molecular -flow factors for a tube with two restricted ends.
(2-89)
9.16i)2
(2-90)
[Taken with permission from L. L. Levenson, K. Milleron, and D. H. Davis, Le Vide 18, 42 (1963).]
which
D is the
Fig. 2-12. Molecular-flow factors for a tube with two restricted ends and a circular blocking plate. [Taken with permission from L. L. Levenson, N. Milleron, and D. H. Davis, in 1960 Vacuum Symposium Transactions
1961).]
^ ZM{^fA
(air)
Slot of Long,
(2-93)
>
C
in
ab^
9
11.6^
(2-94)
\m] L
+ %b
ab^
(2-101)
which
is
C
Long,
2)3
(air)
=
a
30.9
+%b
(2-lOla)
C ^ 3.810(-) C
in
(air)
'pVA
> 6 and L ^b
IT
\M.
(2-96)
9.71
T
~L
(2-102)
12.12
(2-97)
C L
the length of the tube.
(air)
30.9
(2-103)
which
D is the
diameter and
Two
G =
3.810
CHAPTER
(air)
{D^
12.12
a tube with one restricted end and a [Taken with circular blocking plate. permission from L. L. Levenson, N. Milleron, and D. H. Davis, in 1960 Vacuum Symposium Transactions
By combinations of the above formulas the conductances of many comphcated shapes, such as baffle, structures, can be roughly approximated. The Knudsen formulas are
generally
The most important parameter to be measured in a vacuum system the gas pressure. The pressure of interest may be the total pressure, including both the easily condensable and the permanent gas components present, either the condensable or permanent gas components
is
only
approximate
and
may be greater than the true value by as much as 11 per cent. The results of more accurate calculations and measurements for some shapes are given in the text and in
which
1961).]
such as oxygen, nitrogen, hydrogen, carbon dioxide, etc. The range of pressure over which reasonably accurate measurements are of interest extends from atmospheric pressure down to 10"^^ torr or lower. Gauges and techniques have been developed by which any of the various types of pressure mentioned above can, in principle, be measured with the necessary sensitivity; but particularly for values of the pressure below about 10~* torr ambiguity and error arise from parasitic effects within existing gauges which make accurate determination of the
pressure difficult.
REFERENCES
1.
rU
is
3-1.
Liquid Manometers.
liquid
manometer
consists of a
is
2. 3.
4.
5.
6.
M. Knudsen, Ann. Physik 28, 75 (1909). M. Knudsen, Ann. Physik 28, 999 (1909). P. Clausing, Ann. Physik 12, 961 (1932). D. H. Davis, J. Appl. Phys. 31, 1169 (1960). L. L. Levenson, N. Milleron, and D. H. Davis, in 1960 Vacuum Symposium Transactions (Pergamon Press, London, 1961), p. 372. L. L. Levenson, N. Milleron, and D. H. Davis, Le Vide 18, 42 (1963).
tube partly filled with liquid. One end of the to the system in which the pressure is to be measured. The other end
either
tube
connected
volume above
system
Open
illus-
To the
system
Open manometers
are generally
used to measure pressure relative to atmospheric pressure and may be filled with any liquid, of which water,
and mercury are commonly The engineering term gauge pressure and units such as inches of water and millimeters of Hg for the pressure difference grew naturally from the use of open manometers.
oil,
used.
Open
Closed
63
64
65
Closed manometers are more generally used for measurement of pressure small as compared with atmospheric. An exception to this statement is the mercury barometer, which is in fact a closed manometer designed specifically to measure atmospheric pressure in absolute
units,
i.e.,
for
which
it is
manometer responds
directly to the total pressure in the system and is therefore an excellent absolute standard of total pressure. However, since the reading
closed
manometer
is
first
thoroughly evacuated and then filled to the proper level while still under vacuum so that the gas pressure over the liquid in the closed arm is negligible as compared with any pressure to be measured. The open end is connected to the system so that a difference in level or head between the surfaces of the liquid in the two arms will be just
proportional to the total pressure in the system. level h is related to the pressure according to
depends upon the density of the fluid, which in turn depends upon the temperature, some care must be taken to control the temperature of the gauge when highly precise measurements are required. As an example, the density of mercury changes about 0.5 per cent over the temperature range from to 30C.
3-2.
phragm manometer
in Fig. 3-2,
The
difference in
P=gph
in
The principle of the diameasurement of low gas pressures is shown a cutaway view of a gauge manufactured by
is
a very
common
(3-1)
measurement of pressure
"gauge" pressure
many
engineering applications.
measured in grams per cubic centimeter, the density p of the liquid When the liquid is and the head in centimeters, g = 981 cm/sec^. definition equal to level in millimeters is by in mercury the difference mercury at the density of 0C is 13.59 g/cm^, torr. Since the pressure in from we have (3-1) 10 X P(ubar)
P(torr)
ft(mm)
981 X 13.59
7.50
X 10-P
fibaur
(3-2)
750
mm Hg, which
is
and is defined for some purposes as the standard atmosphere. With some care in the arrangement of a mercury manometer, a pressure of 0.1 torr can just be detected and a pressure of 1.0 torr can
be read with the unaided eye with a probable error of about 10 per cent. For lower values of the pressure, differences in capillarity and sticking However, if the tube in the tube tend to produce significant errors. diameter is sufficiently large (---1 cm) and the tube and mercury kept clean, a manometer can give accurate readings down to 10~^ torr by the use of optical means of magnifying small differences in level. The sensitivity of a manometer can be increased by about a factor of 15 by using a diffusion-pump oil instead of mercury. However, organic fluids are much more susceptible of contamination by dissolving gases than is mercury. For accurate reading of low pressures the oil used in a manometer must be purified frequently by- vacuum distillation
pointer, which on a circular dial viewed through a sealed window in the front of the gauge chamber. Standard models of this gauge are available with scale divisions of 0.2 torr and pressure range from
technology, the reference pressure of interest is zero absolute within the range of sensitivity of the gauge. In the Wallace and Tiernan gauge an evacuated beryllium-copper capsule is used as the pressure-sensitive element. The capsule is mounted in the gauge chamber, which is connected to the system within which the pressure is to be measured. Distortion of the capsule due to the pressure is transmitted through a mechanical linkage to a rotating
indicates the pressure
vacuum
to 50 torr.
diaphragm gauges have been developed with diaphragm for measuring much lower pressures than can be measured mechanically. In the manometer of East and Kuhn^ the diaphragm is in the form of a bellows, into the interior of which the pressure to be measured is admitted. Elongation of the bellows due to the pressure is amplified by a light beam reflected by a small mirror which is tilted by motion of the bellows. By this means pressure changes of 5 x 10^* torr were
sensitive
large variety of
means
detected.
and outgassing.i*
* References indicated
using electrical methods of detecting changes in the position of a diaphragm, highly sensitive vacuum manometers have been developed. One such method, as illustrated in Fig. 3-3, depends upon the capacitance between a diaphragm and a fixed electrode. Movements
By
by
superscript
numbers are
listed at the
end of the
chapter.
changes the spacing, and therelore the capacitance, which can be measured with a capacitance bridge or made a part of a resonant circuit, the frequency of which is measured.
diaphragm
of the
66
67
Connecting wire
Insulator
Electrode
of a gauge of this type, in which a plane diaphragm was used, is described by Pressey* and is illustrated in Fig.
3-4.
Diaphragm
2A
about
Diaphragm manometer
from
J.
and the Physical Society by being concentric rings about a central Chapman and Hall, Ltd., London, plane section. By this construction 1964), 2nd ed.] the sensitivity to pressure changes is increased by about a factor of 10 with no significant change in the response to temperature variations so that sensitivity of about 10^* torr
GEARED SECTOR
[Reproduced 3-2. Diaphragm manometer with mechanical indication. through the courtesy of Wallace and Tiernan, Inc., 25 Main Street, Belleville 9,
Fig.
N.J.]
Pig. 3-4. Cross section of diaphragm manometer designed for capacitance measurement of the differential pressure. [Taken with permission from J. H. Leek, Pressure Measurement in Vacuum Systems (Published for the Institute of Physics and the Physical Society by Chapman and Hall, Ltd., London, 1964), 2nd ed.]
68
69
However, in a panel discussion on the subject of vacuum is achieved. gauges held at the 1960 annual symposium of the American Vacuum Society, P. A. Redhead fexpressed the opinion that mechanical manometers have a lower limit of about IQ-^ torr because of effects, such as
that of vapor molecules with dipole
which p is the density of mercury, L the length of the steel tube, and h the length of the steel tube protruding above the mercury. Then
P =
d^^
^d,^
(^
glpJL
gpm\
is
~h)-
p^L]
moments on capacitance
gauges,
which cannot be controlled. Because the sensitivity of the diaphragm manometer depends critically upon the mechanical properties of the diaphragm, such a gauge cannot be regarded as a primary standard in the sense that a liquid manometer may be. For this reason diaphragm manometers must be calibrated with reference It is also necessary to a liquid manometer. of checking method convenient to have some
'
d^-d,^
d,'
lp
(3-4)
The
obtained by setting
f =
in (3-4)
from which
K=
is
(3-5)
Gloss tube
pressure
the length of steel tube protruding above the mercury level at zero and (3-4) may be written as
periodically.
Thin-wall
steel tube
Mercury
brovin gauge ^ is utilizes the displacement of mercury in such a manner as to produce a sensitivity of the order of 10 times greater than that of the simple mercury U-tube manometer. The gauge consists of a glass cylinder partly filled with mercury and a stainless steel tube, closed at the upper end and open at the bottom, floating vertically in the mer-
P=
The
sensitivity of the
d.
d,^
d,^
9Pm(K
h)
(3-6)
(3-6) is
dj^
dP~d^-d,^'^
for
(3-7)
which is to be compared with (3-1) for the mercury U-tube manometer which the sensitivity is l/gp^. The sensitivity of the Dubrovin gauge is thus greater than that of the mercury U-tube manometer by
the factor
\J>
The gauge is prepared for use by on its side so that the open end of steel tube is exposed and evacuating the
cury.
F
d^^
dx'
laying
it
(3-8)
dj^
For a factor
the gauge so that the residual pressure throughout the gauge, including the region Fig. 3-5. The Dubrovin gauge. inside the steel tube, is very low. While still evacuated the gauge is returned to the vertical position with the steel tube floating in the mercury, as shown When gas is admitted through the connection at the top in Fig. 3-5. of the gauge, the steel tube is pushed down by the pressure more deeply into the mercury. For some pressure P in the gauge the balance is reached when the weight of the tube plus the force exerted on the closed end of the tube by the gas pressure is equal to the change in weight of the displaced mercury. If d^ and d^ are respectively the inner and outer diameters of the steel tube and p, its density, then
Tdi2
10 and d^ 1 cm one finds from (3.8) that d^ 1.05 so that the wall thickness of the steel tube must be about 0.025 cm, or 0.010 in. For such a gauge a change in A of 1 cm represents a
F=
cm
^ +7 4
n
{d^
di^)ps9L
=J
d^')Pm9{^
h)
(3-3)
change in pressure of 1 torr, so that pressure changes of 0.1 torr can be detected with ease. With a sensitivity of this order the Dubrovin gauge is a convenient instrument for the measurement of pressure in the range below that easily read on a manometer, but above that normally reserved for the McLeod gauge discussed below. 3-4. The McLeod Gauge. 7 By combining a liquid manometer with means of compressing a sample of gas as is done in the McLeod* gauge, the range over which the pressure can be measured can be extended considerably below the practical limit of about 10-^ torr for the mercury manometer. The essential elements of a McLeod gauge are shown in Fig. 3-6, and consist of a glass bulb with a capillary tube extension on the top, a side arm connnecting to the vacuum system,
raising
70
71
which holds
system.
and
h^ as the
mercury
is
raised in the
mercury, although in a few exceptional instances organic fluids of low vapor pressure have been used.
From
evident that
When the mercury level in the gauge is lowered below the branch point A the bulb of volume V is connected
to the system through side
{K
K){K
1,000
h)-
PV =
const
(3-12)
arm B. The gas in the as that in the pressure bulb is then at the same raised, the bulb level is system. When the mercury of gas sample the is cut off from the side arm and
compressed into the capillary Cj. The capillary C^ is in parallel with a section of the side arm B and has the same bore as C^ so that the surface tension or capillary effect is the same. The difference in level of the mercury in G^ and Cg is therefore due to the pressure difference resulting from compression of the
provided that the pressure is due to a permanent gas as defined by In using a McLeod gauge this point should be periodically (3-9). checked, i.e., the mercury should be raised to two or more levels, the values of A.2 a-nd h^ measured, and the criterion given in (3-12) checked. In an extreme case the remaining gas present in the system may be due to a substance for which the vapor pressure at room temperature is In that case the pressure will increase during compression only P^. to the point at which P = P^, beyond which condensation will occur and the pressure will be independent of the volume of the sample. In that case for a condensable vapor
sample from the large volume V into the small -'volume of Cj above the mercury level. The pressure of the compressed gas in the closed capillary is proportional to (Ag ^i) + ^O' ^^ which hi and h^ are the heights in millimeters of the mercury
in capillaries G^
Fig. 3-6 gauge.
h^-hi = P^ =
const
(3-13)
and
C^,
and P^
P
is
is
the pressure in
McLeod
the system
ratio
is
still
present in Cg.
com-
pared with
hi
A.2
K- The
pressure of the compressed sample of gas is torr within the limit of reading error when
If the
and ^2 are measured in millimeters. manent gas only during the compression
gas law (1-1)
which is in considerable contrast with the criterion in (3-12). If the vapor pressure of the contaminant in the system is fairly low and some permanent gas is also present, a behavior somewhat in between that of (3-12) and that of (3-13) will result. The important point is that if criterion (3-12) is not obeyed, the pressure readings as determined by a McLeod gauge will not be valid. One can then conclude that the system or the gauge itself is contaminated by a condensable material, the room-temperature saturated vapor pressure of which is given approximately by (3-13). Returning to the measurement of permanent gas pressure with a McLeod gauge we find that the pressure from (3-11) is
PV = P'V
in which
(3-9)
P=
A McLeod
level
(^2
hi){hhi)a
(3-14)
and P' are the pressures before and after compression, respectively, V is the volume of the bulb (i.e., the volume of the closed portion of the gauge above the cutoff point A), and V, the volume of the closed capillary above the mercury level h^, is given by
(^0
1,000F
gauge may conveniently be read by bringing the mercury up to the point where ^2 = h^ (i.e., the level in the open capillary opposite the end of the closed capillary) or the mercury level can be set at some standard level h^ in the closed capillary. In the first method
with ^2
(3-10)
hi)a
^0 the pressure
is
F'
1,000
{h,
1,000F
h,Y
k^
= =
ki(M)i^
a/l,000F.
(3-15)
where
is its
the effective height of the closed end of capillary Cj and a cross-sectional area in square millimeters. Then
h^
is
in
In the second
method with
h.
PV =
(^2
hi){ho
hi)a
(3-11)
1,000
^=
a(h
h,)
1,000F
^^'
^^^
^^^^^^'
(3-16)
72
in
73
which the constant of the gauge k^ = a(^o - h^)j\,()()QV. In each method A^ is the difference in mercury level in the open and closed capillary when the mercury level is set in the prescribed manner. The first method leading to the formula (3-15) results in a pressure reading proportional to the square of the reading, whereas the second method leads to (3-16), in which the pressure is proportional to the first power of the reading. The sensitivity can perhaps best be defined from (3-15)
when AA
gauge
is
= 1 mm,
which
is
On
McLeod
P.
=
4,000 X 200
3.9
X 10- torr
(3-17)
is a practical and useful sensitivity for vacuum measurements. The McLeod gauge has a unique role in the measurement of pressure in vacuum systems and is frequently used as the standard gauge for calibrating most other types of low-pressure gauges.* As can be seen from Eqs. (3-15) and (3-16), the cahbration of a McLeod gauge depends only upon the measurement of the volume V of the bulb and the crossThe volume of the bulb can be sectional area of the capillary tube. measured with great precision by inverting the gauge, filling the bulb and tubing up to the branch point A with mercury, and weighing the mercury. The cross-sectional area of the capillary can best be measured by filling a measured length of the capillary with mercury and weighing
which
^0 can more easily be determined and also to avoid an exaggerated tendency of the mercury column to stick whenever the mercury level comes within a millimeter or so of the closed end of the capillary. The effective height of the closed end of the capillary cannot, in general, be determined accurately by eye because of irregularities near the end of the capillary produced in sealing the end. The true value of h^ can be determined by applying criterion (3-12) to the gauge, which is thoroughly trapped to eliminate condensable vapors, and choosing a value of h^ which fits (3-12) best for several values of h^ and h-^. The McLeod gauge is inherently a cumbersome instrument to use in the pressure range from 10-3 to 10- torr, in which it is most needed as an absolute gauge. Since it must be made at least partly of glass, it is a fragile device in which the shifting load of mercury must be carefully supported or disastrous breakage will occur. The interior of the McLeod gauge and the mercury used must be scrupulously clean and particularly free of oil and grease, otherwise readings are meaningless and the mercury sticks in the capillary, refusing to come down when
the mercury level is lowered. The connecting tubing for a portable McLeod gauge is frequently a source of error since to be convenient in use its diameter must be fairly small and its length typically a meter or more.
The conductance of such a connecting tubing is very small, usually not more than 0.1 liter /sec, so that a small leak at the gauge end of the tubing can give rise to an unexpectedly large discrepancy between the and that seen by the gauge. Such an error can and estimated if the gauge connection can be closed off next to the system and the pressure rise in the gauge due to leakage measured for a specific time interval, such as 5 min. The procedure is
pressure in the system
easily be detected
the small sample of mercury. Since capillary tubing is not necessarily of uniform cross section, a length of tubing must be tested and a section of sufficiently uniform diameter chosen. By placing a drop of mercury in the tubing, moving it along the tube, and measuring the length of the mercury column formed at several positions along the tube, the
variations in diameter can be easily determined and an acceptable section found for making both the open and closed capillaries Cj
to take a normal reading P^ with the gauge, then close off the line near the system and wait several minutes and take a second reading P^.
and
C2.
is imExperience has shown that a bore diameter less than 1 practical because of the tendency for the mercury column in a finer capillary to separate, leaving a bead of mercury plugging the closed capillary after a reading has been taken and the mercury level lowered For high sensitivity it is therefore necessary to to empty the bulb. increase the volume F of the bulb rather than to decrease the capillary bore to less than 1 mm. The end of the closed capillary must be sealed off as squarely as possible in order that the zero point of the gauge
mm
appreciable increase of P^ over P^ is an indication that the gauge due to leakage may be serious. To evaluate the gauge error the total volume being filled by the leak must be estimated. This total
error
Any
volume consists of the gauge volume (the bulb and side tube) and the volume of the connecting line up to the cutoff point. As an example, assume that
Pj
P2
<
= =
5
min
300 sec
100
Note: See Sec. 3-13 for discussion of the use of a McLeod gauge with a refrigerated vapor trap for calibrating ionization gauges, and Ref. 51 for a report on observed discrepancies.
*
line)
0.5
cm em
74
75
- L
4
filled
cm^
The leakage
rate
is
the volume
by the
i
time, or
Q^
P =P
Other types of vacuum gauges McLeod gauge is particularly vulnerable because of the low conductance of the line with which it is connected to the system in many applications. In systems built for the calibration of other types of gauges, such as ionization gauges, in the pressure range 10-^ to 10-^ torr no compromises for reasons of convenience should be tolerated in the manner of connecting the McLeod
are also sensitive to small air leaks, but the
{V,
Fe)
torr cmS/sec
Pt-P^
(7 4-7^)10-3
line
is,
torr liters/sec
Set mark
according to (2-82),
G =12.12-^
liters/sec
(2-4)
760
mm
by
Q.
G,{P,
Ps)
C,
AP
is is
-Flexible
in
the pressure in the system when the gauge reading Pj the pressure taken, so that the difference between these values of
which P,
is
which
is
hose
AP =
For the above example
F
^0
C.
q)
McLeod
gauges.
77
0.25
(100)
19.6
cm^
4
9
Q.=
(7
X 10-^
300
(319.6)10-^
9.6
12
125 =
100
is
1.5
X 10-2
Hter/sec
error
gauge The system should be made of fairly large-diameter glass tubing with a high-conductance liquid-nitrogen trap between the McLeod gauge and the gauges to be calibrated, and no questionable connectors with rubber or other organic materials should be used. Several methods, as illustrated in Fig. 3-7, have been devised for raising and lowering the mercury level in McLeod gauges. For consistent readings it is important that the mercury be raised smoothly to the proper level, but not beyond, since there is always a slight mechanical hysteresis in the response of the mercury in the closed capillary due
.
to surface effects.
9-^
1.5
X IQ"" X 10-2
0.64
X 10-3
torr
Thus, for the example given of the reading Pj = 1 X 10-^ torr, the greater part, or 0.64 x 10-^ torr, is the gauge error. Errors of this magnitude and sometimes much greater frequently appear when this
simple test is carried out. The need for truly leakage-free connections to a McLeod gauge and connecting tubing of reasonably large-bore
McLeod
tubing.
for controlling the mercury level is illustrated reservoir filled with mercury is connected to the gauge just below the branch point with a length of rubber
The mercury level in the gauge, riding at about a barometric height above that in the reservoir, is adjusted to the proper height by raising and lowering the reservoir. This simple method has been used for many years although its disadvantages are the contamination of the
75
77
mercury by the
.
considerations.
capillary of 15
the mercury level is case can be made shown in Fig. 3-76 in which the reservoir, which in this extending below the branch of steel, is coaxial with a long vertical tube barometric height of about point a distance slightly greater than the with relatively small 760 mm. The top of the reservoir can be made the mercury is not a clearance at the top so that contamination of and h For sensitive gauges, however, methods a serious problem. reduce to counterbalancing are cumbersome, even when combined with load. mercury heavy the lift to the effort required is located sufficiently below reservoir mercury fixed a In Fig 3-7c for mercury places the level the McLeod gauge that a barometric height to a high vacuum. below the branch point when the gauge is connected loosely fitting wooden The reservoir is in the form of a cylinder with a down to displace most of the or hollow steel plunger which can be thrust fine control can mercury in the reservoir, raising the level. Relatively near the top plunger a threaded collar on the
lifting
from
2.26
to
5.9
X 10- X
150
mm mm
be provided by clamping reservoir housing so arranged to mesh with threads at the top of the by rotating the controlled that the final inch or so of the stroke is
plunger.
achieved by the use In Fig. 3-7d a compact form of McLeod gauge is The level. mercury the of an auxiliary vacuum reservoir to control reservoir the of out mercury in the gauge is lowered by pumping air air, usually above the mercury and raised by admitting atmospheric activated as such through a tube partly filled with a drying agent, convena is valve alumina or silica gel. A two-way stopcock type of reserthe from air removing ient means of switching from admitting to by provided be can Fine control in raising the mercury level voir. of ends both at file notching the inner member of the stopcock with a of development With the the hole through which air is admitted. mercury the as some skill an operator can then reduce the flow rate flow without overshooting level approaches the zero mark and stop the finger or applying a the mark. Tapping the closed capillary with a for the tendency of small mechanical vibrator will help to compensate
the mercury to stick and advance erratically. the The use of a flexible metallic diaphragm for raising and lowering 3-7e. Fig. in illustrated mercury in a compact form of McLeod gauge is The particular model shown is an unusually compact form of McLeod of gauge and has the added feature of a dual range achieved by the use
-,
xu
McLeod gauge.
Measurement in Vacuum
Systems (Published for the Institute of Physics and the Physical Society
Fig. 3-8. McLeod gauge with several scales for an extended pressure
range.
by Chapman and
Ltd.,
Hall,
1964),
London,
2nd
ed.l
In many applications it is desirable to extend the range of the gauge to appreciably higher pressure. One way in which this can be done is to make the closed capillary out of short lengths of tubing of different
bore diameters and for each such section provide a Cg capillary, all in with the side arm, as illustrated in Fig. 3-8. An alternative design due to Romann' is shown in Fig. 3-9 in which a small bulb is inserted between the main bulb and the base of the closed capillary with a second closed capillary tubing connected between the two bulbs.
parallel
two closed capillaries connected to the same reservoir. The pressure range of a McLeod gauge is determined by the length practical of the capillary tube, which is limited to about 15 cm by
The
is
78
79
increased
by about the
Gauges.
K oi
a.
substance
is
H = -Kds
in
(3-18)
which
direction
amount of heat flowing per unit area per second in the parallel to s, and dTjds is the temperature gradient in this
is
the
For rarified gases the kinetic theory of gases provides a derivation* of an expression for the thermal conductivity by methods
same
direction.
from (1-23) and Table 1-3. If the diameter of the filament is very small as compared with that of the tube, each molecule will strike the wall several times before finally hitting the filament and will therefore be in good thermal equilibrium with the wall temperature T^ before hitting the filament. At the low pressure here assumed, however, each molecule which strikes the filament will do so only once before again colliding with the wall. The molecules leaving the filament, therefore, will not be in equilibrium with the temperature T, of the filament, but will be characterized by some lower temperature T'^. Under conditions of temperature equilibrium with both surfaces, the kinetic theory of gases gives for the energy transferred by a monatomic
molecule
The derivation
leads
is
E=
2k{Tf
- TJ
{k
given by
K=
M(9y
5)riC,
(3-19)
which in this case can only be in the form of kinetic energy. However, more generally for diatomic and polyatomic molecules which acquire
vibrational
in
which y
pressure to that at constant volume, and r] is the coefficient of viscosity. But the expression (1-59) for the viscosity is independent of the pressure so that the thermal conductivity is also independent of the pressure over the same range for which the viscosity is given by (1-59). However, as was pointed out in the discussion of that expression, when the
transferred
is
and rotational as well as translational energy, the energy greater by the factor (y + l)/4(y 1) where y = Cj,/C^,
the ratio of specific heats of the gas. Thus, in general, the energy transferred per molecule under the conditions of thermal equilibrium
is
gas pressure is so low that the molecular mean free path is about equal to or greater than the distance between the walls of the containing In that case vessel, the gas is no longer characterized by a viscosity. the expression (3-19) is no longer valid and the conductivity is then found to depend upon the pressure. Therefore, in the pressure range for which the mean free path is comparable with or greater than the dimensions across which the flow of heat occurs, the variation of the
^^)HT,-TJ
(3-21)
However, the situation of interest is one in which the molecules do not come into equilibrium with the temperature of the hot filament so that the energy transfer per molecule is given by a similar expression with T'f substituted for T, where T'f < Tf. Thus, in the case of
interest,
thermal conductivity of the gas with the pressure can be used for the measurement of the pressure. The process of heat flow under these
conditions
is called free molecular conduction. Consider a cylindrical tube with a heated filament of circular cross section running along its axis. According to (1-31) the number of molecules per second striking each square centimeter of the surface of
E
Knudsen
defines the
i.e..
5^7^ ^(^-^J
(3-22)
accommodation
two quantities;
(3-23)
^
f
->
the filament
is
w
is
yinva,T/
_ n/2kTY
2\ 77m;
(3-24)
(3-20)
n/2RoTY'
2\
uM
Combining
(3-20)
81
10^
(0.01 torr) is
from
-I
(100
20)(0.01)
8.87
10-3 watt/cm2
in. (10 cm) long, x 10"^ cm 2 and the gas heat conduction
_
since
gy
2y
from
+ -
-Rq
\^
(rp
^
T )P
'
ergs/sec
cm^
(3-25)
If the filament
is 1
1\27tMTJ
nifc
is
about
7.9
(1-16)
P/T and
the ratio of diameters of the temperature T hitting the filament will be in equilibrium with may be written (3-25) Eq. evaluation numerical of the tube wall. For
in the
by our assumption regarding the tube, the gas molecules and filament the
form
be of the order of 6.9 x 10-* watt. Even under conditions of perfect vacuum (P = 0) in the tube, heat will be lost from a hot filament by thermal radiation. If the surfaces were perfectly absorbing to radiation of all wave lengths, the rate of energy loss would be given by the Stefan-Boltzmann law for "black
body" radiation
in
(3-26)
^^
a
^^^^,
^^,^
^3^^)
(3-29)
which
=
e
..m\^,
in
T^)P
ergs/sec cm''
ized
by the emissivity
which Ao
is
by
energy becomes
If loss of heat
^^
^^^^^^,
^^^^,^
is
^33^^
large as
Ao
Y
2(y
+ 7
Rn
}A
l)\277i/(273)/
llOy
ilf '^
+ -
ergs/sec
1
cm^ C ^bar
watts/cm^ C torr
(3-27)
compared with that due to gas conduction, the latter cannot be used as a means of measuring the pressure because of the large background effect. For the example above, the heat loss by radiation for perfectly "black" surfaces would be
1.47
10-2 y
jfVi
y-1
W,
= =
5.67
X 10-12(373*
293*)
6.80
X 10- watt/cm2
The free molecular conductivity at 0C for a given gas can be calculated the by (3-27) and the resulting value inserted in (3-26). However, depends value its since accommodation coefficient a cannot be calculated and surface not only upon the gas involved but also upon the material For filament. the condition (roughness and adsorbed gas layer) of
the value of a is in the vicinity generally quite low of 0.9, whereas for hydrogen the value of a is For roughened or blackened surfaces a approaches unity (0.2 to 0.5). Since surfaces are generally not highly polished nor
clean metallic surfaces exposed to
air,
result in a
is 7.9 x IO-2 cm2, this would due to radiation of 5.37 x 10-* watt, which is comparable with the loss calculated above due to free molecular conduction at P = 10-2 torr. However, since the emissivities of sm^faces
power
loss
of clean metals at temperatures in the range of to 100C are generally of the order of 0.1, the true loss due to radiation would be of the order of 5 X 10-^ watt, so that radiation loss and gas -conduction loss would
completely free of adsorbed gas layers, the value a for rough approximation. As an example, for air
1.47
0.7
may
be used
become about equal at a pressure of about [(5 x 10-^)/(6.9 x 10"*)] X 10-2 <= 7 X 10-* torr. Inspection of Eqs. (3-26) and (3-30) shows that radiation increases
much
10-2 2.401
1.64
(28.98)'-^
0.401
room temperature
faster with increasing temperature because of the T* dependence than does gas conduction, so that an equality between radiation and gas conduction occurs at higher pressure as the temperature is raised. Therefore, for measurement down to lowest pressure, the filament should be operated at the lowest temperature for which the heat loss due to gas conduction can be measured. Because of the competition with
82
83
radiation loss, thermal conductivity pressure gauges are not normally used for pressure measurements below about 10~^ torr.
a third process by which a hot filament may lose heat to by thermal conduction along the filament to the end mountings. However, this loss can be kept sufficiently small by using a filament of small cross section and heat conductivity. As an
There
is
^Millicmmeter
Thermocouple
Filament
:=: Battery
Microommeter
(low-resistance
type)
Rheostat for
current
adjustment
ID)
Thermocouple gauge; (6) simple electrical circuit diagram. [Taken with permission from Saul Dushman, Scientific Foundations of Vacuum Technique (John Wiley and Sons., Inc., New York, 1949).]
Fig.
3-10. (a)
example nickel has a nearly constant heat conductivity of k <= 0.14 g cal cm/C in the temperature range from to 200C. Exact calculation
of the heat-conduction loss along the wire is a bit tedious because of the temperature variation along the wire. A crude estimate can be made, however, by assuming the central third of the wire to be at the maximum temperature and the third at each end to have a uniform temperature gradient 3(T^ TJjL. By this approximation the heat conducted out both ends to the mountings is
W,
= = =
dT
2(0.239)jfc^
dL
x
10-)
2(0.239)(0.14)(4.9
7.9
3(100
second to watts. For this choice of dimensions the heat lost out to the end mountings by heat conduction along the wire is smaller than that lost by radiation, as calculated above, by a factor of about 750. In general, therefore, the balance between free molecular conduction and radiation is all that needs to be considered as long as the cross-sectional area of the filament is sufficiently small. As we have seen, this balance occurs at a pressure of about 10~^ torr or a bit less for the example chosen. Since these two processes of energy loss are both proportional to the surface area of the hot filament, the balance point is approximately independent of the diameter of the filament. The thermocouple vacuum gauge is a thermal conductivity pressiu-e gauge in which the temperature of the hot filament is measured by a thermoa- 10 couple. The heating current which is passed through the hot filament is kept constant at a standard value independent of the temperature of the filament. As the pressure increases the heat conduction through the gas increases and the temperature of the 16 24 40 48 56 32 filament decreases until the temperaScale reading ture corresponding to the high-pressure Fig. 3-11. Calibration curve for value of the heat conduction through General Electric thermocouple the gas is reached. The thermocouple gauge for Hg, Nj, and Xe. [Taken responds to the temperature of the with permission from Saul Dushman, Scientific Foundations of filament and provides a direct reading Vacuum Technique (John Wiley which can be calibrated against the and Sons., Inc., New York, 1949).] pressure in the gauge tube. The thermocouple type of gauge was first developed by Voege' and has been refined by a number of other investigators. ^"'^^ The thermocouple gauge manufactured for many years by General Electric Company is shown in Fig. 3-10 together with the simple electric circuit frequently used to heat the filament and record the output of the thermocouple. The gauge element consists of a platinumiridium ribbon 0.0234 by 0.0078 cm in cross section and 3.66 6m in length with a Nichrome- Advance thermocouple welded to its midpoint.
20)
10
10- watt
in
the cross-sectional area of the 1-mil wire in square centimeters and the factor (0.239) converts from gram calories per
is
which
The heating current passing through the platinum-iridium element is held constant in the range 30 to 60 mA depending upon the pressure range of interest. In Fig. 3-11 are shown calibration curves of pressure
84
85
plotted against the thermocouple ciirrent for hydrogen, nitrogen, and xenon. About two-thirds of the total deflection occurs in the pressure
range from 0.1 to 0.01 torr. Characteristically the curve becomes very steep at 5 x 10- torr, below which the accuracy of the gauge is
rather poor.
multistation control unit has an auxiliary heating circuit which keeps the elements of all the gauge tubes in the system warm when they are not connected to the pressure-reading circuit, so that there is no delay in obtaining a pressure reading as the control unit is switched from one
tube to another.
Cufrent
Current
Triode
Clevite
CDT1349A
Gauge tube
45 ohms 1wottl7o
Gouge tube
WW
20ohms
Pot
thermocouple Fig. gauge schematic. [Taken with permission from J. M. Benson, in 1956
3-13. Hastings
Equivalent circuit of thermocouple gauge. Hastings [Taken with permission from J. M. Benson, in 19S6 Vacuum, Sym,posFig.
3-14.
WW
1957).]
ium, Transactions
(Pergamon Press,
CCW
London,
1957).]
very successful design of thermocouple gauge manufactured by is described by Benson. ^^ The sensitive element of the Hastings gauge consists of two thermocouples acting in parallel and a third thermocouple in series to compensate for variations in ambient temperature. The gauge elements and circuit diagram are shown in Fig. 3-13. The two thermocouples (A) and (B) are heated in
Hastings-Raydist, Inc.,
Thermocouple gauges are very useful and convenient in a variety of A number of commercial designs are available, of which applications. that manufactiu-ed by the Kinney Vacuum Division of The New York Air Brake Company is of special interest. The gauge tube is of steel construction and is both compact (length of 4 in. and outer diameter of iK in.) and relatively sturdy. The thermocouple elements are Chromel and Cupron, and the heater element is a timgsten wire. A special feature of the Kinney thermocouple gauge is that the tubes are all matched so that the circuit need not be reset for proper calibration when tubes are changed. This is particularly convenient in multistation installations in which a number of thermocouple gaugetubes are connected to a single control circuit through a selector switch. The Kinney control unit is fully transistorized and features a printed
circuit,
by alternating current from a transformer. Thermocouple (C) connected from the midpoint between {A) and {B) to the center tap on the transformer provides temperature compensation. Since thermoseries
is
shown
in Fig. 3-12.
The
and {B) are connected "back to back" in the a-c circuit, they act as parallel sources of electromotive force for the d-c circuit for which the lead from (C) through the d-c meter to the center tap is the common return path. The equivalent circuit for the gauge is illustrated in Fig. 3-14, and a cutaway view of the metal gauge tube revealing the thermocouple arrangement is shown in Fig. 3-15. Multiple thermocouple or thermopile gauges are made for several ranges of
couples {A)
Cutaway view of Hastings gauge showing thermocouple arrangement. [Taken with permission from Vacuum J. M. Benson, in 1956 Symposium. Transactions (Pergamon Press, London, 1957).]
Fig. 3-15.
86
87
pressure determined mainly by the dimensions of the thermocouple wires and the temperature at which they are operated. The ranges for which commercial units are available are 0.1 to 20 torr, 5 to
The constant-temperature
bridge, in
gases are
and 1 to 100 approximately. Calibration curves for several shown in Fig. 3-16. An outstanding feature of the Hastings thermocouple gauge is the speed of response, which is significantly shorter than for most other commercially available gauges. The metal envelope and generally rugged construction are also features of practical
1,000
fj,
fj.,
gauge filament is kept constant by adjusting the bridge voltage to maintain a bridge balance as indicated by zero current through G in In this mode of operation the pressure is approximately a Fig. 3-17.
linear function of the square of the
bridge voltage
P =
in
(3-31)
interest.
10
s.
= "=::
i:r-
"^<
8
6
^
^^:wr
^^.
Acetylene
^^
which Fq is the voltage necessary to balance the bridge when the pressure in the tube is nomThe constant zero. inally
2=-Argo
>
f,
^s
^Carbon dioxide
:^N
Air- *C
Mil
makeup
S s
geometrical
parameters
gauge tube.
2.
-^T-1
pi<-:::='
The
constant-current
10
50
Pressure, microns
100
500 1,000
which the current through the hot filament of the gauge tube is maintained at a steady value.
in
Fig. 3-17. Wheatstone bridge circuit [Taken for Pirani gauge control. with permission from J. H. Leek, Pressure Measurement in Vacuum Systems (Published for the Institute of Physics and the Physical Society
Fig. 3-16. Calibration curves for Hastings gauge of intermediate sensitivity. [Taken with permission from J. M. Benson, in 7956 Vacuum Symposium Transactions
1957).]
About concurrently with the first appearance of the thermocouple gauge, Pirani^* developed a thermal-conductivity pressure gauge in which the resistance of the hot filament was calibrated as a function of the
As the pressure in the gauge tube increases, the thermal conduction of the gas surrounding the hot filament increases, and the temperature of the filament and therefore also its electrical resistance tend to decrease. The usual control circuit for a Pirani gauge is the
gas pressure.
by Chapman and Hall, Ltd., London, balanced when the 1964), 2nd ed.] pressure in the gauge tube is very low, and the imbalance current registered on the meter designated by G in Fig. 3-17 is used as an indication of the pressure. 3. The constant-voltage bridge, in which the voltage across the bridge Because is kept constant, e.g., by means of a regulated power supply. the because and circuitry of the simplicity of the constant-voltage current imbalance the pressure is approximately a linear function of over a limited range in pressure in this mode of operation, the constantvoltage bridge has been widely adopted in commercial Pirani gauge
The bridge
is
Wheatstone bridge, in which one leg of the bridge is the filament of the gauge tube and the other three legs have resistances nearly equal to that of the gauge tube, as shown in Fig. 3-17. It is sometimes advantageous to use two identical gauge tubes in the circuit, one of which is evacuated to a low pressure and sealed ofi". If the sealed-off dummy tube is mounted adjacent to the gauge tube, fiuctuations due to changes in ambient temperature and bridge voltage are to some degree compensated. In the circuit-in Fig. 3-17 the gauge tube is represented by' R and the compensating tube, if it is used, takes the place of R^. The Wheatstone bridge circuit can be operated in any of three ways to provide an indication of the pressure
The bridge circuit is balanced as in the constant-current bridge when the pressure in the gauge tube is very low (below the range of detectable response) and the pressure observed as a function of
circuits.
is
shown in Fig. 3-18. Typical calibration curves" for constant-temperature and constant-voltage-bridge operation are shown in Fig.
Although, with special precautions to ensure a constant ambient temperature for the gauge tube, the Pirani gauge can be designed for operation at much lower pressure, commercial Pirani gauges are useful primarily in the range from 10"^ to 1 torr. A thermistor is a semiconductor element which has a high negative
3-19.
*105
89
10
/
-i.o
/
0.10
/
40
>
B,
/
/
/
; ;
/
(
I /
/
/
/
/
feature,
thermistor-type Pirani
gauges have not until quite recently been successful commercially because of the rather wide variation in propThis producerties of thermistors. tion problem has recently been solved
that thermistors of sufficiently uniform properties can now be obso
tained.
/ / / / /
ll
1/
0.010
1/
20
60
80
100
Fig. 3-19. Pressure -sensitivity curves for {A) constant-temperature operation, and (B) constant-voltage bridge operation. [Taken with permission from A. R. Hamilton, Rev.
Sci. Instr. 28,
ment
693 (1957).]
gauge shown in Fig. 3-20 is now commercially available. The circuit diagram for this gauge is shown in Fig. 3-21, and a calibration curve is given in Fig. 3-22. The relatively linear response
is
is the extension of the useful range to relatively high values of the pressure by the
Insulator
Thermistor bead
Connector pins
Support wires
Fig. 3-18. Typical constant -voltage bridge circuit for Pirani gauge. [Taken with permission from C. M. Schwarz and R. Lavender, Rev. Sci. Instr. 19, 814
(1948).]
88
90
91
negative potential V^ relative to the cathode to ensure that electrons emitted by the cathode or created by collision processes in the annular space between the grid and plate are prevented from reaching
lOOohms lOwatts
Ohmite Brown-Devil
the plate and are essentially all attracted to the grid, which is the most positive electrode. If the grid is made of fine wire, most of the electrons from the cathode miss the grid wires as they are propelled outward by the field and continue toward the plate until
they reach a point in the grid-to-plate region at which the electrical potential is the same as that of the cathode; they are then turned back to oscillate radially, passing through the holes between the wires of the grid repeatedly until they finally strike a grid wire and As the electrons pass through the grid they reach their are captured. maximum kinetic energy of eVg, in which e is the electronic charge (-4.80 X 10-1" esu = -1.60 x 10"" coulomb) and F is the grid potential. The kinetic energy is either in ergs or joules, depending upon the units (cgs or practical) in which e and V^ are expressed. Alternatively, this kinetic energy can be expressed in units of electron volts in which case the maximum kinetic energy of the electrons is equal to Vg electron volts, by which one simply means that kinetic energy acquired by a particle with one electronic charge e falling through a difference of potential equal to F, volts. In atomic and nuclear
Victory
25 A 4 5,000 ohms
or
Fenwal
enclosure of the thermistor element in a metallic cylinder with small The heat flow through the gas from the thermistor element clearances. to the cylinder therefore occurs along a very short path so that the
conditions for free molecular conduction discussed at the beginning of
this section are realized at relatively high pressure.
J^
50
^s
40
k
k.
ohms
/^"/ir^ N^(mAJ
ohms^CLiJ/
3-6.
The hot-cathode
ioni:
zation gauge consists basically of three elements in a gastight tube a thermionic cathode, an electrode usually in the form of a grid for extracting electrons from the cathode, and a positive ion collector or
plate.
30
\\
20
s
volts
ordinary triode-type electronic tube has these elements and can be used as an ionization gauge by opening the tube envelope and sealing on a tubulation by which it can be connected to a vacuum
Any
\
\, \
2
10
system.
N,
V,
The operation of an ionization gauge is illustrated in Fig. 3-23. Electrons from the cathode are accelerated by the electrostatic field through the grid of radius r^, which for this purpose is set at a positive The plate of radius r^ is set at a potential Vg relative to the cathode.
^-2
5
0.001
0.01
0.1
2
Torr
12
10
100
for air.
92
93
commonly quoted
factors are useful
1
the plate or collector are recorded as positive ion current and contribute to a measurement of the density of molecules in the tube. An alternative method of operating a triode-type vacuum tube as an
ionization gauge
is
In this
or conversely
= = = 1 joule = 1 erg =
1.60
1.60 1.60
highly positive
(e.g.,
(e.g.,
somewhat negative
is
25
V).
potentials
an electronic
The
then
6.24
eV
and out through the any are present in
6.24
x 10" eV
if
As shown
heavily space-charge-limited because of the retarding effect of the interposed negative grid. Those electrons which do escape into the gridto-plate region gain energy as they travel outward and strike the plate with their maximum kinetic energy. Ionization occurs mostly in the
region near the plate, and the positive ions are
collected
the tube. If the kinetic energy of the electron at the time of collision with a molecule is greater than the ionization potential of the molecule,
Grid
drawn into and are by the grid. Electronic tubes are traditionally checked after assembly and evacuation by measuring the grid current, which is an excellent indication of the residual gas. This process amounts to
the use of the tube as its own ionization gauge. Because in this mode of operation the electrons do not oscillate back and forth through the grid as they do when the grid is at positive potential, the electron paths
are
an electron
state.
may
Plate
it in an ionized Each such impact decreases the energy of the electron and de-
molecule, leaving
much
shorter
and the
sensitivity as
an ionization gauge
is
poor.
flects
Cothode
Fig. 3-23. Hot-cathode ionization gauge. A typical electron trajectory is shown. The useful region of positive ion production is the shaded area. [Taken with permission from J. H. Leek, Pressure Measurement in Vacuum Systems (Published for the Institute of Physics and the Physical Society by Chapman and Hall, Ltd., London,
1964),
it from its otherwise purely path so that the electron loses energy during its oscillatory motion and becomes more random in its motion. In any case, it eventually falls into one of the grid wires and is captured. Electrons knocked out of
radial
2nd
ed.]
molecules in ionizing collisions may also gain sufficient kinetic energy to cause some additional ionization, depending on the electrical potential at the point at which they are created,
fall
into one
In order to produce ionization by impact with an atom or molecule an electron must have kinetic energy at least equal to the ionization potential, which ranges from 3.89 eV for cesium to 24.6 eV for helium. Of more practical interest in the formation of ions is the probability of ionization Pi, which is defined as the fraction of electrons at a given energy producing an ionizing collision per centimeter of path and per torr of gas pressure. In Fig. 3-24 the probability of ionization is shown graphically as a function of electron energy for several common gases as measured by Tate and Smith, i* Since at a pressure of 1 torr and temperature of 0C the molecular
density
is
of the grid wires because the grid, being positive relative to both the cathode and the plate, is the only electrode which they can reach.
2.69
'
X 10"
760
Those positive ions which are created in the annulus between the grid and the plate are in an electrostatic field directed radially outward. Being positively charged, these ions are driven outward to the plate, where they register as a positive current. Those ions which are formed in the space between the cathode and grid, however, are in a field which accelerates electrons ( ) outward. These ions are therefore attracted to the cathode, where they are captured and are electrically equivalent to electrons leaving the cathode. Only those positive ions reaching
=
the probability of ionization
Pi
in
3.54
X 10" cm-
Wio-j
3.54
lO^cTi
(3-32)
which a^ is the cross section for an ionizing collision by an electron. The number of ions produced by an electron per centimeter of path from
94
(3-32)
95
then
273
273
(3-33)
The use of ionization of gas molecules by electron collisions as a means of measuring the pressure (or more exactly the molecular density) of a gas was first described by Buckley.^'' A simplified circuit for operating an ionization gauge is shown in Fig. 3-25. The positive ion current
i^ to the plate (or ion collector) for a given value of the grid voltage Vg and grid (electron) current i_ is a direct indication of the molecular
20
15
A
lA ^Vl f/^ ^s
'
density and therefore also of the pressure. ionization gauge is defined by the relation
Plate or ion
The
sensitivity s of
an
i^
= si_P
(3-35)
s i_
collector
or
Pi
10
^:
-15 to
volts
"=H
200
-30 :=:
Z^ -^
IVp+P^Vg to
+180
volts
^
J_
400
^:
600
Electron energy, eV
Fig.
in
Fig.
3-25.
Simple
ionization-gauge
C^HalD, 0^(2), ^^(Z), A(4), n^(5), Ne(6), and He(7). [Taken with permission from J. H. Leek, Pressure Measurement in Vacuum Systems (Published for the Institute of Physics
circuit.
Chapman and
1964),
2nd
ed.]
which n is the molecular density corresponding to the pressure P and temperature T of the gas. For an electron stream of current i_ amperes (6.24 x 10^* electrons/sec = 1 A), the positive ion current i^ amperes, assuming all the ions are collected, is thus given by
in
be calculated from (3-34) and the geometry of the gauge tube. However, the average length of path for the electrons and is not easily estimated for tube geometries of practical interest, grid the of value maximum the energy of the electrons varies from the voltage Vg to zero in a complicated way so that direct calibration of each tjrpe of ionization gauge against a McLeod gauge is the only For sufficiently low practical means of determining the sensitivity. contribute to i_ which electrons pressure only a small fraction of the produce two none essentially produce ionizing collisions, and therefore sensitivity the pressure low Therefore at sufficiently or more collisions. is pressure the when However, s is expected to be independent of more produce electrons of the high enough that a significant fraction than one ionizing collision, then the multiplicity will increase with the In Fig. pressure, and the sensitivity is no longer independent of i_. current i_ is electron on the 3-26 the dependence of the ion current ^+ gas nitrogen with gauges ionization shown for three different makes of
and
in principle could
at a pressure of 5
for all three
273
.
(3-34)
The maxima for the Pi curves for most gases occur in the energy range from 60 to 200 eV, above which p^ decreases steadily with increasing electron energy. For the common gases for which curves are shown in Fig. 3-24, p^ reaches maximum values of about 1.5 for helium to about
17 for acetylene so that the
gauges up to values of i^ from about 10"^ to 3 X 10"* A depending upon the characteristics of the gauge tube. Measurement of the sensitivity 5 as a function of the pressure in the range IQ-* to 1 torr shows that s increases with increasing pressure until a maximum is reached and then decreases^" as shown in Fig. 3-27 for nitrogen and helium. The value of the pressure at which the gauge sensitivity reaches its maximum value for helium is seen to be about
a factor of 10 greater than that for nitrogen. governed by the ionization probability Pi, which
is
lO-^ torr.^"
At
maximum
ionization
(Tj
as large for nitrogen as for helium in the electron energy range used in ionization gauges, the pressure difference for the maxima is to be expected. The rise in sensitivity in the vicinity of 10"^ torr for
shown
in the figure.
nitrogen
is
96
VACITtrM SCIENCE
AND ENGINEEEING
PRESSTTBE
97
0.5
and the grid are drawn toward and bombard the cathode, operation of an ionization gauge at pressures much above 10"^ torr greatly shortens the life of the cathode. The upper limit of operation of commercial gauges
is
''
<''^^T
^n
r^
;'/
fp
t/
ff
0.25!
'
therefore usually
2.0
with the result that except near the upper limit of the pressure range, say from 6 x 10"*
set at this pressure
// // it
i
/
1.0
/
of
J\
"
to
is
essentially independent of
lO'
both the
pressure.
10"*
10"
10'
electron current
and the
50
100
150
200
250
Electron current,I_,amp
Fig. 3-26. Ion current as a function of electron current for three different gauges operating on the same pressure of nitrogen gas. [Taken with permission from W. B. Nottingham and F. L. Torney, Jr., in 1960 Vacuum Symposium Transactions
However, for precision measurements detailed calibration of each gauge tube against a IMcLeod gauge over the pressure range from 10~*
to 10-* torr
is
Fig.
3-28. lonization-gauge
sensi-
and neon as a
1961).]
essential.
ionization
mean
free
The
maximum is
to ion-electron or positive-
pends upon the grid potential V^. As shown in Fig. 3-28 for nitrogen (2) and neon (3) the sensitivity for
function of the grid potential. (1) Relative sensitivity, neon/nitrogen; (2) sensitivity, nitrogen; (3) sensitivity, neon. [Taken with permission from J. H. Leek, Pressure, Measurement in Vacuun Systems (Published for the Institute of Physics and the Physical Society by Chapman and
Hall, Ltd., London, 1964),
2nd
ed.]
Since those ions which are formed in the region between the cathode
a given gas increases with the pressure rapidly for low values of Vg and then much more slowly, changing very little with increasing grid potential above 200 V. Because the ionization potential for neon (21.6 V) is much higher than that for nitrogen (14.5 V) the relative
10
20
200
(b)
400
'0-1
1.0
10
100
0.5
1.0
Electron current.mA
p.torr
Collector-v bios
(c)
(o)
Fig. 3-27. Gauge sensitivity at very low electron current as a function of pressure for nitrogen and helium. [Taken with permission from W. B. Nottingham and F. L. Torney, Jr., in 1960 Vacuum, Sym,posium, Transactions (Pergamon Press,
London,
1961).]
Fig. 3-29. Characteristics of typical triode ionization gauge for nitrogen. [Taken with permission from J. H. Leek, Pressure Measurement in Vacuum Systems (Published for the Institute of Physics and the Physical Society by Chapman and Hall, Ltd., London, 1964), 2nd ed.]
98
99
of ionization-
shown
The
circuit constants
and
number
of an ionization gauge for different gases are not fixed values but depend upon the gauge design and the value of Vg. The somewhat complex dependences of the positive ion current potential F^, (6) the grid or accelerating on (a) the electron current
Thus the
relative
sensitivities
gauge tubes of different design are shown in Table 3-1. In this table the sensitivity is given in units of ,aA/micron/mA, that is, the positive ion current i^ in microamperes resulting from a pressure of 1 micron (10-3 torr) of air in the tube when the electron current i_ is 1 milliampere. Table
3-1.
the plate or collector potential F,,, and {d) the pressure are shown in the curves in Fig. 3-29. These four ciurves
(c)
Reference or supplier
F, volts
volts
i_,
mA
/<A/micron/mA
constant but depends upon a number of factors. Commercial ionization-gauge circuits are designed to maintain the
true
parameters (particularly i_ and F^) constant at values for which the sensitivity is approximately
critical
circuit
Jaycox and Weinhart*" Morse and Bowie*^ Bayard and Alpert** Edwards High Vacuum IG.2 Edwards High Vacuum IG.3 * AEI. 29D.15 (miniature) AEI. 29D.2
.
-6
-25
-5 -50 -15
to
to to
20
5
12.5
40
10
5 12.5
5.5
10
to 10
to 10 to 5 to 5
The prob-
NRC NRC*
CVC GIC.Oll* CVC VG.IA
35
lem of regulating the cathode-emission current was first solved by Ridenour and Lampson^i by means of the circuit shown schematically in Fig 3- 30 The
. .
TG
Eital-McCuUock"
lifetime
KIGT Kinney
VAC-NIG, Vactronis Lab.f.
<
.
-25 -20 -40 -45 -25 -25 -25 -25 -25 -30 + 20 to -50
200 200
150 100 200 150 150 150 150 150 160 150
13
to 5 to 15 to 20 to 5
5 5 5 5
7
20
12.5
18
20
4
12
14 10
14.3
10
10
"
supply
Fig. 3-30. Circuit of Ridenour and Lampson for regulating the emission current of an ionization gauge.
between the cathode and grid, so that commercial gauge circuits are not normally made to operate above 10-^ torr. Modern commercial gauge circuits not only provide reguin the space
These data largely taken with permission from J. H. Leek, Pressure Measurement in Vacuum Systems (Published for the Institute of Physics and the Physical
Society by Chapman & Hall, Ltd., London, 1964), 2nd ed. " E. K. Jaycox and H. W. Weinhart, Rev. Sci. Instr. 2, 401 (1931). " R. S. Morse and R. M. Bowie, Rev. Sci. Instr. 11, 91 (1940). R. T. Bayard and D. Alpert, Rev. Sci. Instr. 21, 571 (1950). " B. B. Dayton, Le Vide, p. 349 (1947).
''
[Taken with permission from J. H. Leek, Pressure Measurement in Vacuum Systems (PubHshed for the Institute of Physics and the Physical
Society
by Chapman and
2nd
ed.]
gauge but also means of preconditioning the tube by heating and outgassing the electrodes as well as a sensitive vacuum-tube electrometer circuit with
a selector
* t
Bayard-Alpert construction.
Nude Bayard-Alpert
constrviction.
several ranges controlled by switch for reading the positive ion current conveniently over a wide range on a simple panel meter. A sensitive relay in series with the output meter is also frequently provided as a means of shutting off the gauge circuit and performing other protective functions, such as turning off diffusion-pump heaters and closing valves in the event that
Although the elements of the traditional ionization-gauge tube are superficially similar to those of an electronic tube, the function is quite different and therefore also the details of the design. Since for
applications pressures of 10- torr (10"^^) or lower are to be measured, the positive ion current is of the order of 10 x lO-^ x 5 ^A or 5 X 10-8 A. There is no difficulty in amplifying such a current to the milliampere range for the operation of a sturdy panel meter. However, any electrical leakage in the gauge tube, in the leads, or in the power
many
100
101
supply chassis
current to be measured. The design of ionization-gauge tubes is therefore such as to minimize In Fig. 3-31 is shown the internal electrical leakage to the plate. plate construction of the RCA 1949, which is widely used. The envelope tube support and lead are brought into the through the opposite end to those for the cathode and the grid, providing a very long leakage path from
The gauges are built on flanges for mounting with their electrodes inside the region where the pressure is to be measured.
commercially.
Initial operation of an ionization gauge just after the system has been pumped down from atmospheric pressure results in the heating
of the electrodes and the emission of large quantities of absorbed gases from the surfaces. Unless the gauge elements are heated vig-
other elements of the tube to the plate. The RCA 1949 also has two spiral-wound tungsten wire cathof odes, only one of which is required for operation other the out, burns the tube. When one filament one can be put into service, thus doubling the life of
orously to drive off absorbed gases, the reading remain high as compared with the system pressure for a long time. The grids of several
will
the tube. The design of Morse and Bowie^^ shown in Fig. 3-32 has been popular in industrial applicaThe innovation introduced with this gauge tions. tube is the use of a platinum film deposited on the inner wall of the tube as the ion collector. Not only
in this case
which can be electrically heated. The plate can be heated by electron bombardment by connecting the grid and plate together at the same positive potential. Finally,
filaments
the electrical leakage path to the plate large, but also the plate can be easily heated for outgassing by flaming the gauge tube. Connection to the plate is made by a wire sealed into the side wall
is
sometimes necessary to heat the glass or metal envelope of the gauge tube to the safe limit of temperature. After the gauge tube and elements have been thoroughly outgassed, an opposite effect beit is
form of a spiral filament through which a current can be passed for heating and outgassing. Fig. 3-31. RCA Because all glass ionization-gauge tubes must be ionization1949 [Regauge tube. connected to the system by a tubulation, the conproduced through ductance from the gauge to the system tends to be the courtesy of somewhat constricted. In studying the performance CorporaRadio of diffusion pumps. Blears^* connected to the test tion of America.] dome a normal ionization gauge with its usual tubufrom lation and also mounted inside the test dome an identical gauge elements gauge the which the glass envelope had been removed so that were exposed directly in the test volume. Striking differences were a observed, the "nude" ionization gauge always reading higher by the to due mainly factor of 10 or more when the pressure reading was vapor backstreaming from the oil diffusion pump. This effect was attributed by Blears to the adsorption of the oil vapor on the inside
of the tube.
The grid
is
in the
Because of the very high conductance into the electrode structure of' the nude gauge, the pressure indicated should be very nearly correct even if adsorption does occur to some degree on the plate and grid. Several designs of nude ionization gauges are now available
Because the surfaces within thoroughly outgassed, gas entering the tube is readily adsorbed, particularly on the tube walls. Chemical reactions induced by the hot filament can further enhance the absorption of gas by the gauge. These processes are responsible for the pumping action of well -outgassed ionization gauges reported by a number of observers beginning with Langmuir.24 Riddiford^s has measured a Fig. 3-32. Ionization gauge designed bypumping speed for oxygen of about Hter/sec and Bowie. due to chemical reaction of oxygen with the Morse [Taken with permission hot filament of an ionization gauge. Gas from R. S. Morse and pumping, either due to adsorption or due to R. M. Bowie, Rev. Sci. chemical reaction, will cause the pressure at Instr. 11, 91 (1940).] the gauge to be lower than that in the system. If the pumping effect includes adsorption of gas on the inside surface of the tubulation, the pressure recorded by a conventional gauge is sometimes a factor of 10 less than that recorded inside a test dome by a nude gauge. Furthermore, according to Reich^s the time necessary for pressure equilibrium to be established between the test dome and the gauge tube through the tubulation can be many hours or even days. The adsorbed layer on the inner wall of the tubulation builds up first at
the tube
are
all
comes noticeable.
102
103
the end nearest to the test dome and then progressively grows along the Equilibrium is not estabhshed until the adsorbed layer has tabulation.
The growth
3-7. The Bayard-Alpert Ionization Gauge. In order to reduce the low-pressure limit of the conventional ionization gauge due to photoelectron emission from the plate, Bayard and Alpert^" devised
of high molecular weight, such as oil vapor, are very slow so that the equilibrium time becomes correspondingly long. Investigation of the "Blears effect" by Haefer and Hengevoss^' confirms the original results of Blears^* and leads to the conclusion that if diffusion-pump
a modified ionization gauge as illustrated in Fig. 3-33. The gauge features a cylindrical grid structure with a fine wire ion collector along its axis and a cathode located just outside the grid structure to one side.
vapor is present in a system operating in the ultrahigh vacuum regime (P < 10-^ torr), a normal tubulated ionization gauge indicates little more than the partial pressure of the permanent gases and does not respond appreciably to the organic vapors present due to backstreaming. According to these authors, "this behavior lies in the vastly different conductance of the small connecting tube for oil vapor and permanent gas in conjunction with the cracking of the oil molecules by the gauge." The processes of thermal decomposition and wall adsorption in ionization gauges are not completely understood, but it is clear that these processes do, under some circumstances, lead to large errors in pressure measurement which can be greatly minimized by elimination of the tubulation and the use of the nude type of
oil
Because the plate structure of the conventional ionization gauge is by a fine wire, the total area exposed for X-ray-induced photoelectron emission is reduced by a factor of about 200. The sensitivity of the Bayard-Alpert gauge proves to be comparable with that of the conventional gauge because the electrons in oscillating from the external cathode back and forth through the grid structure spend a large portion of their time in the nearly field-free space (except for the slight field due to the 20 V on the collector wire) within the grid at the kinetic energy (typically 200 eV) at which the probability of ionization 'p^ is near its maximum. The difference in performance between the Bayard-Alpert and the conventional ionization gauge is shown graphically in Fig. 3-34 in which
replaced
is
ionization gauge.
which can be measured with a conventional ionization gauge is determined by a process not anticipated until He found that after all normal it was pointed out by Nottingham. ^^ gauge there remained ionization leakage current was eliminated in an pressure in the system by the a base current to the plate even when ion current of positive produce a other indications was far too low to current to be due plate residual comparable magnitude. He found the striking the grid. electrons to the production of soft X-rays by the energies of photons X-ray When electrons strike a target they produce in turn photons These up to the full kinetic energy of the electrons. release and the grid, go in all directions from the source, in this case conin the Since photoelectrons from whatever surfaces they strike. grid at the produced ventional ionization gauge most of the soft X rays continuously are strike the plate, photoelectrons in significant numbers produced at the plate. The electrostatic field at the plate is such as to attract positive ions and therefore also to repel the photoelectrons, the current of which is recorded by the external circuit as if it were due to positive ions collected by the plate. Results by Nottingham and others demonstrate that conventional ionization gauges never give plate currents less than that corresponding to about 10~* torr even when the pressure is of the order of IQ-i^ torr. Because of the X-ray effect, the lower limit of reliability of the conventional ionization gauge is found to be about 10-' torr.
The lower
limit of pressure
a log-log plot of the collector current as a function of the grid potential shown for each type of gauge for various values of the pressure. At sufficiently high pressure the collector current does not increase signifi-
cantly with grid potential above 200 V, a characteristic also illustrated However, at very low pressure the positive ion current becomes small as compared with the photoemission current. The
in Fig. 3-296.
then increases indefinitely with increasing grid potenbeing proportional to a power of the grid potential lying between 1.5 and 2.0 as indicated by the slope of the log-log curve in the figure. This characteristic can be explained on the assumption that the entire
collector current
tial,
collector current is due to photoemission. For intermediate pressures a characteristic made up of a mixture of the ionization current and photoemission current is obtained. The characteristic of the conventional ionization gauge at P 10-^ torr is a pure power law typical of photoemission with no vestige of any ionization current. The
<
tion
100.
component
by a
factor of at least
Even
at
P<
x lO""
essentially as described
Alpert are
now commercially
available from a
are frequently used on systems even when the operating pressure is not below the limit of the conventional ionization gauge. A. van Oostrom^o describes a modified Bayard-Alpert gauge in which
is
less
104
105
IQ-^
mm) and
a high negative voltage (-200 is comparable the grid, the sensitivity with the small-diameter collector and Stork" with that of the standard Bayard- Alpert gauge. Schuetze
Between the suppressor electrode and the grid is a grounded shield, the function of which is to prevent rays from the grid from striking the suppressor, causing emission of photoelectrons, which would then
be attracted to the ion-collector plate and subtract from the ioncollector current. Schuemann states that this arrangement completely suppresses the photocurrent and thus removes the X-ray limit entirely. Pressures as low as 2 x lO-i^ torr have been measured with the photocurrent-suppressor gauge.
Conventional
Ion
New
Ion
Gouge
Gouge
torr
The Bayard-Alpert gauge displays the same tendencies either of outgassing or of "pumping" by gas absorption as were described for the conventional ionization gauge. According to Redhead,^^ the pumpFilament ing speed of a Bayard-Alpert gauge, (+50 volts) operated at 8 electron current
_
mA
is
about
ie-t-
~^.
put into
action
is
-Sh
Ionization grid
(+200
volts)
100
1,000
100
1,000
Fig.
3-33.
The Bayard-
(A) one of two alternacathodes; {B) the grid structure; (C) the ion
collector.
permission
[Taken with from R. T. Bayard and D. Alpert, Rev. Sci. Instr. 21, 571
(1950).]
Fig. 3-34. Ion collector current as a function of grid potential for (a) conventional ionization gauge and (6) Bayard-Alpert gauge. [Taken with permission from R. T. Bayard and D. Alpert, Rev. Sci. Instr. 21, 571 (1950).]
and chemical pumping. The ion pumping results from ion bombardment of the electrodes and the glass envelope. The chemical pumping is due to chemisorption of gas on the electrode and any metal films which may have formed, e.g., on the inner surface of the glass envelope. Redhead reports that the chemical pump-
Shield (ground)
J
/
/
Fig. 3-35. Schematic of photocurrent suppressor gauge. [Reprinted with permission from The
Macmillan
Company,
from
1962
uum
Society.]
ing ceases for nitrogen after about lO^^ molecules have been pumped. The ion pumping continues at about 0.25 liter/sec until 10" molecules
markedly reduced X-ray limit (5 x lO"" torr) by using an diameter ion-collector wire oflO /i diameter as compared with the usual and by reducing the electron energy to 50 eV. of about 150 A further advance in the suppression of the X-ray photo current in an ionization gauge is reported by Schuemann,*'' whose gauge modifialso report
fj.
suppressor ring electrode cation is shown schematically in Fig. 3-35. located adjacent to the collector electrode is maintained at a high negative potential ( 300 V), imposing a strong electric field such as to drive any photoelectrons emitted by the collector back into its surface.
have been pumped, at which point it decreases rapidly. Chemical changes in gases are produced by the hot tungsten cathode normally operated at a temperature of about 1700C, which is high enough to dissociate water vapor, hydrogen, and hydrocarbons. Carbon impurities in the tungsten cathode react with oxygen to form carbon monoxide and dioxide. Also at this temperature an appreciable fraction of hydrogen molecules incident on the cathode are converted to atomic hydrogen. Since atomic hydrogen is readily absorbed at glass surfaces and reacts chemically with components of the glass envelope and metal electrodes of the gauge tube to produce other gases such as carbon monoxide, any hot-cathode gauge, such as the Bayard-Alpert, has an anomalously high pumping speed for hydrogen. Chemical changes which occur at the surface of the hot filament result in serious errors in pressure measurement when hydrogen, oxygen, water vapor, or hydrocarbons
106
are present.
107
These effects are particularly troublesome in ultrahighvacuum systems in which the predominant gas is hydrogen. These rhenium troublesome effects can be largely eliminated by the use of a emits an ample which (LaB,), boride lanthanum with coated filament (1,000C) than electron current (10 mA) at a much lower temperature that (1,700C) required for a tungsten
filament.
The Bayard-Alpert gauge is much more susceptible to gross error due to the accumulation of an insulating coating on the ion collector than is the conventional ionization gauge. Lauer^* reports that this condition, which can result in gauge readings which are too low by a factor of 10 or more, can be easily corrected by connecting the ioncollector electrode to the grid potential
n
Fig. 3-36. Nude Bayard-Alpert gauge. [Reproduced through the Lab. Vactronic of courtesy
for
collector electrode
by electron bombard-
ment.
Discrepancies between gauge readings and the true system pressure are greatly reduced, as in the case of the
In a study of the operation of Bayard-Alpert gauges, Winters, Denison, and Bills^s conclude that many of the conditions leading to no'nlinearity in response can be greatly alleviated by the simple expedient of reducing the filament temperature and electron emission current. According to Nottingham,^' one cause of anomalous behavior in the original design of the Bayard-Alpert gauge is the loss of positive ions out the ends of the grid structure. He has shown that this defect can be corrected by closing both ends of the grid structure with wire mesh. The uncertainty which develops at low pressure, particularly for P < 10- torr, is due to the fact that the composition of the residual gas cannot, in general, be predicted. Since the gauge sensitivity is a function of the gas composition, varying by a factor of about 5 for the common gases, interpretation of gauge readings in terms of molecular density at low pressure becomes quite uncertain. Because of this obvious difficulty, low-pressure readings taken with a Bayard-Alpert type of ionization gauge are usually quoted in terms of the equivalent air pressure based upon a calibration against a McLeod gauge using a controlled air flow. A much more precise measurement at low pressure involves vacuum analysis, i.e., the measurement of the partial pressures of the gas components present in the system, a topic which will be discussed in a later section.
conventional ionization gauge, when the glass envelope is removed from a Bayard-Alpert gauge and the gauge
unit immersed in the vacuum space. In Fig. 3-36 is shown a flange-mounted nude Bayard-Alpert gauge constructed
Equipment,
port,
Inc.,
East North-
ceramic and metal parts which thorough allegedly will withstand baking at high temperature for strucgauge outgassing. Even if the elements of such an open the with ture absorb, desorb, or chemically react to some degree intergas for gases present in the system, the very large conductance
of
will be change ensures that the molecular density within the gauge gauge the essentially the same as that in the surrounding volume and nude a that reading will be reasonably accurate. Santeler^^ states lO"" about Bayard-Alpert gauge cannot measure pressures less than gauge the torr, however, unless the backing plate which supports elements can be baked at high temperature for thorough outgassing.
Long
Island, N.Y.]
In the origBayard-Alpert gauge the X-ray limit (i.e., the pressure corresponding to the photoelectron current emitted by the ion collector because of X rays emitted by the grid) was reduced by a factor of about 200, compared with the conventional hot-cathode ionization gauge. This was done by reducing the solid angle subtended at any point of the grid by the ion collector, thereby reducing the probability for an X-ray photon emitted by the grid to be intercepted by the ion collector. A further improvement in this respect has been accomplished by Lafferty,38 whose hot-cathode magnetron ionization gauge is shown in Fig. 3-37. According to Lafferty,
inal design of the
3-8.
modified in such a way that the electrons by the positive grid or anode, the probability of them colliding with and ionizing a gas molecule will be greatly enchanced and the sensitivity of the gauge will be improved with no increase in x-ray photoemission. One obvious way of increasing the path length of the electrons is to employ a magnetic field.
is
It
evident that
if
the gauge
is
Also in the nude Bayard-Alpert gauge, space- and surface-charge effects completely are frequently bothersome unless the gauge element is potential. ground enclosed in a metal-screen cylinder connected to
The Lafferty gauge has certain features in common with the magit is named. The hot-cathode filament is approximately on the axis of symmetry of a cylindrical electrode, which is the
Electrons leaving the cathode in the absence of a magnetic
anode.
108
VACXTTJM SCIENCE
AND ENGINEERING
109
the anode. However, when an axial magnetic field is superimposed, the electron paths are bent. Finally, anode-to-cathode as the magnetic field is increased in intensity, with the the anode reaching potential held constant^ the electron current field value magnetic The decreases to a much smaller value.
suddenly
at
which
this
cutoff.
diameter as the cylindrical anode, and is supported over one end of the anode with a sufficient gap to stand off a voltage. The lead for the ion collector is at the end of the tube opposite that at which all the other leads are located, to provide minimum electrical leakage. The lower or opposite end of the anode is closed by a similar disk, referred to as the shield, which catches any positive ions leaving in that direction.
The electrical potentials relative to the cathode are typically as follows anode, 300 V; ion collector, -45 V; shield, -10 V. Typical magnetic field intensity is 250 oersteds, which can be maintained either by a
solenoidal coil placed around the tube. increased slightly beyond the magnetron cutoff value, the length of the electron orbits, and thus also the number of ions produced per electron, is increased by a large factor. The ion current registered on the ion collector is also shown in Fig. 3-38 as a
is
cylindrical alnico
or
by a
-10 volts
V-i,lon collector
Filament
function of the magnetic field intensity. It will be noted that this current is a maximum for a value of the magnetic field only slightly
Ion current
Magnet
XIOO
Anode
Shield
above the magnetron cutoff value. The positive ion current increases by a factor of 10* or 10^ as the magnetic field is changed from less than the magnetron cutoff value of about 100 oersteds to about 150 oersteds, at which the ion current is a maximum. As the magnetic field is
further increased, the electrons emitted by the cathode are confined more closely to the vicinity of the cathode as their orbits are contracted,
and thus sweep a shorter path through the gas, producing fewer ions. The maximum ratio of positive ion-collector current to electron anode
100
200
300
400
500
600
Fig. 3-37. Hot-cathode magnetron ionization gauge. [Taken with permission from J. M.
Lafferty, in 1960
Press,
Fig. 3-38. The anode electron current and positive ion-collector current of the magnetron gauge as a function of the
is thus reached at a magnetic field value about twice as large magnetron cutoff value. These are the most favorable conditions for operation in order that the X-ray limit be as low as possible.
current
as the
is
Vacuum Sym-
applied axial magnetic field. [Taken with permission from J. M. Lafferty, in 1960 Vacuum Symposium Transactions
In Fig. 3-39 the ion-collector current, as a function of the pressure, plotted together with similar data taken with a Bayard-Alpert gauge.
1961).]
Based upon a determination of the X-ray limit of the ion magnetron gauge, which proves to be about 6 x 10"" torr, and a calibration against the Bayard-Alpert gauge at higher pressures, the readings of
the former were assumed to be a linear function of the pressure down to the lowest readings obtained. The Bayard-Alpert gauge readings begin to deviate noticeably from a straight line at lO-^" torr and reach
For higher values of the magnetic field, an electron emitted by the cathode performs a cycloidal orbit which fails to reach the anode and turns back toward the cathode. Only those electrons which are disturbed from this orbit, either by high-frequency electrostatic fields due to electron space-charge oscillation or by collision encounters with gas molecules, ever reach the anode under these conditions. The electron current curve, shown in Fig. 3-38 from Lafferty's paper, decreases by about a factor of 10 as the magnetic field is increased from
zero to 200 oersteds. The ion collector for the Lafferty gauge
is
an asymptotic value corresponding to an X-ray limit of about 3 x lO-^* torr, whereas the ion magnetron gauge indicates collector current as low as that corresponding to a pressure of 4 x 10"" torr. Operating parameters for both gauges are given in the figure. It will be noted that for the ion magnetron gauge, / = 10-' A for the run, where I^ is the emission current read on the anode when the magnetic field is zero. To produce approximately the same ion-collector current at higher pressures, the Bayard-Alpert gauge is operated with /j = 5 x 10-3 A,
no
VACTJTJM SCIENCE
AND ENGINEERING
111
indicating a greater sensitivity for the ion magnetron gauge by a factor 5 X 10*. of 5 X 10-3/10-' At high pressure the secondary electrons produced by ionizing col-
Electron
multiplier
lisions
add
A further improvement in the hot-cathode magnetron ion gauge has been described by Lafferty.^^*'
By drawing the ions produced in
the
First
dynode
Boyard-Aipert gouge
Ep= 300
=
volts
Ep= ISOvolts
Ec
=
-45
volts
volts
-45 volts
5
10-'
Es= -10
H
=
Io=
mA
250 oersteds
10-'
10-'
magnetron gauge out one end by an electrostatic lens system designed to focus the ions on the first dynode of an electron multiplier tube, as shown in Fig. 3-40, the sensitivity of the gauge has been increased by a large factor. The soft X rays from the magnetron anode were prevented from falling on the dynode, with the result that the X-ray level was decreased to such a low level that a pressure of 10"" tonshould be detectable by counting
individual ions
striking
Electrostatic lens
Multiplier
^/shield
cylinder
^
"--I
I
Focusing
cylinder
Aperture
.,
i-^-Alnico
magnet
Ion accelerator
-Magnetron gauge
LaBfi
the
first
Filament
detail
multiplier.
To prevent
10-'
by
LJ
10-'
"I
,0-13
>
I
III
'
ml
10-'
'
i_
io-'2
IQ-"
Pressure, torr
10-'
10"
comparison of the ion-collector current characteristics of the BayardFlQ. 3-39. linear relationship between the Alpert gauge and the magnetron gauge. magnetron-gauge ion current and pressure is assumed at pressures below 10~* [Taken with permission from J. M. Lafferty, in 1960 Vacuum Symposium torr. Transactions (Pergamon Press, London, 1961).]
a lanthanum boride (LaBg)-coated rhenium filament, which produces electron emission at much lower temperatures, was substituted for the tungsten cathode. As shown in the figure, the lowtemperature cathode consisted of a lO-mil-diameter rhenium wire, with
multiplier,
Fig. 3-40. Hot-cathode magnetron ionization gauge with electrostatic lens system to focus ions on the first dynode of an electron multiplier tube. Also detail of low-temperature, lan-
thanum boride (LaBg) coated cathode. [Reprinted with permission of The Maomillan Company from 1962 Vac-
uum Sym.posium
right
Transactions. Copy-
1962 by American
Vacuum
Society.]
an open helix of a 5-mil-diameter rhenium wire wound on it and then coated with lanthanum boride. The lower-temperature cathode should also contribute to a reduction
in the anomalies caused section.
the result that the space charge becomes large enough for oscillations to set in because of cooperative electrostatic forces. The gauge tends to become unstable in operation, with frequent reversal of the ioncollector current because of the ejection of high-energy electrons against the repelling field of the collector. Because of this phenomenon, the magnetron gauge is normally operated at low emission current of
by a hot cathode
3-9. Magnetically Collimated Electron-beam Ionization Gauge. An ionization gauge in which the ions are produced by an electron beam collimated by a carefully aligned magnetic field has been described by Klopfer." As is apparent from Fig. 3-41, electrons
112
113
are emitted from a thermionic cathode K, coUimated through a series of apertures at various electrical potentials, traverse an isolated chamber in which ions are produced, and finally collected by an electron-trapping electrode T. Ions produced within the chamber between G^ and G^ are attracted to the ion-collector plate / at ground potential, which is
beam and
magnetic
field
10'
10
field
1,000 gauss
:10'
10"
30 volts
10-"
IQ-12
/ /
,0-10
l,000gauss
a consequence, the pumping effect of the gauge is determined primarily by the rate of production and collection of positive ions. For nitrogen the pumping speed of the gauge is calculated to be about 3 x 10"^ One side of the liter/sec with an electron beam current of 1 mA. ionization chamber is covered by an open screen providing high conductance to the surrounding space, so that the gauge error due to pumping is believed to be very small. As is illustrated in Fig. 3-42, the response of the gauge is linear from about 10~^ torr to at least as low as 10~i^ torr. The linear response up to such a high pressure is apparently due to the fact that only the electrons coming directly from the cathode and not those produced by The ionization along the beam are effective in producing ionization. electrical potentials are arranged such that the electrons traverse the ionization chamber at an energy of 100 eV, a value at which the specific ionization is near maximum. Electrons formed by collision processes in the chamber, however, have very little energy and no means of gaining energy from the electric fields present in the chamber. 3-10. Cold-cathode Ionization Gauges. The useful life of a hot- cathode ionization gauge is determined by that of the cathode.
10-8
10-6
10-4
,0-:
10"
Pressure, torr
3-41. Schematic diagram of magnetically coUimated electronbeami ionization gauge. [Taken with permission from A. Klopfer, in 1961 Vacuum Symposium Trans-
Fig.
actions
1962).]
Fig. 3-42. Relative ion current vs. nitrogen pressure for the electronbeam ionization gauge of Fig. 3-41. [Taken with permission from A. Klopfer, in 1961 Vacuum Symposium. Transactions (Pergamon Press,
lOcm
London,
1962).]
so that at
all
on the electrode T. The geometry of the gauge is such that X-ray photons emitted by the electron-trapping electrode T because of electron impact cannot irradiate the ion-collector electrode J directly, so that electron emission from the ion collector is minimized. The presence of the magnetic field further reduces the X-ray effect by causing any
electrons emitted at the electrode surface to
Fig.
move
in circular orbits
Penning cold-cathode ionization gauge. [Taken with permission H. Leek, Pressure Measurement in Vacuum Systems (Published for the Institute of Physics and the Physical Society by Chapman and Hall, Ltd., London, 1964), 2nd ed.]
3-43.
from
J.
and return to the electrode. Emission of photoelectrons due to light from the cathode striking the ion collector is avoided partly by the geometry of the gauge, which shields the collector from direct irradiation by the cathode, and also by the use of a low-temperature cathode, such as a lanthanum boride-coated rhenium filament. The use of a low-temperature cathode also serves to reduce substantially the pumping effect due to chemical reactions at the cathode. As
Particularly at relatively high pressures, approaching 10"' torr, chemical attack by some gases and bombardment by positive ions limit the
life of hot cathodes. In 1937, Penning*^ described a cold-cathode gauge, sometimes called the Philips ion gauge or PIG. As shown schematically in Fig. 3-43, Penning's original gauge consisted of two parallel plates P at cathode potential separated by a distance of about
useful
114
2
115
a ring electrode R at anode potential suspended in the midplane between the two plates. A magnetic field of about 500 oersteds, with the lines of magnetic flux perpendicular to the planes of the electrodes, was provided by a permanent magnet with a pole on either When a potential difference of 2,000 V was applied side of the tube.
cm and
creased
end of the cylinder. The sensitivity of the gauge was inby about a factor of 10 to about 1 mA at 10"* torr for a circuit
voltage of 2 kV, the range of linear response extended down to less than 10^' torr, and the amplitude of the erratic changes in the discharge current reduced to 2 per cent or less. With a gauge of similar geometry, but with nickel instead of zirconium cathodes, Leek and Riddoch** obtained the calibration curves shown in Fig. 3-45, which are in good agreement with those obtained by Penning and Nienhuis. The current
between the plates as cathodes and the ring as anode, a current was observed to flow and was found to be
approximately proportional to the pressure for the range 10"^ to 10"^ torr. In the absence of the magnetic field the electrode structure of the Penning gauge
100
4kV
0.3
2ky
i\y
would not
a discharge current pressure, of 10"^ high relatively at except The presence of the magnetic torr or more. field changes completely the character of the
result
in
=1.
50
0.15
discharge because electrons emitted by the cathodes are constrained to move in tight helical paths along the lines of force and are
H
l^
j.-1kV
1
thereby prevented from going directly from the cathode plates to the anode ring. The electrons, instead, oscillate through the ring
2
torr
"
J^
Pressure,
^
10^ torr
TkV
2.5
Pressure,
X 10^
Fig.
3-45.
by Leek and
permission
between the plates and slowly drift outward across the magnetic field toward the ring only as a result of collisions with gas molecules, which are ionized to form positive ions and additional electrons. The positive ions are attracted immediately to the cathodes and, in contrast with the usual gaseous discharges, account for most of the current flow
in the discharge.
Riddoch
range 500 to 1,000 gauss). [Taken with permission from J. H. Leek, Pressure Measurement in Vacuum Systems (Published for the Institute of Physics and the Physical Society by Chapman and Hall, Ltd., London, 1964), 2nd ed.]
is
proportional to the pressure over the useful range of the gauge, i.e., torr. Also the sensitivity is proportional to the anode-to-cathode voltage over the range 1 to 4 kV. At a pressure of
The additional
electrons
contribute to further ionization. Because of the comparatively long electron path, which results from the oscillatory motion, the gas breaks down into a discharge even when the pressure is as low as 10~^ torr.
Penning gauge, characterized by the ring proved to be somewhat unsatisfactory in that the discharge is unstable and the current as a function of the pressure exhibits unpredictable changes of magnitude between 2 and 5 per cent. Below 10~^ torr the discharge becomes erratic and frequently fails entirely at pressures of 10~* torr and below. These difficulties were partly overcome in an improved form of the coldcathode gauge described by Penning and Nienhuis*^ and illustrated in Fig. 3-44. The anode of the improved gauge is in the form of a cylinder of about equal length and diameter, with flat plates as cathodes located
The
tjrpe of
anode illustrated
in Fig. 3-43,
about 10-* torr the current decreases suddenly to a much lower value so that the gauge is not considered reliable below about 10"' torr. Since a ballast resistor must be put in series with the power supply for protection in the event of a short circuit, the gauge is somewhat nonlinear from 10-* to 10-3 torr, above which the current increases steeply with the pressure. The steep increase of the current with increasing pressure above 10-^ torr can be used to protect a system against pressure bursts. A relay in the circuit set to operate in this range can be used
and perform other protective functions event of an excessive gas burst. The Penning cold-cathode ionization gauge cannot be relied upon for pressure measurements of high precision, even with great care in calibration, because of the slightly unstable character of the discharge. However, this type of gauge is sufficiently reliable as a pressure indicator for many vacuum process applications and has a very long
to shut off gauges, close valves,
in the
116
useful
life
117
of the rugged character of the cold cathodes. A cross-sectional view of a commercial Penning discharge type of gauge is shown in Fig. 3-46. The anode in this design is shaped to provide a maximum area while keeping the exterior dimensions of the
gauge tube to a minimum. A shield protects the high-voltage insulator from accumulating spluttered atoms and ions, contributing to a long The gauge tube and life against electrical leakage and breakdown. features the gauge these of Because tubulation are of stainless steel. conditions up to favorable under is said to provide an operating life
Section
AA
Fig. 3-46.
vacuum
Note;S -Fully
CCW
shown
gauge.
Fig. 3-47. Circuit diagram for
50,000 hr. However, as with all cold-cathode gauges, excessive exposure to hydrocarbons will result in contamination due to decomTo remove position products deposited on the electrodes and insulator. washing with recommends manufacturer the the gauge, from oil trichlorethylene, rinsing with acetone, and drying by gentle heating in To remove decomposition products, a cleaning solution made up air. 225 of g of ferric chloride, 500 cm^ of hydrochloric acid (38 per cent HCl) and 1,400 cm' of distilled water is recommended. The gauge tube is placed in a beaker and filled nearly to the top with hot cleaning solution which is maintained at the boiling point for about 15 min;
~^
000
L
-n
">
1 1
/
1
1 1
100
50/iAatO.Ol/i^
1 1 1
N.
y
then rinsed with distilled water followed by acetone, alcohol, the tube or ether and finally thoroughly dried by warming in air. Although the Penning discharge gauge described above is mechanically rugged and is not damaged by exposure to atmospheric pressure
is
11
1
lu
while the voltage is on, observing certain precautions will contribute The gauge tube to greater operating life by avoiding contamination. should not be mounted in a position where it is in a direct line with a source of hydrocarbon vapor. It should not be operated at pressures in excess of 10 torr nor during the pumpdown period. A simplified circuit diagram for the Penning discharge gauge is shown in Fig. 3-47, and the pressure response curve is shown in Fig. 3-48.
/
10"
10'
10"*
10"*
10"
10"
10"'
Pressure, torr
118
119
The gauge
is calibrated over the pressure range from 2 x 10~' to 10 torr, the response above about 2 x 10~* torr being much flatter than at lower pressures as shown in the figure. A further development of the cold-cathode ionization gauge was
carried out by Beck and Brisbane,*^ Haefer,** and Redhead*' and has culminated in a gauge of high sensitivity and reliability. The added
is which E is the electric field intensity in volts per centimeter and is velocity oersteds. Since the drift in strength field magnetic the in H, electrons move E and the both perpendicular to everywhere circular cycloidal paths at a constant average radius from the center. Only upon collision with a gas molecule is the electron disturbed from Each time such a this path because of energy loss in the collision. collision occurs the electron moves into a new circular cycloidal path
With a proper
Auxiliory
cathode
-Anode
choice of the parameters the drift velocity of the electrons is sufficient to ionize gas atoms so that an
start of electron
o o o
oo oo o oo o
ooo
A xiol magnetic
field
Fig. 3-49. Schematic of the Haefer inverted -magnetron type of cold-cathode ionization gauge. [Taken with permission from Hel-
11,
Cutaway view of inverted magnetron cold-cathode ionization gauge. [Taken with permission from P. A. Redhead, in 1958 Vacuum Symposium Transactions (Pergamon Press, London, 1959).]
Fig. 3-50.
mediately to the cathode. By this process each electron emitted from the cathode produces a large number of ionizing events before it finally spirals into the center of the gauge and is caught on the anode. In Fig. 3-50 is shown a cutaway view of an inverted-magnetron
,0-11
,0-9
Pressure, torr
,0-7
feature
is the use of crossed electric and magnetic fields to increase by a large factor the path of the few electrons emitted by the cold cathode, and thus also the efficiency of the electrons in producing positive ions.
schematic representation of the Haefer inverted-magnetron gauge is in Fig. 3-49. The cathode is a cylinder (actually the metal case of the gauge tube) about 5 cm in diameter, and the anode is a small diameter metal rod located on the axis of the cathode. A magnetic field of about 2,000 oersteds intensity parallel with the axis of the tube is maintained by an external coil. A potential difference of several
shown
gauge developed by Redhead*' in Fig. 3-51. lon-current-vs. -pressure which the cathode is surrounded by relationship for inverted-ion-magnean auxiliary cathode outer shell tron gauge. [Taken with permission with cylindrical shields protruding from P. A. Redhead, 1958 Vacuum through the openings into the cath- Symposium, Transactions (Pergamon ode. The auxiliary cathode acts as Press, London, 1959).] an electrostatic shield and protects the edge of the openings through the cathode from field concentrations, thus preventing field emission. The cathode and auxiliary cathode are both grounded, but the current to the cathode alone is taken as the measure of the true positive ion current. The anode rod is typically maintained at 6 kV and the magnetic field intensity at 2,000 oersteds. In the pressure range from 10"" to 10"* torr the positive ion current was found to conform to the relationship
i+
in
kilovolts
is
cP"
electric field is
anywhere
given by
superimposed upon the axial magnetic field. An electron between the anode and cathode will move on a cycloidal path in the E x (azimuthal) direction with a drift velocity
in the region
which n varied from 1.10 to 1.15 and c was a constant. Above 10-3 torr the space charge changes from negative to positive with
v^
lO^E/H)
cm/sec
the result that the characteristics of the gauge change completely. Calibration curves for several models of the inverted-magnetron gauge are shown in Fig. 3-51, together with a similar curve for the
120
121
Table
Gauge
3-2.
voltage,
kV
Bayard-Alpert, 10
A/mm Hg
0.1
Pretreatment of gauge
Pumping
Gas
rate, 1/sec
mA.
Outgassed
Argon Argon
Nitrogen
0.080
5.0
4.0
Baked
at 400C
Anode
5.0 4.0
Oxygen
Operated
hotirs in for several
Argon
Nitrogen
oxygen
Oxygen
Cathode
Fig. 3-52. Cold-cathode magnetron gauge. [Taken with permission from P. A. Redhead, 1958 Vacuum Symposium Transactions (Pergamon Press, London,
1959).]
Reduced-size cold
1.2
Operated for several hours in argon Baked at 400C Induction heated to 800-900C
Induction heated Induction heatedf Induction heatedf
0.46
Bayard-Alpert gauge. As in the case of the hot-cathode magnetron gauge discussed in the previous section, the X-ray hmit of the inverted-magnetron gauge is well below that of the Bayard-Alpert
gauge.
Reduced-size cold
0.3
0.03t
is
an
and designated the coldcathode magnetron gauge by Redhead.** The cathode, as shown in Fig. 3-52, is in the form of a spool consisting of a small-diameter central cylinder and two end disks. The anode is a cylinder with a diameter about equal to that of the end disks and is perforated with many holes to ensure good conductance beinversion of the geometry discussed above
* Taken with permission from T. N. Rhodin and L. H. Rovner, in 1960 Transactions (Pergamon Press, London, 1961), p. 228. f Prolonged treatment with some evaporation of metal. % Ratio of ion current to pressure is not constant.
Vacuum Symposium
for
nitrogen
i^
is
given approxi-
mately by
lOP
10-'
Helium
Pressure, torr
Fig. 3-53. lon-current-vs. -pressure relationship for magnetron gauge in the range 10"* to 10-9 tQj.r. [Taken
tween the regions inside and outside the gauge volume. An auxiliary cathode in the form of an electropolished ring is placed at each end of the cylindrical anode in the gap between the anode and cathode to reduce field emission currents to a minimum. Redhead found that in the pressure range from 10~^ to 10~* torr the cold-cathode magnetron gauge with anode potential of 5 kV and magnetic field of 1,070 oersteds has a linear characteristic
which the pressure is given in It was also observed that at a pressure of about 2 x lO^^" torr there is a break in the response curve so that below this
in
torr.
,y
-->
eio''
value of the pressure the curve is no longer linear but takes the
Run
1-12 10'
Magnetron gauge
operating alone
form
Bayard-Alpert gauge
operating alone
shown for helium in Fig. 3-54. Redhead reports that the coldcathode magnetron gauge has a pumping speed of approximately 0.15 liters/sec. Rhodin and Rovner** have made extensive measurements of the pumping speed of cold-cathode magnetron
as
10
10
10"
Pressure, torr
Fig. 3-54. lon-current-vs.-pressure relationship for magnetron gauge in the range 10~' to 10~l^ torr. [Taken with permission from P. A. Redhead, in 1958 Vacuum Sym,posium Transactions
1959).]
122
123
gauges similar to that of Redhead and report that the principal disadvantage is the high pumping speed of such gauges, leading to some ambiguity in interpretation of the ion current reading. The results of their measurements are summarized in Table 3-2 in which the pumping speeds of the normal size (Redhead) cold-cathode magnetron gauge and one of reduced size are compared with that of a Bayard- Alpert type of hot-cathode ionization gauge. In spite of the high pumping speed and its dependence upon the
Section
1 torr of dry air. An improved version of Vacca'^ is provided with six ranges with full described by the Alphatron scale readings of 10-^, 10"^, 10, 100, and 1,000 torr. The low range is accomplished by an improved electrometer tube and circuit capable of amplifying currents as low as 10-^* A. The higher ranges are obtained by using a second ionization
10-1"
A for a pressure of
The
new gauge
is
said to
A-A
30 to 40
contoined
volts
in
amplifier
ranges.
power supply
Qutput
cathode magnetron gauge is useful 10~ii in the pressure range below the well below torr, a pressure Alpert gauge. Bayardrange of the The high pumping speed is apparently associated with the very high
efficiency of ionization
3-12.
Gauge.
gauge
is
The
two
alternative
known and described^' of the less common vacuum gauges. The basic element of the radiometer gauge consists
of two parallel plates, one of which is heated, separated by a distance which is small as compared with the dimensions Fig.
Housing
by the
elec-
trons
orbits,
in
their
circular
cycloidal
in a high positive ion high sensitivity, with that of compared current, as operating gauge the Bayard- Alpert
which
results
i.e.,
of the
3-56a.
plates,
as
shown
plate
in
is
2nd
ed.]
The unheated
at the
1
same
pressure.
3-11.
2
I
l'Mi|i|i|
II
Any
alpha particles, beta particles, and gamma rays are all ionizing agents, the advantages of which may be considered as possible means of ionizing gas for the
rays,
3-55. The Alphatron gauge. [Taken with permission from J. R. Downing and G. Mellen, Rev. Sci. Instr. 17, 218 (1946).]
Fig.
supported on a sensitive suspension so that a small force acting upon An alternative form is shown in Fig. 3-566 in which it can be measured. the unheated vane is suspended between two fixed plates, one of which The force per unit area on the susis heated and the other cooled. pended vane or plate is given approximately by
chamber
-I
% {^
w^ m
T,V^"
dyne/cm^
(3-36)
purpose of measuring its molecular density. A practical development of this type is the Alphatron (National Research Corporation) gauge of Downing and Mellen'" which utilizes a small source of alpha particles, for example, 0.5-mg piece of an alloy of gold and radium sealed in a capsule. The gauge consists of a source holder and two grid structiires inside a small metallic ionization chamber which serves as the gauge tube (see Fig. 3-55). A diiference in potential of 30 to 40 V is maintained between the two grid structures to sweep out the ions and electrons formed by the ionization process. The ionization current is found to be substantially a linear function of the pressure over a wide range, from 10-* to 40 torr for the first version of the Alphatron, the current being
which T^ and T^ are the temperatures respectively and the vane, T is the ambient temperature of the In the second case walls of the gauge tube, and P is the gas pressure.
for the first case in
/2
dyne/ cm ^
(3-37)
in
which T^ and T^ are respectively the temperatures of the heated and cooled plates and T is the ambient temperature. In this latter case the force on the vane does not depend on its temperature T^. In either
case the force depends directly
124
125
the force per unit area exerted by the gas, with no dependence upon the molecular weight of the gas. In this respect the Knudsen gauge may be considered an absolute pressure-measuring device. A more exact treatment of the theory of the Knudsen gauge, taking into account the accommodation coefficients for the vane surfaces and the inside surface of the gauge tube, leads to much more complicated
expressions for the force per unit area on the vane. Differences in accommodation coefficients at the various surfaces result in responses which differ for various gases, with the response to helium and hydrogen
A needle valve is provided so that any chosen gas can be admitted to the system at a controlled rate to vary the pressure. For calibrating thermocouple and Pirani gauges, usually from 10'^ to large1 torr, only a rather insensitive McLeod gauge (small bulb and diameter capillary) is required, and no serious difficulties are reported. However, for calibrating ionization gauges over a sufficient pressure range in the region of linear response, the greatest McLeod gauge sensitivity
Gas input
through drying tube
being particularly low for some gauge designs. The linear expressions for the response of the Knudsen gauge are valid only in the pressure region for which the molecular mean free path is large as compared with the spacing between the vane and fixed By using the smallest practical spacing and a closed plate or plates. box structure about the vane system, linear response up to a pressure
of 10- torr can be obtained. At higher pressures the response is always less than the linearly extrapolated value and eventually begins to decrease with increasing pressure. The useful range of the radiometer gauge thus tends to be from about 10"* to 10-* torr. practical designs of the Knudsen gauge pressures as low as 10-^
Gauges
to be
coiibrated
Severe
McLeod gauge
locations
around
chamber
In
Since a sensitive suspension is required, all designs of the Knudsen gauge thus far developed are too cumbersome and Many special adaptations have been fragile for most applications. made and successfully applied, however, when the unique features of
Liquidnitrogen trap
ffiK^'\K.'\K<^K
I
Liquidnitrogen trap
Liquid-nitrogen and
water-cooled baffles
the Knudsen type of gauge are important, such as freedom from chemical decomposition of heavy organic molecules, a process which
Diffuion
pump
does occur in all types of ionization gauges. 3-13. Calibration of Vacuum Gauges. For many years the accepted standard for calibrating other vacuum gauges in the pressure range below that easily accessible to the simple mercury U-tube manometer has been the McLeod gauge. The limitations of the McLeod gauge and the precautions necessary to obtain consistent It is clear from that discussion that results are discussed in Sec. 3-4.
the calibration of other vacuum gauges is limited to gases which obey Boyle's law up to the maximum pressure to which it is compressed Accepted practice has been to in the operation of the McLeod gauge. provide a glass- or metal-walled chamber evacuated by a liquidnitrogen-trapped diffusion pump to which the McLeod gauge and the
vacuum gauges
against a
McLeod
gauge.
is
the lowest pressure which can be rehably measured with sufficient precision by means of a McLeod gauge is about 10-5 torr. Thus ionization gauges are normally calibrated from 10-5 to 10-^ torr, in the upper portion of which range the response of
required.
Even
so,
is
no longer
linear.
The range of
linear
response available for calibration in this manner is a factor of 30 in pressure for determination of the gauge constant. Recently, however, a much more serious objection to this method of
calibration of ionization gauges has been raised by Ishii and Nakayama,^* who find a pronounced dependence (up to 25 per cent) of the calibration
The use of liquidthe ionization gauges is essential to protect from mercury vapor from the McLeod gauge and also to protect the McLeod gauge from contamination by hydrocarbon vapor from the
cooled trap, somewhat as
shown
in Fig.
3-57.
nitrogen-cooled traps
upon the ambient temperature. They attribute this effect to the pumping action of the mercury vapor stream from the McLeod gauge
into the liquid-nitrogen-cooled vapor trap.
By
cooling the
McLeod
126
127
gauge at the cutoff point with dry ice before the mercury is raised, thereby reducing substantially the vapor pressure driving mercury vapor into the cold trap, this particular gauge error was eliminated.
S,=
Gi
S2
G1S2
G,
The
accepted method of pressure measurement involving large temperature differences when condensable vapor pressure dominates in some portion of the vacuum system.
G2
(3-39)
as long as Cj
<
S^.
AP is
(3-40)
To (reference) gouge
Gouge to be
calibroted
Gas
inlet
Q ^^ = ^-^- = ^.-C,-fC;.
The pumping apertures are most conveniently in the form of which the conductance given by (2-89) is
G-fi^
circular
holes for
C =
2.861
\^r
11)2
liters/sec
and is accurately determined as long as the diameter of the aperture and the ambient temperature are known. The method also requires an accurate measurement of the gas flow Q, alternative methods for which are discussed in Chap. 7. This type of system may be extended by the addition of several pumping chambers with apertures between
Fig. 3-58. Apparatus for calibrating vacuum gauges at low pressure by gas flow through apertures. [Taken with permission from J. R. Roehrig and J. C. Simons, Jr., in 1961 Vacuum Symposium Transactions (Pergamon Press, London,
1962).]
down
method for calibrating ionization gauges has been by Roehrig and Simons^^ and results compared with calibrations against a McLeod gauge. The apparatus as illustrated in Fig. 3-58 consists of a test chamber and an adjoining pumping chamber connected by an aperture of small conductance Cj. The pumping chamber is evacuated by a high-speed trapped diffusion pump through a second aperture of conductance C^ which is small as compared with the pumping speed Sj, of the trapped diffusion pump. The pumping speed out of the pumping chamber is then, according to (2-8),
alternative
An
developed
Co
S^ Go
s^ Go
(3-38)
Gauges to be calibrated are connected to the test chamber, and the gauge responses are compared with a series of pressure increases AP. The authors state that results obtained in this manner on a four-stage system were in agreement with results obtained by calibration of the same gauges with a McLeod gauge within 2 per cent over the pressure range from 10"^ to 10""^ torr, with poorer agreement below 10"* torr because of the inaccuracy of the McLeod gauge in this range. This agreement appears to be inconsistent with the discrepancies reported by In any case, the Ishii and Nakayama^* for McLeod gauge calibration. action of pumping aperture method is free of criticism based upon the apparently is a liquid-nitrogen trap associated with a McLeod gauge and the most reliable method yet developed for ionization gauge calibration. This method has the further advantage of being applicable, according to Roehrig and Simons, for accurate calibration of gauges down to 10-9 torr. Precautions which seem to the author to be important are
to ensure
1. That the diameters of the apertures between the test chamber and the pumping chamber and between the stages are very small compared with the dimensions of the chambers
which
differs only slightly from G^ if Cj < S^. Thus the pumping speed out of the pumping chamber is relatively constant and independent of small changes in S^. The pumping speed out of the test chamber
128
2.
129
That the actual pumping speed of each diffusion pump is large as compared with the conductance of the pumping apertures 3. That the base pressure P, is small as compared with the pressure P at which calibrating measurements are made
Requirement
1
is
essentially
due to dynamic flow will be negligible. Requirement 2 allows the use of the approximate expression in (3-40) involving the conductances only. Otherwise the value of *Si given in (3-39) must be estimated and substituted for Cj in (3-40).
static so that directional pressure effects
determined by the X-ray effect, from lO-" to about 10-" torr by mtroducing an additional electrode and superimposing a voltage modulation. Torney and Peaks*' have used this technique to extend the range of comparison between the modulated Bayard- Alpert and the cold-cathode ion magnetron gauges, confirming the results reported by Redhead" down to 10-" torr and observing ion current readings
corresponding to a nitrogen pressure of about 3 x 10-" torr. Results such as these suggest that, in spite of the difficulties listed above, reasonably reliable pressure measurements can be made even in the range below lO-" torr if the proper precautions are observed. These processes have been discussed in relation to relatively
static
For precision calibration S^ should be at least of the order of lOCj. Requirement 3 ensures the elimination of subtle questions regarding the propriety of simply subtracting P^ from P for determining the pressure rise P due to the gas flow Q.
3-14.
urement
used in
General Remarks on Ambiguities of Pressure Measin Vacuum Systems. In describing the instrumentation measuring the pressure in vacuum systems some of the pre-
systems in which the pumping speed is relatively small so that directional effects are not important. In this case the isotropic pressure is given by the familiar P = QjS, where Q is the total gas load due to leaks and outgassing, and 8 is the pumping speed. In
i.e.,
systems,
cautions required to obtain consistent results have been mentioned. However, the extent of the ambiguity in measurement at low pressure (e.g., less than 10~* torr) has perhaps not been stressed sufficiently.
order to attain a very low pressure in such a system (as in the usual ultrahigh vacuum systems), the gas load Q is reduced by eliminating all leaks and outgassing the internal surfaces by baking the system at
high temperature.
in ionization gauges
and contribute to
Surface adsorption (gauge pumping) and desorption Chemical decomposition, dissociation, or reaction with the hot filament
2.
3. 4. 5.
Electrical oscillations
Radiation (the Nottingham X-ray limit) Electrostatic effects due to surface and space charges
Whether a vacuum gauge meets the requirements of a given appliupon its sensitivity, but also upon the extent
to be measured.
without excessive distortion the conditions where The distortions due to sorption processes are greatly minimized by making the conductance into the sensitive portion of the gauge from the region of interest as high as possible. The nude ionization gauge represents the logical extreme in this regard. Proper electrostatic shielding, use of a low-temperature cathode, thorough outgassing of the gauge components all contribute further to minimizing the discrepancy between the gauge reading and the true pressure. Specialized techniques have contributed to an extension of the range of measurement with improved assurance of reliability down to exceedingly low values of the pressure. Redhead^* has extended the lower limit for a Bayard- Alpert type of gauge, as
the pressure
An alternative approach to the attainment of low pressure is to provide the highest possible pumping speed. The additional problems arising in this instance because of directional effects have been discussed by Santeler.58 In space-simulation systems the objective is to expose a test body to an environment as nearly like that of outer space as possible. Since such test bodies cannot generally be baked at high temperature, the simulation must be accomplished as well as possible in the presence of surface outgassing so that the gas load Q remains high. In outer space, molecules leaving a space vehicle fly off into space and very few ever return. Thus true simulation of the outer space environment would be accomplished by surrounding the test body with perfectly absorbing walls. If this could be done, then at low pressure at which the mean free path exceeds the dimensions of the system, gas molecules desorbed from the test body would travel on straight-line paths to the wall of the vacuum vessel and there be absorbed, whereas
no gas molecules would be received by the surface of thg test body. The pressure would then be highly directional. An ionization gauge located in a recess in the wall of the vacuum vessel would receive a flux of molecules which could produce an ion current corresponding to a
pressure reading. An ion gauge located in a recess in the test body shielded from direct access by molecules from the surface of the test body would receive nothing, as would be the case in outer space. In practice no such perfectly absorbing surfaces exist. Liquidnitrogen-cooled or cryogenically cooled surfaces condense some gas
130
131
molecules more or less completely, but not generally on a single bounce. Evaporated surfaces of active metals absorb certain gas molecules very Other gases must be well, but again only after several encounters. the ratio of the total as defined is If A pumps. diffusion pumped by exist if all the walls would which that system to of the speed pumping
The extent to which the techniques described by Santeler can be used to correct for the directional pressure effect in systems of very high pumping speeds is difficult to assess. However, the directional
effect exists
and
results in large
were perfectly absorbing, then, as Santeler shows, the directional pressure effect is given by 1/(1 - A). An ordinary ionization gauge
Liquid-nitrogen
extension
REFERENCES
C. D. Hickman, Rev. Sci. Instr. 5, 161 (1934). Wallace and Tiernan, Incorporated, 25 Main Street, Belleville Absolute Pressure Indicator Type FA- 160. 3 H. G. East and H. J. Kuhn, J. Sci. Instr. 23, 185 (1946). 4 D. C. Pressey, J. Sci. Instr. 30, 20 (1953). 5 J. Dubrovin, Instruments, 6, 194 (1933).
1
.
K.
9,
N.J.,
Gouge
Electrical
electrodes
leads
6
Perforations
7,
8,
H. McLeod, Phil. Mag. 48, 110 (1874). M. P. Romann, J. Sci. Instr. 3, 522 (1948). E. H. Kennard, Kinetic Theory of Gases (McGraw-Hill Book Company New
Chamber wol
Chamber liquid-nitrogen
surface
9.
W.
10,
Fig. 3-59. Gauge mounting designed to compensate for directional pressure [Taken with permission from D. V. Santeler, Rev. Sci. Instr. 33, 283 effect.
(1962).]
and H. Klumb, Physik Z. 37, 440 (1936). G. C. Dunlop and H. G. Trump, Rev. Sci. Instr. 8, 37 (1937). R. J. Webber and C. T. Lane, Rev. Sci. Instr. 17, 308 (1946).
T. Hasse, G. Klages,
14.
located in the wall of the test vessel will then receive a flux of molecules times the flux received by the surface of the test body. 1/(1 In order for a gauge located in the wall of the vacuum vessel to produce a reading corresponding to the flux arriving at the surface of
15. 16.
A)
17.
the test body, Santeler developed the gauge arrangement illustrated in Fig. 3-59. A Bayard-Alpert nude gauge is mounted in a cylindrical extension from the liquid-nitrogen-cooled inner surface of the chamber. The liquid-nitrogen-cooled surface is perforated with an array of holes through which gas is admitted from the center of the test chamber. The ratio of the area of the holes to the total area of the cooled surface
is
19,
20.
21.
made equal to the wall pumping efficiency A Because the pumping speed for permanent gas
of the
is
main chamber.
M. Benson, in 1956 Vacuum Symposium Transactions (Pergamon Press, London, 1957), p. 87. M. Pirani, Verhandl. deut. physik. Ges. 8, 686 (1906). C. M. Schwartz and R. Lavender, Rev. Sci. Instr. 19, 814 (1948). A. R. Hamilton, Rev. Sci. Instr. 28, 693 (1957). J. A. Becker, C. B. Green, and G. L. Pearson, Trans. A.I.E.E. 65, 711 (1946); and Bell System Tech. J. 26, 170 (1947). P. T. Smith, Phys. Rev. 36, 1293 (1930); Phys. Rev. 37, 808 (1931); J. T. Tate and P. T. Smith, Phys. Rev. 39, 270 (1932). 0. E. Buckley, Proc. Natl. Acad. Sci. U.S. 2, 683 (1916). W. B. Nottingham and F. L. Torney, Jr., in 1960 Vacuum Symposium Transactions (Pergamon Press, London, 1961), p. 117. L. N. Ridenour and C. W. Lampson, Rev. Sci. Instr. 8, 162 (1937). R. S. Morse and R. M. Bowie, Rev. Sci. Instr. 11, 91 (1940). J. Blears, Proc. Roy. Soc. (London) A188, 62 (1947). 1. Langmuir, J. Am. Chem. Soc. 37, 1139 (1915).
J.
small both for the gauge and for the main chamber, the pressure due to permanent gas is isotropic and essentially uniform throughout the test chamber and the
25. L. Riddiford, J. Sci. Instr. 28, 26. G. Reich, in 1960 Vacuum 27.
375 (1951).
Press,
However, only the fraction of condensable vapor which returns from the surrounding walls to the test object is admitted to the gauge recess. Thus the gauge reads the full value of the permanent gas, all of which the test object receives, but only that portion of the nonisotropic pressure of the condensable vapors which the test object
gauge
recess.
28. 29.
30.
31.
receives.
London, 1961), p. 112. R. A. Haefer and J. Hengevoss, in 1960 Vacuum Symposium Transactions (Pergamon Press, London, 1961), p. 67. W. B. Nottingham, J. Appl. Phys. 8, 762 (1937). R. T. Bayard and D. Alpert, Rev. Sci. Instr. 21, 571 (1950). A. van Oostrom, in 1961 Vacuum Symposium Transactions (Pergamon Press, London, 1962), p. 443. H. J. Schuetze and F. Stork, in 1962 Vacuum Symposium Transactions (The Macmillan Company, New York, 1962), p. 431.
132
32.
33.
London, 1961),
34.
p. 108.
E.
J.
Lauer,
Lawrence
Radiation
Laboratory,
Livermore,
private
35.
36.
37.
38.
39.
40.
41.
communication D. J. Santeler, Rev. Sci. Instr. 33, 283 (1962). H. F. Winters, D. R. Denison, D. G. Bills, Rev. Sci. Instr. 33, 520 (1962). W. B. Nottingham, in 1954 Vacuum Symposium Transactions (Committee on Vacuum Techniques, Inc., Boston, 1955), p. 76. (Pergamon Press, J. M. Lafferty, in 1960 Vacuum Symposium Transactions London, 1961), p. 97. Transactions (Pergamon Press, J. M. Lafferty, in 1961 Vacuum Symposium London, 1962), p. 460. Transactions (The Macmillan J. M. Lafferty, in 1962 Vacuum Symposium
CHAPTER 4
The measurement of the "total" pressure as represented by the response of an ionization or thermal-conductivity gauge is a very useful indication of the state of a vacuum system, but is insufficient as a guide
toward further improvement. Measurement of the partial pressures (or molecular densities) of
Cover
cut away
1962), p. 438.
42. F.
London, 1962), p. 439. M. Penning, Physica 4, 71 (1937). 43. F. M. Perming and K. Nienhuis, Phillips Tech. Rev. 11, 116 (1949). 44. J. H. Leok and A. Riddoch, Brit. J. Appl. Phys. 7, 153 (1956). 45. A. H. Beck and A. D. Brisbane, Vacuum 2, 137 (1952). 46. R. A. Haefer, Acta Phys. Austr. 7, 251 (1953); 8, 213 (1954). 47. P. A. Redhead, Can. J. Phys. 36, 255 (1958). Elec48. P. A. Redhead, Report on the 18th Annual Conference on Physical
tronics,
the
component
provides
much
more which
ment.
insight
limit
into
the processes
MIT
Press,
(1958);
and
in
Transactions (Per-
London, 1959), p. 148. 49. T. N. Rhodin and L. H. Rovner, in 1960 Vacuum Symposium Transactions (Pergamon Press, London, 1961), p. 228. 50. J. R. Downing and G. Mellen, Rev. Sci. Instr. 17, 218 (1946). 51. R. H. Vacca, in 1956 Vacuum Symposium Transactions (Pergamon Press, London, 1957), p. 93. 52. M. Knudsen, Ann. Phys. 31, 633 (1910). 53. J. H. Leok, Pressure Measurement in Vacuum Systems (Published for the Institute of Physics and the Physical Society by Chapman & Hall, Ltd., London, 1964), 2nd ed.. Chap. 4. 54. H. Ishii and K. Nakayama, in 1961 Vacuum Symposium Transactions (Pergamon Press, London, 1962), p. 519. 55. J. R. Roehrig and J. C. Simons, Jr., in 1961 Vacuum Symposium Transactions (Pergamon Press, London, 1962), p. 511. 56. P. A. Redhead, Rev. Sci. Instr. 31, 343 (1960). 57. F. L. Tomey, Jr., and F. Feakes, Rev. Sci. Instr. 34, 1041 (1963). 58. D. J. Santeler, in 1959 Vacuum Symposium Transactions (Pergamon Press, London, 1960), p. 129; and Rev. Sci. Instr. 33, 283 (1962).
gamon
Measurement of
means
sys-
of detecting leaks in
vacuum
tems more quickly and with greater sensitivity than by other methods.
4-L
Magnetic- deflection
The mass spectrometer ana[Taken with permission from A. Guthrie and R. K. Wakerling (eds.), Vacuum Equipment and Techniques (McGraw-Hill Book Company, New York, 1949).]
Fig. 4-1.
lyzer.
measurement or vacuum
standing and improving vacuumsystem performance, to the author's knowledge, was that
carried out at the University of California Lawrence Radiation Laboratory in connection with the development of the electromagnetic method (Calutron) for separation of uranium isotopes in 1943 and 1944. The
apparatus developed by Backus^'* was an adaptation of the simplest form of mass spectrometer due to Dempster,^ as illustrated in Fig. 4-1.
*
chapter.
134
135
Ions of the various gases present in the system are produced in a Penning (PIG) type of discharge, drawn out of the discharge through a narrow source sht by means of an electric field, and deflected through approximately 180 by a magnetic field normal to the direction of motion of the ions. Por appropriate values of the applied voltage and magnetic field all ions of a given charge-to-mass ratio zejm are refocused at the 180 point of a receiver slot, through which they pass to impinge on a collector electrode, and can be recorded. The kinetic energy of the
ions issuing from the ion source
is
{z=l,M
1.008)
would be
6.35
Vh{p
cm)
16.0
X 1,560
1.008
24,700
which would be somewhat impractical. Rather than attempt to apply such high voltages, a second receiving slot was provided in the Backus analyzer at a distance 2p = 5.7 cm from the slit of the ion source so that ions of p = 2.85 cm were also recorded. From (4-4)
y^mv^
108
zeV
(4-1)
if\144.5/
so that
where
m= V =
= z = V = c =
e
Gc p2 for fixed
values of
M and B.
=
/2 85\^
I
Thus
1,560
cm/sec
carried
Fo(p
2.85 cm)
^
85\^
by the ion
and
315
V
(2
^24,700
6.35/
'
an ion
in a
magnetic
field is
5,000
cmv
cm
Combining
(4-1)
(4-2)
which
B is
and
A wide range of masses could thus be focused on one or the other of the two slots with a reasonable range of the accelerating potential. Ionic masses from 1 to 40 could be scanned by varying the accelerating
potential from 5,000 to 625 V. horizontally on an oscilloscope
(4-2) yields
By
cm /2zeV
108
zeB\ mc
slots vertically, a trace showing current peaks roughly proportional to the partial pressures of the residual gases in the chamber was dis(4-3)
1.12
X 10 ,,hI_Y-
range.
Since the mass of an atom of unit atomic weight is 1.66 x 10~^* g, the mass of an atom of atomic weight if is m = 1.66 x 10"^* M, so that (4-3) becomes
played by sweeping the accelerating voltage repeatedly over the desired In Fig. 4-2 is shown a schematic diagram of the mass spectrometer vacuum analyzer together with its circuit diagram.
144.5/ifF\^
B
As an example,
gauss
for a singly
\~
(4-4)
charged atomic oxygen ion [M = 16), an and a magnetic flux density of 3,600
p{M
so that the focus
is
= =
16)
6.35
cm
the
slit
mass spectrometer which the central orbit of the ion beam and orbits diverging by a radians from the central orbit are shown. If the source is a narrow slit parallel with the magnetic fleld, which is perpendicular to the plane of the figure, the central ray reaches the base line drawn through the source and the center of curvature of the central ray at a distance x^ = 2p from the source. Prom the figure it is apparent that both the + a and a rays return to this same base line at a distance x^ = 2p cos a = 2/3(1 ol^j2) =
can be readily determined by reference to Fig.
4-3, in
The
2p
located 2p
12.7
cm from
angle.
pa^ since cos a 1 <x.^l2, approximately, when a is a small The width of slot to accept all orbits from +a to a is
With a choice of
this dimension for the location of the receiving slot, the accelerating potential necessary to record an atomic hydrogen ion
therefore
Ax
Xg
x^
poL^
(4-5)
136
137
From
M
\
beam
is
144.5/
V
32
^
so that
144.5/
and
^M
Ap = 2 2 X
p
or
(4-6)
Aa;
In the instrument the beam was limited by metal vanes at the 90 point to a total angular width of 20 ( 10). The receiving slot was located 12.7 cm (p 6.35 cm) from the ion source and had a width equal to 6.35/32 fn 0.2 cm. Ionic mass
Magnetic field
perpendicular
to paper
peaks of If = 15, 16, 17 were then completely resolved, and ionic masses up to about 50 were reasonably distinct.
Phase-shifting
produces a wide variety of ions slot from the residual gas in a vacuum system. Atomic ions such as H+, C+, N+, and 0+, and molecular ions such as H2+, N2+, O2+, CH+, CH3+, CH,+, C0+, CO2+, and B.fi+ Fig. 4-3. Quality of focus for 180 magnetic -deflection mass spectrometer. are nearly always observed. Some of the ions result from dissociation and ionization of atmospheric gases present because of small leaks in the system. Water is usually present and produces not only the H2O+ ion but also H+, H2+, and 0+ ions. Decomposition of hydrocarbon diffusion-pump fluid and of oil left on inner surfaces of the vacuum system results in C+, CH+, CH2+, and
PIG
supply
network
Fig. 4-2. Schematic circuit diagram for the vacuum analyzer. [Taken with permission from A. Guthrie and R. K. Wakerling (eds.), Vacuum Equipment and Techniques (McGraw-Hill Book Company, New York, 1940).]
since x
vacuum
for Jf
For acceptable resolution of ion masses of interest in Ax was made to equal a half mass unit 16 (oxygen). Thus the slot width must be
2p.
CH4+ ions as well as additional H+ and H2+ ions. The relative magnitude of the peaks corresponding to these ions on the oscilloscope trace changes as the pressure in the vacuum system decreases and the internal surfaces outgas. The if = 18 (II2O+) peak usually dominates for some time after the system has been pumped down from atmospheric pressure. Later, if the system is pumped by trapped oil diffusion pumps, the hydrogen, carbon, hydrocarbon, C0+, and CO2+ peaks become predominant and the water peak decreases and finally disappears. If the = 16 and = peaks remain high, there is a
A.=^^ 16
2
Po
32
PoK
(4-7)
strong possibility of a leak admitting air into the system. If this should be the case the mass spectrometer vacuum analyzer can be used very effectively for locating the leak by directing a jet of gas other than air in turn over the flanges and other parts of the system vulnerable to developing leaks. When the jet falls on the leak, the air leaking into the system is mixed with probe gas, and ion peaks appropriate to the probe gas appear or increase on the oscilloscope trace. Because
138
139
there
normally no if = 4 peak on the oscilloscope trace, helium 4 makes an ideal choice as a probe or leak-hunting gas since the ilf = strikes jet helium the when quickly line very zero peak rises from the
the leak. The use of helium gas in leak hunting was found to be so effective as shown in that a simplified version of the mass spectrometer analyzer, California of University the at Loevinger^ by Fig. 4-4, was developed
trometer in
leak
vacuum
analysis
and
Electron torget
^^\m% of mass
detection,
many
improveand a
^^12 or greoter
Plate
Electrons
P
Mognetic
field
Accelerator
are
mercially.
Hydrogen
To pumps
Hippie* substituted a hot-filament type of ion source for the PIG and
generally refined the electronic
cuits,
cir-
Filament
Helium ions
in resolution, sensitivity,
bility.
and
sta-
Fig. 4-5. Ion source, beam focusing, and collector for 180 magnetic-deflection helium mass spectrometer. [Taken with permission from D. E. Charpentier,
(Pergamon
Press,
London,
1957).]
beam
focusing,
and
collector
Removable
arrangement used by Charpentier^ in a simplified mass spectrometer type of helium leak detector. The author states that the sensitivity of his improved instrument for relatively rapid detection of leaks is
Cathode
faceplate
Filament
Kovor seal
/
t
The mass spectrometer helium leak detector. [Taken with permission Techniques from A. Guthrie and R. K. Wakerling (eds.), Vacuum Equipment and
Fig. 4-4.
New
A
Section
York, 1949).]
Magnetic
field
A-A
Lawrence Radiation Laboratory. The ion source used was a PIG one receiving similar to that in the Backus vacuum analyzer, but only reduced to was beam the of width angular the slot was provided, and intended was detector leak the Since focus. increase the sharpness of
4 beam due to helium, the amplitude of the to respond only to the If potential was only large enough accelerating a-c sweep applied to the 4 on the oscilloscope. The pulsating Jlf to cover the full width for and applied to the vertical amplified was current to the ion collector
ollector
Baffle'-
Collector slit-
plates of
meter through a tuned amplifier. vacuum system all self-contained a with The design of the apparatus, mounted in a compact cabinet on casters, made the helium leak detector a convenient and effective instrument for vacuum-system
an oscilloscope and
also to a
Suppressor^
100,000 megohms^
diagnostics.
mass spectrometer designed for helium leak detection. [Taken with permission from Saul Dushman, Scientific Foundations of Vacuum Technique (John Wiley and Sons, Inc., New York, 1949).]
140
10-8
141
atm cm^/sec
or 7.6
x lO-"
torr liter/sec,
magnitude lower
use of the instrument is exercised. For the accurate measurement of the masses of the isotopes of the Hght elements a 60 magnetic- deflection mass spectrometer was developed by Nier^ several years before initiation of the Manhattan Project work reported above. Subsequently a vacuum analyzer and helium leak detector of high resolution was developed for the Manhattan Project operations at Oak Ridge under contract with the University
of Minnesota
when the accelerating voltage is set at about 270 V and the magnetic field at 900 oersteds. The preamplifier tube for the ion current to the collector is mounted inside the spectrometer tube to ensure high resistance and to minimize leakage currents. The vacuum system for the mass
spectrometer tube consists of an oil difi"usion pump and dry-ice-cooled vapor trap with a resultant pumping speed of about 30 liters /sec. The authors claim a sensitivity of about 1.5 x 10~^ torr liter /sec of helium. A 90 magnetic-defiection mass spectrometer especially designed for ultrahigh-vacuum applications has been developed by Davis and Vanderslice. 1" The instrument is of unusually compact design, consisting of an electron-bombardment type of ion source and an electronmultiplier tube connected by a 90 elbow of welded stainless steel as shown in Fig. 4-7. After bakeout is completed, a magnet is put in place with poles on either side of the 90 elbow to bend the ion beam The ion detector is a 10-stage commercial photointo the detector. multiplier tube with the glass envelope removed and the unit remounted For ions incident upon the first dynode in a stainless steel housing. multiplier the gain is 10^ at 300 V/stage before bakeout the electron of 10' after bakeout at 425C. Slit dimensions and becomes about and alignment in the ion source are so chosen that the mass peak width is just equal to the separation due to one mass unit at an ion mass of 140. Partial pressures of the order of 10-^* torr can be measured with the instrument. Fast electronic sweeping of the ion-accelerating voltage, using a sawtooth sweep generator, provides a continuous oscilloscope trace of the gas components present. A double-focusing magnetic-deflection mass spectrometer, the beamfocusing scheme for which is illustrated in Fig. 4-8, is described by The ion beam passes through a 90 deflection magnet and is Peters. ^^ focused on the interstage slit, which becomes the source 8^ for an identical second-stage deflection magnet which refocuses the beam on the collector slot. The author states that by the double-focusing technique the signal-to-background ratio is improved relative to the The backtypical single-focusing apparatus by about a factor of 100. ground signal in the normal mass spectrometer consists mainly of ions, other than that for which the instrument is focused, entering the collector slot either because the normal breadth of the focus overlaps the mass separation or because ions which are well separated initially are scattered and enter the receiving slot. In the double-focusing instrument the intensity of the Jf = 4 beam is not changed significantly, but the background due to random ions other than ilf = 4 is reduced by a large factor. The author claims a sensitivity to helium of 10-^* cm^/sec of helium at atmospheric pressure, or slightly less than 10-^^
torr liter/sec.
by Hustrulid and Nier' and was later perfected for commercial manufacture in the form described by Nier, Stevens, Hustrulid, and Abbott.8 An instrument of the Nier type, designed specifically as
a helium leak detector by Worcester and Doughty,^ is illustrated in 4 peak is focused on the collector slot In this design the Jf Fig. 4-6.
Fig. 4-7. Sensitive 90 magnetic-deflection mass spectrometer for ultrahighvacuum applications. [Taken with permission from W. D. Davis and T. A Vanderslice, in 1960 Vacuum Symposium Transactions (Pergamon Press, London,
1961).]
142
LEAK. DETECTOBS
143
-i-s,l'"
Moss spectrum of propane (C3Hg
H
St
anolyzer
Interstage
slit
2nd anolyzer
pole piece
pole piece
Fig.
eter.
4-8.
Ion
beam
with
trajectories
mass spectrompermission
[Taken
from
J. L. Peters, in
19S9 Vacuum
Fig. 4-9. Ion trajectories in a cycloidally focused mass spectrometer. [Taken with permission from G. D. Perkins and D. E. Charpentier, in
1957
actions
1958).]
Vacuum,
London,
I960).]
Robinson and Hall^^ describe a cycloidally focused mass spectrometer, the operating principles of which are illustrated in Fig. 4-9 and Fig. 4-10. The structure of the instrument consists of three parallel, equally spaced plates. An electric field is maintained between the
plates
by application of voltage across the outside positive and negative field plates, and a magnetic field is applied parallel with the plate The region between the field plates is thus one of crossed structure. The center plate has a source slit, an electric and magnetic fields.
ion-resolving
independent of the ion energy, and (3) unusually compact structure compared with other magnetic-deflection mass spectrometers. The complete ion mass spectrum obtained with a similar instrument developed by Perkins and Charpentier^* is shown in Fig. 4-11 and demonstrates the resolution of individual
and a collector slit, all running parallel to the magnetic field. Ions are formed by an electron beam and projected through the source slit by application of a substantial voltage to a repeller electrode. The resulting beam of Positive field plate ions is roughly focused on an4
slit,
Ion
mass units in the range 16 to 44. The sensitivity of the instrument is reported to be about 1.7 x 10"^
torr partial pressure of nitrogen.
Signal
Path
of ions
generotor
at resonance
collector
slit
upon the
varies
and
at
resolving ^
slit
is
Ion collector
h-
V
Trapping
voltage
the
The
twice
the
nitrogen
is
sensitivity.
End view
Side view
lectron
node
Negot
field
Fig. 4-10.
Cutaway view of
ion source
and beam-focusing scheme for cycloidally focused mass spectrometer. [Taken with permission from G. D. Perkins and D. E. Charpentier, in
1957 Vacuum Symposium, Transactions (Pergamon Press, London, 1958).]
such that ions of a given charge-to-mass ratio all come to a sharp focus at the collector slit, even though the spread in energy gained from the repeller is large. The advantages of the design are (1) high resolving power due to the double-focusing feature resulting from the use of the resolving and collector slots, (2) high ion efficiency because the focus is
cycloidal focusing
The spectrum
maintained on an
at the
The omegatron of Sommer, Thomas, and Hippie. [Taken with permission from H. Sommer, H. A. Thomas, and J. A. Hippie, Phys. Rev.
Fig. 4-12.
an almost continuous record of 82, 697 (1951).] the gas components in the system. 4-2. The Omegatron Mass Spectrometer.
Sommer, Thomas, and Hippie^* reported a precision determination of the charge-to-mass ratio of the proton by measurement of the cyclotron resonance frequency in a device illustrated in Fig. 4-12, which has since been referred to as an omegatron. A singly charged ion moving in a direction perpendicular
144
145
moves
Bev
V
mv^
(4-8)
orbit in centimeters
is
>tt\
/
/
/
P,
mcv
J?
(4-9)
15
C:n^
*\
where
m=
= = V B= e =
c
mass of the
particle
'
''
/^
\\^
^v ix^
x^ _
, Ro=10
277i?
=
is
mc
27
>
sec
Be
'
=
x
=
2TTCm
e
<
cycles/sec
'
'
-^
-
13
4.80
and
m=
1.66
10-^*
M,
in
which
x
is
IQ-i" esu, c
lO^"
the mass
number of the
Fig. 4-13. Spiral orbits of ions in an omegatron. [Taken with permission from D. Charles and R. J. Wamecke, Jr., in 1959 Vacuum Symposium Transactions (Pergamon Press, London, I960).]
1.53
10=
Ion collector
R-f electrode
Electron beam
cycles/sec
(4-10)
Slit
.
field
and the
in a
an ion
uniform magnetic
field increases,
according to (4-9), but the frequency of the motion remains constant. The cyclotron type of particle accelerator is based upon this principle; thus the term cyclotron frequency is used to refer to the rotational frequency given in (4-10).
> RN
L->-To electrometer
Electron-source
filament
The omegatron is somewhat similar to the cyclotron in that ions are produced at the center of the device and spiral outward as shown in Fig. 4-13 under the influence of a high-frequency field which is perpendicular to the steady magnetic field. The body of an omegatron is in the form of a metal box, along the axis of which and parallel with the magnetic field a beam of electrons is directed as shown in Fig. 4-14. The electrons are drawn from a hot filament by an accelerating potential pass through holes in opposite sides, of the box, and are collected on an
ground point
from
W. R. Watson, R.
146
147
Ions produced by ionization of the residual gas molecules in the metal box are extracted from the If the electron stream by an applied high-frequency electric field. frequency of the applied field is equal to the cyclotron frequency for a particular type of ion formed in the electron column, ions of this type
electrode on the far side of the box.
by Watson, Wallace, and Lech" is shown in Fig. 4-15. As the frequency of the applied field is varied, the ion-collector current varies from some low background value to peak values occurring
whenever the
oscillator
Variation of
magnitude
of
peak
R-f
supply
^^
'^/iA
44
OS a function of
the amplitude Eq
of the electric field
for various
values
of the ionizing
current
+015volt
-0.48volt
-With
1,0 volts
2,820gauss
P = 410"^torr
221/2
volts d-c
90
volts
dic
Filament
supply
Electrometer
Recorder
1.5
20
2.5
to
cmp
Eo, volts/cm
Oto
2 volts o-c
Fig. 4-15. Electrical circuit for the omegatron. [Taken with permission from W. R. Watson, R. A. Wallace, and J. Lech, in 1960 Vacuum Symposium Transactions
1961).]
Fig. 4-16. Influence of the amplitude of the alternating field on the magnitude of the peaks for an omegatron. [Taken with permission from D. Charles and R. J. Warnecke, Jr., in 1959 Vacuum Symposium Transactions (Pergamon Press,
London,
I960).]
gain transverse energy and spiral outward until they strike the ioncollector electrode at the side of the box. If the frequency of the applied field is not equal to the ion cyclotron frequency, ions spiral out from the
electron
beam a
short distance
and then
spiral
back
in again, oscillating
Such ions are end plates of the box. At high gas pressure nonresonant ions tend to accumulate and produce an excessive space charge near the electron column and reduce the effectiveness of the applied high-frequency field in extracting ions. At very low pressure insufficient positive ions are formed to neutralize the spacein radius but never getting out to the ion collector.
lost along the lines of force to the
ions
charge potential of the electron beam with the result that resonant must be extracted from a potential well by the alternating field. To counteract these effects a weak positive d-c trapping voltage is applied to four sides of the box relative to the two sides across which the alternating voltage is applied. A schematic diagram of the circuit
frequency of some type of ion present. The height of the current peak for a given partial pressure is a measure of the sensitivity of the instrument and has been examined by Charles and Warnecke. i In Fig. 4-16 are shown their results on the influence of the amplitude of the = alternating electric field on the magnitude of the current peak for 44 (CO2+). When the alternating-field amplitude is very small, extraction of ions from the negative space charge of the electron beam is inefficient, and furthermore the spiral path for ions to reach the collector is long as compared with the collision mean path. As a result, ions are lost to the walls of the box by scattering along the magnetic field before they reach the collector and the collector current is low. As the alternating field is increased from about 0.5 to 1.0 V/cm, the magnitude of the peak current increases rapidly as the spiral path to the collector becomes progressively shorter. However, beyond 2.0 V/cm the magnitude of the peak current is observed to decrease again, and the authors
148
149
250-
Ic of the ionizing
current
|U--f(Ic)|
for different volues of the
amplitude Eq
Vpi=t0.15volt
200
with
Vp2 = -0.48volt
B P
Vo
=
l.57volts/cm
2,820 gau
= =
4xlO'Sn
Vo
Eo=1.256volts
=90
150
0.942 volt/cm
Eo 100
2.I98volts/cm
1.884 vol ts/ci
En
\Eo
50
0.628 volt/cm
0123456789
Press,
Eo= 0.314
volt/cm
44 on the partial pressure current. It is noted that the peak current is proportional to the partial pressure up to about 4 X 10-* torr for small electron -beam currents (1 to 3 //A) and up to about 2 X 10-* for an electron-beam current of 5 /lA. These authors report that under favorable circumstances and with an electron-beam current of 10 /lA a partial pressure of the order of 5 X lO-^^ torr can be detected. The corresponding ion-collector current is about 5 X lO-is A. Charles and Warnecke also report that the sensitivity of a similar omegatron designed specifically as a leak detector using argon as the probe gas is 2 x lO-^^ torr liter /sec in a dynamic system and as low as 2 X 10-1* torr liter /sec in a well-baked system using an accumulation process. The authors state that for the omegatron leak detector, argon, because of its relatively high ionization probability, is preferable to helium as a probe gas. In a detailed discussion of the potentialities of an omegatron as a leak detector Nicollian^' reports results in agreement with those of Charles and Warnecke, 1* using argon as the probe gas. However, he finds that the alternating-field gradient can be increased tenfold when using helium as the probe gas without causing broadening and splitting
Fig. 4-17. Influence of the electron-beam current on the magnitude of the positive ion peaks for an omegatron. [Taken with permission from D. Charles and R. J. Warnecke, Jr., in 7959 Vacuum Symposium Transactions (Pergamon
London,
I960).]
n7^(p
for different values of the
state
and broadened.
that the ion-ciirrent-versus-frequency peaks become distorted This latter condition develops when the increase in
is
60
.50
ionizing current Iq
Vp|= 0.17 volt
some of the resonant ions take the outer orbit shown in Fig. 4-13 and miss the collector entirely. Thus an optimum alternating-field amplitude was found to be about 2 V/cm, at which the collector current was maximum and the current peaks (as a function of frequency) were sharp and undistorted. The influence of the electron-beam current on the magnitude of the ion peaks as measured by Charles and Warnecke is shown in Fig. 4-17 for various values of the amplitude of the alternating field. For an alternating-field amplitude of about 1 V/cm or more the ion-collector
radius per turn in the spiral path
so large that
,
Vp2=-0.48volt
B =2,820 gauss
,40
=0.94volt/cm
=
Vo
Vo =90 volts.
30
20
/
la)
......
lc
=
10
^.
0'
l/iA
current
approximately proportional to the electron-beam current as is expected. At lower values of the alternating-field amplitude the curves are badly distorted, probably because of failure of the small alternating field to extract ions from the negative space charge of the electron beam. Finally, Fig. 4-18 shows the results of Charles and Warnecke for the
is
10
15
20
25
30
35
40
P 10'^ torr
= 44 on the partial Fig. 4-18. Dependence of the peak ion current for pressure of COj for various values of the electron -beam current for an omegatron.
Symposium Transactions (Pergamon
[Taken with permission from D. Charles and R. J. Warnecke, Press, London, I960).]
Jr., in
1959
Vacuum
150 of the
151
i peak. By taking advantage of this effect NicolHan obtains about the same leak-detection sensitivity with hehum as with argon. In either case a Hmiting partial-pressure sensitivity of about
10-12 torr
M=
sensitivity of
hydrogen gas is 30 to 50 times greater than for H+, the frequently not seen on vacuum-analyzer traces.
Table
4-1.
M=
peak
is
Neon*
Omegatron sensitivities calculated from ionization-gauge "pressure" readings and the heights of the principal peaks of the respective gases, that is, ikf = 16 = 20 for neon. for methane and
Hydrocarbon
fractions
Methane
Pressvire, torr
5.8
6.5 6.2
X 10-8
X 10"
6.2
ICeon
Pressure, torr
Sensitivity, div/torr ....
Sensitivity,
(i_,_/'J_)/P
. .
2.4
6.1
X 10-5
X 109
6.1
3.9 X 10-'
8.1
X 109
8.1
* Taken with permission from W. R. Watson, R. A. Wallace, and J. Lech, in 1960 Vacuum, Symposium Transactions (Pergamon Press, London, 1961), p. 421.
12 13 14 15
26 27 28 29 30 3132
4041424344
Mass number
Mass spectrum during bakeout of reflex klystron, (a) Just before' development of an air leak; (6) just after development of a small air leak. Temperature 320C, total pressure 5 x 10-^ to 1 x lO-^ torr by ionization gauge. [Taken with permission from D. Lichtman and A. Hebling, in 1960 Vacuum Symposium Transactions (Pergamon Press, London, 1961).]
Fig. 4-19.
Watson, Wallace, and Lech^^ have investigated the sensitivity in terms of (i^/O/P,,, in which i^ is the peak current for a given ion, i__ the electron-beam current, and P^, the partial pressure of the gas from which the ion current is produced. Table 4-1 gives some typical values The gases for the sensitivity obtained by Watson, Wallace, and Lech. used for these measurements were neon and methane. The "pressures" quoted are the nominal readings of an ionization gauge calibrated against
nitrogen.
the processing of ultrahigh-vacuum power tubes is discussed by Lichtman and Hebling." In Fig. 4-19 are shown the residual gas spectra obtained (a) during the advanced stage of bakeout of a reflex klystron and (b) just after the development of an air leak. In (a) the residual gas is made up mainly of hydrogen (H2+), water vapor,
Since the sensitivity of ionization gauges, i.e., the ratio about four times greater for nitrogen than for neon, the actual pressure of neon present was greater by about a factor of 4 than
{i+liJ)P,
is is
not
known
to the author.
The conversion factor for methane (CH*) The decrease in sensitivity for neon at
methane, and hydrocarbon fractions. When the leak develops in (&), the carbon dioxide, argon, oxygen (O2+ and 0+), nitrogen (N2+ and N+), and neon peaks become dominant.
is consistent with the general observation that the sensitivity of the omegatron decreases noticeably for pressures in excess of lO^^ torr. This trend is illustrated most clearly in Fig. 4-20 from the paper of Watson, Wallace, and Lech, in which the response to carbon monoxide is plotted as a function of pressure, giving a
2.45
152
153
low-pressure sensitivity of about 9.5 x 10^ div/torr for the Jf 28 peak and dropping off rapidly with increasing pressure beginning at about 10~^ torr.
Omegatron
Carbon monoxide calibration
Oiv.mass28
o ^
Sensitivity
^^q-9
Torr pressure
,_^o9.*o%
oo-
/:o?^o^
o
atomic mass discrimination. One advantage of these systems is that a magnetic field is not required to effect mass separation. The linear accelerator scheme was adopted by Moody^' in the design of a mass spectrometer helium leak detector and vacuum analyzer. The arrangement of ion source, r-f accelerator, and ion detector is illustrated schematically in Fig. 4-21. Ions are produced in a chamber by a beam of electrons which are emitted from a filament outside the chamber, pass through small collimating holes in the walls, pass completely through the chamber, jywvA*o+ Resolution and are stopped by an electrode adjustment outside the chamber on the far
side.
o o
o o
(fe
by the
of the
o o o o
cP
electron
o o
.-,.
-
rf
t^
D1V12
column by an extracting voltage between the ionization chamber and the focusing electrode. The extracted ions are focused on a collector through apertures in a series of plates which serve as
Ion
R-f
analyzer
Kinetic-energy
chamber
selector
Fig. 4-21. Schematic of linear mass analyzer. [Taken with permission from R. E. Moody, in 1956 Vacuum
(Pattern coefficient,
m/e
12)
(100
OiV28
i
1
10-7
10"6
10-5
Pressure, torr
10"
FiG. 4-20. Sensitivity of omegatron as a function of partial pressure of CO as = 28 peak. The pattern coefficient ratio between indicated by current in the = 12 (C+) and the = 28 (CO+) ion currents remains nearly constant as the sensitivity decreases at pressures above 1 x 10~^ torr. [Taken with permission from W. R. Watson, R. A. Wallace, and J. Lech, in 1960 Vacuum Symposium Transactions (Pergamon Press, London, 1961).]
are remarkably high. A problem which has not to the author's knowledge been satisfactorily resolved arises from the fact that the omegatron element is almost completely closed and has an enormous ratio of surface to volume. Both absorption and desorption effects must be present so that pronounced differences in the partial pressures of component gases must generally exist between the system being tested and the interior of the omegatron. 4-3. Linear High-frequency Mass Spectrometers. Linear resonant systems which depend upon the selective processes familiar in
high-frequency linear accelerator design or upon the differences in time of flight for ions of different mass have been utilized effectively for
electrodes for applying a high- Symposium Transactions (Pergamon frequency accelerating field. Alter- Press, London, 1957).! nate elements of the high-frequency electrode system are driven 180 out of phase by an oscillator. Actually, one set of alternate plates is grounded and the other is driven by the oscillator. The electric field in the gaps between any two adjacent plates varies as (EgjG) sin 2TTft, during the positive half-cycle of which ions coming from the source will experience an accelerating pulse. During the negative half-cycle, ions will be decelerated. Thus, if the ions pass through the series of high-frequency gaps in random phase, the net gain or loss in energy will be very small. However, if the spacings G^ between the electrodes increase in the proper sequence, ions of a particular ejm ratio and initial velocity entering the first gap in the proper phase will arrive at each succeeding gap at the proper time to receive an accelerating pulse at each gap. Such ions gain energy about proportional to the number of high-frequency gaps and attain considerably higher energy than those ions which are not so synchronized. By changing the applied frequency, ions of other e/m ratios can also be accelerated. Sweeping the frequency of the oscillating voltage through a sufficient range brings different ions into synchronism in succession. The ion beam leaving the high-frequency section is electrostatically deflected. Those ions which are not in synchronism, and thus have
154
155
not gained much energy, are bent sharply by the deflecting field, whereas those which are in synchronism and do gain maximum energy are only In order to discriminate slightly deflected and enter the ion collector. between the true ion current and the background current, the ion beam is modulated by imposing a 15 cycle/sec voltage to an electrode between the ion source and the accelerating electrode and detecting only the a-c component of the collector current through a tuned amplifier. A block
f/j^, and the r-f grid structure at negative d-c potential. Between the ion source and the r-f filter is a grid at high negative potential so that ions are drawn out of the source by a high accelerating electric field then decelerated somewhat to attain an energy eUg (as indicated on the diagram) on entering the r-f filter. Only those ions which gain an energy AW >
e?7j(,t ill passing through the r-f filter will reach the partial collector and be recorded as a partial pressure. The highly negative grid between the ion source and the r-f filter is designated the total collector since it intercepts a uniform fraction of ions of all masses and thus provides a current indication which is proportional to the total pressure as read by an ionization gauge. The grids of the r-f filter are equally spaced, and are alternately connected to the opposite polarities of a variable-frequency r-f oscillator so that successive grids are driven 180 out of phase. Each group of three adjacent grids constitutes a sorting structure of the Bennett type.^^ In passing through the sorting structure, ions, in general, experience a succession of accelerating and decelerating impulses and on the average gain or lose kinetic energy. The change in energy AlFy for an ion depends
Ion
gouge
(21
-^
-^
i_
,..,.,
.'.,
In |U
Filament
regulator
(2)
375
volts
power
supply
(4) -
15modulotor
(2)
^^-i
R-f
oscillator
250
volts
1-
power
supply
(4)
(4)
/
"a
z'
+
Preomplifie
(2)
\
A-c
amplifier
(2)
-^
Fig. 4-22. Block circuit diagram for linear r-f helium leak detector. [Taken with permission from R. E. Moody, in 7956 Vacuum Symposium Transactions
1957).]
the device is shown in Fig. 4-22. In the vacuum analyzer version the applied high frequency is swept over the range necessary to bring into = 100. In Fig. 4-23 is 2 to synchronism ion masses from
M=
vacuum
filter,
analyzer.
An
of Bennett stages in
the
entering
accelerating
electrode
structure.
was originally proposed by Redhead^" for use as a vacuum analyzer. Ehlbeck et al.^i have discussed the theory of this type of mass spectrometer and given results of measurements on the resolving power and Ions are produced sensitivity as a function of operating parameters. by electron bombardment in the ion source, accelerated through an r-f filter consisting of {2N + 1) precision-made grids, decelerated by a retarding grid, and finally selectively recorded on the partial-collector
electrode, provided a particular type of ion has gained sufficient energy
potential
the amplitude
cp
of the r-f
and the
where
[Taken with permission from R. E. Moody, in 1956 Vacuum Symposium, Transac(Pergamon Press, tions London, 1957), and through
the courtesy of Beckman Instruments, Inc.Fullerton,
CaUf.]
2nfd\2-U
surmount the potential applied to the retarding grid. Between the ion-retarding grid and the collector is an additional grid at high negative As potential to prevent any electrons from reaching the collector.
to
The transit angle is the phase interval of the applied r-f which a particle would spend in traversing the distance d between two adjacent grids of the ion sorter at the velocity v = [2(e/m)f7] at which it enters the sorting structure. The entering phase
interval
shown in the d-c potential plot in Fig. 4-24, the filament and partial ion collector are at ground potential, the ion chamber at a positive
99
positive
and the
over which the change in energy ^- after stages is fractional gain in energy AW^I^U^ are both critical
AW
157
Table
4-2.
functions of the ion e/m ratio when the remaining parameters (d, U, U, f) are held constant. Alternatively, if all other parameters are held constant and the radio frequency varied, then ions of different e/m ratios receive the maximum energy gain at discrete values of the frequency.
Analyzer*
electron current. Total yield of ion source at 4 Half width of energy distribution of ions Total current sensitivity (signal to total ion collector) Partial current sensitivity at U (r-f amplitude) = 140 volts Resolving power at U = 140 volts Upper limit of pressure at which the partial current is proportional to the partial pressure
.
mA
x 10~^ A/torr
4.5
eV
Pump
100
5 x 10* torr,
approx
Taken with permission from H. W. Ehlbeck, K. H. Loecherer, J. Ruf, and H. J. Schuetze, in 1960 Vacuum Symposium Transactions (Pergamon Press, London, 1961), p. 407.
*
Resolving power of 100 could only be attained by using such a highA mass spectrogram obtained with the precision grid structure.
instrument at a total pressure of 1 X 10"^ torr and resolving power of about 100 is shown in Fig. 4-25. Assuming that a partial-collector current of 10-1* A can be detected above 3.10""amp background, the minimum partial pressure detectable with the sensitivity of 2
X 10- A/torr
is
X 10-
can probably be reduced somewhat, at least for the lower range of mass
values.
oLLuIIL
10 12 14 16
18
JL
28
22
al.^i
shown
power
is
defined as
R=
/m
100
compact and relatively simple type of r-f mass spectrometer, called the Farvitron, has been described by Reich. 23 The electrode system for the Farvitron is shown schematically
in Fig. 4-26 together with the axial
Fig.
total
pressure
r-f sorter
type of
r-f
spectrometer
X(f2-A)
potential
in which /max is the radio frequency for a given e/m value at which the collector current is maximum and f^ and /a the frequency values below' and above /max at which the collector current reaches half the maximum value. The operating parameters of the instrument are given in Table
distribution.
Because
of
the geometry of the electrodes and the d-c voltages applied, the axial potential
of Ehlbeck et al. [Taken with permission from H. W. Ehlbeck, K. H. Loecherer, J. Ruf, and H. J. Schuetze, in 1960 Vacuum Symposium Transactions (Pergamon
Press,
London,
is
1961).]
distribution
is
99
parabola, that
is,
The authors emphasize that the transparency of the grid structure They found is a critical feature in determining the resolving power. that a grid structure in which each grid was a square mesh of 5 X 10"* cm diameter molybdenum wire with a spacing of 0.05 cm transmits only 7 per cent of the incident ions, whereas a structure in which each grid
4-2
.
between the two end electrodes and the central ring electrode. An ion of charge-to-mass ratio e/m injected into such a field experiences an
axial oscillation of frequencj^
C!\-
consists
35 per cent.
158
159
C =
4/7rL
electrical potential
frequency / is satisfying the above frequency relation will resonate and gain sufficient energy to escape from the potential pocket.
the distance between the end electrodes at 0. If an alternating potential of g9 superimposed upon the d-c potential, an ion of e/m
is
and
frequency is varied periodically by a 50 cycles/sec wobbler signal over the range of 0.13 to 1.8 megacylces/sec. For the dimensions and d-c voltages chosen by Reich the resonant frequencies are given by
/
where
= 2AM-
megacycles/sec
is
D-c supply
r-'r-i-ii
I"-!-"!
:
I
Tube
R-f
amplifier
The frequency swing imposed by the wobbler can be varied in breadth anywhere over the available frequency range so that either the full mass range from ilf = 2 to ilf = 250 can be displayed, or a much narrower mass range can be chosen and expanded to the full width of
the oscilloscope trace.
i^i'-r-'.
1
Demodulator
R-f generator
0,l3-l8Mc
The Farvitron is a relatively simple and compact form of r-f mass spectrometer which can be conveniently constructed for baking out at high temperature. The sensitivity is apparently limited, at least in the form described by Reich, to partial pressures not less than about 10~^ torr. The high scanning rate of 50 cycles/sec makes the Farvitron
particularly useful in following rapidly changing conditions in a
Wobbler
50cps
D
(Pergamon
Oscilloscope
vacuum
LflMfiiJ
Fig. 4-26. Schematic diagram of the electrodes and of the axial potential distribution of the Farvitron mass spectrometer. [Taken with permission from G. Reich, in 1960 Vacuum Symposium Transactions
Fig. 4-27. Circuit diagram for the Farvitron mass spectrometer. [Taken with permission from G. Reich, in 1960 Vacuum, Symposium, Transactious
1961).]
system.
4-4.
discussion
Press,
London,
would not be complete without mention of the halogen leak detector based upon the enhanced positive
ion
(Pergamon
1961).]
Press,
output
of
halogen-sensitive
Air flow-
London,
diode.
In the Farvitron the ions are produced by accelerating a regulated current of electrons from a tungsten filament axially into the electrode on the left, the end of which is a wire mesh. The electrons start from a cathode potential of 100 V, as shown in the schematic circuit diagram in Fig. 4-27, and will therefore penetrate the parabolic field to a depth of 100 V, producing positive ions by collisions with any molecules present. These ions oscillate in the parabolic field, most of them not having sufficient energy to reach the cup-shaped electrode on
the far end. However, when an r-f voltage is applied to the electrode on the left, ions of the e/m corresponding to the above frequency relation gain in amplitude of their motion and escape to the collector
electrode
diagram of halogen leak detector. [Taken with permission from W. H. White and
J.
S.
(1948).]
work function of the surface White greater than the ionization potential of the gas molecule. and Hiokey^" utilized the greatly enhanced production of positive ions
which occurs when a gas containing any one of the halogens (fluorine, chlorine, bromine, and iodine) comes in contact with a hot (^^900C) platinum surface as the basis for a leak detector. Their detector consists of a platinum cylinder mounted on a ceramic-clad heating element placed centrally within a larger platinum cylinder, as shown
on the right. The r-f current to the collector electrode is amplified and then rectified to produce a d-c voltage which is applied to the vertical deflection electrodes of an oscilloscope. The radio
The heated inner cylinder is made positive and the ion current is read on a microammeter, as shown in the diagram, or by means of an
schematically in Fig. 4-28.
(100 to 500 V) relative to the outer cylinder,
160
amplifier.
VACUUM SCIENCE AND ENGINEERING The halogen detector is most effectively used as a leak by placing it inside the vacuum system and probing the
fine jet of
161
detector
system with a
Freon-12 or other halogen-containing gas. Torney^' has made a study of the optimum conditions for operating a halogen detector to ensure stability and sensitivity. The platinum diode produces a background current of positive ions even when no halogens are present. The background current due to this effect varies with the gas pressure and the temperature
of the
element
^C;
Amplifier
relaxation
oscillator
ond
power
supply
Requloted
voltages
,.
inner
element.
Fig. 4-30. Block diagram of circuit for halogen leak detector. [Taken with permission from F. L. Torney, Jr., in 1957 Vacuum Symposium Transactions
The
signal
ionization in the
1958).]
halogen- containing gas also depends upon the gas pressure (of
air).
re-
two
though this change is small as compared with typical changes in the background ion current. One feature of halogen leak detectors which can cause difficulty is the relatively long "memory" of the detector once it has been exposed
to a surge of halogen gas.
rent
20
40
100 200
400
ipOO
To
reSensing head
Pressure,;!
Halogen leak detector backgroand positive ion current for 1.60-A and 1.75-A heater current and signal
Fig. 4-29.
on the pressure in the system is shown graphically in Fig. 4-29. The operating range (70 to 200 fi)
crosshatched in the figure is so chosen that the background current
is
duce the memory period, Torney^' devised a mounting for the detector which provides for the purging
of the detector
as
is
by the introduction
The unit
system as
From system
under test
current for standard leak of 10~* cm^/sec as a function of the pressure in the system. [Taken with permission from F. L. Torney, Jr., in 1957 Vacuum Symposium Transactions (Pergamon
Press,
shown
in Fig. 4-31.
To vacuum pump
relatively independent of
London,
1958).]
forevacuum
shown
The
an
in parallel as
shown
in Fig. 4-316.
Coble to
control unit
optimum
Torney also observed that the background positive ion current changes slowly with time provided that the pressure and circuit parameters are steady, whereas the signal due to the introduction of a halogen gas rises much more rapidly. Utilizing this difference in response, Torney developed a circuit which facilitates discrimination between background fluctuations and signals due to a leak. The circuit, a block diagram of which is shown in Fig. 4-30, contains a network between the detector and the amplifier which constitutes a bandpass filter which bypasses through C2 the high-frequency noise generated in the detector, is unresponsive to the very low frequencies associated with changes in the background ion current, but transmits an intermediate band of frequencies typical of changes in the signal due to detection of a leak by use of a halogen gas. Subsequent amplification of the signal beyond the bandpass filter then permits the sensitive detection of the enhanced positive ion current due to the application of a halogen gas to a leak even
series arrangement is the resulting low conductance for gas flow. The parallel arrangement in Fig. 4-316 may be permanently installed in a system without impairing pumping
pump
From system
under test
performance. According to Torney, 2' leak rates of 2 X 10~* atm cm^/sec will produce a full-scale deflection on his
version of the halogen leak detector, and leak rates as small as 2 x 10"^
Fig. 4-31. Methods of connecting halogen leak detector into a vacuum system. [Taken with permission from F. L. Torney, Jr., in 1957 Vacuum. Symposium Transactions
(Pergamon
Press,
London,
1958).]
atm cm^/sec can be detected when proper precautions are observed. 4-5. Leak-detection Techniques. Leakage through flange seals,
welded or soldered joints, and flaws such as cracks and porous sections of metal is an important cause of vacuum-system failure. The degree
162
to
163
which leakage must be eliminated in vacuum systems is far greater than that required for pressure and most other vessels in common engineering experience. Because of the importance of eliminating leakage, methods of detecting and localizing leaks constitute an important element in vacuum practice. Larger leaks in vacuum systems
relatively crude methods.
shown
in Fig. 4-33.
baffle
pump. When the pressure has reached a nearly steady value during pumpdown, the trap is cooled by liquid nitrogen removing the condensable vapor, after which a base pressure is reached depending upon the outgassing and leak rates. The valve between the diffusion pump and
may be detected by any of several The system may be pressurized slightly by closing the valve to the pump and
connecting a tank of nitrogen to the system through a regulating valve set at a gauge pressure of a few pounds per square inch, carefully avoiding the risk of applying
is
rise
Burner
Copper plote
an unsafe overpressure. Gas will then flow out through the leaks, the larger of which can then be
located
by
painting
suspicious
and lookI
Punnping
storted
Liquid nitrogen
eokoge
rote
from A. Guthrie and R. K. Wakerhng (eds.), Vacuum Equipment and Techniques (McGraw-Hill Book Company,
can be detected most easily if the gauge pressure is kept very low. Alternatively, the system may be Now York, 1949).] pressurized with a halogen-containing gas such as Preon-12 and a sniffing method used to detect the halogen gas coming out through the leaks. A hahde torch such as that illustrated in Fig. 4-32 is convenient for this purpose and reasonably sensitive. The air-intake hose shown in the figure is used to explore the system for leaks. When the inlet end of the hose sniffs the halogen gas, the flame in the torch turns green. Small components of a vacuum system can be
separately sealed, connected to a compressed-air supply, and immersed
in water.
Pump
volved off
introduced
trap
from system
Time
>
vacuum system with a significant leak present. [Taken with permission from W. F. Briggs, A. C. Jones, and ,T. A. Roberts, in 195S Vacuum. Symposium Transactions (Pergamon Press, London, 1959).]
Fig. 4-33. Pressure vs. time for a
followed in time.
is
Since the outgassing rate diminishes with time, the pressure-rise curve typically has a decreasing slope as long as the curve
dominated by outgassing. However, if the curve becomes a straight some time, the pressure rise may be assumed to be dominated by a leak, the value of which \s Q = V dPjdt, where V is the volume of
line after
According
the system.
to localize the source.
to Guthrie
and Wakerling,^* the pressurizing methods are limited in sensitivity to a leak rate of the order of 10^^ atm cm'/sec, which is entirely adequate for locating the larger leaks that would prevent pumping a system down to the region of ionization-gauge operation. When the leak rate in a vacuum system is low enough that the diffusion pumps can be put into operation and a pressure less than 10~^ torr attained, more sensitive methods are required to locate the remainIn this case ionization gauges may be operated in the portion of the system and heat-conductivity gauges in the forevacuum section. The behavior of a vacuum in this condition has been described by Briggs, Jones, and Roberts^^ in terms of the pressure
ing small leaks.
fine
has been determined that a leak is present, the next step is The procedures which may be used are many, but only some of the most efficacious will be mentioned. Briggs et al.29 describe the use of a null method in the circuit of a cold-cathode (PIG) type of ionization gauge as shown in Fig. 4-34 to detect with high
it
When
sensitivity
any change
in the
system pressure.
leak
present, the system pressure remains fairly steady at a value determined by the leak rate and the pumping speed of the system.
is
vacuum
Under these conditions the steady reading due to the system pressure can be balanced out as shown schematically in the figure and any changes in pressure, up or down, detected with increased sensitivity.
164
165
the
leak
rate
now be
thus roughly localizing any leaks present. The is very widely applied since it facilitates localizing the leak within a small area. A tank of helium with a regulator valve and a hose terminated by a small nozzle is used When the probing gas jet hits to explore the vacuum system in detail. the leak, the helium leak detector responds in a time depending upon the capacity of the system and the size of the leak. Most leak detectors produce an audible signal, the sensitivity of which can be set for detection of small or large leaks. The sensitivity of a helium leak
in
D-c power
supply
I
Ionization
gouge
d^
changed by squirting water or other liquid on suspicious parts of the system to cause a momentary
decrease in the leak rate or using some gas other than
hoods of plastic
foil,
gas probe
method
by
air,
Sr
To
vacuum system
Shunt
Null indicotor
such as Freon, COj, or helium, to cause a change in the gauge response, the balance in the gauge circuit will be disturbed and the of a leak indicated. location Methods such as these are generally capable of detecting leak
rates
detector
is
C3
Reference voltage
J.
of the
order of lO""
atm
is
cm'/sec.
sensitivity
As was
respond when air is replaced by pure helium at atmospheric pressure. In the earlier sections of this chapter the sensitivities of several helium leak detectors are given on this same
instrument
will
S7
rn
Equipment
Leak detecto
Roughing
under test
/?=^
^^|,^^
pump
hunting. [Taken with permission from W. F. Briggs, A. C. Jones, and J. A. Roberts, in 195S Vacuum Symposium Transactions
leak
already stated, this device is installed in the forevacuum line of the vacuum system and has a
sensitivity of
basis.
However,
in
in searching for
leaks
ditions
Fig. 4-36. Gas probe method of apply[Taken with ing helium leak detector. permission from W. F. Briggs, A. C. Jones, and J. A. Roberts, in 1958
(Per-
1959).]
about 2 X
lO"**
atm
cm*/sec.
atm cm^/sec
By far the most sensitive and versatile of the leak detectors is the mass spectrometer helium leak detector, several types of which are
described earlier in this chapter.
detector,
In application the helium leak which has its own complete vacuum system, is connected into the forevacuum of the system being tested through a control valve as Equipment under test\ illustrated in Figs. 4-35 and 4-36. In Fig. 4-35 the system to be leak tested is enclosed in a hood into which helium is injected so that the system is surrounded by a mixture This method is of air and helium. particularly effective if the problem Envelope containing is to determine whether a vacuum helium-air mixture device has a leak greater than some Fig. 4-35. Hood method of applying specified value, but it does not
help to locate the leak.
and detection of leaks as small as 10^1" atm cm^/sec is entirely possible under good conditions. No matter what probe gas is used in leak detection, precautions must be taken to avoid excessive flooding of the system and its surroundings with probe gas. The objective is to determine the precise location of the leak, not simply to determine whether one is present. If there is an
usually relatively straightforward,
excessive
amount of probe gas about the system, the leak detector will continue to respond for some time, whether or not the gas probe is A fine gas directed at a leak, so that time is lost in localizing a leak. jet which is turned on only for brief intervals and then turned off again
is
best.
will
and
The leak detector can then be kept operating at high sensitivity respond when the leak is struck by the gas probe with whatever
delay
is characteristic of the system. In Chap. 9 the operation of getter-ion pumps is discussed in some detail. Ackley et al.^" describe how the current drawn by a Vac-Ion type of pump may be used as a sensitive indicator for leak detection.
very large systems the hood method can be applied to sections of the system by enclosing portions of the system
On
helium leak detector. [Taken with permission from W. F. Briggs, A. C. Jones, and .T. A. Roberts, in 1958 Vacuum Si/mposium Transactions (Pergamon Press, London, 1959).]
One property of this type of pump is that for a given type of gas the current drawn by the pump is proportional to the throughput. Thus
for gas of type a
Qa
S,xPa
in the system,
(4-11)
and
component gas
where
7 is the
166
167
current
drawn
for a given
Since the measurement depends upon the change from the substitution of the gas at
of the getter-ion
pump
I
for that
=h
SPh
(4-12)
'
the leak, the limit of sensitivity for leak detection by this method depends upon how small the fluctuations in the getter-ion pump current may be before making the substitution. The authors state that by taking proper precautions the fractional change in the getter-ion pump
current can be of the order of 1/2,000.
pump
current determined
where Qg represents the internal outgassing load and Q^ is the leakage throughput of gas of type 1. If, now, at time t = 0, gas of type 1 is replaced by gas of type 2, then after a time t the change in the getter-ion
'r.\BLE 4-3.
pump
current
is
Relative Values of Pumping Speeds, Leak Rates, and IjP Factors Used in Determining the Change in Getter-ion Pump Current Due to Substitution of One Gas for Another*
Probo gas
Mit)
lit)
(IlnjillPh
1.25
SJS,
0.834
0.30 0.50
1.73
Q2IQ,
0.85
2.7 2.7 3.8 3.8
Ai//t
(^b' (^h.
since
-exp(^^^)_
-
A
He He
(4-13)
(with increased (with increased
0.5
0.167
-1-0.5
pumping speed)
0.167
0.50
-exp(-^<)_
remains constant.
Ha Hj
Oa
+ 0.1
-0.1 -0.5 -0.5
pumping speed)
0.50
1.0
2.12
1.25
0.95
presumably
The
CO,
detailed form of the current change with time depends critically on the
relative
sufficient
* Taken with permission from Ackley, Harrington, Francis, Jopsen, Lothrop, and MandoH, in 1902 Vacuum Symposium Transactions (The Macmillan Company,
Xcw
I
York, 1962),
p. 380.
[IIP). S, Q,
1
(4-14)
(IIP), S, Q,
for the parameters (IjP)2l{IIP)i, S^jS^, and Q^jQi and the observed values for A/// are given in Table 4-3 for several gases and +1 are observed conditions. It is noteworthy that values of A7/7 in response to the substitution of one gas for another, providing an excellent sensitivity for leak detection. The value of IjSP for air was measured and found to be about 20. Using the data from Table 4-3 and the above parameter for air, one finds from (4-14) that by substituting helium for air at a leak in the system, the change in current drawn
The data
due to outgassing, the smaller the minimum detectable leak rate. The consequence is that the minimum detectable leak rate is generally in the range from 10^" to IQ-^^ torr liter/sec, depending upon the condition of the system. Whatever the limiting value may be, it is comparable with that achieved using a mass spectrometer type of leak detector and is very convenient in systems in which getter-ion pumps are used.
REFERENCES
1.
by the
getter-ion
pump
is
M=
lOQair
2. 3.
in
amperes when Qair is the leak rate for air in torr liters per second. Ackley et al. claim that since a simple electrometer circuit can easily measure currents of lO^^^ A, the corresponding minimum detectable leak rate is about 10^^' torr liter/sec, and that even smaller leak rates may be detected by using a more sensitive
is
which A/
in
4.
R. Loevingor and A. Guthrie, in A. Guthrie and R. K. Wakcrling (eds.), Vacuum Equipment and TccliHiques, National Nuclear Enei-gy Series (McGraw-Hill Book Company, New York, 1949), pp. 207ff. A. J. Dempster, Phys. Rev. 11, 316 (1918). Locvinger and Guthrie, in op. cit., pp. 212ff. H. A. Thomas, T. W. Williams, and J. A. Hippie, Rev. Sci. Instr. 17, 368
(1946).
5.
London, 1957),
p. 114.
6.
A. O. C. Nior, Rev.
212 (1940).
108
7.
8.
9.
10.
p. 417.
CHAPTER 5
12. C. F.
13.
14.
15.
16.
17.
London, 1960), p. 94. Robinson and L. G. Hall, Rev. Sci. Instr. 27, 504 (1956). G. D. Perkins and D. E. Charpentier, in 1957 Vacuum Symposium Transactions (Pergamon Press, London, 1958), p. 125. H. Sommer, H. A. Thomas, and J. A. Hippie, Phys. Rev. 82, 697 (1951). W. R. Watson, R. A. Wallace, and J. Lech, in 1960 Vacuum Symposium Transactions (Pergamon Press, London, 1961), p. 421. D. Charles and R. J. Warnecke, Jr., in 1959 Vacuum Symposium Transactions (Pergamon Press, London, 1960), p. 34. E. H. Nicollian, in 1960 Vacuum Symposium Transactions (Pergamon Press,
18.
19.
20. 21.
London, 1961), p. 80. D. Lichtman and A. Hebling, in 1960 Vacuum Symposium Transactions (Pergamon Press, London, 1961), p. 187. R. E. Moody, in 7956' Vacuum Symposium Transactions (Pergamon Pr(>ss, London, 1957), p. 119. P. A. Redhead, Can. J. Phys. 30, 1 (1952). H. W. Ehlbeck, K. H. Loocherer, J. Ruf, and H. J. Schuotzo, in 19G0 Vacuum Symposium Transactions (Pergamon Press, London, 1961), p. 407.
Functions of Mechanical Pumps. Mechanical vacuum variety of functions in vacuum systems. Tiie first requirement is that of removing most of the atmospheric air from the system to some acceptable operating pressure; this operation is sometimes referred to as roughing out the system. The final operating level
5-1.
pumps perform a
as far as the mechanical roughing pump is concerned is in many cases the forepressure required for safe operation of a diffusion pump. The roughing pump must also maintain a satisfactory operating pressure
22.
W. H.
23. G.
24.
Bennett, J. Appl. Phys. 21, 143 (1950). Reich, in 1960 Vacuum Symposium Transactions
(Pergamon Press,
London, 1961), p. 396. I. Langmuir and K. H. Kingdon, Science 57, 58 (1923); Phys. Rev. 21, 380 (1923); Proc. Royal Soc. (London) A107, 61 (1925). K. H. Kingdon, Phys. Rev. 23, 774 (1924). W. H. White and J. S. Rickey, Electronics 21, 100 (1948). F. L. Torney, Jr., in 1957 Vacuum Symposium Transactions (Pergamon
Press,
London, 1958),
p. 115.
28. A.
Guthrie and R. K. Wakerling (eds.). Vacuum Equipment and Techniques, National Nuclear Energy Series (McGraw-Hill Book Company, New York, Chapter 5 of this book provides an excellent review of leak-detection 1949). instruments and techniques developed by the Manhattan Project during
World War
29.
II.
F. Briggs, A. C. Jones, and J. A. Roberts, in 1958 Vacuum. Symposi^im Transactio7is (Pergamon Press, London, 1959), p. 129. 30. J. W. Ackley, A. E. Barrington, A. B. Francis, R. L. Jepsen, C. F. Lothrop,
W.
and H. Mandoli, in 1962 Vacuum Symposium Transactions (The Macmillan Company, New York, 1962), p. 380.
whatever gas evolution occurs during the operation these two functions are essentially oil-sealed gas compressors in which the inlet pressure is that of the system and the outlet pressure is that of the atmosphere. Two other types of mechanical pumps which are effective in vacuum systems when backed by a roughing pump as a second stage are the positive displacement mechanical booster pump (sometimes called a blower) and the molecular-drag pump. The function of these latter two types of pumps is to provide high pumping speed at low pressure. 5-2. General Features of Oil-sealed Mechanical Pumps. Any mechanical pump capable of maintaining a high pressure ratio when used as an air compressor may be used as a vacuum roughing pump. Of the many possible types those which are most successful are oil-sealed rotary pumps of positive displacement, i.e., pumps which isolate at each revolution a specific volume of gas at the pressure in the system, compress the sample, and exhaust it to atmospheric pressure. In Fig. 5-1 is shown a cross section of a rotating plunger type of pump in several phases of operation from suction to exhaust. Figure 5-2 illustrates one type of vane pump in which an eccentric cylinder rotates within a hollow, cylindrical casing with a reciprocating vane mounted in the casing and maintained in contact with the eccentric rotor to provide a seal between inlet and outlet ports. Another type of vane pump in which vanes are mounted in a balanced, rotating member is shown in Fig. 5-3. Pumps of all these types are commercially available
in the presence of
of a system.
169
170
171
Inlet port
Rotor
II
II
Seal
I
Fig. 5-2. Vane type of |3ump in which an eccentric cyhnder rotates within a hollow, cylindrical ca.sing with a reciprocating vane mounted in the casing and maintained in contact with the eccentric rotor to provide a seal between inlet
compound
or double-
stage
(a)
(b)
Kinney
KC pump
is
Two important characteristics to be considered in rating mechanical roughing pumps are the pressure ratio which the pump can maintain between inlet pressure and the exhaust pressure, and the pumping speed with which the pump removes gas while decreasing the pressure from atmospheric pressure down to the limiting pressure of which the pump is capable. As a mechanical pump removes air or other gas from a tight system, the pressure decreases with the time until the pressure reaches the ultimate vacuum of the pump. As long as the system is free of leaks, the ultimate pressure is the atmospheric pressure divided by the compression ratio capability of the pump. As compared with
(c)
Fig. 5-1. Cross section of a rotating-plunger oi- Kinney type of pump beginning of the suction stroke, (6) at an intermediate position, and end of the exhaust stroke.
(a)
(c)
at the
at the
Fig. 5-3. Vane type of pump in which \-anes are mounted in a balanced, rotating member.
pump.
'
172
173
ordinary air-compressor performance, the compression ratio required of mechanical vacuum pumps is very high, typically from 10^ to 10'. In the design of pumps to be used for vacuum service great care is exercised to ensure that during each cycle the entire volume of gas which is taken in during the suction stroke is exhausted at the end of the cycle. Any small sample of gas which is not forced through the exhaust valve will be added to the subsequent displacement volume and tend to limit the ultimate pressure of the pump. Single-stage mechanical vacuum pumps are typically capable of maintaining an ultimate pressure of 5 x 10~^ torr partial pressure of air as indicated on a McLeod gauge. This performance represents a compression ratio of over 10^ and in practice can be achieved more effectively by a rotary rather than
This quantity multiplied by the rotational the displacement speed 8j). Mechanical pumps generally are rated in terms of the displacement speed rather than the true pumping speed, which is the displacement speed multiplied by
pump.
is
speed of the
pump
Table
5-1.
Model no.
Pump
rpin
cfm
liters/min
liters/sec
rc?jhr
by a
reciprocating compressor.
Because mechanical vacuum pumps must have positive clearances between the operating parts which communicate directly between the exhaust port and the inlet port, a sealing medium is required to prevent leakage through the clearances and ensure the maintenance of a high compression ratio. Because the closely mating parts require lubrication, the most satisfactory sealing medium is lubricating oil of low vapor pressure and high lubricity. In addition to sealing the pump against blow-by from the exhaust into the intake, the proper flow of oil through the pump is essential to ensure lowest attainable ultimate vacuum. When the pressure on the inlet side of the pump is low, the gas admitted to the pump during the early portion of the cycle is compressed into a tiny bubble by the time it reaches the exhaust port. If there is a sufficient flow of oil through the pump, this small bubble of gas is carried out through the exhaust valve with the slug of oil admitted during the cycle. Because the oil in a mechanical vacuum pump serves these two separate functions, the rate of flow and the distribution of Unfortuoil through the pump are important features of the design.
nately, the processes of sealing the clearances in a rotary
KD-30 KD-850
450 360 360 530 444 440 555 535 630 476 880 796
13 27 47 31
369 765
1,330
6.14
12.7
22.1 45.9
22.2
14.6
80
52.7
850
65 80 131 154
880 24,100
1,840 2,260 3,710 4,360
401
30.6 37.8 61.9
1,445 110.5
136
221.1
72.6
250
301
520
262 425
510.2
885
755
1,135
2
3 5 8
57
0.944
1.42
3.4
5.1
630
1,000
525 500
1,725
15
2.36 3.78
7.08 21.7
8.5
13.6
25.3
78.1
46
21.2
600
10
36
pump, the
entrainment of air in the oil stream, and the final ejection of the slug of oil with entrained gas bubbles are not understood in detail so that the design of mechanical vacuum pumps depends much more upon experimental development than upon analytical procedures. 5-3. Pumping Speed of Oil-sealed Mechanical Pumps. In Sec. 2-2 the pumping speed of a vacuum pump is defined as the volume of gas, measured at the inlet pressure, which the pump removes from the system per unit of time. In the case of positive displacement pumps the pumping speed can be defined as the product of two performance factors, displacement speed and volumetric efficiency. The displacement per revolution of such a pump is a purely geometrical quantity and is the free volume exposed to the inlet port at each
2.3 3.2
65.3
1.09
1.51
3.91
4.4 6.8
91 124.2 192
5.44
7.47 11.6
2.04
3.2
is, Sj,
eSjj.
vacuum pumps together with their displacement speeds at designated rotational speeds. For convenience the theoretical displacements are given in cubic feet per minute (cfm), liters per minute, cubic meters per hour, and liters per second, all of which units are commonly used in designating the pumping speed of mechanical pumps. (See Appendix IV for conversion factors.)
Kinney rotary
oil-sealed
174
175
varies from a atmospheric speed at displacement value which is nearly equal to the performance The pump the of pressure to zero at the ultimate pressure curve of a mechanical vacuum pump consists of a graph showing the actual pumping speed as a function of the pressure from atmospheric pressure down to the ultimate pressure of the pump. Such curves can be obtained only by detailed measurement of the pumping speed over many decades in the pressure by methods such as those described in detail in Chap. 7. Air is bled through a needle valve at a constant rate into a small, vacuum-tight vessel connected to the inlet of the pump and the equilibrium pressure read on a McLeod gauge. The flow rate of air at atmospheric pressure is measured by any of several methods,
.
pumping speed of the pump but is only an indication of what is pump as a sealing medium. The pumping speed of Kinney oil-sealed mechanical pumps depends upon the inlet pressure in a manner indicated by the graphs in Figs. 5-5 and 5-6, in which the pumping speed is plotted as a function of the permanent gas pressure. It will be noted that, whereas the pumping
put into the
speed of the single-stage pump falls to practically zero just below 10"^ torr, the pumping speed of the compound pump is greater than 50 per
19,824
depending upon the magnitude of the flow at each point. rate is then
The flow
16,992
Q =
time
r^ in
mm
ft'
torr
cim
liters
torr liters/sec
(5-1)
time in sec
when atmospheric pressure is 760 torr. The reasons for the particular choice
3) for these
of the McLeod gauge (see Chap. measurements are of some interest. One reason is that the McLeod gauge is an absolute gauge for permanent gases, which obey the general gas law PV = RT, and can therefore be calibrated by straightforward measurement of gauge dimensions. The other reason is that the McLeod gauge measures only the permanent gas pressure and does not respond significantly to condensable vapor pressure unless the vapor pressure is much higher than occurs under the conditions here The permanent gas pressure rather than the total pressure described. (permanent gas plus vapor) is also the only logical pressure to use in the method described above. The gas flow which is measured and enters the pump is that of the permanent gas, such as atmospheric air. Therefore the equilibrium partial pressure of the permanent gas determined by this flow is the pressure which should be measured to determine the pumping speed. The vapor pressure present during these measurements is mostly that of the sealing oil used in the pump and has nothing directly to do with the mechanical pumping action being measured. Moreover, the vapor pressure can be varied over a wide range, depending upon whether the oil is vacuum-processed or ordinary
lubricating
oil,
0.001
0.005 0.010
0.050 0.100
0.500
500
1,000
Pressure, torr
pumps
as
fall
to
zero until the pressure reaches the range of 1 to 2 x 10^* torr McLeod gauge reading or about 10~* torr on a liquid-nitrogen-trapped ionization
gauge.
without noticeably affecting the pumping speed of the The vapor pressure is thus not an indication
Equation (2-1) defines the gas flow through a pump as Q = P-m^j,^ which is equivalent to Eq. (2-18) in which the gas flow is defined as the volume flow dVjdt multiplied by the pressure. This quantity is sometimes called the throughput and is the same at all points in a system consisting of pumps and conductances, as long as there are no leaks. The throughput of a pump as a function of the pressure is a useful representation of pump performance for the design of vacuum systems.
177
226,4
KCV-7
Displocement;/ cfm (198 liters/mm)
ot
650 rpm
169.8
KCV-5
Displacement
:
at
113.2
liters/min)
KCV-2'
Displocement; 2.3 cfm (65 liters/mini
ot
56.2
479 rpm
10 50 100
0.0001
00005
0.001
0.005 0.010
0.050 0.100
Pressure, torr
0500
10
(a)
QOlO
0.050 0.100
Pressure, torr
0.500
50 100
2832
(c)
Fig. 5-6 (continued). Pumping speed of Kinney vacuum pumps as a function of inlet pressure.
2264
compound
oil-sealed
mechanical
169.8
In Fig. 5-7 the throughput for typical single- and double-stage mechanical roughing pumps is shown as a function of the pressure from "blank-off" to atmospheric pressure. The use of throughput curves in vacuum-system design is discussed in Chap. 8.
5-4.
The
Effect
of
113.2
After the pressure has been reduced to the operating level, a mechanical vacuum pump removes the gas from a vacuum system at a very low pressure and compresses it by a large factor to somewhat above atmospheric pressure during ejection through
Pump
Performance.
56.6
For permanent gases this process entails no serious complications. However, condensable vapors are frequently present. These materials tend to condense during compression and seriously
impair pumping efficiency. The condensable materials of concern in the operation of mechanical vacuum pumps are those which have a vapor pressure which at moderate temperatures is high as compared with, or at least comparable to, the desired total pressure in the system being evacuated, but low Such a material enters the as compared with atmospheric pressure.
O005 QOlO
0.050 0.10
Pressure, torr
(6)
0.500
10
50 100
Fig. 5-6.
pumps
compound
176
oil-sealed
mechanical vacuum
178
179
vacuum pump
pression.
as a vapor but condenses inside the pump during comPerformance of the pump is adversely affected in two ways: (1) That portion of the condensed material which is not ejected through the exhaust valve with the slug of sealing oil is added to the gas in the subsequent displacement volume and reexpands, thereby reducing the capacity of the pump for the subsequent stroke; (2) that portion which is carried out in the ejected slug contaminates the sealing oil in the
pressure decreases correspondingly, and the pump capacity occupied by the vapor is even larger. Thus, aside from questions of contami-
10^
10^
^^
10'
10
10'
...,.,_^
pump
10"
"
y ^ y
f^
^ ,^
''
_^^
pumping of condensables such as water disadvantageous because of the extremely low density of the material as it passes through the pump as compared with its much higher density as a liquid or solid in the system. Whereas the condensable material which most persistently causes trouble in mechanical vacuum-pump applications is water, other materials, such as the Freons encountered in the reconditioning of refrigerator units and the various solvents encountered in vacuum processing of many materials, also cause difficulties. Because of the wide range of contaminants encountered in vacuum processes and the wide range of requirements as to desired operating pressure, no one
nation of the
in a
pump
sealant, the
is
vacuum system
solution to the problem of contamination of vacuum-pump sealing oil has been found, but a variety of remedies applicable to specific situations have been developed.
10-2
10-3
>
^ ^^
10
^
-3
y"
/
:/
1-4 10"
10^
lO""
pumps
Fig. 5-7. Throughput for typical single- and double-stage mechanical roughing as a function of the pressure.
The contaminated oil eventually reenters the vacuum pump and further impairs pumping efficiency. In some cases the contamreservoir.
inant also causes sludge formation in the reservoir with the result that the oil does not flow properly and is no longer an effective sealant. In
contaminated by a condensable no very effective cleanup action in the standard operation of the pump, and the contaminated oil must be replaced by fresh oil to restore good pumping efficiency.
case, once the oil is excessively
is
any
material there
An important factor contributing to poor performance of a vacuum system, in which an appreciable quantity of water or other condensable material is present, is the fact that the condensed material expands by a very large factor in evaporating and thus occupies a disproportionpump capacity. As an example, water at 72F expands by a factor of 50,000 in changing from the liquid to the vapor phase, so that 1 lb of water becomes about 800 ft^ of vapor, provided the temperature of the source of vapor is maintained at the initial value. In practice, because of the high latent heat of vaporization, the temperature of the source of vapor drops significantly, the equilibrium vapor
ately large
In connection with the problem of contamination of vacuum-pump oil by condensable materials a word of caution is needed on the Such installation of an exhaust pipeline to carry away the discharge. lines, unless steam or electrically heated, normally provide very large condensing surfaces. If the pump discharge contains water vapor or other condensable material, condensation will occur in the exhaust line. This line should not, therefore, run directly upward from the outlet of the vacuum-pump separator tank because the condensate formed in the line would then drop back into the oil reservoir and recontaminate the sealing oil. An appropriate "drop-out" tank or trap with a drain should be installed beside the separator tank and the exhaust line taken up from the drop-out tank. Any condensate from the line is then caught where it can be drained periodically without The interconnection between getting back into the pump reservoir. the separator tank and the drop-out tank should preferably be heated to eliminate condensation at a point where flow back into the reservoir would be possible. This precaution should be taken on any installation involving an exhaust line, no matter which of the several methods for dealing with condensable contaminants is adopted. 5-5. Gas Ballast. The term gas ballast was applied by Gaede^-* to a method of preventing condensation in mechanical vacuum pumps invented by him and first applied to pumps manufactured by E. Leybold's Nachfolger of Cologne, Germany. The principle of gas ballast is to admit sufficient air into the cylinder during the compression stroke to prevent condensation of any vapor
sealing
*
chapter.
180
181
In the adaptation of Kinney pumps for gas ballast a through the head at the end of the cylinder in such a position relative to the eccentric cam that the hole is uncovered only during a portion of the compression stroke when the body of air being compressed in preparation for ejection is sealed off from the intake by the position of the piston. A valve is provided, which controls the
being pumped.
is
hole
drilled
temperature of the vacuum system is 68F (20C), the exhaust temperature is 140F (60C) and dry air is admitted as gas ballast at a flow rate equal to a fraction / of the displacement speed >S^, then from the general gas law (1-9)
where T^
Outlet
T^
Gas-bollast
line
= 273 = 273
20
60
= =
fSi,
X 760
needleodjustment
valve
293
F.,
_ ~
Fex
X 900
(5-2)
333
333^93
X 76^00 XfSjj
0.96 /<S^
(5-3)
Gas-ballast
air inlet
For the above example the exhaust volume V^^ must be no less than (l/15)Sjr where Sj) is the initial volume per unit of time of the body of vapor being pumped. Thus
Fex=^'S^ =
Admission of gas ballast
0.96/^^
(5-4)
Compression
harge to atmosphere
so that
0.069
(15)(0.96)
(5-5)
180
360
ot rotation
540
-
720
Amount
Amount
of
Rotation
is
This means that for the assumptions in the above example the gasdry air flow must be about 6.9 per cent of the displacement of the pump in order to prevent condensation when water vapor at 10 torr
ballast
shown schematically.
vapor pressure
pump through the hole from zero up to about 10 per cent of the displacement of the pump. The arrangement is shown schematically in Fig. 5-8. Under conditions of gas-ballast flow the exhaust temperature, taken
flow of atmospheric air entering the
is being pumped. The ultimate pressure of a single-stage mechanical vacuum pump is impaired by the introduction of gas-ballast air because the average gas pressure across the seals between intake and exhaust is greatly in-
creased with a corresponding increase in internal leakage past the seals. The curve shown in Fig. 5-9 illustrates the performance of a gasballasted
From
these
tests
it
conveniently as the temperature of the sealing oil in the vicinity of the exhaust valve, may, for example, be 140F (60C), at which temperature the vapor pressure of water is 150 torr. Thus in order to prevent condensation when pumping water at a vapor pressure of 10 torr, the
appears that a single-stage, duplex Kinney pump will blank off" at approximately 1.0 torr with a full gas-ballast flow and 5 x 10-=* torr with gas ballast turned off".
of about 15/1 bled into the cylinder during the compression stroke to prevent the volume of gas mixture during e:khaust from being less than 1/15 that of
volume by more than a factor without causing condensation. Sufficient air must be
in
Gas ballast is even more attractive for two-stage mechanical vacuum pumps. To prevent condensation in the sealing oil, only the second or
backing stage requires gas ballasting.
of
so that there
no danger of condensation
full
volume.
3
:
the displacement. 2 The exhaust pressure is about 900 torr, determined by atmospheric pressure plus the load of the valve spring. If the
182
183
120
.^
100 6 sb alia St
off
V
7 f/
/
/
1
.*-'
/
/
80
60
/
40
20
/
/
1
L 1z% ^"
gas ballast
^
t
/
5%gc s
bollos
/
0,001
0.01
-f 1.
0.1
r
1
10
100
Pressure, tcrr
Fig. 5-9. Performance curves for a gas-ballasted Kinney pump with zero, 2 per cent, and 5 per cent gas-ballast flow.
KDH-130 vacuum
some interstage condensation is possible during the early phase of pumping down a system which has been exposed to humid atmosstage,
pheric pressure.
Since the ultimate pressure at the intake of such a pump is not sensitively dependent upon the interstage pressure, the interstage pressure can be raised considerably by gas-ballast air injection into the second stage without seriously impairing the ultimate
typical gas-ballasted Kinney KC-15 pump pressure at the intake. blanks off at 2 x lO^^ torr with the gas-ballast valve open. In Fig. 5-10 are shown the performance curves for a KC-15 Kinney vacuum pump with zero and full gas-ballast flow (10 per cent of the displacement
A problem of some concern in the application of gas ballast is the water content of the injected air. Although dry air is obviously preferred, in most cases it is inconvenient or impractical to use anything but regular atmospheric air. Consider a gas-ballast flow of 10 per cent displacement with air at 78F and 80 per cent relative humidity. Since saturation water vapor pressure at 78F is 24 torr, the water vapor partial pressure is about 20 torr in the gas-ballast air. The water vapor injected into the pump with 10 per cent gas-ballast flow of humid air is thus equal to that which would be pumped at an inlet vapor pressure of about 2 torr. Since a dry air gas-ballast flow of 10 per cent would normally permit pumping water vapor at 10 torr, the use of air at 78F and 80 per cent relative humidity would limit satisfactory operation to 8 torr water vapor pressure, about a 20 per cent reduction in the amount of water vapor which the pump can handle without condensation in the sealing oil. For the worst conditions of temperature and humidity likely to be encountered, the capacity of a gas-ballasted pump for water vapor should not be decreased to less than about 3^ of its capacity with dry gas-ballast air. A single-stage mechanical pump with adequate gasballast flow performs with good pumping efficiency with water vapor inlet pressure in the range 5 to 20 torr. However, the pumping efficiency of a single-stage pump with adequate gas-ballast flow is poor at pressures below 5 torr. For operation with gas ballast in the range 0.02 to 5 torr a compound pump is required for satisfactory pumping
efficiency.
speed).
vacuum pumps make somewhat more noise when the near the ultimate attainable by the pump. The hydraulic noise due to action of the exhaust valves can be eliminated by opening the gas-ballast valve slightly without impairing the inlet pressure to the
All mechanical
is
pressure
~1
16
14
pump
5-6.
significantly.
--'
10
/ / /
_} ?7
/gas
t
/X
'
^
(
"/
/[
?*
Other Methods of Preventing Contamination by ConAlthough gas ballasting is a satisfactory method of preventing contamination of vacuum-pump sealing oil by condensables in many applications, other methods are more effective in some applications. One disadvantage of gas ballast is the impairment in pump performance, especially in the case of single-stage vacuum
densables.
/
/
-10%
gas ballast
pumps.
ballast
/
0.0001
0.00)
1
0.01
0.1
Pressure, torr
Fig. 5-10. Performance curves for a KC-15 Kinney 2 per cent, and 10 per cent gas-ballast flow.
zero,
Furthermore, gas ballast is particularly applicable to batch operations in which condensables are pumped for only a brief period followed by a period of relatively vapor-free operation during which the pump can rid itself of contamination. However, for continuous pumping of saturated vapor, such as water, and in very large installations other methods, some of which are briefly described below, are more
effective.
184
185
of the most obvious methods of preventing concondensable material in any system is to maintain of a densation so high that the saturation vapor temperature the system a throughout
Hot Pump.
One
pressure is in excess of the maximum gas pressure. Since the pressure at the discharge of a mechanical vacuum pump is typically about 900 torr, a temperature throughout the pump and separator tank of 230F (110C) is normally sufficient to prevent condensation of water
vapor.
pump
temperature water vapor should pass through the were a permanent gas. Kinney has developed a hot-pump installation and has applied it The hot pump is successfully in a number of large installations. particularly applicable to large installations involving very long running
At
this
Chap. 8 the use of refrigerated traps in this connection is discussed in some detail. When systems of very large volume must occasionally be let down to atmospheric pressure, the vapor load on condensers and traps can frequently be greatly reduced by drying the air bled into the system. Such a precaution can in some installations save hours of outgassing time in restoring the system to operating pressure after each letdown to atmospheric pressure. As can be seen from Fig. 5-6, 5-7. Mechanical Booster Pumps.
as
if it
periods and saturated water vapor, such as large freeze-drying and vacuum-cooling installations and water deaeration, and for maintaining
consist of a lagged
pump and
compound vacuum pumps consisting of two conventional, oil-sealed pumping units connected in series are generally capable of maintaining good pumping efficiencies down to pressures of the order of 10"^' torr and of maintaining limiting pressures of 10~^ torr or lower. However, in such a two-stage vacuum pump the compression ratio maintained by the first or high- vacuum stage is very low. Thus the useful work done by the first-stage unit in pumping the low-pressure gas is very small indeed, whereas the power required for operation is large because of
In spite of doing little useful work, the viscous drag of the sealing oil. a conventional, mechanical oil-sealed pump as a high-vacuum stage is limited to operating at low rotational speeds in part because of the viscous drag of the sealing oil and in part because of mechanical
limitations.
separator tank with thermostatic control at 230 to 235F connected to a drop-out tank, as previously described. An oil of higher viscosity must be used in place of the standard sealing oil so that the viscosity at the elevated running temperature will be in the correct range to seal the
pump
effectively.
In some installations a method of Air Stripping. consisting of blowing dry air into the discharge pipe between the exhaust valve and the separator tank has proved to be effective in
Knox Method
preventing contamination. This method was devised by F. A. Knox' at Oak Ridge during World War II and applied to large Kinney pumps The method has been independently discovered by in the Y-12 plant. others, particularly for application to small single-stage pumps where the Freons are troublesome sealing-oil contaminants. One advantage of the Knox method over gas ballast is that gas is not introduced into the pump interior so that the ultimate pressure capability of the pump For optimum effectiveness the air stream should be is not impaired.
injected just above the exhaust valve assembly.
Oil Purifier Systems. large mechanical
The net pumping action of a mechanical pump may be considered as pumping speed or displacement speed minus the rate at which gas migrates back through the pump because of internal leakage If the forward pumping speed is large as compared or other causes.
the theoretical
In many large installations involving several vacuum pumps, such as cable vacuum impregnation
and purification systems has proved to be advantageous. Oil from the discharge of all the pumps is pumped through a purification system and thence into a reservoir for purified
plants, the use of oil circulating
oil.
reservoir, rather
oil for all the pumps is supplied directly from this than from the separator tanks on the pumps. Condensers and Vapor Traps. In some installations, particularly for high-vacuum systems, it is preferable to prevent condensable In materials from passing through the mechanical vacuum pumps.
The
sealing
with the reverse migration, then the volumetric efficiency will be high even with a high pressure ratio between outlet and inlet, and the compression ratio which the pump can maintain at zero fiow (no fiow into the inlet) will be large. For operation at or near atmospheric pressure positive displacement gas compressors of a variety of different types are used successfully without the use of a sealing oil. Such pumps are built with wellbalanced interleaving rotors and definite clearances between the moving parts so that operation at high rotational speeds is common practice. Thus compressors or blowers of this type have high displacement speed for relatively modest dimensions. In normal use, however, the pressure ratio against which compressors of this type are required to pump is small. The loss in pumping speed with increasing pressure ratio, moreover, is very rapid even in the case of axial flow compressors, which are designed specifically to extend the range of operation into
the region of higher compression ratios. To adopt an unsealed, positive displacement compressor for use as a
good
186
187
with a high pressure ratio is a normal requirement service. However, the hmitation on the attainable comfor vacuum displacement blowers is the reverse flow ratio for positive pression difference maintained by the pump. due to the pressure slip called
pumping
In any multistage vacuum pumping system the throughput determines the pressure at the inlet of each successive stage in accordance with the continuity equation (2-1) of Chap. 2, which for our present purpose we shall write as
This reverse-flow rate is given by the pressure difference across the pump multiplied by the average conductance of the clearance slots between the internal parts of the blower. As we have seen, gas conductance is proportional to the pressure in the viscous-flow range and at reduced pressure finally reaches the comparatively low value char-
Q
where
^nPn
const
(5-6)
Q =
S^
gas throughput
= =
several stages of
pumps arranged
The effectiveness of the positive disacteristic of molecular flow. placement type of blower in terms of maintaining a high compression ratio should therefore improve as the outlet pressure is decreased until
the transition pressure range, discussed in Sec. 2-6, is reached. Below the transition pressure the conductance through the clearance slots in the pump is constant, and the compression ratio of the pump should As the first-stage unit of a twotherefore be expected to remain high. stage vacuum pumping system, a positive displacement blower does
therefore have considerable potentiality.
5-8. Analysis of Mechanical Booster-pump Performance. The pumping speed of a typical single-stage, oil-sealed vacuum pump
P
This
is,
stages in series
additional gas
stages.
is
which the
pump
sustains
is
(5-7)
Pi
Thus,
if
S9.
the first-stage
falls
below
torr.
At 5 X
pressure the mean free path for air [see Eq. (2-57)] is 10~3 cm Thus for mechanical clearances of several 2 X 10-3 j^. thousandths of an inch, which are tjrpically maintained in positive
this
pumping speed, resulting in an overall economy in the size and cost of the two -stage system. The net forward pumping speed of a positive displacement booster
^^^
Pi
^^^
displacement blowers, the maximum compression ratio is attained ^.gll within the range for good performance of a single-stage backing
pump may
be stated as
Oj
Si
Q
S,-
(5-8)
Pi
where
Sj)
= =
displacement speed
flow of gas into the
loss in
pump.
An example
of the type of
pump
Q =
S^
pump
at inlet pressure
P^
to internal leakage
backward
When
pump
inlet, i.e.,
that
commonly
when Q
0,
the
referred to as a
Roots blower, as illustrated in Fig. 5-11. Because of the high rotational speed at which pumps of this type can be operated, very large displacement speeds can be obtained
Fig. 5-11. Schematic cross section of a mechanical booster pump. [Taken with permission from C. M. Van Atta,
in 1956
the limiting value determined by the internal leakage parameter and the interstage pressure, i.e., the pressure Pj maintained at the discharge by a backing pump. Under these conditions of zero net flow the forward
Sjj is
completely balanced
Sr
(5-9)
relatively
small
by the
1.
internal leakage,
which
is
actions
1957).]
been analyzed in some detail by Van Atta and Sylvester,* Van Atta,^ Ziock,'' and Winzenburger.'
The flow back through the clearances of the pump due to the Pi) maintained by the pumping action. This
188
189
properly be referred to as the pump slippage, as a result of which an amount of gas (5-10) Q, = (P, - Po)C,
may
now
is
Pj instead of Pj.
By
combining this
result with
Eq.
(5-6)
we have
P.Sr,
Q = P^S, =
where S2
is
Q,
P,S,
(5-16)
back into the inlet side of the pump from the discharge side through the clearances. Here P^ respresents the inlet pressure under conditions of zero flow and C^ is the conductance through the clearances in the pump mechanism. 2. The reverse pumping action due to the existence in the pump mechanism of small pockets of gas which get trapped and carried back from the discharge port to the inlet port. If S^ is the volume per unit of time of gas at discharge pressure which is thus carried back, the amount of gas due to this cause which must be again pumped out is
leaks
pressure Pj.
the speed of the backing pump which maintains the foreSince P^ is now the inlet pressure, by analogy with
Qi __ 2
"1
JL.
^V
(5-17)
so that (5-16)
becomes
PiSn
from which
{P2
Pi)Os
s,
c,
PA = P,S,
+
s^
(5-18)
Qr
Pj'S'r
(5-11)
P,
+
is
(5-19)
The combined effect of these two mechanisms accounts for the entire pumping speed. Thus the quantity of gas which flows back from the discharge to the inlet because of these two internal processes
loss of
is
at pressure
P^
then
S,
Qi
(5-20)
^Qs +
Qr
(5-12)
from which
From Eqs.
(5-9)
loss in
to this
pump
internal leakage
o ^i
it appears that the pumping speed of the mechanical booster a magnification of that of the backing pump. Moreover, since from Eqs. (5-14) and (5-19)
is
^D =-B-^
Qi
= Qs+Qr 5
-^
p
-^0
p
(5-13)
*
Pi
Sj,
(So
C,
'
K
is
(5-21)
K
Po
flow
is
K=
+ +
pump
1
C,
(5-14)
S^
S^
Sj,
(5-22)
c.
In order for the compression ratio to be high, the two quantities in the denominator must be small, i.e., both the gas conductance C^ through the clearances of the pump and the reverse pumping speed S^ must be
small.
pression ratio,
The above discussion applies to the compressive action of the pump against the forepresssure when no net gas flows into the intake. Now consider the result of admitting a gas flow Q, as in Eq. (5-8), which by reference to Eq. (5-13) becomes
S,
a Qi
(5-15)
obtained from Eqs. (5-20) and (5-21). Thus if K, the zero-flow comis very large and therefore both 0^ and S^ are very small, the pumping speed S-^ is nearly equal to the displacement speed Sj). By a similar analysis Winzenburger' arrives at an expression slightly
different
from
namely
Pi
= T^ + K SD
S.
by
Px
considering the reverse flow through the clearances to be of the nature of gas flow through a nozzle in which the pressure ratio exceeds
190
191
the
For
mass flow
PiS,
is
given by
which the
is
first
cAP, T^^
when Pav
sufficiently large.
a quantity somewhat
(2-37)
less
RT^
than unity as has been shown for the case of a tube in Eq. (5-25) can be written in the form
and
(2-38),
This assumption appears to be valid for P^ > Pc defined in (5-30), i.e., in the region of viscous flow, in which (5-21) reduces to a simflar exHowever, for P^ < P^ the pression to that given by Winzenburger. and an expression of the molecular, flow through the clearance slots is form given in (5-21) is then valid. In the foregoing discussion the conductance through the clearances of the pump G, and the reverse pumping speed 8^ due to imperfections in the rotor contours have been tacitly treated as constants independent
of the pressure. This is true only for limited ranges of the pressure. In Sec. 2-6, the Knudsen formula for the conductance of a tube of circular cross section in the transition pressure region is discussed. By analogy with Eq. (2-39) the conductance of a slot of long, narrow
cross section in the transition pressure region
is
C where
in
Co
C,(^P..
62
^) 61 Co
62
-1
av
^
4
(5-26)
(5-27)
Ci
of the form
which the constant 4 is arbitrary so far as the consistency in this is concerned but is chosen to conform with the definition of the An approximation to the transition pressure P^ discussed in Sec. 2-6. conductance of a tube to replace the more exact expression given in (2-43) leads to an expression like (5-26) in which the transition pressure Equation (5-26) approximates the true is that given in (2-55). conductance (5-23), which has the same general form as Eq. (2-43), shown graphically in Fig. 2-3 by two straight lines, a horizontal
step
C =
in
&lPa
C,P^
(5-23)
KP: 2-* av
by (5-24) for Pav Pt/4 extending the molecularfiow value up to the transition pressure and a straight line proportional P/4. to the pressure for the viscous-flow region for Pav
straight line given
<
>
the average pressure through the slot, Co is the low pressure or molecular-flow conductance of the slot similar to that given in (2- 102), 6^ and 62 are constants similar to k^ and Ajj used in (2-39), and Ci is a constant which replaces the constant TrD*ll28rjL of (2-39)
which Pav
is
Since
is
by
slot
Pav
^^^
approx
(5-28)
The clearance slot, the conductance of which we wish to represent in general form, is one for which the width is a minimum at the line of near contact between the rotors or between rotor and cylinder and increases with the contour of the parts on either This slot can be represented by one side of this minimum clearance. of a uniform width equal to the minimum clearance and of a length in the direction of flow, which depends on the details of the geometry. The conductance is in any case given by an expression of the general form of (5-23), although the exact values of the constants Cq and C^ which depend upon the geometry must be determined empirically. Note that for very small values of the pressure Eq. (5-23) reduces to
in the case of a slot.
then if the operating conditions are such that Pj is very small as compared with P2 the former may be neglected in (5-28) so that
-t
Pav
^r^
(5-29)
and Eq.
(5-26)
becomes
C.
~r
-^ (P2
p.>4'
(5-30)
C =
1
'
in
(5-24)
which P^
Co
Pav
"<
-^av "^ 7~
Several rather gross approximations are made in the transition from the exact expression (5-23) for the conductance through a slot to the
Co
CjPa
61
(5-25)
62
combination of Eq. (5-24) for Pg < P(/2 and Eq. (5-30) for P2 > Pf/2. For an analysis of the performance of a mechanical booster type of vacuum pump, however, these approximations lead to a sufficiently accurate representation of the conductance through the clearances
192
since in practice
193
is 10 or more times P^ of interest. range (the inlet pressure) over most of the pressure narrow slot for air at low low-pressure The conductance of a long, pressure is given by Eq. (2-102) of Chap. 2. For our present purposes
when the interstage pressure Pg = P(/2 = P^, at which pressure the mean free path is twice as long as it is at the pressure P^. Thus at
the value of the pressure Pa at which the sharp break appears in the approximate conductance curve
we
d
(5-31)
62,
^2
C,
in
Pc
free
(5-35)
For
air at
path
(5-36)
which the constant k depends upon the units used. The clearance between the rotors of a Roots type of compressor, however, is not of uniform width for a clearly defined depth z, but is defined by curved surfaces. Nevertheless it can be seen that the low-pressure conductance of such slots will have the form
Co
= -'
in. is
where
torr.
is
/Ij
x 10-^ cm
x 10~^
the
mean
free
path at
P =
Thus at the
critical
mean
or
free
path as follows
k^d^
(5-32)
where d is the root mean square of the mechanical clearance between the rotors or between each rotor and the cylinder wall, and fcj is in part a geometrical constant averaged over all orientations of the rotors. What is of main significance is that the low-pressure conductance
varies as the square of the radial clearance d.
=^ d
vacuum
(5-37)
in
it
is the average rotor clearance. the above discussion of the conductance of gas through slots can be seen that the conductance through the mechanical clearances
which d
From
high values of the pressure the conductance through a slot is proportional to the pressure. Also in the case of a slot the conductance at high pressure is proportional to the cube of the slot width. Thus it is clear that in Eq. (5-30) as applied to the conductance through the radial clearances of the pump, Cj must be of the form
(5-30), for sufficiently
booster
fol-
pump may
be represented with reasonable approximation by the lowing expressions, each applied to its proper pressure range
C,
k^d^
for
P,
<
6Ai
(5-38)
Ci
in
fc^rf*
(5-33)
and
C^
k^d^
k^d^
which k^ is in part a geometrical constant. Note that in the above discussion no mention is made of conductance through the clearances at the ends of the rotors. The reason for this omission is that the flow path \z in Eq. (5-31)] for the end clearances is very long between flat surfaces. Thus for very adequate end clearances this conductance is negligible as compared with that through
the radial clearances.
(--^)
for
P,
>
'-^
(5-39)
pump
The
The transition pressure Pj has been related to the mean free path arid diameter in the case of gas flowing through a tube. In Eq. (2-56) it is seen that at the transition pressure the diameter of the tube D is about 11 times the mean free path 1< of the gas molecules. Approximately this same factor applies to a slot in which case the width of the slot d at the transition pressure is
d
12L
approx
(5-34)
in the
now be seen from Eq. (5-20). independent of the pressure until the interstage pressure reaches the transition value and then increases linearly with the pressure above this value, as given in Eqs. (5-38) and (5-39), is of fundamental importance. When the gas flow Q is zero and the interstage pressure is at the limiting pressure of the backing pump, that is, 82 = 0, the pumping speed 8-^ of the booster pump will also be zero. At this point the limiting value Pq of the inlet pressure will be determined by the zero-flow compression ratio as in Eq. (5-14). The pumping speed will then rise rapidly as the gas flow and pressure increase, primarily because the pumping speed of the backing pump is increasing. When the pumping speed of the backing pump reaches its normal plateau value at an interstage pressure of about 0.1 torr, the pumping speed of the booster pump will also reach a plateau value.
as a function of the pressure can
is
fact that C^
T
194
195
As the pressure
will
pumping
then be maintained until the interstage pressure reaches the speed 5' P^, which in practical cases may be in the range 1 to value critical Above this point the pumping speed is expected to decrease with torr. increasing pressure, primarily because C the internal leakage through the clearances, is increasing rapidly with the pressure. Thus one expects a very broad pumping-speed curve which rises from zero at the ultimate pressure to a flat plateau value and then decreases as the
interstage pressure
the preceding test and Eq. (5-14) amounts to a determination of the parameter S^. In addition, the value of the interstage pressure at which begins to decrease as the pressure increases is identified as
f I
14
h-i
.,
70
12
^~60 Q_
increased beyond the critical pressure. If the zero-flow compression ratio given in Eq. (5-14) is high as compared with the staging ratio between the displacement of the booster and its backing pump, the plateau pumping speed should be very nearly equal to the
is
^"110
--J
-/
50
''
r"
.
._.. ._..
Hj /[
Average
=
1.J
^
<L>
30
20
10
10"'
10"^
10"'
8.0 cfm
pumping speed
by the staging ratio. 5-9. Computed Performance Curves for Mechanical Booster Pumps. By; choosing reasonable values for the parameters C^ and function 8^ and knowing the pumping speed of the backing pump as a
of the backing
pump
multiplied
"
I O
45
90
135
10
100
of the pressure, the performance curve of a positive displacement compressor used as a mechanical booster pump can be computed point by point from Eq. (5-20). The performance of the first experimental
Fig. 5-12. Curve for obtaining experimental value of conductance through pump clearances. [Taken with permission from C. M. Van Atta, in 19S6
Fig.
for
Compression -ratio curve mechanical booster pump. [Taken with permission from C. M. Van Atta, in 19S6 Vacuum, Sym,5-13.
mechanical booster pump was predicted in this manner and was later confirmed in its general features by pumping speed measurements. However, for a more accurate test of the theory and a better understanding of mechanical booster-pump performance subsequent calculations of pumping-speed curves have utilized the results of two preliminary tests designed to measure directly the parameters C,
Vacuum
Symposium
Transactions
1957).]
Press,
(Pergamon
The significant dimensions of the positive displacement type of mechanical booster pump, on which extensive calculations and tests were carried out, were as follows
Cylinder length: 16 Cylinder bore 9-M
:
in.
and K:
1.
in.
:
positions throughout a complete revolution, the conductance through the clearances of the pump was determined by admitting a measured flow of air into the inlet with the forepump in operation and measuring the pressure at the inlet P^ and that at the interstage P^. By appli-
Radial clearances, d 0.008 in. average Displacement speed at 1,740 rpm, Sj^: 1,230 cfm Displacement speed of backing pump: 130 cfm
Q =
C,{P,
P,)
to these observations for a small gas-flow rate and averaging these results, the low-pressure value Co of the conductance C, through the
clearances
2.
in operation but without any flow into the was varied from the limiting pressure of the backing pump up to several torr, and the pressures at the inlet and interstage were measured. The ratio of these pressure readings, K = P2IP1, is the zero-flow compression ratio, which combined with C^ from
inlet the forepressure
measurements on the conductance through the clearances as described in the first of the two preliminary tests outlined above are given in Fig. 5-12. From these measurements the average value of Co = 8 cfm. The dependence of the compression ratio on the Although these latter results interstage pressure is shown in Fig. 5-13. show an unanticipated droop in the compression ratio at the lower limit of the pressure range, an average value for the low-pressure range is taken to be ii' = 50. These values of Co and K substituted into Eq.
The
results of the
(5-14) yield S^
16.8 cfm.
results of the second test also yield a value for the critical pressure. Since begins to decrease sharply at an interstage pres-
The
0.008
in.
196
197
The remaining undetermined constant is C^ appearing in Eq. (5-30). The value of this constant has been arrived at by trial-and-error fitting of the high-pressure end of the experimental performance curve. The value chosen by this procedure is Ci = 2.8 cfm/torr.
experimental performance curve of the 130-cfm backing pump used. Dotted lines connect points on the booster-pump performance curve with those on the backing-pump curve from which they were computed.
The staging
repetition
combination
is
1,230 to 130, or
From
these results
we have
for
Eq. (5-30)
very nearly 10 to
C,
8.0
1.4(P2
The curve plotted in Fig. 5-15 as Case I is a of the computed performance curve shown in Fig. 5-14.
1.5)
cfm
1,200
1,000
\,
r
'i"
\
\ \
wn
\ \
\
\ \ \
;;'K
I
800
>
1 \
\
/ \'
600
>
\ \ \ \
\ \
I
Q.
/.
\
^
\
\
\
\ V \
1
f:
\
\
\
V
*
400
\
I
\\
\
>
\\
1 .
\
\ \
\ V
200
/
S
\
\
s.
10"
;^
^^
-^
^
10-
1 ^Backing pump
',J
1
_i..l
10-"
10"
10"
10
100
torr
Fig. 5-14. Pumping-speed curve for a mechanical booster pump. [Taken with permission from C. M. Van Atta, in 1956 Vacuum Symposium Transactions
Pressure
(McLeod
(Pergamon
Press,
London,
1957).]
With the numerical values of the above constants measured or assumed and the pumping speed 8^ of the backing pump as a function of the interstage pressure Pg known from previous measurements, the
Fig. 5-15. Pumping-speed curves showing dependence of mechanical boosterdisplacement of the backing pump. [Taken with Atta, in 1956 Vacuum Symposium Transactions (Pergamon Press, London, 1957).]
pumping speed S^
of the booster
is
illus-
Thus we have
1.5 torr
trating the expected effect on booster-pump performance of doubling the staging ratio, i.e., decreasing the displacement of the backing pump
-;;;
8^
S, TT^ 24.8
for
Pj
<
1,238
1.4(P2
1.5)
8^
24.8
1.4(P2
8,.
for
P,
>
1.5 torr
1.5)
is
by a factor of 2 from 130 to 65 cfm. For simplicity it is assumed that the pumping speed of the smaller backing pump would be just half that of the measured value for the standard backing pump at each value of the pressure.
Conversely, the performance curve
illustrates the
shown
^2
Pi
8,
The calculated performance curve for the standard combination of parameters given above is shown in Fig. 5-14 together with a typical
on the booster-pump performance of decreasing the staging ratio by a factor of 2, that is, by increasing the displacement of the backing pump from 130 to 260 cfm. Figure 5-16 illustrates the effect of radial clearances on the performance of the mechanical booster pump. The calculated performance
effect
expected
198
199
.,,
"*
..
1,000
/:^
800
1
y' "
,.
\^
s
^^
'
CoseE
/
/
Case
600
/
1
/
Cose
Rotor
(
\,
s
'''
mechanical booster pump of 1,234-cfm displacement speed backed by a roughing pump of 130-cfm displacement speed is shown in Fig. 5-17. The shaded area where the two curves join represents the changeover from the booster-pump operation to the booster bypass. 5-10. Measured Performance Curves for Mechanical Booster Pumps. The pumping speed at any point in a multistage system may be defined in accordance with Eq. (5-6) as
*s
\
Q_
//
400
//
1
i
IJ
[
0.008 0.004
0.01 6
Case!
\,
30cf I 230cfm
P.
200
'.
\
in
which Q
is
the throughput admitted as a steady flow at the inlet to P is the resulting pressure at the point of interest.
definition of pumping speed
is
n
10"^
10"'
10"-'
However,
10"^
10
if this
to bear
any
relation-
100
in the previous
Fig. 5-16. Pumping-speed curves showing the effect of radial clearances on the performance of the mechanical booster pump. [Taken with permission from C. M. Van Atta, in 195(i Vacuum Symposium Transactions (Pergamon Press,
London,
1957).]
curve for Case I with standard clearances of 0.008 in. is shown for comparison with similar curves calculated for radial clearances of 0.004 Note that for this latter case in. (Case IV) and of 0.016 in. (Case V). curve falls off pumping-speed the and the plateau has disappeared that at the peak. below and pressure above rapidly for values of the of a consisting combination typical a The throughput curve for
pumping speed is to be measured is that due to the gas admitted at the inlet. Under the usual conditions of test, the pressure of permanent gas in the system is due to air and the remaining pressure is due to condensable materials originating, for example, in the backing pump. Since a McLeod gauge measures the pressure due to the permanent gas and is very little affected by the vapor pressure present under these circumstances, the pressures used for pumping speed measurements are McLeod gauge readings. The role of condensable materials backstreaming from the backing pump is a separate matter and will be
the
discussed later.
10
,10'
^
10*
^^--'
-^ 'i^ T
,-<<^
10
jio^
10'
,^ ^ y^
0.010
s o
Backing
pump-KOH-130
10*
10^
0.001
0.001
0.100
1.0
10 15
100
1,000
Experimental results for the pumping speed of a 1 ,230-cfm mechanical pump backed by a 130-cfm forepump are shown in circles in Fig. 5-14. These results compare favorably with the calculated pumping-speed curve for which the basic parameters are in good agreement. Experimental pumping-speed results are also shown as circles in Fig. 5-15 for the 1,230-cfm booster pump backed by a 220-cfm forepump. These results should correspond fairly closely with the calculated curve designated as Case III of Fig. 5-15, although the backing speed is not quite as high as that assumed for the calculated curve. Comparison between the experimental and calculated pumping speeds shown in Fig. 5-15 indicates that the theory developed for the operation of a positive displacement rotary compressor as a vacuum booster pump is approximately correct. However, examination of the experimental results reveals minor deviations in behavior from that
booster
predicted.
Fig. 5-17. Throughput curve for a typical combination consisting of a mechanical booster pump of 1,234 cfm displacement speed backed by a roughing pump of 130 cfm displacement.
ance
The greatest and most fundamental deviation from expected performis exhibited by the zero-flow compression ratio. From Eq. (5-14)
'
200
it is
201
apparent that the compression ratio should be independent of the pressure from the lowest attainable pressure up to the point where the interstage pressure equals the critical value. Above this point the slippage conductance G^ through the rotor clearance is no longer constant, but increases linearly with the pressure. Measurements, as shown in Fig. 5-13, show that the compression ratio decreases as expected as the pressure is increased above the critical value. However, as the pressure is decreased from the critical value, the compression ratio, instead of remaining constant at its maximum value as
expected, drops off appreciably.
Tests carried out by Dobrowolski^ show a pronounced dependence of the zero-flow compression ratio of a mechanical booster pump on the mechanical clearances and surface finish of the rotors. Three booster pumps of 1,300-cfm displacement speed but with different rotor
known to be
for
tests
and to be much
Since
more pronounced
above
rough than
is
pump C
had
so, its compression ratio at low surfaces than either pump polished pressure is much better owing to the with smaller clearances pump B, A, with about equal clearances, or
poor.
Even
60
s ee Fig. 5
40
with a 220-cfm backing clearances and surface finish were tested, booster pumps was the of the in two pump. The rotor svirface finish in the case of the third surface whereas the standard machined surface
all
20
pump was
The
clearances
and rotor
10"
10'
^ =^ i ^ ^
>' ^A
10"
-13-
\\
s.
10"
F^ ^^ ^
10
\^
10'
Pump
Slippage
conductance
18.2
(Cg),
cfm
Rotor
finish
Fig. 5-18. Zero-flow compression ratio as a function of rotor clearances and surface finish.
A B
C
0.012
0.008 0.0115
8.35
16.8
Machined Machined
64-microinch polished
The pumping-speed curves also show a minor deviation from the expected shape. The predicted pumping-speed curve rises very
sharply from zero for values of the pressure slightly in excess of the ultimate value for zero flow. Although accurate measurements on the steep part of the performance curve are difficult, the results indicate that the rise in pumping speed with increasing pressure above the ultimate value is not as steep as predicted. Furthermore, the experimental results shown as circles in Fig. 5-14 are not quite as good a confirmation
In Fig. 5-18 the measured zero-flow compression ratio is shown as a function of the interstage pressure for the three pumps. The curves show the anticipated marked increase in compression ratio with decreasing clearances in the high-pressure region above the maximum compression point. However, for pressures less than that for the
maximum,
pressure for
falls
off rapidly
less
with decreasing
rapidly for
pump
C.
The
compression ratios is 0.27 for pump A, 0.21 for B, and 0.67 for C. The very marked improvement in low-' pressure compression ratio due to improved surface finish largely explains the discrepancy between the theoretically predicted performance curves and the measured curves. The zero-flow compression ratio at low pressure is theoretically expected to be independent of the pressure. When the rotor surfaces are rough they may be expected to carry gas back from the high-pressure outlet side of the pump to the lowpressure inlet side in the form of adsorbed gas. The dependence of this effect upon pressure cannot be predicted except that outgassing effects
ratio of low pressure to
maximum
drop in compression ratio in the low-pressure range on the pumping speed was partly compensated by the fact that the performance of the
of the predicted performance as first appears.
The
effect of the
130-cfm backing pump used in the tests was somewhat better at low pressures than had been assumed for the calculations. The minor differences between the observed and predicted performance noted above can best be explained by assuming that the reverse pumping parameter, by which some gas at outlet pressure is carried back to the inlet side of the pump, is not a constant but increases as the inlet pressure decreases. Experimental study of this effect has
particular
is partly due to an outgassing process, that of alternating absorption and reemission of gas
202
203
by the inner surfaces of the pump, which are alternately exposed to the interstage and to the inlet pressure. 5-11. Overheating of Mechanical Booster-pump Rotors. Compression of a gas in the process of pumping, as occurs in all mechanical vacuum pumps, involves doing work on the gas. Unless there is some process by which heat can flow easily from the gas to a heat
In the case of mechanical sink, the temperature of the gas increases. roughing pumps, oil is circulated with the gas stream and the exhaust gas bubbles through oil in the reservoir. The result is that heat is removed from the gas and rather efficiently distributed throughout the pump. Because of this process, cooling the pump as a whole dissipates
the heat satisfactorily. Thus small roughing pumps are cooled by air convection and large roughing pumps by water circulated through portions of the pump housing. However, in the case of mechanical booster pumps there is no oil present for distributing the heat generated throughout the pump structure. Furthermore, the rotors are not in immediate contact (except indirectly out through the shafts to the
pump, the test results of which are given in detail in the pumping power as computed from (5-42) with Sjj = 1,230 HO cfm at an inlet pressure P^ 10 torr 690 cfm, S2 cfm. Si = 5.47 hp, which is in good agreement watts 4,080 be IT out to turns this pump operating at an inlet For power input. measured the with A is therefore put into the gas fiow. pressure of 10 torr, about 4
Sec. 5-9,
kW
portion of this power is dissipated in the pump rotors which heat up and expand. If the pump is operated at inlet pressure above 10 torr for an extended period, the expansion of the rotors exceeds the available
pump
will
pump housing and therefore tend to overheat when the pressure difference across the pump is too great. As is pointed out by Noller^ the power required for pumping the gas is
bearings) with the
which the rotors are cooled by oil circulated through hollow shafts. He also describes the use of an
interstage cooler at the outlet of
Fig.
5-19.
Exhaust
gas
cooler
in-
W = SniP, in
Pi)
(5-40)
which
Sjy is the
From
the
the mechanical booster pump to cool the exhaust gas and therefore Figure 5-19 indirectly the rotors.
illustrates the
Vacuum
V, 25
arrangement of an
PA = P,S = PA
where
S-^
installations in
and S2
its
are the
pumping speeds
e
pump and
S^ISj^
vacuum melting of steel and similar which the operating pressure is high. Because of the limit on pressure differential imposed by the expansion of the rotors due to overheating, mechanical booster-pump
installations frequently include a pressure-sensitive switch set to operate at 10 to 50 torr, depending upon the characteristics of the particular
of the booster
pump.
W = S '^'ik~^)
(5-42)
From this equation there are obviously two ways of limiting the power requirements of the booster pump: (1) The upper value of the inlet pressure P^ during operation can be limited to some maximum value riot to be exceeded in operation or (2) the pumping speed S^ of the backing pump may be increased as needed to tolerate a larger value of the pressure. In practice either remedy is applied, depending upon the requirements of the system. In the pressure range below 1 torr the power requirements of mechanical booster pumps is determined almost entirely by the frictional
;
booster and backing-pump combination. At pressures greater than the set value the pressure switch holds a valve in a bypass connection open so that the gas flows directly into the backing pump from the system and the booster-pump power is turned off. When the pressure becomes
less
significant.
bypass valve and system so equipped, when started from atmospheric pressure, will pump down initially through the bypass with only the backing pump in operation and will then switch over to pump as a two-stage system as soon as the pressure passes through the Figure 5-20 illustrates such a two-stage selected switch-over value. around the mechanical booster connection bypass with the system valve is an obvious alternative the bypass of Manual operation pump.
than the
pump.
which
is
some
installations.
204
205
made on
pump
lonization-gauge readings are consistent with this expectation. at the inlet of the two-stage system consisting of a 1,230-cfm booster pump backed by a 130-cfm single-stage backing pump show limiting plate currents corresponding to a pressure of 5 x 10"* torr on an air
A very small two-stage mechanical booster pump of only 30-cfm displacement backed by a conventional compound forepump produced McLeod gauge readings of 1 X IQ-" torr or lower and untrapped ionization-gauge readings corresponding to 8 x 10~^ torr. Since the true calibration of the ionization gauge for the vapor concerned is not known, it can only be concluded that the true total pressure is considerably below the latter value. Another observation of interest is that the limiting pressure indicated by an ionization gauge in a system evacuated by a mechanical booster pump is not changed by putting dry ice in a trap situated between the
calibration.
the gauge tube. However, substituting liquid nitrogen for dry ice in the trap results in the ionization-gauge reading dropping to the pressure indicated by a McLeod gauge. This observation demonstrates that the vapor compression action of the mechanical booster pump is such as to decrease the vapor pressure in the system
booster
Fig. 5-20. Mechanical booster
pump and
due to backstreaming from the forepump below that corresponding to the equilibrium at dry ice temperature. Noller' confirms that the vapor pressure "is 1 or 2 powers often lower behind the blower than it is
in front of it."
5-12.
of a
Mechanical Booster
pumping-speed
Pump.
and referred to throughout in this connection is the partial pressure of the permanent gas which is admitted in controlled and measured amounts at the inlet of the pump. This procedure is justified on the grounds that only in terms of the pressures at various points in the system of the gas admitted as a measured flow at the inlet can one hope to understand the performance of a multistage system. The condensable vapors present are generally progressing backward through the system and therefore have nothing directly to do with the processes by which the permanent gases are pumped. The behavior of condensable vapors originating from the sealing oil used in the backing pump, however, is determined directly by the zerocurves, the pressure measured
In 1912 Gaede^^ introduced a 5-13. Molecular-drag Pumps. type of mechanical pump which does not operate on the positive displacement principle but upon the principle of imparting momentum to gas molecules preferentially In the molecin the direction of the desired flow. ular-drag pump there is an open passage from the inlet to the outlet, between which a pressure difl'eris maintained by the high-velocity motion of one side of the passage relative to the housing of the pump in which the inlet and outlet are located. In Fig. 5-21 the principle of the molecular-drag
ential
Fig.
5-21. Prin-
pump
is
illustrated.
cylindrical
member
rotates
flow compression ratio. Since the mechanical booster pump itself can be clean and free of sources of volatile materials, the condensable vapors of concern are those incident on the interstage side of the booster
pump.
Therefore the vapor pressure on the high-vacuum side of the will be that in the interstage region due to backstreaming from the forepump divided by the zero-flow compression ratio of the booster pump.
booster
pump
ular-drag pump. within a casing with a radial clearance h between them. At the top of the cylinder the clearance space is blocked by a projection of the cylinder wall which reduces the clearance At either side of the projection the clearance locally to essentially zero. passage opens into a closed volume. If there are no leaks in the system the total amount of gas in the system remains constant, but some gas is shifted by the motion of the rotor with a reduction of the pressure Pj
206
207
and increase in the pressure P^. The equiUbrium relationship between P^ and Pa depends upon the rotational velocity of the rotor and upon
the pressure regime in which the
pump
is
operating,
large
i.e.,
whether
(du\
TjW
0X1-I
\dy/y+dy
is
(Pi
P-Sl"^ is
enough so that
is
the
mean
free
<
A,
rjW
bx-\u
du
dy ^y)
(5-45)
Assuming that the layer of gas at each surface is at rest with or moving with the surface, then the gas in contact with the outer cylinder is at rest and that in contact with the rotor is moving with the peripheral speed rw, in which r is the radius and w is the rotational speed in
dominated by viscous behavior.
r, the radians per second. If A curvature of the annular space can be ignored and the problem reduced to that of two plane surfaces
Equilibrium will occur when the force due to the pressure difference given in (5-43) balances the sum of the viscous forces in (5-44) and (5-45), so that
IV
bP
by^F, +
riw bx
F,
<
by
(5-46)
y//////////////A^7>my//.
dy^
which yields
d^v __
1
7}
with the upper surface moving with respect to the lower surface with
separated by a distance
A,
bP
(5-47)
dy^
dx
velocity v
is
of the form
rio,
as illustrated in
The peripheral disFig. 5-22. tance between the inlet and outlet in Fig. 5-21 is L, which is the length ofthelower or stationary plate in Fig. 5-22 in the plane case. Following a procedure analogous to that in Sec. 2-3, the gas contained in a thin layer of thickness by at height y above the stationary plate, and of length bx in the direction of the motion, experiences a force opposite to the direction of motion of the upper plate given by the crosssectional area w by, where w is the width of the plates perpendicular
to the plane of the figure, multiplied by the pressure difference which occurs in the distance bx, so that
Ay^
+ By + C
(5-48)
By
bx
is
at rest,
at y == 0,
bP
C =
its
0.
velocity
is
v at y ==
Ji.
Putting these
F
At equilibrium
=wbP by
(5-43)
1
2,7]
bP
h^
Bh
(5-50)
bx
1
this force is balanced by the difference between the viscous forces from the gas above and below the thin layer under The component of viscous force from the gas below consideration. the layer can be written by reference to the definition of viscosity given
so that
^-P,
(5-51)
2rj
bx
in (1-54) as
and C
I
F,
= -.,^s'^=-yjwbx[^^^
dy
(5-44)
ir^y bx
2r]
^P
/v
bP
bx
2r]
h y
(5-52)
in
which u
is
the velocity of the gas in the sample layer, since the area
which
is
a parabolic form.
=w
bx.
layer
is
The negative sign arises since the gas below the sample moving more slowly and therefore retards its motion. The
The net volume flow of gas from the region at the pressure Pi to that at Pj is given by integrating the flow from y = Since to y = h.
208
209
level y the volume flow in the layer of thickness by at the total flow is
dV
'dt
wu
v=o
by
1
=W
Jy=i Jy=0
bP
bx
dy
\h
2ri
.2rj
bx
wvh ~^
w bP
h^
The compression ratio P2IP1 for the pump described above large, therefore, unless the outlet pressure were only become not would The simple molecular-drag pump here 1.71 torr. than greater slightly as the first stage of a two-stage only effective therefore is described is fairly low. By decreasing pressure interstage the which in system difference pressure maintained the channel h, pumping of the depth the also the therefore backing and regime, viscous-flow the in pump by the substantially can be ratio, compression large for a required pressure
cylinder.
(5-53)
Ur] bx
increased since the pressure diff'erence depends inversely on h^. If the molecular-drag pump is backed by a pump which maintains
Under equilibrium conditions the pressure difference (Pj such a value that the net flow is zero. Thus from (5-53)
wvh ~2~
so that If the length of the channel
-Pi)
has
w bP
(5-54)
12r]
bx
bx
(5-55)
QrjV
bP
h^
is
the interstage pressure P^ at such a low value that the mean free path of the gas molecules is long compared with the dimension h of the pumping channel, viscosity no longer plays a role and the relationship In this (5-56) between the inlet and outlet pressures no longer holds. and stationary the with alternately collide molecules gas the regime moving surface. Consider the flow across an element of length bx of the channel. 12 Each molecule which strikes the moving surface of
area
L, then
^^
by
v,
bx
Each molecule which strikes a stationary surface, either opposite to the moving surface or at the two sides of the channel, receives a The resulting average drift velocity is the velocity zero-drift velocity. times the ratio of the area of this surface to the surface moving of the
total surface of the channel element of length bx,
QrjvL
h^
(5-56)
wv
(5-57)
"
independent of the pressure. The molecular-drag pump operating in the regime of viscous flow is thus expected to maintain a pressure difference between inlet and outlet under conditions
since the viscosity
is
2(w
is
h)
to this drift
motion
directly proportional to the peripheral velocity and proporthe length of channel between inlet and outlet and inversely pressure this for order In depth. channel tional to the square of the necessary difference to give rise to a large compression ratio P2IP1, it is
is
^^-^^^^
= 5X
IV%V
2{w
+
h
torr cm^/s^c
h)
10-Pis
w%v
torr liters/sec
(5-58)
that P2 not be
Consider
inner cylinders
rotational speed
0.2
in.
the pressure difference (P^ - Pi)clearance between the outer and 4 in. = 10 cm; the 0.5 cm, r
If a pressure difference
bP
and
x 10^ radians/sec 10,000 rpm, so that w 1.83 x 10"* rj 20C air at 10^ for Since cm/sec. 1.04 x
= = 1.04
a flow will occur in the opposite direction because of this pressure The counterflow difference through the conductance of the channel.
is
given by
is
^1
2.28
10^ /xbar
34.4/ T\'^
w%2
2{w
^p
h)
bx
torr liters/ sec
(5-o9)
103
X 750 X
10-
1.71 torr
inlet
assuming a distance of 50
cm between
and
outlet ports
9 71
the
\M/ w
h bx
210
211
by
= +
Q^, so that
9-^1 hr7
\MJ w
dP
~P
h dx
10*
w%v w
(5-60)
or
_5x
~~
lO-^i^V
9.71
\t} h
dx
P=
P^ to
P=
P^ and x
to
InP,
\nP^=k
h
vL
From the foregoing calculation the zero-flow compression ratio for a simple molecular-drag pump is predicted to be so very large when operating in the molecular-flow regime that the limitation in a real pump is due to factors not specifically considered. In the simple pump described, leakage from the outlet region back into the inlet through the clearances at the ends of the rotor and the imperfect sealing between the rotor and cylinder, where the radial clearance is assumed to be zero, would prevent the attainment of the theoretically predicted compression ratio. Even
Gaede reports and Dushman^^ confirms compression ratios of the order of 10^ attained by a multistage molecular pump based upon the prinso,
or
P P
5
(5-61)
where
10-
T~\t)
(5-62)
above.
number
of alternative designs
low that the flow is molecular, the compression ratio maintained by the simple molecular pump described above is independent of the pressure and depends exponentially on the quantity vLjh, which is made up of the parameters of the
if
Thus
the pressure
P^, is sufficiently
for molecular-drag
pump.
For
air at
3.181
k
9.71
(3.181)
1,62
X 10-5
(5-63)
conductance leakage path from outthrough the running clearances of the pump. The second is to vary the depth of the pumping channel to provide a decreasing channel depth as the gas is comlet to inlet
becomes
pressed so that the cross section of the channel at the inlet of the pump
will
Fig. 5-23. Cross section of molecular-drag pump design due to S. Siegbahn with pumping channels in the form of Archimedes' spirals cut
Y =exp(l.62 X 10-s^j
Taking as an example the same values for parameters of a pump,
v,
(5-64)
L,
and h
vL
T^
P2/P1
1.04
10* (15
X 50
1.04
10"
[Taken with in the two flat sides. be as large as possible to ensure permission from S. Von Friesen, good pumping speed, but still to Rev. Sci. Instr. 11, 362 (1940).] ensure that this depth will be small relative to the molecular mean free path over as much of the comIn Fig. 5-23 is shown a cross pression range of the pump as possible. section of a design due to S. Siegbahn^* in which pumping channels in the form of Archimedes' spirals are cut in the two flat sides of the housing, within which a disk rotates at high rotational velocity. The clearance between the disk surface and the flat section of the end
plate between the adjacent spirals
is
made
exp exp
(1.62
(16.8)
X 10-5 X
1.04
lO")
free rotation.
=
which
is
lO'-^"
surprisingly large.
periphery of the disk and the discharge at the hub. In the unit shown, three spiral grooves are cut in parallel, starting 120 apart, providing three times the pumping speed of a single channel.
The
inlet is at the
212
213
Pi
cP,
c is
a constant of the
The performance of a pump somewhat analogous to that of Siegbahn The rotor in this case is (see Fig. 5-24) is described by Beams.^^ induction through the vacuum driven magnetically and by suspended material bearings. seal and of shaft the problems wall, eliminating 10* which is about cm/sec, of 1.4 of the order x Peripheral speeds typical for the are room temperature, air molecules at for one-third Vav sealed and is is completely the unit Since model tested." preliminary trap, liquid-nitrogen-cooled with a diffusion pump an oil Ijacked by
the forepressure can be very low. A composite curve of the observed compression ratio P2/P1 for various values of the forepressure Pg and the rotational speed is shown in Fig. 5-25.1^ A theoretical curve for the
ically
J.
Fig. 5-24. Molecular pump of Williams and Beams. Rotor is suspended magnetand driven by induction. [Taken with permission from C. E. Williams and
in
1961
Vacuum Symposium
compression ratio as a function of rotational speed is shown for comparison. It is evident that P2/P1 departs further from the theoretically predicted value as the forepressure Pg is decreased, indicating the influence of outgassing from surfaces at the lower values of Pj attained during the test. As an example, with a forepressure of 4 x 10"' torr the untrapped ionization-gauge reading at the inlet was 2 X 10"* torr, yielding a compression ratio of only 200 compared with the predicted
value of nearly 3,000 for
01/277
Because the computed compression ratio for a pump of this description is tremendous when the internal leakage is ignored, the pump acts as though it has a forward pumping speed which is independent of the pressure shunted at intervals of pressure between Pj and Pj t>y leakage conductances. Examination of this model leads to the conclusion that the zero-flow compression ratio should have the form
10-^
olxlO'^torr
A
10
-
6x10"^
/
,-'-'
,--'
1.2x10
''
s 4x10-^ X6xl0"^
- = exp ^'iri
in
(5-65)
P310Theoretical
y: ,-'-^9'xio>^
mass 30
y^
'^^'\,''^''
^^^^
410-'"
which w
is
A;
is
y^<''5<'^^
X^?-:
^^^^
pump
design, and a is the clearance between the rotating disk and the end plates through which the internal leakage flows. In comparing this expression with (5-64) it must be realized that a in this expression represents the internal leakage clearance, whereas the h in (5-64) corresponds to the depth of the pumping channel. The pumping speed of a pump of this description with disk diameter
/^ '^y^
1
jk:>
100
Theoretical moss 8
200
27r
300
400
of 54 cm is reported by Eklund^^ to be as high as 80 liters/sec at 8,300 rpm and to increase proportionally with the rotational speed. The inlet
Fig. 5-25. Composite curve of observed compression ratio for various values of the forepressure and the rotational speed. [Taken with permission from C. E.
Williams and
Press,
J.
W. Beams,
1962).]
in ld6l
Vacuum Symposium,
Transactions (Pergamon
pressure P^
to
London,
214
215
including variation of the rotational speed and of the forepressure, indicate that the zero-flow compression ratio is only in very rough
when Pj
is
when Pj
is
equal to
torr.
agreement with
outgassing.
(5-65),
may
be due entirely to
The compression
outlet pressure.
ratio for
hydrogen
is
same
pump
are freedom
from contamination by condensable vapors, high compression ratios, and short startup time. One additional feature which is useful in some
a higher pumping speed for gases of high molecular weight, which is the inverse of the performance of diffusion pumps. One obvious disadvantage is the low pumping speed, considering the
applications
is
io''i
:
1
= nil :
nil
III
nil
10-2:=:
'-'-'-
.1
J 3
==^ ?=::::
-
hazard associated with rotational speeds of 10^ rpm and more, which are required to ensure good performance. Although molecular-drag pumps have proved to be effective in certain special applications, particularly in Europe, they have not been widely used in the United States, either
size
device.
There
is
10"'
-^H ;=:|-:
10-^::=
-J --^^-:
=4
-.
=:
'---izl:
Kf'-^'
10-' =
=
== :==
-1
-: :--: -. = :: = = :=:;:= =
::
5-14. Axial-flow
10-'::;;:=:
.!_.
l-\zz-.
t-.:
--
'--
10-8=-:: = -=
'
i
-=-.
A. Pfeiffer, GmbH, of Wetzlar, German Federal Republic, which may be described as an axial-flow molecular turbine pump.^' Figure 5-26 shows the general arrangement of the pump, and Fig. 5-27 illustrates the details of the design. Rotating disks all mounted on the central shaft are disposed alternately with stationary plates mounted in the
housing.
hx 1 eoi-2] io-'::;i:Z:;;|=;;:=:
^,o~~V" r' A
10-'
10'^
I0-'
::::::
.1
1
10
P2,torr
Fig.
tails
5-27.
De-
The
disks
that gas molecules caught in the slots of the moving disk are proFig. 5-26. General arrangement of the axial-flow molecular turbine pump. [Taken with permission from Willi Becker, Vakuum-Technik 7, 149
(1958).]
of the rotor and stator plates of the molecular turbine pump. [Taken with permission from Becker, Willi
and
Vakuum-Technik
7, 149 (1958).]
the refrigerant Freon-12 for the molecular turpump. [Taken bine with permission from Willi Becker, VakuumTechnik 7, 149 (1958).]
be-
mm, which
The pumping speed of the molecular turbine pump as a function of is shown in Fig. 5-29 for the same three gases. The pumping speed characteristic of the forepump used in these tests is
pump.
The
rotational
having a rotor diameter of about 17 cm is 16,000 rpm, giving a peripheral speed of 1.56 x 10* cm/sec, about one-third
speed for a
pump
room temperature. The observed dependence of the inlet pressure P^ on the outlet pressure P^ is shown graphically in Fig. 5-28 for hydrogen, air, and the
Wav for air molecules at
refrigerant Freon-12.
not given in the paper so that the relationship between the curves in these two graphs is not clear. The pumping characteristics of an axial-flow, bladed turbine pump have been investigated theoretically and experimentally by Kruger^" and by Kruger and Shapiro. ^^ A portion of the vane structure of a rotor of the type used in the high-vacuum turbine pump is shown in Fig. 5-30, illustrating the probability of transmission of molecules
216
217
Ha
Au.
molecular flux v^ incident on the area A of the rotor blade from upstream, then the net flow in molecules per second is
Q = HAvy =
Freon-12
Avj:^^
J-VaSai
(5-66)
I
10"^
10"'
where v^ is the molecular flux incident from downstream, the temperature is the same on both sides of the rotor,
10"^
1010-
However,
if
IQ-"
10"^
P,, torr
(5-67)
"1
the FiG. 5-29. Pumping speed of the molecular turbine pump as a function of from permission with [Taken Freon-12. and inlet pressure for hydrogen, air, Willi Becker, Vakuum-Technik 7, 149 (1958).]
^1
from Eqs. (1-31) and (1-16), where n^ and n^ are the molecular densities and Pi and Pj the pressures respectively upstream and downstream. Then from (5-66) and (5-67), the compression ratio is given by
(5-68)
through the rotor blade for the case in which the blade speed is large as compared with molecular speed (a) when the molecules are incident from upstream (side 1), and (6) when they are incident from downstream The difference between these probabilities, S12 and S21. (side 2).
(the Ho determines the net pumping speed. If we designate by coefficient) the ratio of the net molecular flow through the rotor to the
^1
^21
is
When
pump
and the
zero-
J) vPi/e=o
It is also evident that the
^
2.21
(5-69)
compression ratio across the rotor blade should decrease linearly as the flow Hv^ is increased. This prediction has been confirmed experimentally by Kruger and Shapiro. ^^ Also from (5-68) it appears that for no pressure rise across the blade, Pj = Pi, the net flow through the rotor blade is
Qi
(S12
^2\)A
(5-70)
In order for the compression ratio to be high it is important that Sia be large compared with Hji. However, for the pumping speed [8 = QjP) to be large it is necessary for S12 to be large in an absolute
sense.
pump
^Molecules incident
from side (T)
Fig. 5-30. Probability of transmission of molecules through the rotor blade of an axial-flow turbine pump for the case in which the blade speed is large as compared with molecular speed, {a) Molecules incident from upstream; (b) molecules incident from downstream. [Taken with permission from C. H. Krugor and A. H. Shapiro, in 1960 Vacuum Symposium Transactions (Pergamon Press, London,
1961).]
According to Kruger and Shapiro, ^^ in the design of a multistage of the axial-flow turbine type it is possible to design the blades of the first few rotors for large pumping speed and low compression ratio and later stages increasingly for high compression ratio and low pumping speed. The increase in pressure toward later stages permits a lower pumping speed to accommodate the flow. By carrying out a series of Monte Carlo calculations of the motion of individual molecules through rotor blades, Kruger 20 has determined the compression ratio
and pumping speed for rotors with various values of the pitch angle, spacing, and length of the blades. To be effective the velocity of the rotor blades must be two or three times the quantity (27^T)'^, but increasing the blade velocity beyond this value does not result in a
218
significant
gain in performance. Although variation of the pitch angle varies the zero-flow compression ratio and pumping speed, a pitch angle of 20 appears to be a good compromise for many applications. Since a compression ratio per stage of about 5 can be achieved,
a pump having 9 stages should maintain a zero-flow compression ratio of the order of 5" <= 2 x 10''. The pumping speed of a well-designed
axial-flow turbine pump is comparable with that of a diffusion pump of the same diameter but has the advantage of being free of hydro-
CHAPTER
carbon vapors. The disadvantages of the axial-flow turbine pump are (1) the hazard of very high rotational speeds, (2) the comparatively great weight for the pumping speed, and (3) the very high cost in terms of dollars per liter per second.
6-1.
The Steam
Ejector.
REFERENCES
1.
2.
3.
4.
Gaede, Z. Naturforsch. 2A, 233 (1947). B. D. Power and R. A. Kenna, Vacuum V, 35 (1955). F. A. Knox, U.S. Patent No. 2,551,541, May 1, 1951. NSA 5, No. 3578. C. M. Van Atta and R. L. Sylvester, in Proceedings of the Vacuum Metallurgy Symposium of the Electrochemical Society, Boston, Mass. (Electrochemical
W.
5.
6. 7.
8.
9.
10.
11. 12.
13.
14. 15.
16.
17. 18.
Symposiwm Transactions (Pergamon Press, London, 1957), pp. 62-70. K. Ziock, Vakuum-Technick (Rudolph A. Long Verlag, Berlin, 1957). E. A. Winzenburger, in 1957 Vacuum Symposium Transactions (Pergamon Press, London, 1958), pp. 1-5. Z. C. Dobrowolski, Lab. Rep. No. 2290, Kinney Vacuum Division, The New York Air Brake Company, Dec. 20, 1961. H. G. Noller, in 1956 Vacuum Symposium Transactions (Pergamon Press, London, 1957), pp. 57-61. R. Thees, Vacuum V, 25 (1955). W. Gaede, Ann. Physik 41, 337 (1913). Robert B. Jacobs, J. Appl. Phys. 22, 217 (1951). Saul Dushman, Phys. Rev. 5, 224 (1915). S. Von Friesen, Rev. Sci. Instr. 11, 362 (1940). S. Eklund, Arch. Math. Astron. Phys. (Roy. Swed. Acad.) 27A, No. 21 (1940), and 29 A, No. 4 (1942). J. W. Beams, Science 130, 1406 (1959). C. E. Williams and J. W. Beams, Bull. Am. Phys. Soc. Ser. II 5, 286 (1960)'. C. E. Williams and J. W. Beams, in 1961 Vacuum Symposium Transactions (Pergamon Press, London, 1962), pp. 295-299.
C.
volumes down to pressures of the order of about 1 torr was first demonstrated by M. Leblanc,i* steam ejectors have been used successfully in a wide variety of rough vacuum applications. A typical steam ejector is illustrated in Fig. 6-1 and consists of (a) a steam chest in which the pressure and temperature are maintained at the proper values, (6) the nozzle through which the steam flows to form a jet, (c) the mixing chamber through which the steam jet passes and entrains gas admitted to the chamber through {d) the inlet port, and (e) the diffuser through which the jet carries the entrained gas to (/) the discharge. Under normal operating conditions the pressure in the mixing chamber is very low as compared with that in the steam chest and at the discharge port so that the steam is expanded in passing through the nozzle by a large factor and then compressed in passing through the diffuser. Since the cross-sectional area of the steam chest is large as compared with that of the nozzle, the directed or drift velocity in the steam chest is small as compared with that through the nozzle. The random energy of thermal motion of the steam is therefore converted in passing through the nozzle into directed kinetic energy with the formation of a supersonic jet, i.e., one in which the directed velocity of flow is large as compared with the average random velocity of the molecules determined by the temperature.
The flow of steam from the steam mixing chamber, and out through the
chest,
Becker, Vakuum-Technik 7, 149 (1958). H. Kruger, "The Axial-flow Compressor in the Free -molecular Range," Ph.D. Thesis, Department of Mechanical Engineering, Massachusetts Institute of Technology, Cambridge, Mass., 1960. 21. Charles H. Kruger and Asoher H. Shapiro, in 1960 Vacuum Symposium Transactions (Pergamon Press, London, 1961), p. 6.
19. Willi
20. Charles
flow of a compressible fluid through a tube of varying cross section. Thermodynamic analysis of the flow through a converging-diverging nozzle, assuming isentropic behavior (no heat exchange with the walls,
that
is,
PV^^
Pjpi
const), leads to
chapter.
219
'^
220
221
minimum
cross section),
nozzle,
According to (1-28) the velocity of sound in the gas at the nozzle throat under critical flow conditions is given by
y
1. The flow through the nozzle increases with decreasing pressure P3 beyond the nozzle until a critical value P^ is reached beyond which the flow is independent of P3 and P^ = P^2. The flow velocity at the throat of the nozzle is Mach 1, that is, is equal to the local sound velocity, when P3 < Pj.
Pc
2y
r
Pi
(6-4)
1
Pi
Thus
v^lVgj
at the throat
is
Mach
1.
isentropic,
'pU/y /PX''
/ I
9 2
\l/(y-l)
V''
'
(6-5)
The
Steam
inlet
calculation
shows that
v/(,-l)
\r \y
1^1/ + 1/
If
^1
again by use of (6-1). The mass flow rate through the nozzle
(^-^)
is
dW = cp.v^A^
dt
-cp,v,D^^
g/sec
(6-6)
as
P3 <
Pj.
the
gas
flowing
Mixing
chamber
through the nozzle is a perfect diatomic gas for which y = 1.40 (see Table 1-4), the result is P, = O.535P1. However, for steam y = 1.32 and thus for steam sufficiently superheated so that condensation does not occur in the throat, P^. = O.545P1. For steam not sufficiently superheated, however, condensation does occur and the effective value of y is such that P^ == 0.575Pi, approximately. Associated with the
pressure P^
is
where D^
c is
is
coefficient, which is a number generally in the range of depending upon the geometry of the nozzle and the flow conditions (particularly the Reynolds number). By substituting (6-3)
the nozzle
1.0,
0.95 to
and
is
~dt
=
is
1.30, the
mass flow
for
steam
d.W,
critical
flow
v^
"df
when Pi
Pi
is
0.524c
(9P'
g/sec
(6-8)
is
measured
V^ in
This
If the pressure
velocity
is
given by
measured
becomes
g/sec
(6-8a)
Fig. 6-1. Cross section of typical steam ejector. [Taken with permission from V. V. Fondrk, in 1957 Vacuum Symposium Transactions
2y
y
P,
Pi
-'
(i-i)/y-l
dW^
19.14c
dt
Elf D,
vj
(6-2),
1958).]
where p^ is the density of the gas upstream from the nozzle (e.g., of the steam in the steam chest). By reference to (6-1) the stream velocity at the critical pressure is given by
2y
most engineering applications Pi is in pounds per square inch, V^ in cubic feet per pound, and D^ in inches, the expression
Finally, if as in
becomes
dW,
892.6c
dt
Pj
I
2y
y
/PiV^
Ib/hr
(6-86)
Pi
PiFi
I
(6-3)
From
rate
may be
c
where V^
l/p^
is
that
the steam conditions upstream from the nozzle the mass-flow calculated approximately from (6-8a or 6-86) by assuming 1. Engineering steam tables give values for the specific
223
1-sq-in.
nozzle
throat
11
Saturation
w^
10 9
s ~
8 7
~4 00
T 3tal remperatureh
1
50f rv
600 F-
6
5
s s
3 2
1
/
fr
^ A ^
^ Am
^ ^4 f
ho 0F
lUU U''F
V' ^ ^ '^
/
"~i
f^
Additional curves of this type as well as the procedure for determining the value of the nozzle coefficient are given in the reference. As an example, for the steam
conditions given as
^700
800 F
Pj
Ti
= =
Fi
random molecular energy of the steam in the steam chest is converted by the converging-diverging nozzle into very low random energy plus a high directed energy. Since the drift velocity is then large as compared with the random molecular velocity, the jet stream in the mixing chamber has a high Mach number, typically in the range Because the temperature of the steam jet is low in the mixing 2 to 4. chamber, the vapor pressure is also very low as compared with that at room temperature, so that the water vapor pressure seen at the gas inlet port is correspondingly low. When the steam jet enters the difii'user, a process of isentropic compression occurs. The steam together with the entrained gas being pumped through the inlet port is compressed from the low pressure in the mixing chamber to the exhaust pressure. The maximum exhaust pressure attainable is that for which the flow velocity of the steam The design of the jet is completely converted to random velocity. must be such that this conditions steam diffuser and the imposed jet will be unable to the Otherwise minimum requirement is met. allowing exhaust down, break differential and will sustain the pressure
gas to flow back through the system. The pumping speed of a steam ejector
is
n
20 40 60 80 100 120 140 160 180
Pressure, psig
9.06 ft3/lb
mass-flow rate for steam through a convefgingdiverging nozzle of 1 sq in. throat area assuming a nozzle coefficient of to 620 1.00 for the pressure range
Fig. 6-2.
critical
The
sq
in.
design of the nozzle and diffuser. The design of the nozzle is based upon sound thermodynamic principles and can be specified with considerable precision for a given set of requirements. The design of the diffuser is much more empirical, for which experience in a great
=^.
2,900c
in.
psi
psia) for
gauge pressure (14.7 to 634.7 steam temperatures up to 1000F. [Reprinted from the Standards for Steam Jet Ejectors, 3rd ed. Copyright 1956 by the Heat Exchange Institute, 122 East 42nd Street, New York, X.Y. 10017.]
IS
dW, ~df
c is
vacuum applications provides a sound basis for predicting performance. The pumping capacity of a steam ejector is generally given in terms of the number of pounds of dry air removed
variety of industrial
Ib/hr
in one hour (lb air/hr).
Pair
Since
which than 1.
in
number
slightly less
1
lb
dry
1
(68F
20C)
and
lb
dry
air (68C)/hr
pressure P^
allow for the expansion of the steam in its free expansion from the = P at the throat to the design value for the pressure in the mixing chamber where gas at pressure Pg is to be mixed with the steam jet and pumped out through the diffuser. The external pressure
= = = = =
1.205
7.52
168.5 torr
The capacity of a
the inlet pressure speed in Fig. 6-4.
is
shown
in Fig. 6-3,
on the steam jet as it traverses the mixing chamber is P^, the inlet gas pressure, which at the design point for the system is just balanced by the transverse pressure in the steam jet because of the random thermal
molecular motion. Since in practice the design inlet pressure is quite low, the intrinsic pressure in the steam jet is also low, corresponding to a low temperature frequently well below the freezing point. The
For high-vacuum pumping, steam ejectors are used in multiple combinations, such as the three-stage system illustrated in Fig. 6-5. A water-cooled condenser is placed between the second and third In such a system the stages to decrease vapor load on the final stage. first two stages are referred to as booster stages, in which the steam conditions are such that very low temperatures are reached by the
224
225
steam-water or steam-ice
120 100
vapor
80
60
/ f
)
1 1
/ /
/^
s
*>,
correspondingly very low. The ultimate vacuum attainable with steam ejector systems is consequently surprisingly low as compared with the vapor pressure of water at room temperature (17.5 torr at 20C). Multistage steam ejector systems may consist of as many as seven stages with interstage condensers
40
20
/
two or three stages. With such systems base pressures as low as 10^^ torr are attainable with capacity of the order of 6 Ib/hr of dry air or pumping speed of 100,000 cfm at 10~^ torr.
after the first
200 500
1,000
50
100
200
500
1,000
20
50
100
Fig. 6-3. Capacity of a typical singlestage steam ejector as a function of inlet pressure. [Taken with permission from V. V. Fondrk, in 1957 Vacuum Symposium Transactions
Fig. 6-4.
Pumping speed
of a typical
1958).]
steam ejector as a function of inlet pressure. [Taken with permission from V. V. Fondrk, in i957 Vacuum Sym,posium, Transactions (Pergamon Press, London,
single-stage
1958).]
2-stage steam
ejector evocuatinq
common surface
condenser (C)
at
torr
4-tons/min
Suction chomber
Third stage
Common
surtace
units (8, ,82)
4 2-stage
Steom
inlet
torr and
handling goses
of pouring rote 4
tons/min
of
torr
Discharge
v4
Intercondenser
hogging steam
ejectors (A)
62N62
Water dischorqe
range
30
in
to V2 torr
2 min
Fig. 6-5. Layout of three-stage steam ejector system. [Taken with permission from V. V. Fondrk, in 1957 Vacuum Symposium Transactions (Pergamon Press,
London,
6-ia diophragm
volve (Vj'
1958).]
Fig.
6-6.
system.
226
227
steam ejector systems may consist of several booster ejectors in parallel, backed up by one or two In the system illustrated in Fig. 6-6 there are four twofinal stages. stage boosters in parallel, backed by two final stages with interstage
beyond
Table
6-1.
in
Inlet
Capacity,
torr
lb air/hr
Throughput
torr
Pumping speed
cfm
liters/sec
and steam flow back through the system into the vacuum vessel. The steam conditions and load must be regulated in such a manner as to avoid this type of blowback since the consequences in most cases would be serious. The problem of backstreaming of water vapor into the vacuum vessel does not present a serious problem unless the flow of gas from the vacuum vessel to the mixing chamber falls off to a value which is considerably less than the design point for the system. However, at
result that discharge gas
cfm
torr Hters/soc
1.0
2.5
3.0
4.0 5.0
674 843
zero throughput serious backstreaming will occur. It is sometimes necessary to provide means for introducing air, steam, or other gas at a controlled rate into the system between the steam ejector inlet and the vacuum vessel through a needle valve to prevent the inlet pressure from dropping to the point where backstreaming would
condensers.
system
is
Fig. 6-8.
The performance of this four-stage multiple ejector shown in Table 6-1 and shown graphically in Fig. 6-7 and The inlet pressure for which the system was designed was
900
1
5.0
4,0
3.0
2.0
1,0
/
/
/
/
y y
y
cu
800
r-' /^
423
400
350
300"^
g 700
"-600
S.500
,|'400
J
250:|
become serious. Steam ejectors operate generally in the pressure range of single-stage mechanical vacuum pumps. However, for processes involving the evolution of large amounts of water vapor (as vacuum cooling or dehydrating), steam ejectors usually have a distinct advantage in spite For processes in which of a high rate of consumption of steam. corrosive gases must be pumped, ejector systems lined throughout with graphite or other corrosion-resisting material are used. The nozzle and diffuser parts can be made of any material which can be machined or ground with precision. The term diffusion pump is normally 6-2. Diffusion Pumps. applied to jet pumps which utilize the vapor of liquids of comparatively low vapor pressure at room temperature and which provide base
pressures significantly lower than those easily attainable with oil-sealed mechanical vacuum pumps. The development of the modern diffusion
200 S
150 g'
300
D-
pump
detail.
200
100
100
I
Q_
/
03 04 Q5 06
0.7
has a complicated history which is too involved to trace in Those interested should refer to Dushman's book^ for the
50
LL
0.8
;140
1 1
Langmuir
o'
1.0
01
0.2
09
1.0
or 02 03
Fig. 6-8.
0.4
Inlet
05
0.6
07
the diffusion
pump and
0.8 0.9
Inlet pressure.torr
pressure.torr
Pumping speed
beginnings the modern The cross section of a typical diffusion pump is shown in Fig. 6-9. Such a pump normally consists of a cylindrical housing within which is a jet assembly and at the bottom of which is a boiler for the working fluid. The nozzles which form the jets are generally annular and
As the inlet pressure is raised, the throughput of a steam ejector system increases and the discharge pressure at each stage is correspondingly increased. A limit is reached when the discharge pressure in one of the interstage regions equals the maximum which the steam can
attain in the diffuser during compression.
If the flow
is
increased
arranged so that the vapor streams from them are directed downward and outward. The housing, particularly in the region where the vapor jets impinge, is cooled to ensure condensation. The vapor from the boiler passes up through the chimney formed by the jet assembly and out through the annular slits, which act as nozzles for directing the
228
229
pump at the inlet is given momentum vapor streams. the vapor streams downward by and forced out the discharge, where appropriate forepressure is maintained by a backing pump, usually in an
the form of an oil-sealed mechanical vacuum pump. The annular space between the jet assembly and the housing must be completely sealed by vapor at sufficiently high pressure so that the back pressure from the discharge cannot break through the jets and flow backward through the pump inlet. For any design of diffusion pump there is a fairly
(low pressure)
'
Pump
fluid
Gas molecules
Jet assembly
y-.
)-<-Foreline
below which the pump is effective and above which pump fails because of breakdown of the jets. the Typical diffusion-pump jet assemblies consist of three or four annular nozzles, as shown in Fig. 6-9, or three annular nozzles and an ejector type of nozzle located in the discharge port, as shown in Fig. 6-10. The downwardly directed vapor stream from each annular nozzle entrains gas molecules incident from above and gives them momentum downward toward the discharge port. Each annular jet is capable of maintaining performance against a specific forepressure, which is relatively low for the first jet, for which the radial clearance is large,
critical forepressure
ejector jet.
(Pump
Foreiine
outlet)
"'
c
High forepressure
baffle
^;-;
3
3
:i'u.<---::- 3
(Ejector type)
Pump
Electric heoter
fluid
pump
and
is
small.
In a
pump
of
optimum
work has gone into the study of diffusion-pump design and performance, an exact understanding of the mechanisms of jet formation, gas entrainment, and pumping has not yet been attained.
Consider the enlarged view in Fig. 6-11 of the vapor jet issuing from downward and radially across the annular space between the jet assembly and the housing, and condensing on the wall. The vapor stream cannot be regarded as a well-defined jet with a sharp
the nozzle, directed
pump
jets.
boundary, particularly if the surrounding gas pressure is very low. The vapor stream in leaving the nozzle will tend to expand, causing some molecules of the working fluid to acquire net upward velocities opposite to the desired direction of flow, with the result that two undesirable effects are introduced. The backward-directed vapor molecules in colliding with gas molecules impart momentum in the wrong direction so that some of the gas molecules are expelled from the jet region and thus fail to be pumped out of the system. Also some of the backward-directed vapor molecules continue their contrary flight upward and out through the inlet, resulting in backstreaming of working vapor, which constitutes a major source of contamination in vacuum systems. During the past few years significant steps have been taken to decrease the backstreaming and improve the pumping
230
231
with resulting increase in pumping speed and decrease in the rate of contamination of the system by the working fluid. Serious attempts have been made*"* to understand in detail the physical processes involved in vapor-jet vacuum pumping with the conviction that optimum design can only be realized on this basis. There are two principal performance characteristics of vapor- jet pumps which need to be explained. These are (1) the compression ratio P2/-P1 which the pump can maintain under conditions of zero flow and (2) the pumping speed under conditions of steady gas flow into the inlet.
Vapor
flow
is Pj and the molecular density t^^ cm"". Below the surface permanent gas pressure is Pg and the molecular density ?^2 cm-^. All gas molecules which become entrained in the vapor stream quickly
of the vapor.
is
number of gas molecules per second and per square centimeter swept downward is n U. However, since P^ is greater than P^ as a result of the pumping action, there is diffusion of gas molecules in the reverse direction given by D{dnjdy), where is the diffusion coefficient of gas molecules in the vapor. Under conditions of zero flow these two quantities must be equal, so that
n^ cm-^, the
vapor
wwwwww
Gas
flow
When,
as
rig
is
very small
compared with the density n of vapor molecules, then the diffusion coefficient is inversely proportional to the vapor density and can be
written as
D=
so that
,,j||||
^
nJJ_
(6-10)
Do drig n^ dy
or
dP
~P
dy
(6-11)
exp
from diffusion-pump [Taken with permission from nozzle. B. D. Power and D. J. Crawley,
jet
issuing
(f4
(6-12)
where
pump.
L is the distance from the surface a-b to the surface c-d. Since the value of Z> is given approximately by the expression^
Vacuum
6-3. Theoretical
Compression Ratio
for a Vapor-jet
Pump.
in
8\2/
which
^^
77(1,
+ |j4
(6-13)
original work of Gaede,* Jaekel/" and Noller' the following derivation of the compression ratio for a single-stage jet pump is provided. Figure 6-12 is an idealized version of a vapor-jet pump in which everywhere below the surface a-b the cross section of the
and
| are respectively
molecules,
pump
assumed to be uniformly filled with vapor moving with a velocity U downward to the surface c-d, where the vapor is condensed and removed. Above the surface a-b the permanent gas
barrel will be
the temperature
As an example let us consider the case of vapor for which the temperature in the jet
pumped by mercury
typically be about
may
232
100C.
233
The constants
T^
for
= ig = 1^ = M, = M, =
D.
373K
3.72
10-8
4.26
X 10-8
cm cm
U is small compared with the velocity of the gas bacK through the vapor. If this were not the case, the rate of back diffusion would be appreciably diminished by the flow of the vapor stream and the compression ratio would be even greater
of the vapor stream
28.96 g
200.6 g
4
8
X 10"
229.56
(8.315
10')(373)
(27t)'^
(7.98)2 L(28.96)(200.6)
= =
3.32
X 10"
is
Since the pressure in the core of a mercury vapor jet 67 /ibar(dynes/cm2) 0.05 torr
of the order
N
P,
in
n^kT,
67 fih&T
(c) (d)
which k
is
n"
=
(1.38
67
10-i
=
X
373)
1.3
10^^
molecules/cm*
issuing
typically about
Mach
1,
steam would
be as given
in (1-28)
Fig. 6-13. Configuration of a vapor jet for various values of forepressure. Oper{A) Pumping aperture; (B) nozzle; (C) nozzle aperture; ation of the vapor pump. (D) vapor inlet; (E) gas inlet; (F) vapor jet; (G) pump casing; (H) water or air (a) Forecooling; (L )forepressure outlet; (M) to forepump; (JV) shock wave, pressure about 0.001 torr; (6) forepressure about 0.02 torr; (c) forepressure about 0.04 torr; (d) forepressure about 0.1 torr. [Taken with permission from N. A.
Florescu,
Vacuum
10,
250 (I960).]
1.66
and
Pv
= wm^ =
nM
6.023
1-3
1023
1.67
X 10-' g/cm^
Florescu* approaches the analysis of vapor-jet pumping from a The bedifferent point of view from that given above. havior of a vapor jet for various values of the forepressure is illustrated When the forepressure is very low, in Fig. 6-13 from Florescu's paper.
somewhat
giving
/
--
1.66
\1.67
2.6
10* cm/sec
the jet spreads on leaving the nozzle, as shown in Fig. 6- 13a, and completely fills the body of the pump from the pumping aperture A down to the forepressure outlet L. The full length of the pump is sealed against backward flow of gas which enters at E and is ejected
Substituting the above approximate values of n^, U, and Dg into (6-12) 10 cm, the exponent and taking for the length of the vapor column L
at L.
in (6-12)
becomes
nU
so that
L =
1.3
1015
2.6
10*
3.32
,102
X 10"
10
102
As the forepressure is increased (Fig. 6-136), a point is reached such that the vapor stream no longer persists for the full length of the pump by the barrel but is terminated in a shock front as indicated at dashed line. With further increase in the forepressure the shock front
]^Q(102)(0.434)
^_ lO**-^
In the above crude calculation almost any reasonable set of numbers vapor density, pressure, and temperature in the vapor stream results in a very large theoretically predicted compression ratio. The
for
moves up close to the pumping aperture, as shown in Fig. 6-13c. Beyond the shock front the vapor and gas densities are appreciably greater than in the vapor stream. The molecules in the jet stream above the shock front have net directed velocity downward, whereas those below the shock front are randomly directed.
r
234
235
Further increase in the forepressure results in a change in form of further, the jet stream. Since the shock front would tend to retreat walls of the the reach longer no can the result is that the vapor stream The outlet. and inlet between seal complete pump barrel and effect a
jet
is
everywhere, according to
(6-14)
stream then becomes narrow, as shown in Fig. Q-I3d, and gas from the jet the forevacuum region can then flow freely backward past through the inlet. Florescu then considers the case corresponding to that shown in
Fig. 6-136, in
P,
where
v^ is
}4n^m,v^^
since the density of the gas molecules increases with the distance below
sure
is
by the
(6-15)
pump
barrel
dPg
}/3mgV^^ drig
stream.
sisting
in the volume between the isobars within a unit area (1 cm^) is n^dy. Since n^ decreases with distance below the nozzle as the vapor stream spreads, assume as a convenient
velocities
directed downward in the pump barrel, subjects the gas molecules present in this region to a series of impacts, driving them downward. This action establishes and maintains a gradient
model that
n
N (-1)
(6-16)
of gas
concentration
increasing
outlet.
nr
6-14.
toward
Pressure, torr
the
forevacuum
Also,
Density distribution of Fig. permanent gas along a diffusion-pump barrel from inlet to exhaust pressure
entering molecules gas through the pumping aperture are continually captured and driven
Bit y = L, the length of the vapor stream. such that n^ = For gas molecules of most probable thermal velocity Vj, passing through the vapor of average thermal velocity Wav, the probability for a given gas molecule to have one collision per second with the vapor molecules is
naua,v
2
(D(x)
r+^;jo^"
dz
due to the pumping action of the jet. {A) Gas inlet; (B) vapor inlet; (C) forepressure outlet. [Taken with permission from N. A. Florescu, Vacuum
10, 250 (I960).]
(6-17)
from encounter impacts vapor molecules the gas molecules are given a high velocity in the direction of the vapor stream. By this action the density of gas molecules should decrease sharply just
where
1
and
^{x)
(f.
f.)
below the pumping aperture and then increase because of the pressure gradient maintained by the vapor stream toward the forevacuum. In
er^
2x
'
-).r xl Jo
dz
shown the jet pump configuration and the pressure (or density) of the permanent gas being pumped as deduced by Florescu from measurements made by Alexander.^" The gas pressure throughout the pump barrel can be visualized in
Fig. 6-14
is
free
is
1/0, the mean Since the average time between collisions is then t path for the gas molecules between collisions with vapor molecules
v V
terms of a
which
for convenience
may
be
V^r
(6-18)
assumed to be plane surfaces perpendicular to the axis of the barrel. Between two successive isobars separated by a distance dy at a distance
y from the nozzle the pressure increase dP^
results
from
momentum
In an infinitesimal thickness dy the probability of a given gas molecule The number of gas suffering a collision with a vapor molecule is dy]!^.
236
237
molecules crossing a unit area of thickness dy in one second is WgWav so that the number which suffer collisions with vapor molecules in this infinitesimal volume per second is
n^v^y
Mach
u^
at
373K
for
mercury vapor.
2.86 3.24
Correspondingly,
= -^-UgNaunv
O
(l
and
a;
ttav
= =
(2/y)'-^C/
(2/77'-^)Wj,
= =
X x
dy from
If
(1-30) and (1-23). we assume the temperature
-|)(D(x)%
vapor
jet is
pumped
is
300K,
since
from
(1-23) Wav/
S/tt'-^.
the rate of change of square centimeter over the layer in question, which is equal to the change in pressure over the distance dy, is
transfer per collision
m^U and
/3A;TV-^
\
=
I
/3
\
1.38
r-^i
x 10"" x x
1
SOOV-^
/
nig
4.81
777^^ 10-23
5.1
10* cm/sec
and
from
(1-22).
Vj,
/2kT\^i
I
12^Ir
dP^=^^U^^\l-l)^(.)dy
But
this pressure difference is the
(6-19)
in (6-17)
is
same
2 4.15
1.45
Mav
tt'-^
3.24
/1-45
}/3
m^v^
dtig
= mJJ^'^"?-^(l-|)0(.)<Z,
ZNaniMun
tt'^
from which
cl)(l.45)
e-^"
(2.90
0.69)
Jo
e-^'dz
related
dP, and
P
The
^j<^{x)dy
(6-20)
integral
f*
\
_
e
'
is
Jo
mt;
P =
P2 and from y
to y
n(x)
Pi
in
The
result
is
that
n(1.45)
ri-i
=
=
0.960
'"
'A
which the ratio of the molecular weights MJM^ is substituted for the ratio of the atomic masses m^jm^. If the same conditions are assumed in evaluating this expression as were assumed in evaluating (6-12), we note that
so that
Jo
e-
dz
0(1.45) --'--' 2
^A
(0.960)
0.848
and
(D(1.45)
0.122
(3.59)(0.848)
3.16
M^
(for
mercury)
(for air)
Mg
iV^ (for
mercury)
(I,
= = = =
= =
200.6 g/mole
28.96 g/mole
1.3
x 10" molecules/cm^
p^=exp
/3
1.3
015
5.01
10-16
277'/^
|,)2
(4.26
3.72)2
10-
3.2x10* -
X3.16 X 10
=
5.01
2.6
X X
10-15
cm^
exp (392)
10(392)(0.434)
10170
[/
10* cm/sec
which again is a very high compression ratio, even higher than that obtained for the same example by the method of Jaekel and Noller. ^m.
238
239
jet,
simple approach to the problem of the compression ratio for a jet pump is to assume that the velocity of the gas molecules is high compared with that of the vapor jet, which can then be considered essen-
pumped
is
dn
dt
Gas molecules penetrate the jet from the forevacuum tially at rest. density is Wj and the pressure Pj. If we assume that the side, where which experience a collision with a vapor molecule molecules those gas the jet stream and are carried to the forevacuum in are entrapped those gas molecules which pass completely through then only region, region without collision with vapor forevacuum from the the jet high vacuum side of the vapor jet. If to the penetrate molecules
(6-26)
is
Hn2Va,ye-^'^
y%niVa.v
so that
Pi
(6-27)
^1
i =
10 cm,
(f.
cr
^.r
(6-22)
5.01
10-15 cm2,
and
?i
1.3
the collision cross section for gas molecules on vapor molecules, then the mean free path for the gas molecules between collisions with vapor
is
1.3
1015
X 5.01 X
10-15
cm
6.52
and
L -=
10
6.52
65.2
molecules
is
X=
vapor
jet
(6-23)
The compression
ratio
on
this simplified
^ _ I __
,
model
,
is
thus
W(T
Pi :
Pi
g65.2
]^()(6S.2)(0.434)
2
IQii
from
(1-31),
1 1028.3 = =-
where
n,^ is
2
X m%0z,v
(6-24)
the density of gas molecules in the forevacuum region and The number of these which av is their average thermal velocity. will penetrate a distance L through the vapor jet and thus reach the
again a very large number. Thus for the simplified model of a vapor jet which has been adopted at the outset of this discussion any reasonable theoretical treatment
leads to a very high compression ratio P2/P1 against the forevacuum gas when the gas flow into the inlet is zero. However, the expression
for the compression ratio in each case is of the
is
form
dn
dt
dn 2
dt
g_i/A
Hn^Vs.^e-^"'
(6-25)
no other gas molecules entering the system on the highvapor jet, then when a steady state is reached just this same number of gas molecules (per second and per square centimeter) is being pumped by the jet from the high- vacuum region. The rate at which molecules strike each square centimeter of the exposed area of the jet from the high-vacuum side is WiWav/4, according to (1-31). However, because of the turbulent properties of the boundary of the jet, with some vapor molecules moving in the opposite direction of that desired, not all these gas molecules will become entrained in the vapor jet. If we assume that about one-half this number are actually
If there are
in
vacuum
side of the
which the quantity /S is different for each method of derivation. The predicted compression ratio is thus in each case critically dependent on the average value of the vapor density n^ in the jet length L over
which
this density persists. In the typical diffusion pump the annular nozzles produce vapor jets in the form of conical sheets. The vertical thickness through the dense part of such a jet may only be of the order of 1 cm instead of 10 cm taken for the simple model with purely axial
Furthermore, the first-stage jet in modern diffusion ticularly those using organic fluids rather than mercury,
flow.
pumps, paris
generally
may
be
less
than the-
240
241
values assumed above by a factor of 10 or more. Finally, since a depends upon the molecular diameter of the gas being pumped, a pronounced Since helium and hydrogen difference between gases is to be expected. the compression ratio for diameters, particularly small molecular have
these gases
pumping permanent gases. There is a bonus, however, in that for the pumping of condensable vapors the effective pumping speed due to
condensation on the baffle surfaces may be considerably greater than that of the diffusion pump. During the past 25 years or more the fluids most commonly used in diffusion pumps are vacuum-distilled hydrocarbon oils and a variety of synthetic liquids, all of which have much lower vapor pressures than mercury at room temperature.
be expected to be lower than for air. It is found in practice that some commercial diffusion pumps are so marginal in vapor density for the first-stage jet that the compression ratio for air In some is quite adequate, but that for helium and hydrogen is poor. the pumping light gases for that indicates cases the performance determined is pumping speed the that first-stage jet is so ineffective This deficiency comes about in an entirely by the subsequent jets. speed for air while avoiding excessive pumping effort to obtain high from the first-stage jet backfluid diffusion-pump backstreaming of
may
of these fluids
inlet.
Working Fluids
for Diffusion
Pumps.
In the above
theoretical discussion of the compression ratio at zero gas flow for vapor- jet pumps, mercury has been assumed to be the working fluid
vapor traps
applications.
is
acceptable for
Generally,
many
,lO-
systems
because it is the only chemical element widely used for this purpose. Since it is an element, mercury is not subject to decomposition. However, because of its relatively high vapor pressure (about a micron at
successfully as a working a vapor trap, usually in the form of a baffle system maintained at low temperature, is installed between the In this case the diffusion pump and the vessel being evacuated. effective pumping determined the is by for permanent gases pressure of condenwhereas the pressure the trap, the pump through speed of
using the better grades of diffusionpump oils can be operated down to pressures of the order of 10~^ torr, and sometimes as low as 10~*
torr, using
many applications
only
if
only water-cooled baffles to impede the backstreaming of working fluid vapor into the evacuated vessel.
10
15
20
25
30
35
,C
40 45 50 55
Temperature
sable vapors, including that of mercury, throughout the system determined by the efficiency and temperature of the trap. Thus,
is
if
pumping speed of a pump (normally quoted for air at room temperature) is 8 and the conductance of the vapor trap is C^, then the effective pumping speed 8^ of the combination for air as given by (2-8) is given by
the
J,
Latham, Power, and Dennis^^ have given the results shown graphically in Fig. 6-15, in which the vapor pressures of a number of
diffusion-pump oils are plotted as a function of the pressure. The curve for mercury is shown for comparison. From the discussion in the previous section one might assume that the ultimate pressure attainable in terms of the partial pressure of the
(2)
Arochlor 1254;
(3)
Narcoil 10;
Oil;
(4)
(5)
Pump
Oil 6; (6)
=^ ^ ^e ^v
Apiezon
7T
^t
-
(6-28)
,
tri-TO-cresyl
phosphate;
In order to maintain a sufficiently low vapor pressure of mercury beyond the baffle system and at the same time not cause rapid accumulation of all the mercury from the diffusion-pump boiler on the baffle system, the trap design becomes rather elaborate and the resultant conductance lower than one would like. It is generally difficult to design an effective trap within an acceptable volume which has a conductance greater than the pumping speed of the diffusion pump. Thus even with good trap design, typically 8^ ? 8^,12 for
ethyl hexyl sebacate; (13) Octoil S; (14) tri-xylene phosphate; (15) Apiezon C. [Taken with permission from D. Latham, B. D. Power, and N. T. 33 (1957)].
permanent gas being pumped would M. Dennis, Vacuum II, be so low that the observed base pressures would be independent of the working fluid. This
is
is,
in fact,
competent. essentially true, as long as the design of the diffusion pump due to the pressure However, what is normally observed is not the base
total
242
all
243
causes existing on the high- vacuum side of the diffusion pump. As diffusion pumps are normally operated with oil as the working fluid, the
oil
decomposes because of local regions of high temperature in the and the catalytic effect of the metal parts of the boiler chamber. Some products of decomposition are of lower vapor pressure than the original fluid and tend to remain in the boiler without evaporating. Accumulation of tarlike, low-vapor-pressure decomposition products will eventually clog the jet system. Other products of decomposition are of high vapor pressure, so much so that they evaporate from the jet region of the pump and escape condensation on water-cooled baffles. The base pressure measured under these conditions represents the equilibrium density of decomposition products on the high-vacuum side of the pump rather than the partial pressure of the gas being
boiler
(Dow-Corning Chemical Co.) and OS-124 high-temperature and lubricant (Monsanto Chemical Co.). The use of the latter as a diffusion-pump fluid was first suggested by Hickman, i* Silicone 704 and OS-124 have been tested by Batzer'^ in a multistage
functional fluid
Table
Ultimate Pressure Produced by a Diffusion Pump with Various Oils* Expressed in Terms of Equivalent Nitrogen Pressure at 20C; P in 10"* torr
6-2.
Measured values
Oil
Nonfractionating
Octoil-S
6.4
Fractionating
2.9 6.6 9.2
0.01-1.0
1.0
14.0 17.0
0.1-5.0
10.0
consist of ionization-gauge readings near the inlet gas flow has been reduced to zero. Since ionization gauges are normally calibrated for dry air and the readings being recorded in such measurements are due to the ionization of heavy and complex organic molecules from oil decomposition, no knowledge of the true pressures or molecular
densities corresponding to the ionization-gauge readings
is
45.0
. .
.
19.0
225 310
260
J. Blears, Proc.
100 100
(1946).
obtained.
However, measurements taken in this manner with different working fluids, or on different diffusion pumps, are still significant since an ionization-gauge reading measures a property of the high-vacuum region which is related to the electrical breakdown and is of primary importance in electronic and electronuclear applications.
Blears^* has investigated the base pressure attainable using various
In these diffusion pump with housing diameter of about five inches. 90 could mainbaffle which be one-bounce a elbow served as tests a
tained at about 20C
cooling coil, or at a nitrogen. The pressure was liquid using much lower temperature connected to ionization gauge of Alpert type measured by a Bayard-
Table
6-3.
commonly available diffusion-pump fluids. The measurements were made in a metal test dome having a diameter about three times that of
with a set of water-cooled baffles over the pump inlet to prevent direct backstreaming of hot vapor from the jet region into the test volume. A standard ionization gauge with the usual long tubulation was connected to the test volume and a high-speed or nude ionization gauge, consisting of a standard ionization gauge with the glass envelope removed, was mounted inside the chamber. The ultimate pressure recorded by the nude gauge was always higher than that recorded by the standard gauge by about a factor of 10. Blears used both a, fractionating and a nonfractionating version of the diffusion pump and compared the base pressures obtained with various pumping fluids with those previously published. He attributes the much lower base pressures reported by others to the adsorption effects within the standard gauges. His results are given in Table 6-2. To this list of pump fluids must be added Silicone No. 704 and
the diffusion
Ultimate Pressure Produced Using Silicone 704 and OS-124* Expressed in Terms of Equivalent Air Pressure at 20C
pump
Oil
cooled
..
Elbow by
Ultimate pressure,
torr
0.5 X 10-8 1-2 X 10-9 0.5-1 X 10-9
Silicone
704.
OS-124 OS-124
Water Water
Liquid nitrogen
* Taken with permission from T. H. Batzer, in 1961 Vacuum Transactions (Pergamon Press, London, 1962), pp. 315-319.
Symposium
the end of the elbow furthest from the diffusion-pump inlet. The ultimate pressures shown in Table 6-3 were observed. In order to obtain optimum performance with OS-124 it was necessary to increase the power input into the pump boiler by nearly a factor of 2 over the rated power input and to allow the lower end of the pump barrel to operate at a temperature of 90 to 100C. The fact that the ultimate
244
245
pressure decreased
by only about
x 10-
torr
when
cooled with liquid nitrogen indicates a very low rate of production of condensable vapor from the fluid. Whether the results quoted in
Table 6-3 are directly comparable with those in Table 6-2 is doubtful because of the many differences between the conditions of measurement. Hickman" has also discussed- the use of the polyphenyl ethers as diffusion-pump fluids. The chemical structure of these compounds BL-10 consists of a number of phenyls linked together by oxygen. available commade ethers is a refined mixture of 5-ring polyphenyl pressure at vapor The mercially for use as a diffusion-pump fluid. but later IQ-^ torr, 25C is reported by Hickman as (1.3 0.3) x samples were predicted to have substantially lower vapor pressure. Hickman reports attainment of base pressures of 2 to 3 x 10"^ torr using BL-10 in commercially available diffusion pumps with a roomtemperature baffle. He recommends the combination of two baffles in series, one at slightly above ambient temperature and the other cooled by chilled water for the attainment of partial pressure due to backstreaming vapor of less than lO"" torr. The advantages of the polyphenyl ethers are claimed to be due to the very high bond energies,
runs back into the pump, and the second baffle at a much lower temperature (such as that of liquid nitrogen), depending upon the acceptable mercury vapor pressure in the system. The performance of large mercury diffusion pumps has been investigated by Power, Dennis, and Crawley." The pumps were of mild
Table
6-4.
Number
Fluids
1st
of cycles
run
2nd run
205
160
Octoil S* Di-2-ethyl hexyl sebacate Tri-xylene phosphate ....
Pumps
residual
tC"areoil
10+
*.
Pumps
failed
because of
Dow-Corning 703
which are among the strongest in organic chemistry. Latham, Power, and Dennis^'' tested the ruggedness of a number of
backing pump.
oils
Dow-Corning 702
500
1,51611
system of a few liters volume and then letting in air with the diffusion pump still hot but the heater turned off. The sequence was repeated
until the oil
diffusion
pump
failed.
Vacuum
t
Taken with permission from D. Latham, B. D. Power, and N. T. M. Dennis II, 33 (1952).
The fluids are given in Table 6-4 in the order of increasing ruggedness. The choice between mercury and oil as the working fluid for a However, the choice diffusion pump is generally not difficult to make. and the many synthetic between the vacuum-distflled hydrocarbons used for practically any Mercury may be fluids is not so easily made.
application for which diffusion
Vacuum-distilled hydrocarbon oil. Di-2-ethyl hexyl sebacate. Methyl polysiloxanes (silicones) If In the case of Dow-Corning 702 the fluid never broke down, but the tests were discontinued when half the original charge had been lost into the
J
forevacuum.
pumps
vapor baffles at sufficiently low temperature are used to condense the mercury vapor which would otherwise be present at the room-temperature vapor pressure of about 2 x 10-^ torr throughout the high-
or stainless steel construction of conventional multijet design and of 2M to 24-in. barrel diameter. Since the vapor pressure of mercury is considerably higher than that of most diffusion-pump oils, the density of vapor backstreaming from the
first jet is
vacuum
much
oil.
For long periods of pumping, the vapor baffles avoid accumulation of frozen mercury in such quantities that the contents of the boiler eventually are condensed on the baffles and the pump fails for lack of working- fluid. The difficulty can be avoided by dividing the baffle into two sections, the section nearest the diffusion pump operating at a temperature just above the melting point of mercury ( 38.87C) and arranged so that the condensed mercury
must be designed to
In order for gas molecules entering the inlet to reach the first jet, they must diffuse through this cloud of mercury vapor, the molecular density of which is generally much greater than that of the gas being pumped. In the absence of a vapor trap the pumping speed is found to decrease with the distance above the mouth of the pump, as shown in Fig. 6-16. These curves were taken for three different temperatures of the condensing surface (25, 19, and 5C), i.e., the upper portion of the
246
247
pump
By mildly refrigerating the condensing surface the throughout the region above the jet is reduced, pressure mercury vapor The the resistance to gas flow into the pump. reducing also thereby the cooling more or by of 3 factor a speed is increased by pumping of advantage take et al. Power 25 5C. surface from to condensing
the dependence of pumping speed on the distance above the inlet
Table
6-5.
Wall condition
Clean
Slight contamination
Back -migration
rate,
Pumping
speed,
cm^/hr (liquid)
1.05
4.3
. . .
liters/sec
Heavy contamination.
Condensing
wall
12.4
temperature
[I]
25C,[2] I9C,[3]
5%
* Taken with permission from B. D. Power, N. T. M. Dennis, and D. J. Crawley, in 1961 Vacuum Sym,posium Transactions (Pergamon Press, London,
4,000
8,000
12,000
16,000
1962), p. 1218.
contaminated, and heavily contaminated walls; all taken at a wall temperature of 14C, using a 9-in. diffusion pump. Back-migration rate was measured as a function of wall temperature with and without a water-cooled copper cover with a long skirt reaching down to intersect a portion of the jet stream, as described by Vekshinsky, Menshikov, and Rabinovich.^* The results of these measurements are shown in
Fig. 6-17.
Power
et
al.
recommend
barrel,
4.5
pump
cleaned
or
4.0
walls
13.0
6
or 2.5
Fig. 6-16. Variation of air-pumping speed with distance above inlet flange for three different cooling-wall temperatures. Curves plotted to scale for pump of 60 cm diameter. [Taken with permission from B. D. Power, N. T. M. Dennis, and D. J. Crawley, in 1961 Vacuum Symposium Transactions (Pergamon Press,
diameter mercury diffusion pumps made of mild steel. Chupp observed that thoroughly cleaned pumps sometimes fail to exhibit
2.0
1.5
London,
1962).]
placing the combination of a refrigerated (.~ 30C) and a liquidnitrogen trap above the umbrella nozzle with as little vertical clearance
by
pump
inlet.
any appreciable pumping speed, but after a day or two of operation become "conditioned" so that mercury is efficiently condensed and the performance finally comes up to standard. Once conditioned, the
1.0
0.5
-5
10
15
20
25
30
Performance of mercury diffusion pumps has been found by many investigators to depend critically
pump
upon the
wall and of the pump housing generally. Mercury condensation is seriously impeded by wall contamination, resulting in poor or even negligible
pumping speed and a higher than normal mercury vapor pressure above the jet. Table 6-5 gives some data by Power et al. showing the mercury back-migration rate and pumping speed for clean, slightly
operates at a high performance level indefinitely. An alternative precaution found by Chupp to be effective is to copper plate the inner surface of the pump barrel where the mercury vapor should condense. Apparently the slight
Fig. 6-17. Variation of mercury back migration rate with temperature of the pump condensing wall. [Taken with permission from B. D. Power, N. T. M. Dennis, and D. J. Crawley, in 1961 Vacuum Symposium Transactions
1962).]
248
249
amalgamating
for
which
that of liquid nitrogen. In any case there are oil is the preferred diffusion-pump fluid
many
applications
pumping speed
disadvantage
for
pump.
Whether
this
is due to an inherently lower Ho coefficient (see Sec. 6-5) mercury jets, or whether it is merely a reflection of the greater Even effort expended in developing oil diffusion pumps is not clear. with this disadvantage there are applications for which mercury is the
semiconducting films. 2. In the evacuation of vessels in which high electric-field gradients must be maintained. At the University of California Lawrence Radiation Laboratory large (32-in.-diameter) mercury diffusion pumps are used on many particle accelerators and certain other electronuclear machines in preference to oil diffusion pumps. Surface contamination
In general -purpose vacuum -pumping systems for laboratory and small production applications. 2. For the evacuation of most types of transmitting electron tubes, klystrons, and magnetrons. 3. For the evacuation of particle accelerators in those cases in which the voltage gradient requirement is not too severe. Most cyclotrons, betatrons, and electron and proton synchrotons are oil-pumped, but with the addition of liquid-nitrogen-cooled vapor baffles. 4. In vacuum coating and sputtering systems. 5. In vacuum molecular stills and fractionating columns. 6. In the vacuum purification of metals, as in vacuum melting and casting, electron beam melting, and vacuum zone refining.
1.
6-5.
Pumping Speed
of Diffusion
Pumps.
In the normal
by
oil
operating pressure range the action of a diffusion pump can be compared with that of a somewhat imperfect hole into a region of zero pressure. This behavior can be understood in terms of the incidence of gas
probably because of the decomposition products remaining on the surfaces after repeated vacuum sparking. 3. In systems in which a very high forepressure is an advantage, either because of unusually large throughput or because of the choice of the backing pump, e.g., a steam ejector. 4. For pumping highly reactive chemical vapors which might decompose diffusion-pump oils. Mercury, being an element, is the most rugged of the diffusion-pump fluids in the sense that it cannot be
molecules on the surface of the vapor jet and the process by which some fraction of them are trapped by the jet and carried into the forevacuum region. Each gas molecule that approaches the vapor-jet surface experiences collisions with heavy vapor molecules. If the pressure at the inlet of the pump is low as compared with the vapor
pressure of the jet because of its local temperature, the jet will expand on leaving the nozzle. The boundary layer of the jet consists in part of
decomposed by overheating or by chemical action. 5. For evacuation of mercury vapor electron tubes, such as rectifiers, For these applications a vapor trap is not ignitrons, and thyratrons. required since the finished tubes contain pools of mercury to be partly
vaporized during operation.
Oil diffusion
pumps have
vapor molecules which diverge outward and upward from the main body of the jet and impede penetration of the jet by gas molecules. However, some fraction H of the gas molecules incident on the jet surface penetrate the boundary layer and are captured by the jet. The value of depends upon the detailed structure of the boundary layer. If A is the pumping aperture (i.e., the surface area through which gas molecules enter the vapor jet from the inlet) then the rate of incidence of gas molecules on this surface is, according to (1-31),
the
applications in which the desired operating pressure caiT.be maintained with a simple water-cooled baffle and no refrigerated trap. Operating pressures down to 10"^ torr can easily be attained in this manner, and with care in selection of the fiuid down to about 10~^
torr.
many
vA
Hn.VavA
(6-29)
jet is
The
thus
rate at
arrangement sometimes used is similar to that described for mercury, the baffle nearest the pump maintained at a temperature of 30 to 40C by a refrigerator, and the second baffle at a much lower temperature.
cooled baffle
The
= HvA =
HHn^v^vA
(6-30)
baffle
pumped
is defined as the volume of gas at inlet pressure per unit of time. If Q denotes the gas-flow rate as defined in (2-18), then the pumping speed is obtained by dividing the molecular
250
251
by the molecular
density, so that
(6-31)
when the pressure substituted into (6-31) is the partial pressure of the gas admitted at the measured flow rate Q. The performance of a typical four-stage diff'usion pump of modern design when pumping air is shown graphically in Fig. 6-18. The solid curve represents the
true
pumping speed
which Pj represents the partial pressure at the inlet of the gas being admitted at the flow rate Q. The existence of other gas components which contribute additional background pressure at the pump inlet, such as backstreaming vapor from the jet and outgassing from the walls of the vacuum chamber, has nothing directly to do with the pumping process and should, therefore, be ignored in the determination of the pumping speed. By substituting the value of Wav from (1-23)
in
each point
on the curve was obtained by substituting the measured flow Q of air and
the corresponding pressure Pj in the
equation
r
(6-34)
1.600
'^
/
/
i. 1,200
^
1
pumping speed
where Pj is the inlet pressure measured with either a McLeod gauge or an ionization gauge when the flow rate is Q, and Pq is the base
pressure of the
800
/
V\
\
10''
10'^
10'^
400
/
F^,
= =
which
for air at
3.64
TV \0^H[ A
I
\
pump when
the flow
is
10"
10'"
10'
cm^/sec
rate
is
zero.
made up
of backstreaming diffusionoil
'Mm) ^
8^
liters/sec
(6-32)
pump
tion of the
(20C) becomes
liters/sec
Fig. 6-18. Pumping speed as a function of inlet pressure for a typical 6 -in. oil diffusion pump.
room temperature
\\.%EA
(6-33)
on the inlet side of the pump. The difference between Pj and Pq is due to the flow of air admitted at the inlet and is therefore equivalent to Pi in (6-31).
when
the
pumping aperture
is
measured
in square centimeters.
Because the surface of the vapor jet is diffuse and its area therefore difficult to define, the cross section of the annulus between the nozzle
in Fig. 6-18
is
and the pump housing is normally taken as the pumping aperture A. The coefficient R appearing in the above equations is a measure of the Asefficiency of the pump and is referred to as the Ho coefficient. ^^ weight molecular the of independent is suming that the Ho coefficient of the gas being pumped, Eq. (6-32) implies that the pumping speed
of a diffusion pump should be inversely proportional to the squall root of the molecular weight of the gas. This proves to be approximately true for some pump designs, but is by no means consistently
=Q
(6-36)
the case, particularly for multistage pumps with closely spaced jets. For example, results reported by Noller, Reich, and Bachler^i obtained for the pumping speed of a multistage diffusion pump, with a liquidnitrogen-cooled vapor trap, for hydrogen and air yield a ratio of 2.25 instead of 3.8. Stevenson^^ reports a factor of only 1.2 between the
pumping speed
for hydrogen and that for air. Equation (6-32) also implies that the pumping speed of a diffusion pump is independent of the pressure. This is indeed very nearly the
which has generally been used in commercial practice, instead of the pumping speed. The apparent pumping speed given by (6-35) becomes zero at the base pressure Pq. However, the true pumping speed obtained by using Eq. (6-34) is found to be independent of the pressure for as small a value of the leak for which the pressure diff"erence Pj Pj can be reliably measured. The validity of Eq. (6-34) in giving the true pumping speed is emphasized by the measurement of the pumping speed through a liquid-nitrogen-cooled vapor trap installed at the inlet of the pump. Figure 6-19 shows the conflguration of the trap and Fig. 6-20 the resulting performance curve. The base pressure of the pump is now reduced by a large factor by the condensation of the condensable vapors from the pump oil. It is therefore possible to carry the measurement of the pumping speed of the pump plus the vapor trap
true
252
to
253
lower values of the leak rate. The expected pumping speed for air is that of the untrapped pump combined with the conductance of the trap
1
1 1
much
real or instrumental
slight decrease with decreasing pressure a valid question in view of the difficulty of the at pressures as low as 10~* torr.
8
or
8p
C,
=
8^
C,
(6-36)
between the measured pumping speeds of that implied by (6-32) is that the pumping speed is not necessarily inversely proportional to the square root of the molecular weight of the gas. The pumping speed for hydrogen should
difference
pumps and
air by a factor pumps the factor is considerably smaller. One may conclude that
combined
diffusion
pump
the
Ho
coefficient // is a function of
and
cold trap as
shown
in Fig.
6-20 reaches a peak value at about 10~^ torr and then decreases with
the molecular weight of the gas and decreases more specifically that
600
,200
value which is then independent of the pressure to values of the pressure substantially less than the base pressure of the pump without the trap.* This would not be the
decreasing
pressure
to
with molecular weight. There is some evidence that the difficulty arises because of an insufficient quantity of vapor flowing through the flrst-stage nozzle. Since for hydrogen the molecular velocity is greater and the collisional cross section smaller than for
air,
800
^y
\ \
\
400
^
10"'
_j
10"='
10"^
10'"
10"^
10"'
Inlet pressure.torr
if the true pumping speed for were zero at the base pressure for the untrapped pump. Consistent with the theoretical results
case
air
pump
(Fig.
6-18)
with
jet
that a very high zero-flow compression ratio for the gas being
mounted on
in
6-in. diffusion
pump.
pumped by a
jet
pump
is
to be
terms of the partial pressure of air due to migration back through is apparently always low as compared with the background pressure due to outgassing and other factors. This being the case, the true pumping speed should be essentially independent of the pressure for many decades in pressure. Noller, Reich, and Bachler^^ have measured the pumping speed for air and hydrogen of a multistage diffusion pump with a liquid-nitrogen-cooled trap over the range from 10~^ to 10~" torr and observed a decrease in pumping speed of only about 20 per
the jet
* The peak in the curve at relatively high pressure arises from the fact that the annular clearances in the vapor trap are of the order of 2 in., which is comparable with the mean free path at a pressure of 10~* torr. As the pressure
increases in this range, the conductance for the trap is no longer that for true molecular flow but increases with the pressure as expected (see Chap. 2) in the
transition range.
be good for air and poor for vapor trap (Fig. 6-19). hydrogen as is implied by the remarks In this case the pumping speed for hydrogen at the end of Sec. 6-3. would be primarily determined by the second-stage jet for which the aperture is smaller. Since commercial diffusion pumps are usually rated on the basis of maximum pumping speed for air with minimum backstreaming of oil vapor, the optimum jet design tends toward minimum effective vapor flow through the first-stage nozzle compatible with high pumping speed for air. This criterion may automatically ensure insufficient vapor flow for pumping gases of low molecular weight, such as hydrogen and helium. Results reported by Normand^^ on the pumping speed of a commercial diffusion pump show a normal, smooth performance curve for air with a pumping speed of about 320 liters/sec over the range from 5 x 10"^ to 2 x lO"' torr, but an erratic behavior when pumping hydrogen. The pressure fluctuations were such that the pumping speed varied rapidly over the range from about 320 to 480 liters/sec. The Ho coefficient is not necessarily the best figure of merit for diffusion pumps, as has been pointed out by Stevenson. ^^ The area A referred to in (6-32) is the cross-sectional area of the annular space between the first-stage nozzle and the pump housing. A design which requires a large diameter nozzle may have a high Ho coefficient in
may
254
255
pumping speed
terms of the small pumping aperture which remains but a small for the diameter of the pump. Stevenson suggests that a better criterion is the speed factor defined in terms of the pumping speed for air as S^ M'^
3.64J
= PS
J,)
is
shown graphically
in
10"
pump
T
for air at 20C
(6-37)
shown in Figs. 6-18, 6-20, and 6-21. The zero-flow limiting forepressure independent of the pumping is speed of the forepump and is therefore a useful standard for
10"
=
1,440
torr
iters/sec
comparing
of a
Q = 144
tor
liters/sec
U.6A,
similar to the definition of the
diffusion
pumps.
forepressure
^ 10"
Ho
coefficient
in (6-32)
and
(6-33)
The
limiting
except that A^ is the cross-sectional area of the pump housing rather than the annulus. This definition is consistent with the Knudsen and Clausing conductance factors discussed in Chap. 2. For the particular pump developed by Stevenson the Ho coefficient was found to be 0.51 and the speed factor as defined by (6-37) was 0.45. 6-6. Limiting Forepressure for Diffusion Pumps. In Fig. 6-13 the behavior of a vapor jet as a function of the forepressure is illustrated. When the forepressure is sufficiently high, the shock front, which is the boundary between the jet composed of directed vapor molecules and the randomized region of forevacuum vapor and gas, contracts to the point where the jet fails to bridge the pumping
aperture.
effective
diffusion
a 10'
number
following
the
Q = 14.4
torr
liters/sec
may
be listed
Q=
10"
torr
liters
/sec
forepressure
2.
10'
ing fluid
0,100
0.200 0.300
0.400 0.500
When
this condition
is
reached, there
is
direct
communiand
giving
the
higher
limiting
fore-
Forepressure, torr
pressure
diffusion
Q through pressure (e.g., by admitting an increasing flow of gas into the forevacuum region) with a constant or zero flow of gas into the diffusion-pump In the former method the pumping speed of the pump is found inlet. to be independent of the pressure until the forepressure limit is approached. The pumping speed then decreases rapidly with increasing' flow and finally reaches the pumping speed of the forepump. The curve in Fig. 6-18 shows the decrease in pumping speed as the forepressure reaches the limiting value. This effect can be presented more dramatically by plotting for a constant throughput Q, the inlet pressure as a function of the forepressure controlled by admitting air into the forevacuum region, as is done in Fig. 6-21. In this latter method the inlet pressure is found to be independent of the forepressure over a wide range and then to increase rapidly as the forepressure is increased beyond a critical value. The value of the forepressure at which the inlet pressure has increased above its normal value by 10 per cent can be used as an arbitrary definition of the limiting forepressure. The
increasing the flow rate
limiting forepressure as a function of the flow rate or throughput
Power input into the boiler; the greater power input giving the
3.
pump.
Optimum design and operating conditions for a multistage diffusion pump are a compromise between these factors and others, such as the
rates of decomposition
fluid.
In
Fig. 6-23 are curves of inlet pressiu-e as a function of forepressure with lo.6
5
0.5
|o.4
^>
0.3
0.2
constant throughput for various values of the power input. It will be noted that for power input appreciably above that recommended for normal operation the
pumping speed
_
,-2
for
lo.i
2
3 5
7 ]Q-I
5 7
1
Iff
10
Throughput, torr
liters/sec
decreases as the power input and limiting forepressure increase. Increase in the power input,
Fig. 6-22. Limiting forepressure as a function of throughput for 6-in. diffusion pump.
256
257
of the jet in leaving the nozzle and therefore more vapor molecules adversely directed on the surface of the jet exposed to the high vacuum. The result is a greater proportion of gas molecules knocked away from the jet by backward-moving vapor molecules before penetrating to the
core of the jet where they can be propelled toward the fore vacuum.
10"
The
first-stage jet
must withstand
o
0\J
o
r<S
CO
T3-
o o
not only the back pressure due to the gas being pumped, but also the higher vapor pressure from
the subsequent jets. The result is a change in the jet form which contributes to the condition men-
10"
oO
o
oo
larger Pj is, the smaller need 8^, be to meet this requirement. In Fig. 6-23 on the left is the throughput as a function of inlet pressure, showing the maximum reached at that value of the inlet pressure for which the forepressure is equal to its limiting value. From this characteristic of the diffusion-pump performance and a knowledge of the maximum throughput to be expected the required capacity of the forepump can be determined from (6-38). Alternatively, in a batch-processing system in which a short roughing-down time is important, the high limiting forepressure permits transfer from the relatively low pumping speed of the mechanical pump to the high speed of the diffusion pump at an earlier point in time. For systems of high throughput or of frequent pumpdown a high limiting fore-
The
- o,
o
10"*
o o ro
1
* * * oo o
pressure contributes to
economy and
efficiency.
J J J
Q.torr liters/sec
o c
tioned above. Also, in the case of organic fluids, as contrasted with mercury for producing the working vapor, the diecomposition rate increases rapidly with temperature.
^10
o o o 5
Some
of the components of
10"'
o O O O o_ 7 ;
oo _0 77
C\J
^ ?
the decomposition are permanent gases in the sense that they do not
Q,torr liters/sec
condense at any convenient trap temperature, and others are of comparatively high vapor pressure. These contaminants contribute further to the decrease in
10"
0,100
Q200
0.300
0.400 0.500
the
In order to increase further the maximum throughput without increasing the size of the mechanical backing pump, an additional vapor-jet pump may be inserted at the exhaust of a diffusion pump. Two types of vapor- jet pumps using working fluids of higher vapor pressure than those normally used in diffusion pumps have been developed for this purpose, the booster diffusion pump and the oilvapor ejector pump. The former is similar in design to a diffusion pump except that the jet clearances are smaller and the number of stages is usually only two. The latter resembles the steam ejectors described at the beginning of this chapter, having a converging-diverging nozzle and a diflfuser. The designs of booster diffusion and oil ejector pumps are optimized for fluids such as butyl phthalate, butyl sebacate, highvapor-pressure hydrocarbons, chlorinated hydrocarbons, and ethyl or propyl phthalate. Limiting backing pressures are typically of the
order of 0.5 torr for booster diffusion
Forepressure, torr
boiler temperature
increased.
is
pumps and
vacuum
power input
for
6-in.
diffusion
applications such as
vacuum
materials and
pump.
low pressure with mechanical backing pumps to meet the requirements at low gas flow. A high limiting forepressure permits the continued operation of a vacuum system at the full pumping speed of the diffusion pump for a high gas flow or throughput with a backing pump of modest capacity. Since the throughput is Q = PS^,, then at the limiting forepressure for the diffusion pump the forepump speed must be
(6-38)
vacuum degassing
6-7.
of metals.
Factors Contributing to tlie Ultimate Pressure of a Diffusion Pump. The ultimate pressure of a diffusion pump is the inlet pressure which the pump maintains with zero gas flow. This definition is incomplete because the conditions of measurement are not defined. The use of low-vapor-pressure oil instead of mercury as a working fluid makes it possible under many circumstances to use
water-cooled baffles without a refrigerated trap for obtaining a sufficiently low base pressure for many applications. Because of the convenience of this type of operation it has become customary to define the ultimate pressure of an oil diffusion pump as the ionization-gauge reading obtained in a test dome protected from the direct backstreaming
is
258
259
vapor from the pump by a simple water-cooled baffle. For such measurements the air calibration of the ionization gauge is normally used, even though the residual gas remaining in the test dome is not air but is a mixture of gaseous and vaporous products of decomposition from the diffusion-pump fluid, as was briefly mentioned in Sec. 6-4. Some of the factors which influence the ultimate pressure as defined above are
e.
through the annular space between the nozzle structure and the pump housing from the free surface of oil in the reservoir. The evaporation of oil vapor from the region of the pump housing where the first-stage jet impinges and is intended to
condense.
The leakage of vapor from source (a) was easily eliminated by providing an unperforated cap and attaching it to the center post from below.
Wetting of the
to result
lip
of the umbrella
(6)
1.
was found
pump.
2.
Decomposition of the working fluid with the evolution of components which in part consist of condensable vapors of high vapor pressure and of permanent gases. Release in the vapor jet of forevacuum gases dissolved in the working fluid after condensation
boiler
from condensation of the working fluid on the bottom surface of the nozzle cap due to the
continual loss of heat
by
radiation.
3.
Heating the umbrella cap was found to eliminate "wet running" and reduce the backstreaming substantially. They found that evaporation and spitting from the region of the oil reservoir {d) could be
by proper design of the The remaining sources of boiler. backstreaming (c) and (e) from the
eliminated
region of the
4.
Outgassing of surfaces on the high-vacuum side of the pump. Some aspects of the backstreaming of diffusion-pump
fluid
of
its
Fig.
6-25.
Fig.
Sources of vapor backpump. in a diffusion streaming [Taken with permission from B. D. Power and D. J. Crawley, Vacuum IV, 415 (1954).]
6-24.
in Sec. 6-4.
The
results of a
comprehensive study of the backstreaming problem have been published by Power arid
diffusion
Working on the hypothesis Crawley, Vacuum IV, 415 (1954).] that the boundary layer of the jet adjacent to the nozzle surface consists largely of vapor molecules somewhat randomized in direction because of collisions with the nozzle surface, as illustrated in Fig. 6-11, they devised a water-cooled guard ring,
one form of which
is
shown
Crawley.2*
pump
of typical design
randomly directed vapor When the bottom rim of the cap extends
of
at the time of their investigation the sources of backstreaming vapor could be identified as indicated in Fig. 6-24 as
a.
6.
c.
d.
The leakage of vapor around the nut with which the "umbrella" for the first jet was secured to the center post. The wetting of the umbrella of the first jet by condensation of vapor on the lip followed by evaporation. The scattering of some oil molecules upward toward the inlet of the pump by collision processes in the first-stage jet. The evaporation of oil vapor and spitting of oil droplets up
the nozzle to intercept a significant portion of the jet, the backstreaming is found to be reduced by a factor of 10 to 30, depending upon the design of the pump tested. Except in pumps of small bore,
not reduced by addition of the cooled guard ring. Vekshinsky, Menshikov, and Rabinovichi* described the use of a water-cooled cap with a long skirt placed over the first-stage nozzle and reaching down far enough to intercept the boundary layer of the The funcjet to reduce the backstreaming, as illustrated in Fig. 6-26. described by ring guard tion of the cap is identical with that of the Power and Crawley.
the
pumping speed
is
260
261
The thermodynamic principles involved in the performance of a diffusion-pump boiler and nozzle system in producing vapor jets were examined theoretically and experimentally by Smith. ^^ Observation of the temperature of the oil in the boiler and that at various locations on the vapor stack and nozzle system showed that the stream issuing
from the nozzles in a typical commercial oil diffusion pump is a mixture of vapor and condensed fluid due to contact with relatively cold surfaces The "wet running" referred to by in the stack and nozzle assembly. Power and Crawley above was therefore attributed to condensation on
these cold surfaces.
vapor in the boiler to maintain the pressure required by the nozzles of the diffusion pump. It is noted that the boiler temperature for the particular diffusion pump tested increased more than it should for a
given increase in power input if the Clausius-Clapeyron equation for change of state and the Langmuir equation for the rate of vaporization both apply to the fluid. The conclusion reached as a consequence of these tests was that the pump boiler did not provide adequate surfaces of proper geometry to promote the generation of vapor at a rate high enough to supply the jets at the normal rate of evaporation. Increasing power input instead of increasing the rate of evaporation by a
temperature of the oil by more than the appropriate amount. By providing a tubular radiant heater down through the center of the jet assembly above the level of the oil and a separate heater attached to the top umbrella, the nozzle assembly could be kept at higher temperature than the oil in the boiler and the vapor thus superheated. As a consequence of these changes Smith found that the peak pumping speed of the pump was about one-third greater and the base pressure typically reduced by a factor of 2.5 as compared
signiflcant factor raised the
Enthalpy
6-26. Water-cooled cap diffusion-pump nozzle to repress backstreaming. [Taken with permission from S. A. Vekshinsky, M. I. Menshikov, and I. S. Rabinovich, Vacuum 9, 201 (1959).]
Fig.
over
Vacuum Symposium
Transactions
I960).]
with the performance of the pump without the axial superheater and top nozzle heater in operation. Another approach to the problem of jet formation is that described by Florescu,* who has developed a special coaxial nozzle illustrated in Fig. 6-28. The nozzle consists of a central cylindrical section which, because of the lack of any expansion, produces a dense, low-velocity jet core and an outer de Laval type of expanding nozzle which produces a supersonic jet of low density. According to the author, the coaxial nozzle reduces backstreaming substantially but at the same time
provides the high-density core necessary for effective pumping against a high forepressure. The coaxial nozzle had not been incorporated
in
vapor to be everywhere abovfe the vapor-mixture line on the enthalpy diagram as represented by point 3 in Fig. 6-27, corresponding to superheated vapor. The amount of superheating desired is such that in expanding through the nozzle the vapor temperature does not drop below that corresponding to point 1 in Fig. 6-27 on the vapor-mixture line. Cooling to a lower temperature corresponding to point 2, which is in the mixture zone, during expansion through the nozzle will result in some condensation with the formation of droplets of oil in the jet stream and a deposit of a film of liquid oil on the nozzle surfaces. Another requirement discussed by Smith is the generation of sufficient
for the oil
In two excellent papers Hablanian and his colleagues^*'^' have reported results on the backstreaming rates and the sources of backstreaming vapor for diffusion pumps of improved design. They measured the backstreaming rate for a 6-in. diffusion pump with the apparatus shown in Fig. 6-29. The backstreaming oil vapor was caught on the walls of a chamber mounted on the pump with a flange cut conically at the top to admit all oil molecules leaving the rim of the top nozzle and flying directly out of the pump inlet. The oil ac-
cumulation drained into the calibrated collector was measured periodduring each run. With the flrst-stage nozzle, as illustrated in Fig. 6-29, the backstreaming rates shown in Fig. 6-30 were obtained.
ically
262
263
the chamber was thoroughly cleaned before the run, the curve was obtained; whereas when the inner surfaces of the chamber were dehberately wetted in advance with diffusion pump oil, the curve labeled wet dome resulted. It may be seen that about forty hours of operation were required to reach a steady rate with the
labeled dry dome
When
wet dome, but nearly twice that long was required with the dry dome. The authors emphasize the need for long periods
ments of
Bockstreoming
foctor
the collect-
IT
05
gouge
Collector'
Fio.
6-28.
Coaxial diffu-
sion-pump nozzle. [Taken with permission from N. A. Florescu, Vacuum 10, 250 (I960).]
Fig. 6-29. Apparatus for backstreaming measurements. [Taken with permission from M. H. Hablanian and H. A. Steinherz, in 1961 Vacuum
London,
1962).]
backstreaming rate on the length of connecting pipe was measured, with results shown in Fig. 6-31. The decrease in pumping speed due to the added lengths of connecting pipe is also shown. The insertion of a length of pipe equal to 1.5 diameters of the pump reduces backstreaming to only about 2.5 per cent of the original value, whereas the pumping speed is reduced to about 63 per cent. Thus for the simple,uncovered firststage nozzle used in these tests the backstreaming rate could be reduced by a large factor by insertion of a length of connecting tubing without
excessive loss in the resulting
03
IT
0.5 to
02
to
015
O02to0.01
pumping
speed.
.0.04
V
05
Dry
1
Speed
\ "
;0,02
dome
50
100
D
Bockst reoming
shapes added to Steinherz, in 7967 Vacuum Symposium, the original simple first-stage nozzle is shown in Transactions (PergaFig. 6-32. For the type of water-cooled cap mon Press, London, adopted as standard the reduction in backstream- 1962).] ing rate was a factor of 0.02 to 0.01. Applying this factor to the asymptotic results shown in Fig. 6-30, one obtained 0.024 to 0.012 mg/cm^ hr as the backstreaming rate for the pump with optimized cold cap, which is said not to impair pumping speed. This backstreaming rate is of the same order as, but slightly smaller
effect of various cold-cap
The
2D
Time.hr
Distance from
pump
inlet
by Hickman^"
for a
pump
of similar improved
Fig.
6-30.
Backstreaming
rates
Fig.
6-31. Effect
of
length
of
measured by two different techniques showing eventual agreement between the methods after sufRciently long running time. [Taken with permission from M. H. Hablanian and H. A. Steinherz, in 1961
connecting pipe on backstreaming rate. [Taken with permission from M. H. Hablanian and H. A. Steinherz, in 7967 Vacuum Sym,posium Transactions (Pergamon Press,
Hablanian and Steinherz^" have also reported results on the backstreaming rate as a function of inlet pressure which show a rapid increase with pressure as the pressure exceeds lO^^* torr. The results
are shown in Fig. 6-33, in this case using a 32-in. diffusion pump and a different tjrpe of collecting system. The change in backstreaming rate is negligible with increasing pressure until the inlet pressure
London,
Vacuum Symposium
1962).]
Transactions
1963).]
reaches 10-^ torr; then the rate increases by more than a factor of 1,000 as the inlet pressure increases from 10^^ to 10-^ torr. In systems
264
265
requiring contaminant-free operation the need for avoiding highpressure operation of the diffusion pumps is emphasized by these results, which are more quantitative but entirely consistent vi^ith
previous results, such as those of Power and Crawley.^* In a study of the remaining sources of backstreaming in a diffusion pump with a water-cooled cap over the first-stage nozzle, Hablanian
vapor flow required by the jets determines the power input necessary The power density must be relatively low in for effective pumping. order to avoid local hot spots resulting in a high decomposition rate and eruptive boiling. Erratic pressure surges, excessive backstreaming, and poor average base pressure are observed in the performance of most of the older types of oil diffusion pump. Stevenson^^ describes a major advance in boiler design, in which the heat-transfer area is not only large as compared with that of previous designs but is in the form Glass tape O.OI4in of hot baffles which extend up above
I
100
/
:
/
/
Stevenson reports a much lower of decomposition, broader pumping-speed curve, higher peak
rate
pumping
speed,
and higher
limiting
forepressure
10 r
= /
/
/
/
than were obtained with heater systems previously used. A further advance in heater
Nichrome
xO.OOl
in.
"
10"^
Inlet
0.1
10"^
10"^
10"'
10"2
10"'
pressure, torr
-6-in, diffusion
pump (HS-6-1500)
Fig.
6-33.
Backstreaming rate
vs.
a 32-in. diffusion pump (fluid: Narcoil 40). [Taken with permission from M. H. Hablanian and H. A. Steinherz, in 19C)1 Vacuum Symposium TransacHofis
inlet pressure for
Fig. 6-34. Experimental arrangement for detection of backstreaming source. [Reprinted with permission from The
1962).]
Macmillan Company from M. H. Hablanian, in 1962 Vacuum Sym,posium Transactions. Copyright 1962 by American Vacuum Society.]
Fig. 6-35. Diffusion-pump heater of design for diffusion pumps is reported by Milleron and Levenson.^* Milleron and Levenson. [Taken with permission from X. Milleron and L. Their heater consists of a corruL. Levenson, in 1961 Vacuum Symgated strip of Nichrome wound posium Transactions (Pergamon into a spiral with a strip of glass Press, London, 1962).] tape to provide insulation between adjacent turns. The heater, as shown in Fig. 6-35, is operated only partially submerged at the surface of the oil. Even with a high rate of evaporation from the surface, the oil below the heater remains
relatively cool.
uses a pinhole camera technique, as illustrated in Fig. 6-34. His conclusion is that the remaining major source of backstreaming is the region of the jet around the edge of the cold cap.
Stevenson^' has demonstrated a pronounced effect of the shape of the upon the backstreaming rate of a diffusion pump. Backstreaming from the first-stage vapor jet may be thought of either
first-stage nozzle
always occurs at operating temperature, the decomposition rate depending upon a number of factors. In Sec. 6-4 various fluids are rated in terms of their relative ruggedness. In all cases, as one should expect, the decomposition rate increases rapidly with the boiler temperature. The heat transfer area in contact with the oil is an important design parameter which has recently received critical study. Several observers have noted the occurrence of eruptive boiling resulting from
fluid in
an
oil
diffusion
pump
an
for the
some of the oil molecules toward the pump expansion of a coherent supersonic vapor jet into a region of relatively low surrounding gas pressure, both of which processes are illustrated schematically in Fig. 6-36 from Stevenson's paper. The distribution of backstreaming vapor from the jet was measured by catching the condensed vapor as it drained down the wall of the extended pump casing into a series of gutters arranged at several heights above the pump nozzle, the first gutter being located at the same height as the lip of the pump nozzle, as shown in Fig. 6-37. The apparatus also provided means of draining off the oil vapor
as a
random
scattering of
266
267
Directional scottering
from
this
curves shown for the various nozzle configurations tested. The dotted curve shows the correction to the 45 curve due to subtraction of backstreaming from the lower jets. From these results Stevenson
Random scattering
concludes that the rate of backstreaming of oil diffusion pumps can be by improving the configuration of the nozzle. The paper does not give any data, however, on the effect on pumping
substantially reduced
o
c
o D
speed of various gases such as nitrogen and hydrogen caused by changmanner described. If the improvement in backstreaming results in a decrease in vapor-jet density, it may also result in a poorer ratio of pumping speed for hydrogen relative to that for air, which would be a distinct disadvantage for some applicaing the nozzle configuration in the
tions.
o o
The
is
rate
of decomposition
increased
by the
catalytic effect of
o
Return
line to boiler
1.0
Fig. 6-36. Nature of backstreaming. [Reprinted with permission from The Macmillan Company, from D. L. Stevenson, in 1963 Vacuum Symposium Transactions. Copyright 1963 by American Vacuum Society.]
Fig. 6-37. Schematic diagram of test apparatus for measuring amount and distribution of backstreaming. [Reprinted with permission from The Macmillan Company, from D. L. Stevenson, in 1963 Vacuum Sym,posium, Transactions. Copyright 1963 by
1 ' 1
' 1
' 1
1
>.
r1
1
: -
-30
1
*I5 -15
American Vacuum
Society.]
condensed on the top of the test chamber. The oil flow from each of the collecting surfaces was conveyed by a tube to a separate buret so that the accumulation during a specified period of time could be measured. The backstreaming total rate was measured as a function
of the angle of the lower
1.0
\\
I
NX.
A
member
of the nozzle, as
shown
Correction
in Fig. 6-38.
^
B
_ bosed
on effect of
The angle of the conical surface of the lower member of the nozzle was varied from +15 (protruding) to 45 (receding) relative to the^ vertical. Curve A was obtained with normal heater power input and curve B with heater input reduced about 18 per cent. The dotted curves are corrections to curves A and B due to a measurement of the backstreaming contributed by the lower jets of the pump. It is
apparent from these results that the total backstreaming rate is reduced by a factor of about 5 by changing the angle in question from + 15 to
installation of a baffle
.1,1,1
0
-15
+ 15
-30
-45
20
40
Angle, deg
60
Angle
of inner wall of
nozzle
Fig.
6-38. Relative
for
backstreaming
configurations.
Fig. 6-39. Distribution of relative rates of backstreaming as a function of the angle above the plane of the
The distribution of relative rates of backstreaming as determined by readings on the separate collection burets is shown in Fig. 6-39. The rapid decrease in backstreaming rate as a function of the
average angle above the plane of the nozzle
lip is
45.
Vacuum
Copyright
Symposium
Transactions.
1963
by
American
Vacuum
Society.]
diffusion-pump nozzle. [Reprinted with permission from The INIacmillan Company, from D. L. Stevenson, in 1963 Vacuum Symposium Transby actions. Copyright 1963
American Vacuum
Society.]
268
269
If this flow
is
work on the
radially inward
reservoir.
has been reported to permit a detailed discussion of the subject. However, the decomposition rate in glass pumps is apparently sigHot aluminum in nificantly less than in comparable metal pumps. However, undesirable. as regarded with the working fluid is contact are so fluids used of the properties differences chemical individual in
great that generalizations are not valid. The products of decomposition of diffusion-pump fluids consist of materials of both higher and lower vapor pressure than the original
fluid.
impeded by barriers with small openings, the fluid is heated substantially while it is still near the outer portion of the reservoir so that high-vapor-pressure constituents are boiled off near the outside. As the fluid flows toward the center it is further heated and lowervapor-pressure components are vaporized. The nozzle stack is constructed of concentric tubes arranged such that each nozzle receives
Those of
sufficiently high
gases in the sense that their vapor pressures are so large that they are
not condensed on liquid-nitrogen-cooled baffles. Other products are heavy liquids and solids of very low vapor pressure which accumulate in the boiler and eventually clog the nozzle system with a dark deposit. Because an appreciable decomposition rate is typical of diffusion-pump
operation, the ultimate pressure, even with good liquid-nitrogen-cooled baffles, is limited by the rate of decomposition and production of high-
vapor-pressure products which migrate into the high-vacuum system beyond the baffles and must then be pumped out again. Best results
in terms of very
pump
is
low ultimate pressure are obtained when a diffusion run with low power input and with a fluid of greater than
normal
stability.
and Purging. From the time high-boilingintroduced by Burch^" in 1928 for use instead of mercury in diffusion pumps, the need for continual purification to eliminate high-vapor-pressure components initially present in the oil, or produced during operation by decomposition, was realized. Hickman'i and his collaborators were largely responsible for the systematic study of decomposition and contamination of diffusion-pump fluids and the development of specific mechanisms for purging the pump6-8. Fractionation
first
Fig. 6-40. Two-stage fractionating glass diffusion pump. [Reproduced through the courtesy of ConsoHdated Vacuum Corp., Rochester, N.Y.]
specific
boiler.
The backing
and separating the remaining Figure 6-40 shows a two-stagb glass diffusion pump with boiler compartments to separate the fluid roughly according to the vapor pressure of the constituents and catchment lobes on the exhaust arm of the pump for elimination of high-vapor-pressure components into the backing pump. Large horizontal pumps of metal construction based upon the glass fractionating designs were developed but have not proved to be as convenient
boiler
of
undesirable
constituents
from the outer portion of the boiler where the vapor pressure is highest, and the first jet receives vapor from the central section where the vapor pressure of the fluid is the lowest. One advantage claimed for the mechanism of fractionation is that the high-vacuum jet is supplied only by the relatively low-vapor-pressure constituents of the working fluid, contributing to a lower backstreaming rate and vapor pressure at the inlet of the pump. Another advantage
or flnal jet receives vapor
claimed is a higher forepressure tolerance because of the relatively high vapor pressure of the constituents forming the backing or final jet.
in
in practice as
pumps
of vertical design.
have been incorporated into the design of metal vertical pumps. One of many such designs is illustrated in Fig. 6-41. Fluid returning from the jets to the boiler flows
principles of fractionation
The
However, these advantages have not been as clearly demonstrated commercial diffusion pumps of metal construction as in glass pumps
of the type fllustrated in Fig. 6-40, at least in part because of the process of reverse fractionation discussed by Hickman. ^^'^^
270
271
disadvantage in the construction of a metal fractionating pump is the low conductance for vapor flow to the nozzles inherent in the concentric tube design for the nozzle stack. Because of the relatively small gain
by the introduction of fractionathe limitations imposed by the presence of the concentric tubes against significant improvements to the
in ultimate
vacuum
actually achieved
pumps and
to gain
and jet system, the design trend has been away from fractionation some of the features described in the previous section. All
recent commercial
pump
and
Fig. 6-41. Three-stage fractionating oil diffusion pump of metal construction. Effective fractionation can only be obtained by careful separation of the respective boiler zones. This is obtained by shaping the component resting on the pump base plate as shown (shaded). The lower jet cap is at the same height asthe lower end of the cooling jacket. As a result, the oil flowing down the walls is
components according to vapor pressure is the problem of purging the pump fluid of high- vapor-pressure components by ejection into the forevacuum. Hickman^^ and Latham, Power, and Dennis^^ have demonstrated that complete ejection of the more volatile constituents of the working fluid from the pump is more effective than fractionation. The rate of ejection of volatiles in a design such as that shown in Fig. 6-9 is influenced by the vertical spacing from the bottom nozzle to the liquid level in the reservoir and the temperature of the pump housing on which the oil condenses. The temperature of the pump housing near the top must be cool to ensure efficient condensation of the fluid from the first jet. In some applications the top few turns of tubing are separated from the rest of the cooling coil and either cooled by chilled water just above the freezing point or by a mechanical refrigerator to reduce further the vapor pressure at the inlet of the pump. However, the wall temperature should preferably increase from the region of condensation of the first jet to a considerably higher temperature below the bottom jet so that higher- vapor-pressure components which are condensed near the top are evaporated as the
fluid into
warmed,
facilitating degassing.
Nollei-,
fluid drains
Vacuum
V, 59 (1955).]
The partial condensation of the vapor on the inner walls of the tubes supplying vapor to the nozzles in fractionating pumps, such as that shown in Fig. 6-41, is held to be responsible for a reversal of the desired direction of fractionation. Furthermore, the separation of the fluid into constituents according to vapor pressure is not as well controlled or efficient in the commercial metal pumps as in the glass fractionating pump, such as that shown in Fig. 6-40, since the concentrically divided boiler of the vertical metal pump is not the equivalent of the separated boiler compartments of the horizontal glass pump. Another serious
down the housing wall toward the backing jet and pumped out with the permanent gas into the forevacuum. The forevacuum section of the diffusion pump must also be allowed to run warm so that the more volatile constituents of the effluent will not be condensed and permitted to flow back into the boiler. The optimum temperature distribution is a compromise which allows a sufficiently high rate of ejection of volatiles from the pump without
permitting an excessive rate of loss of pump fluid into the forevacuum. For a given pump design the stability and vapor pressure of the fluid are factors which determine the optimum temperature distribution along the pump housing and forevacuum connection. An extreme example mentioned in Sec. 6-4 is OS- 124 (Monsanto Chemical Company)
272
for
273
which best performance was obtained by Batzer" when the lower end of the pump housing was allowed to run at 90 to 100C. his6-9. Resume of Diffusion-pump Performance. Although upon was based torically the original development of diffusion pumps
eff'ort mercury as the working fluid, since about 1930 the far greater diffusion has been devoted to understanding and improving "oil" oil pumps. With a few important exceptions listed in Sec. 6-4, nearly used on diffusion pumps instead of mercury diffusion pumps are certain inherent of spite In systems. electronuclear and industrial all
4.
5.
6. 7. 8. 9.
factor advantages of mercury, such as chemical stability, the speed mercury greater than for (6-37) for oil diffusion pumps is significantly
diffusion
pumps.
fluids,
usually referred to as oils, now available for use that about equal to in diffusion pumps range in vapor pressure from vapor mercury down to such low values that the room-temperature
Organic
Saul Dushman, Scientific Foundations of Vacuum Technique (John Wiley & ons, Inc., New York, 1949). W. Gaede, Ann. Physik 46, 357 (1915); Z. Tech. Physik 4, 337 (1923). R. Jaekel, in Proceedings First International Congress on Vacuum Technology, 1958 (Pergamon Press, London, 1960), p. 21. R. Jaekel, Kleinste Drucke (Springer-Verlag, 1950), pp. 140-197. H. G. Noller, Vacuum V, 59 (1955). N. A. Floresou, Vacuum 10, 250 (1960). E. H. Kennard, Kinetic Theory of Oases (McGraw-Hill Book Company, New York, 1938), p. 108, p. 194. P. Alexander, J. Sci. Instr. 23, 11 (1946). Smithsonian Physical Tables, 9th rev. ed. 1954, p. 40. D. Latham, B. D. Power, and N. T. M. Dennis, Vacuum II, 33 (1952). 188, 62 (1946). J. Blears, Nature 154, 20 (1944); Proc. Roy. Soc. (London) K. C. D. Hickman, Nature 187, 405 (1960). T. H. Batzer, in 1961 Vacuum Symposium Transactions (Pergamon Press,
16.
high-temperature pressure can only be estimated by extrapolation from hydromeasurements. Narrow cuts of petroleum oils, chlorinated been have carbons, and a wide variety of synthetic organic fluids high of In the development of synthetic fluids used.
successfully
17.
18.
London, 1962), p. 315. K. C. D. Hickman, in 1961 Vacuum Symposium Transactions (Pergamon Press, London, 1962), p. 307. B. D. Power, N. T. M. Dennis, and D. J. Crawley, in 1961 Vacuum Symposium Transactions (Pergamon Press, London, 1962), p. 1218. S. A. Vekshinsky, M. I. Menshikov, and I. S. Rabinovich, Vacuum 9, 201
(1959).
an molecular stability and low vapor pressure there would appear to be stable Highly opportunity for continued improvement in the future. to meet an fluids with a wide range of vapor pressures are needed extreme range in performance from high pumping speed at low ultimate pressure to high throughput at high backing pressure. The performance of diffusion pumps can be judged in terms of the
forebase pressure, backstreaming rate, speed factor, and limiting Several industrial vacuum firms have demonstrated difpressure. fusion pump designs for which
1.
19.
Calif.,
private
20.
21.
22.
23.
24.
25.
is
26.
using Silicone 705 or OS- 124 fluid. g/cm^ min. 2. The backstreaming rate is of the order of lO"" the order of 0.5. of is (6-37) 3. The speed factor as defined in Eq.
4.
27.
28. 29.
The
limiting forepressure
is
System designs should be based upon the assumption that diffusion pumps meeting approximately the above performance specifications
can be obtained.
30.
31.
32.
REFERENCES
1.
33. J.
communication T. L. Ho, Physics 2, 386 (1932). H. G. Noller, G. Reich, and W. Bachler, in 1959 Vacuum Symposium Transactions (Pergamon Press, London, 1960), p. 72. D. L. Stevenson, in 1959 Vacuum Symposium Transactions (Pergamon Press, London, 1960), p. 134. C. E. Normand, Oak Ridge National Laboratory, private communication. B. D. Power and D. J. Crawley, Vacuum IV, 415 (1954). H. R. Smith, in 1959 Vacuum Symposium Transactions (Pergamon Press, London, 1960), p. 140. M. H. Hablanian and H. A. Steinherz, in 1961 Vacuum Symposium Transactions (Pergamon Press, London, 1962), p. 333. M. H. Hablanian, in 1962 Vacuum Symposium Transactions (The Macmillan Company, New York, 1962), p. 384. Norman Milleron and L. L. Levenson, in 1961 Vacuum Symposium Transactions (Pergamon Press, London, 1962), p. 342. D. L. Stevenson, in 1963 Vacuum Symposium Transactions (The Macmillan Company, New York, 1963), p. 134. R. C. Burch, Nature 122, 729 (1928). K. C. D. Hickman, J. Franklin Inst. 221, 215 and 383 (1936). K. C. D. Hickman and J. J. Kinsella, in 1956 Vacuum Symposium Transactions (Pergamon Press, London, 1957), p. 52. K. C. D. Hickman, Rev. Sci. Instr. 22, 141 (1951).
M. LeBlanc, in L. Dunoyer, Vacuum Practice, trans, by Nostrand Company, Inc., New York, 1962), pp. 41-42.
London, 1958),
p. 88.
H. Smith
(D.
Van
2.
275
of surfaces reach a balance in a relatively short time, particularly for permanent gases, so that if Q (the admitted flow) is kept constant,
Qf,
CHAPTER
vacuum pumps,
the
pumping speed
decreasing pressure near the ultimate pressure Pq of the pump. Thus the ultimate pressure is determined not so much by the outgassing
in the
system as
it is
Pumping
is
Speed.
The
important
(or
whether there
0,
is
an
vacuum system
by
the
(7-1)
pump so
that
0.
Eq.
(7-1)
becomes
PS
in
dP
dt
Qo
dP
'dt
Qo
V
p -u9if
;7-4)
which
is
the pressure measured at some particular point in the the pumping speed at that same point, V is the volume of
or
S is
the effective
pumping
speed,
put
is
(e.g., in torr liters/sec) of gas flowing into the system, and Qg the gas flow due to interior surface outgassing. Thus, in general,
Observation of the pressure several times during the pressure rise which is Qo/ V. By putting the result-
the
pumping speed
is
^^-P^
VdP
Qo
and dPjdt 0, the ing value of Qo into (7-1) with Q speed at the ultimate pressure can then be determined from
On
If the
Sf, is
pumping
P
a
In order to measure the pumping speed, conditions are imposed on the system so that a simplified form of Eq. (7-2) is applicable. For example, if the outgassing rate is negligible {Qq = 0) and the system is operated at constant pressure with a steady flow of gas {Q = const) entering the system, then dPjdt = and from what remains of (7-2)
-;^
^0
(7-5)
system is free of accidental leaks then, for mechanical pumps, generally very small as compared with the value of S at higher
pressures
-I
as given in (2-1).
(7-3)
This expression is sometimes used as the definition of pumping speed and is a valid basis for measuring the pumping speed in the pressure range of most oil-sealed mechanical pumps, provided
the pressure measured actually corresponds to the gas admitted to the system, as discussed in Sec. 5-3. If air is admitted to the system at a measured flow rate Q and the pressure measured with a McLeod gauge, then because of the compression effect of the McLeod gauge, as described in Sec. 3-4, the pressure reading will be just the partial pressure of the permanent gas (in this case air) admitted to the system at the
Also, in the pressure regime of oil-sealed mechanical pumps the sorption processes (adsorption and desorption)
and can be neglected. In the case of diffusion pumps, however, the situation is quite different. As discussed in Sec. 6-5, the pumping speed of a diffusion pump is essentially independent of the pressure over the range in which measurements are normally carried out. The ultimate pressure is then not the result of the pumping speed approaching zero but of a limitation on the attainable pressure due to the outgassing rate Q^. The situation is frequently such that a base pressure is soon reached which then
changes only very slowly with the time because the outgassing rate becomes nearly constant. Returning to (7-1) for the case in which the system is operated at constant pressure, we have
PS = Q +
and
Qo
(7-6)
274
276
277
of (6-34)
(7-7)
which is unambiguous only if the pumping speed is essentially independent of the pressure. In this case it also follows that
In the pressure regime of diffusion-pump operation (7-9) and (7-10) do not generally apply because the outgassing rate is an important However, in this case the outgassing factor near the base pressure. rate rather soon reaches a nearly constant value, changing slowly enough that it may be regarded a constant during the period required In this case Eq. (7-1) can be written for a pumpdown test.
dP_
S
Thus
if the
PS +
Q,
(7-11)
P2-P1
Q
is
(7-8)
dt
equilibrium pressure
of the gas flow Q, in sequence, and then the slope of the curve is the
Fig. 7-1.
Still
pumping speed of a diffusion pump is independent of the since for periods of interest the outgassing rate Q may and pressure constant, the value of Q^ can be substituted from (7-6) regarded as be
Since the
so that
pumping speed
dP
speed can, in principle, be
another way in which the pumping determined is by observing the pressure as a function of the time as the system is pumped down. If the and the outgassing gas flow Q = rate Q^, is negligible, then from (7-1)
F
where
yields
P,, is
^ = -S(P at
Po)
Jp,
dP
Sdt
(7-9)
so that
In
that
re-
and
^1)
P1--P0
logi
(7-12)
^1
V
2.30
login
^
(7-10)
P,
When
the assumptions leading to Eq. (7-11) are vahd, then Eq. (7-12)
real situations to
which
this ideal
equation applies. In Fig. 7-1. Graph of throughput the regime of mechanical roughing against resulting pressure for determination of the pumping speed. pumps, the pressure usually changes so rapidly during the pumpdown operation that the outgassing rate is not negligible and is changing with time. Only if the volume of the system is very large as compared with the displacement speed of the pump, so that the pressure changes slowly during pumpdown, will observations based upon Eq. (7-10) check approximately with the known pumping speed of a mechanical vacuum pump. In this case the walls remain nearly in equilibrium with the pressure in the system and outgassing does not play an important role.
pumpdown
provides a basis for pumping-speed measurement which has certain Measurement of the gas flow is not required, and distinct advantages. the gauge constant cancels out, provided that the gauge is linear over the pressure range of the measurement. The volume of a system can generally be determined with fair precision from external dimensions. The procedure is simply to pump down the system to a steady base
pressure, then let in
enough air (preferably through a drying tube) to bring the pressure back up by a factor of 10 or more, and flnally to read the pressure at several specific values of the time as the system pumps down again. This procedure works well on large systems,
particularly
is
not so
7-2.
when the rate of change of pressure during pumpdown rapid as to make pressure readings difficult.
of
Measurement
Gas Flow.
Pumping-speed measurements
for determining
vacuum-pump performance are predominantly carried out under conditions of constant flow. The full range over which gas
must be controlled and measured for the routine measurement of pumping speeds of commercial vacuum pumps extends from about 10"
flow
278
279
is normally admitted through a an appropriate test dome mounted on the pump and the flow rate controlled to a value at which the desired pressure is maintained. The gas flow must be measured in such a way that the volume per second and the pressure are both known, so that In those cases in which the throughput Q = P{dVjdt) is determined. no significant pressure drop occurs in the flow-measuring device, the
pound. The parameter c is the nozzle coefficient and generally has a value close to unity. Since from Eq. (1-8) the gas density is
W
V
MP
R,T
is
progress.
is the barometric pressure when the test is in In those cases in which there is a pressure drop, the value of the pressure at which the flow dVjdt is measured must be determined. Some of the devices commonly used for throughput determination in
pressure of interest
dt
p dt
=
by
substituting R^
air,
4904c|
Ml
Z)2
cm^/sec
1-2.
(7-15)
various flow ranges will be briefly described. Calibrated Orifice. For large flow rates a calibrated orifice connected to the test dome through a gate valve is convenient and
=
dY
dt
For
20C
for
which
M = 28.98,
15.6
at a
atmospheric pressure or any desired gas maintained at a controlled pressure in a tank upstream from the orifice flows into the system at a rate determined by the diameter of the
reliable.
The surrounding
air at
=
=
103cZ)2
cm^/sec
(7-16)
and the temperature, average molecular weight, and pressure of the gas upstream of the orifice. The flow rate is independent of the pressure downstream from the oriflce, provided that the pressure is less
oriflce
15.60c7)2
liters/sec
when
D is
measured
in centimeters, or
than the critical pressure given in Eq. (6-1), as in the case of the flow of steam through an ejector nozzle. For the common diatomic gases y = 1.40, so that the critical pressure from (6-1) is
P,
dV
dt
213.3cZ'2
cfm
(7- 16a)
0.535Pi
(7-13)
where P^
The
critical
given by
the pressure upstream from the orifice. mass flow through an oriflce under these conditions Eq. (6-7), which for a diatomic gas becomes
is
is
D is measured in inches. For example, the approximate critical flow of atmospheric air through an orifice of 1 in. throat diameter under the above conditions is 100.6 liters/sec or 213.3 cfm, obtained by setting c = 1 in the above equation. If atmospheric pressure is 760 torr, then the throughput for air through
when
a
1-in. -diameter orifice is
dM =
HI
when the
pressure
is
0.538c(Pipy^'D^
g/sec
(7-14)
= P =
dt
760 X 100.6
7.65
measured
760 X 213.3
1.62
in centimeters
dW
or
dt
-^ =
19.64c(Pip)'^^Z2
g/sec
(7-14a)
In practice the values of the temperature and pressure at the time of the measurement must be substituted in (7-15). The exact value of the nozzle coefficient depends upon the flow For testing the conditions and the detailed shape of the orifice.
"capacity" or pumping speed of steam ejectors, the Heat Exchange Institute! * has developed a standardized long radius orifice, the design
when
the pressure
is
measured
in torr.
Finally,
dW
11
in the usual engineering
915c(Pi/Fi)'^i)^
Ib/hr
(7-146)
of which
is
shown
in Fig. 7-2.
pumping-speed
inch,
form in which D is in inches, the pressure is in and F^ = 1/p is the specific volume in cubic
chapter.
280
281
measurements is shown in Fig. 7-3. The above reference also gives tables and graphs of flow rates in pounds of air per hour for a sequence of standard nozzles ranging in throat diameter from 0.0625 in. to 1.0 in. Table 7-1 contains a sample of the data given in one section of the above reference with the flow rates in pounds per hour as in the original, and also for convenience
with the flow rates converted to torr cubic feet per minute and These to torr liters per second. values differ very little in general from those calculated from (7-15)
-Manometer
*-
Control volves
by assuming that
1,
indicat-
ing that for flow in the range of interest for vacuum applications
the nozzle
coefficient
for
is
the
shape
very
Fig. 7-3. Arrangement of standard orifice for critical air-flow tests. [Reprinted from the Standards for Steam Jet Ejectors, 3rd ed. Copyright 1956 by the Heat Exchange Institute, 122 East 42nd Street, Xew Yorlc, N.Y. 10017.]
nearly equal to 1. The tables and curves permit accurate determination of the critical flow rates
Table
7-1.
Critical
7-2. Proportions of standard Fig. long-radius flow nozzle recommended by the Heat Exchange Institute for [Reprinted gas-flow determination. from the Standards for Steam Jet Copyright 1956 by Ejectors, 3rd ed. the Heat Exchange Institute, 122 East 42nd Street, New Yorli, N.Y. 10017.]
Dry Air
at 70F
30.0 in.
Hg =
762 torr
Density
Orifice
0.07510
flow
Ib/ft^
1.203 g/liter
Throi] ghput
torr
listed in
is
of standardized orifices Table 7-1. Also in Ref. 1 a detailed method given for computing the nozzle
set
Volume
in.
Mass
flow Ib/hr
3.65 8.18
14.6
33.1
HEI
no.
1
diameter,
0.063 0.094 0.125 0.188 0.250 0.313 0.375 0.500 0.750 1.000
1.250 1.500
cfm
0.810
1.82
liters/sec
cfm
coefficient for
an
orifice
of any
0.382
0.857
1.53
throat diameter, but of the HEI standard shape. The flow values
2
3
3.24 7.35
13.0
4
5
6
7
given in Table 7-1 are extended to orifices of larger dimensions than those of the HEI series. In using standard HEI orifices for controlling and determining flow rates it is most important to note whether the pressure in the system just downstream from the orifice satisfies
58.8
9.94 X 103
1.56 X 10* 2.25 X 10* 4.01 X 10*
20.5
29.5
P3
from
(7-13),
<
P^
0.535Pi
9 10
52.6 119
24.8 56.0
100.1
212
332 476
956
1497
21.50
where
P,
(^
760 torr
156 224
If this condition is not fulfilled, then the flow is subcritical and depends on the pressure difference across the orifice instead of simply the upstream pressure. The expression for subcritical flow is
* Reprinted from the Standards for Steam Jet Ejectors, 3rd ed. Copyright 1956 by the Heat Exchange Institute, 122 East 42nd Street, New York, N.Y.
dW =
10017.
t Mass flow data for Pig. 20, p. 23.
19.64c(Pi
P^y-^'p^'^D^
g/sec
(7-17)
HEI
standard
orifices
Xos.
through 10 given in
ibid.
282
283
when
is
in torr.
915c
D^
V,
Ib/hr
(7-17a)
when
is
the pressure
is
measured
pounds per square inch, V^ = l/p cubic feet per pound, and D is measured in
in
with a rotating float inside. The height of the float is determined by the gas-flow rate upward through the tube. Rotameters are calibrated by the manufacturer and can be obtained in a wide range of sizes. The throughput range from about 150 torr liters/sec to about 1 x 10* torr liters/sec can be covered conveniently with overlapping ranges with a set of four units. Care must be exercised to ensure that the rotameter is accurately vertical or the float will rub on the side of the tube, producing erratic readings.
inches.
1.0 [-
0.99 -
0.98
t-
-5^ 0.96
0.95
r
-
-^ U.94
nn
Nozzle dio
Z.0.5003
.--'
:= __ = =
,
"""
= ^" ^
Inverted Buret. This term is used here to designate any of several versions of a simple device which in its original form consisted of an
inverted buret thrust into a beaker of diffusionpump liquid, as shown in Fig. 7-6. The outlet at the top of the buret may be connected to a T in a
..^
'/
>
y'
,0^
0^
'
f
.^
c->
^
OJ
0.91
-f-
o^-
S o
0.89
0.88 0.87
nflfi
/
/
J
tube leading to the needle valve, the other leg of the T opening to atmosphere through a stopcock, so that when the stopcock is closed the gas flow sucks oil up into the tube. To a first approximation the rate of rise of the oil level in the tube multiplied by the cross-sectional area of the tube
a direct measure of the flow rate. However, in this simple arrangement the pressure on the gas above the oil level is continually decreased by the difference in head of oil in the buret and in the
is
/
100
"""
200
300
400
500
600
700
800
900
1,000
1,100
1,200 1,300
Flow, pounds per hour per square inch of nozzle throat area
Fig. 7-4. Nozzle coefficients for standard orifices. Dry air at 70F. [Reprinted from the Standards for Steam Jet Ejectors, 3rd ed. Copyrigiit 1956 by the Heat Exchange Institute, 122 East 42nd Street, New York, N.Y. 10017.]
must be made.
Since the density of a light oil is only about Ks that of mercury, the reduction in pressure of the gas entering the pump when the oil level in the
The expressions
is
for critical
and
buret
orifice
at least four times the throat diameter of the orifice. Values of the nozzle coefficient c have been determined for the standard orifices by a series of tests conducted by the Heat Exch3,nge
7-4.
that in the reservoir is only about 20 torr in a normal atmospheric pressure of 760 torr, or a pressure reduction of about 2.6 per cent. The pressure of the gas entering the
is
30
cm above
The rotam-
pump
a
tall
Some results of these tests are shown graphically in Fig. Inspection of this set of curves and comparison with the flow rates shown in Table 7-1 for critical flow will reveal that the range of values of the nozzle coefficient of interest for most vacuum work is 0.0625 in.) to 0.99 for the from 0.94 for the smallest orifice {D
Institute.
can be kept constant by using as a reservoir cylinder which is moved up around the
flowmeter. [Reprinted through of courtesy the Fischer and Porter Co., Warminster, Pa.]
Timing the rise in level between two marks oil level rises. and manipulating the oil reservoir to keep the levels inside and outside the buret the same is somewhat inconvenient, but can be managed. If
buret as the
this is
largest
shown on the graph (D = 1.00 in.). Rotameter. The rotameter (see Fig. 7-5), which is a variable orifice device, is a vertically mounted tube of tapered precision bore
the
not done, then the pressure of the entrapped gas decreases as such that
poil
(7-18)
png
284
285
large systematic
where P
h
pjjg
poii
= atmospheric pressure = difference in height of the oil levels = density of mercury = density of the oil in the buret
and use of the simple expression (7-22) will lead to a error in the measurement of the flow rate.
after each
V represents the volume of gas at atmospheric pressure initially trapped above the oil level at the instant the bypass valve is closed, then when the oil level has risen h cm, the volume is
If
In the use of the inverted-buret flowmeter the bypass valve is opened measurement and the oil level in the measuring tube drops back down to that of the reservoir. Before closing the valve and starting another measurement it is essential to wait long enough for
the
initially left
V
in
=Vo-hA
(7-19)
the cross-sectional area of the buret. The total quantity of gas in the apparatus above the oil level is the product of (7-18) and
which
is
(7-19),
To needle valve
which
is
PV
PaVo
PAh
poll
(Voh
Pag
-h^A)
(7-20)
on system
then
e=^(PF
PA
i*:
PV)
Poll
{V,-hA)
(7-21)
PHg
iisi;
down from the wall of the tube. Since the surface film behind on the tube wall can, in effect, reduce the crosssectional area of the tube significantly, large errors in the flow measurements can result from this effect, particularly when the diameter of the measuring tube is small. This type of error can be reduced considerably by using a liquid of low viscosity. Since air cannot have a larger content than that corresponding to 100 per cent relative humidity (Ph o ^ 18 torr at room temperature or ~2.4 per cent of atmospheric With pressure), water is for some purposes a better choice than oil. as throughput a 0.2-cm3 buret tube calibrated in units of lO-^^ cm^ a accuracy. acceptable low as 10"* torr liter/sec can be measured with The importance of variations in room temperature in measuring small flow rates by the inverted-buret method is emphasized by Dayton. A change of temperature of 3C results in a 1 per cent change in volume of a body of gas at constant pressure. During the measurement of a small throughput the rise in the oil level may be completely masked by a change in volume of the gas entrapped between the oil level and the needle valve. An example given by Dayton is quoted below
oil
to drain
Light
WM
inverted-
If h
is
measured
t
in centimeters,
in square
centimeters,
Fig.
7-6.
in seconds,
is
The
resulting flow
For example, a 50-cc buret is a convenient size for testing 4-inch pumps having peak speeds of about 200 liters per second at pressures from 10"^ to 10-2 mm. But in the range from 10"* to the ultimate vacuum of about the leak rate drops below 1 cc in 10 minutes. While the oil 3 X 10"*
mm
meters per second. If the volume hA displaced by the ofl is term in (7-21) becomes very small correction the nearly equal to V^, so that
Q
This
is is
(7-22)
the case
of the buret
is
to the 1-cc mark in 10 minutes, if the room might normally rise from temperature increases by 3C and about 50 cc of air are trapped above the The oil, the expansion wiU force the oil back down the tube by 0.5 cc. measured leak rate is thus 0.55 cc in 10 minutes, which is just one-half the true leak rate. A rise of only 0.3C in 10 minutes will cause an error of Of course, the ambient temperature of the buret about 5% in this case. could be held constant by a water jacket, but the author has found it more convenient to hang a thermometer near the buret and note the temperature
and the volume of the connecting tubing However, if hA < Fo, small.
and stopping the stop watch. Usually the room temperature can be adjusted so that the final temperature is within O.TC of the initial temperature.
at the times of starting
Q
In
this case,
PA + AA
^
PHg
F,
(7-23)
.)
which
is
more
term
may be
quite large
The pipe organ is a compact multiple-tube arrangement, based upon the principle of the inverted buret, designed and built at the Kinney vacuum laboratory for measuring gas flow over a wide range.* The apparatus shown in Fig. 7-7 consists of two
286
28/
The volume of the connecting manifold is small so that the liquidlevel compensation makes a perceptible difference only when the smaller
Selector volve
(typical)
Ven1 valve
measuring volumes are being used. By applying a correction of the type given in (7-23) to these smaller measuring volumes, the compensation system can be eliminated entirely, thus simplifying somewhat
the operating procedure. In Fig. 7-8 is shown a sketch Constant-pressure Gas Flowmeters. of a constant-pressure flowmeter described by Stevenson'* in which a
To pump under
test
Sight glass
Calibrated
measuring
cylinder (typical)
Oil or
other low-
vapor -pressure
fluid
group of five calibrated tubes of widely differing volumes are used for measuring the flow rate. The novel feature of this instrument is the upper reservoir and overflow arrangement by which the tube in use is filled by oil flowing in near the top instead of by suction from below. The level in the upper reservoir is maintained by pumping oil from the lower to the upper reservoir, from which it returns to the lower reservoir by way of an overflow. The spout for supplying oil requires the oil level to rise only a limited amount to spill over into the measuring
tubes.
Drain plug
Section A-A
The
in the
static
Maximum
correction throughout
groups of concentric, vertical metal tubes with several valves at the top so that the volume of each central tube and each of the several annular spaces between the tubes can be separately, or in combination, connected to the manifold at the top. A large rectangular reservoir at the base provides an ample volume of the fluid. Two glass sight tubes are mounted on the front of the instrument. One runs from the reservoir in the base to the manifold at the top and is connected in parallel with whatever combination of tubes is being used The liquid from the reservoir rises in this for measuring the flow rate. volume and provides a means of measuring tube as well as in the transparent tube protrudes into other The observing the liquid level.
is open at the top. By the atmosphere, normally open to which is means of a two-way valve, the volume above the liquid surface in the reservoir can be connected to a compressed-air supply and the pressure raised as necessary to maintain the liquid level in the open sight tube the same as that in the measuring volume. When all measuring volumes are connected into the manifold, the total volume displaced is about 2.3 liters, which enables the operator By using only to measure throughput up to about 200 torr liters/sec. the liquid-level indicating tube as a measuring volume, flow rates as low as 0.1 torr liter /sec can be measured.
measured flow due to pressure difference does not exceed 0.5 per cent on any of the burets. The practical range of throughput measurement is from about 50 to about 3 x 10-^ torr liter /sec. As in the simple inverted buret, an oil film remaining on the inner surface of a measuring tube due to a previous operation can cause significant
error, especially in the use of the smaller-diameter tubes.
Sufficient
Overflow
column
Stevenson, in 1961
1962).]
Vacuum Symposium
288
289
time for the oil to drain back into the reservoir must be allowed between measurements. Also reasonably constant room temperature is essential to avoid error due to expansion. Another form of a constant-pressure flowmeter, which is particularly effective for the measurement of small flow rates, is shown in Fig. 7-9. This instrument* consists of a glass capillary tube of 0.1 to 0.3 cm bore in an approximately horizontal position with a drop of mercury If a small inserted at one end. previously been estabflow has gas Needle volve Colibroted copillory lished through the tube, the merBaffle, Air ond cury drop moves along the tube mercury inlef at the flow rate of the gas with Test dome
connecting to the vacuum system. When the system is pumped down against the air leakage through the capillary from atmosphere, the leak
rate can be adjusted
by breaking
is achieved. The system must have a valve at the inlet to the vacuum pump, and the volume isolated from the pump by the valve must be known. Then by closing the valve and noting the rate of pres-
sure
rise,
Controlled leok
very
McLeod
gouge
Mercury trop
little
frictional
is
drag.
The
Mercury pellet
Monometer
Vacuum pump
Fig. 7-9. Gas flowmeter consisting of a calibrated glass capillary with a traveling mercury pellet.
mission from C.
Vacuum
(Pergamon
expanded into a funnel at one end into which a drop of mercury can be inserted with an eye dropper whenever a measurement is to be made. The other end of the capillary is expanded into a normal-diameter
capillary
can be adjusted and calibrated on a laboratory vacuum system and then used for pumping-speed measurements. By this technique capillaries of leak rates as low as 10~^ torr liter/sec can be prepared. The
principal
1^
Metering
tube
Liquid
|_Jlon
gauge
Hz
Pz
i-2-in.pump
4-
in.
pump
To forepump
To forepump
drawback
in the use of
is
calibrated capillaries
their ten-
1.27cm
D,=
)0.5cm
L= 30.5 cm
D2=5.3 cm
by dust
particles.
By
enclosing
London,
1957).]
small reservoir on a T, in which to accumulate the mercury droplets which come through the tube. Although the pressure difference required to move the mercury drop in the capillary is small, the true pressure of the gas entering the system can easily be measured by means of a manometer connected to the connecting T near the mercury trap, as shown in Fig. 7-9. The crosssectional area of the capillary can be checked for uniformity by noting any variation in length of a mercury drop in passing along the tube and can be measured by weighing several mercury fillings of a marked length. Flow rates as low as about 3 x 10~* torr liter/sec can be measured with acceptable accuracy. Calibrated Capillary Leaks. The trick of preparing very fine capillaries is known to many familiar with glassblowing techniques. A short length of glass tubing is softened in a flame and allowed to thicken and contract. When the central part of the tube is nearly a solid rod, the bore having contracted to a fine line, the tube is removed from the flame and quickly stretched out at arm's length. The whisker of glass connecting the two ends of the tubing will then be a capillary of small diameter and microscopic bore. The fine capillary can be broken off at one end but left attached at the other end of the normal-diameter tubing from which it was drawn to be used for
each capillary in a rugged glass Fig. 7-10. Arrangement for determining gas flow by measuring the sleeve plugged by a filter of glass pressure drop across a known conwool, the tendency to clog up is ductance. [Taken with permission reduced and the capillary is pro- from A. A. Landfors and M. H. HabEven lanian, in 1958 Vacuum Symposium, tected against breakage. with the best protection, however, Transactions (Pergamon Press, London, 1959).] the leakage rate of a capillary can change because of clogging so that a periodic check of the leakage rate is essential to ensure
reliability.
Known
Conductance.
The measurement
most
effectively accomplished
difference across a known conductance form of a tube or aperture.''* The gas entering the test dome of a vacuum pump flows through a tube or aperture of known dimensions from an auxiliary chamber, as in Fig. 7-10. A needle valve for admitting the test gas and a separate diffusion pump are connected to the auxiliary chamber so that the pressure Pi can be adjusted to any desired value. If the pressure is sufficiently low that the molecular mean free path is not too small compared with the diameter, the gas flow through the interconnecting tube into the test chamber is then, ac-
cording to (2-4),
Q = C{P, -
P,)
290
291
C ^ 3.810y
when
the diameter
If,
Z)3
L +
%D
hters/sec
centimeters.
and the length L of the tube are measured in instead of a tube, the conductance is in the form of an
Pg) is less than the error in either pressure reading alone. At very low values of the pressure, such as lO"" torr as observed in Ref. 6, however, significant gauge errors, even for the pressure differences, are to be expected. Errors of this type due primarily to sorption and decomposition effects in the gauges can be
significantly reduced by using the noble gases such as helium, neon, and argon, for which these effects are far less bothersome than for the
C
in
3.641
^J A
Ml
Hters/sec
the area of the aperture in square centimeters. or aperture is accurately known from dimensions which can be measured with great precision. The principal error in determining the throughput by this means is in the determination of Pj, the pressure in the auxiliary chamber, since generally Pg Pi for cases of interest. The uncertainty can be
which
is
Over most of 7-3. Mechanical Pump Speed Measurements. the operating pressure range of mechanical vacuum pumps the molecular mean free path is short as compared with the diameter of the
Leak valve
To
fl owmeter
-~. To flowmeter
2
min
I
I
"Tt
Mini D
<
10 D max
reduced substantially if the parameters are so chosen that P^ is in any case greater than 10"* torr so that an accurate determination using a McLeod gauge is possible. However, in order to maintain the advantage of using the simple conductance formulas given above, the parameters must be chosen to ensure that the maximum value of P^ does not exceed the lower limit Pj of the transition pressure range, as discussed in Sec. 2-7, for which values for air at room temperature This requirement is approximately equivalent are given in Table 2-1. to the requirement that the molecular mean free path not be less than
the diameter of the metering tube. Thus, for example, the metering tube diameter should not exceed 0.5 cm if the pressure Pj is to be as high as lO-^ torr. If the length of the tube is 30 cm and its diameter is 0.5 cm, then from (2-97) the conduct-
Vacuum gauge
min
I
-Vacuum
min
gouge
L_i_:
Pump
(a)
Pump
(b)
Fig. 7-11. Arrangements for measuring the pumping speed of oil-sealed mechanical vacuum pumps, (a) For inlet larger than 2 in. inside diameter; (6) for inlet smaller than 2 in. inside diameter.
The 4.94 X 10-2 liter/sec for air at room temperature. C highest pressure for which this conductance value is valid is lO^^ torr, 4.94 x 10~* torr so that the maximum flow rate measurable is Qmax
ance
Gas flow is therefore viscous and the conductance into the compared with the molecular-flow value. The conductance of any reasonably proportioned test dome connected to the pump inlet for the purpose of making pumping-speed measurements will therefore be very large over most of the pressure range of interest and will not seriously affect the results. However, some precautions
inlet.
pump pump
large as
liter/sec,
provided
Pj
< Pj.
The
lowest
flow
accuracy using a McLeod gauge (Pj = 10"* torr) is torr liter/sec using this same metering tube. By using ionization gauges instead of a McLeod gauge much lower flow rates have been measured by this method, as described in Refs. 5 and 6. Although reliance upon ionization gauges for absolute measurements is open to criticism due to the difficulties described in Chap. 3,
the measurement of the throughput, described above as applied to pumping-speed determination, is a relative measurement, Pj relative to Pj in a sequence of observations. Since in such a system the base pressure for zero flow is nearly the same for both gauges, the error in the
Typical arrangements for measuring the pumping speed of a mechanivacuum pump with constant gas flow are illustrated in Fig. 7-11. The system consists of a test dome of the same diameter as the pump inlet and of height equal to at least 1.5Z). The connection for attaching a McLeod gauge should be perpendicular to the wall of the test
cal
dome
at a distance of
inlet.
1 in.
or 0.25i), whichever
is
larger,
above the
pump
test
at a distance
dome
pump inlet near the closed end of the pump inlet through a small tube bent so
that the air flow is directed toward the closed end of the dome. This precaution is necessary only when the pressure is less than 1 torr to
292
293
avoid the possible formation of a supersonic gas jet directed into the pump inlet, the occurrence of which may result in fictitiously high values of the pumping speed. For pressures above 1 torr this precaution is
However, the gas inlet should not be oriented directly not necessary inlet. the gauge toward Oil-sealed mechanical pumps all tend to splash or spit some oil out the inlet into the test dome. In some cases this condition can be so troublesome that the gauge connection becomes clogged and the gauge
Table
7-2.
Table 7-3. Instruments should be chosen to provide some overlap in ranges to minimize errors in transition from one instrument to the next as the throughput is changed. In testing mechanical pumps by the constant-flow method it is found that consistent results are most readily obtained by first pumping down the test system to an acceptable base pressure and then increasing the throughput from zero upward, taking pumping-speed readings at
7-3. Flow-measuring Means Recommended for Use in Determination OF Pumping Speeds of Oil-sealed Mechanical Vacuum Pumps*
Table
Pkbssube Gauges Recommended fob Use in Determination of Pumping Speeds op Oil-sealed Mechanical Vacuum Pumps
Pressure range, torr
1
Flow range Q
Flow-measuring device scfm
0-0.2 0.1-100
torr
Type
oj gauge
cfm
Inverted buret
X 10-*-0.5
10-2-5.0
0-150
76-7.6 X 10*
15,200 and
0.1-100 1.0-760
20 and up
up
of
oil.
some baffling in the test dome between the pump inlet and the
* Orifices
critical flow
(i.e.,
test
dome
to prevent
oil
accumulation
torr).
P >
from interfering with pressure measurements. Pressure measurements should preferably be made with closed mercury U-tube manometers and McLeod gauges connected directly The ranges to the test dome without a low-temperature vapor trap. of the pressure gauges should overlap and have full scale readings no greater than the upper range of the pressure limits given in Table 7-2. As has been discussed in Sec. 5-3, the objective of the pressure measurement is to determine the partial pressure of the gas admitted as the measured throughput Q, unaffected by the vapor pressure of the sealing Total pressure gauges, such as a precision U-tube oil in the pump. manometer, should not be used for pressures below about 1 torr, since the pressure of a poor quality or contaminated oil can easily be as great as 0.05 torr without affecting the pumping speed of a mechanical pump significantly. The McLeod gauge (see Chap. 3) has the distinct advantage of responding perceptibly only to the permanent gas under these conditions and should therefore be used for all values of the pressure less than 1 torr. For all but the largest flow rates, air should be admitted to the system through a properly chosen needle valve after passing through the flowmeter. When calibrated orifices are used, a gate valve should be installed between the test dome and the orifice mounting to facilitate changing orifices without shutting down the pump being tested. Approximate ranges for various flow-measuring devices are given in
rate should be
computed
as outlined in Ref.
1.
pressure
is
it is
by
By this procedure the base reasonable and the system is the time the preliminary pump-
completed the outgassing rate will have dropped to a low enough value not to affect the results measurably. A typical set of data of this type is shown in Table 7-4 and shown graphically in Fig. 7-12. A smooth curve is drawn through the data points as representative of pump performance. The combined uncertainty in the measurement of the gas throughput and the attending pressure result in errors of the order of 5 per cent in pumping-speed deis
down
termination, when all reasonable precautions are taken. 7-4. Measurement of the Pumping Speed of Diffusion
Pumps.
pressure range of interest for diffusion-pump operation is generally less than 10-* torr, the pressure range in which the molecular mean free
path is of the order of, or greater than, the diameter of the pump barrel. In this pressure range the geometry and dimensions leading into the pump inlet affect the resulting pumping speed critically. For example, adding a tubular extension of the same diameter as the pump barrel and of length equal to three times its diameter to the inlet of a diffusion pump will reduce the net pumping speed to about half that measured
294
295
Table 7-4. Typical Set of Data for Determining Pumping Speed as a Function of Inlet Pressure for a Single-stage Mechanical Vacuum
Pump
Inlet
Throu ghput
torr
P umping
cfm
speed
pressure,
torr
directly at the pump inlet. The configuration of the test dome connected to the pump inlet, the location and orientation of the gas inlet into the test dome all influence the measured value of the pumping speed more or less critically. In a significant paper Dayton^ has discussed directional effects in
cfm
torr liters/sec
liters/sec
m^/hr
pumping-speed measurements.
An
is
method
0.007 0.010
0.021
criticized
by Dayton
illustrated in
0.149
0.821
2.01
0.0703 0.389
0.95
1.97
15.0
7.08
18.5
25.5 66.6
39.2
67.0 83.5
31.2 39.4
4.18
6.50
3.07
93.0
96.6
43.9
9.66
21.9 55.3
4.56
10.3
45.6
51.8 52.2
109.5
26.2
117.3
3.0
5.0
7.0
10
436 604
1,244 1,866 2,967
114 142 158 164 186 189 199 212 234 224 217 224 224 214 221 217 222 221
80 70
and
compared with a cosine pattern (dotted [Taken with permission from B. B. Dayton, in 195G Vacuum, Symposium Transactions (Pergamon Press, London, 1957).]
L = 5D
arrangement there
inlet flange of the
Air inlet
dome, but simply a flat plate bolted to the with connections for vacuum gauges and the gas inlet. Gas molecules flowing from the inlet tube form a molecular beam in which the molecular velocities are far from randomly oriented. The angular distribution of molecular velocities resulting from gas flow through a tube of length equal to five times its diameter, as calculated by Dayton,^ is illustrated in Fig. 7-14, which must be
is
no
test
pump
10"'
10"'
10
10'
10^
Diffusion punnp
compared with the normal cosine distribution (dotted curve) charrandom molecular motion. The angular distribution is even more forwardly directed if the length of the inlet tube is many
acteristic of
Fig.
7-13. Incorrect
measurements
on
an
oil-sealed
measuring [Taken with permission from B. B. Dayton, Ind. and Eng. Chem. 40, 795 (1948).]
The gas molecules preferentially directed into the effectively removed from the system than are randomly directed molecules. The result is that the pseudo pumping speed measured by this incorrect method is invariably much greater (such as a factor of 2) than that obtained by recommended methods
times
its
diameter.
pump
jet are
more
296
297
which do not involve beaming of the gas flow into the pump inlet. As Dayton^ has also pointed out, pumping-speed measurements can
easily be in error pessimistically if
or that at
The
't^'T^'t:^
150 liters/sec
280
liters/sec
1
180
liters/sec
"^
or
230
liters/sec
the gas inlet is so oriented that a gas jet is directed into the gauge opening. The position and orientation of the gauge opening also has a striking
effect
Fig. 7-15. Effect of position and orientation of the gauge opening on the measured pumping speed of a
speed as illustrated in Fig. 7-15 from Dayton's paper.^ This striking effect arises because the flux of molecules the number per second per (i.e., [Taken with perdiffusion pump. square centimeter) through a small mission from B. B. Dayton, Ind. and Eng. Chem. 40, 795 (1948).] opening placed above the inlet to the pump is not independent of the The distribution of the molecules is not orientation of the opening. isotropic because they are coming generally from above, where they are
admitted, and disappearing below, where by the diffusion-pump jet. Particularly in the pressure range below 10-* torr, where the molecular
240
lifers/sec
opening oriented directly away from the pump inlet, will receive the same flux of molecules as does the pump itself and will therefore produce a pressure reading equivalent to that received by the pump. However, the question is not so much the pressure incident on the pump inlet, but the pressure associated with the random motion of the molecules just above the pump inlet, uninfluenced by the added increment of flux due to the downward flow. Dayton reasoned that the gauge connection at B, with the plane of its open end parallel with the direction of flow, would receive no additional
flux (positive or negative)
flow,
molecules,
just the correct flux to produce a reading properly characterizing the pressure above the pump inlet. In addition, a nude gauge C of the type
many
of
them
are
removed
recommended by Blears (see Sec. 6-4) was also included in the test dome as an alternative means of pressure measurement. The signiflcant differences between nude and tubulated ionization gauges are discussed in some detail in Chap. 3. Because of the critical dependence of the measured pumping speed of a diffusion pump on the arrangement of the test dome and its accessories, adoption of a standardized test dome and procedure is
needed so that the performance of different models of diffusion pumps can be specified unambiguously. A preliminary step toward meeting this need was taken by the Committee on Standards and Nomenclature of the American Vacuum Society." The committee recommended that the test dome to be used for this purpose have
1.
mean
path is very long, this pronounced. Since the pressure measured by the gauge is proportional to the flux of molecules entering the gauge opening, the gauge reading reflects the nonisotropic distribution, giving a higher reading if the end of the gauge tube is turned upward toward the top of the test dome and a lower reading if it is turned
free
is
effect
2.
Inner diameter D equal to that of the pump barrel. Height (face of diffusion-pump flange to the closed end of the
least 1.5Z).
dome) of at
3.
Exit of gas-inlet tube located on the axis of the test dome 1.0Z>
As
of pumping-speed measurements
Fig.
dome arrangement for speeds. diffusion-pump measuring [Taken with permission from B. B. Dayton, Ind. and Eng. Chem. 40, 795
Fig. 7-16. Test
(1948).]
above the face of the pump flange and oriented directly toward the top of the dome. 4. Gauge-connection tube oriented so that the plane of the open end is parallel with the axis of the test dome and located so that the open end is just above the top surface of the pump and protruding inward from the wall of the test dome about 0.25 in. to avoid fouling by diffusion-pump fluid condensed on the wall. Such a test dome is illustrated in Fig. 7-17.
7-16 for testing diffusion pumps. The gas flow was admitted at the
top of the test dome through a series of circular baffles so placed as to break up the molecular beam from the inlet tube and distribute the flow randomly. The question arose whether the gauge connection
The committee also recommended that the gauge not be connected through a low-temperature vapor trap and that the gauge be calibrated before and after each series of measurements by inserting a
298
299
standard conductance between the pump and the test observing the resulting change in pressure.
dome and
The principal difference between the recommendations of the committee" and those made earlier by Dayton^ is in the choice of the
diameter of the test dome.
Although the recommendations contained in the committee's report have not been officially adopted by the American Vacuum Society, the testing of diffusion pumps in commercial practice generally follows
the
to the four items listed Choice of gauge, use of refrigerated traps, method of flow measurement, and the like are still questions left to individual choice. However, the procedure as practiced at the Kinney Vacuum Laboratory may be considered as an example and is illustrated in some detail in
By
Sloped top
choosing the diameter equal to that of the pump barrel the problem of the entrance conductance (equivalent to the end effect in the case of
the conductance of a tube, as discussed in Sec. 2-11) is avoided since the test dome becomes a uniform extension of the pump If the test dome is conbarrel.
siderably larger in diameter than
above.
Gas-inlet
line
Fig. 7-17.
Leak vole
Input meter
Test dome
Vacuum gauge
the
Fig.
Diffusion
pump
7-16,
barrel, as illustrated in
pump
for
gas
the
diffusion
pump would
be somewhat similar to the entrance flow pattern illustrated in Fig. 7-14, showing that some of the molecules are reflected
There are two gauge-connection tubes dome, both installed as prescribed in item 4 above. For pressure greater than 10"* torr a McLeod gauge is used, connected to the test dome through a liquid-nitrogen-cooled vapor trap. The use of the vapor trap with the McLeod gauge is essential to prevent mercury vapor from the gauge from contaminating the test dome and to prevent backstreaming oil vapor from the diffusion pump from contaminating the gauge. The McLeod gauge in any case responds only to permanent gas, which is appropriate for the measurement of the true pumping
in the test
Pressure Measurement.
Fig. 7-17. Test dome and accessories for measuring the pumping speed of a diffusion pump consistent with the recommendations of the Committee on Standards and Nomenclature of American Vacuum Society.
back from the entrance, resulting in an impedance to the flow. Because no entrance conductance
is
involved in the use of a test of a diameter equal to that of the pump barrel, the pumping speed measured by the procedure recommended in the committee's report will be somewhat greater than that which would be obtained by using a test dome of larger diameter. In estimating the net speed of a diffusion pump when installed on a system this effect must be taken into account by combining the
dome
specified
if
on a valve body, manifold, or vapor effort to be more precise on this point is hardly justified, however, since the geometry of a diffusion pump together with its associated components is such that only rather rough calculations of the net pumping speed can be made using the simple formulas given in Chap. 2. In most cases the net pumping speed can be calculated accurately only by the Monte Carlo method described in Sec. 2-12. This method has been applied by Pinson and Peck" to a number of practical combinations of vacuum components.
the
to be installed
pump
An
than 10"* torr two ionization gauges are used, one connected directly and the other connected through a liquid-nitrogencooled vapor trap. The pressure read by the trapped ionization gauge is the partial pressure of permanent gas only and is again appropriate for substitution into (6-34) for the true pumping speed. The untrapped ionization gauge provides a reading dependent upon the "total pressure," including condensable vapor, which can be used for determination of the apparent pumping speed according to (6-35). Gas Flow. The gas flow into the test dome passes through a flowmeter and needle valve. Since the throughput must be constant for appreciable periods of time in order to establish equilibrium pressure in the test dome for each reading, the needle valve must be of suitable design to ensure steady and controllable flow at the low flow rates used for diffusion-pump tests (typically from about 10"^ to about 10* cm^/min at atmospheric pressure). All routine pumping-speed tests are made with air, although measurements for other gases, such as hydrogen, helium, and argon, are also made for the purpose of underless
For pressure
standing pump performance. Various types of flowmeters suitable for measuring the throughput are described in Sec. 7-2. In the diffusion-pump range some form of the inverted buret, such as the pipe organ shown in Fig. 7-7 or the constant-pressure version shown in Fig. 7-8, is most convenient in use and is relatively free of systematic error when used with proper precautions.
300
301
The pumping-speed curve shown in Fig. 6-18 was obtained substantially as described above. As will be apparent from the deviations of the experimental points from the smooth curve, the probable error in individual determinations of pumping speed is about 5 per cent because of uncertainty in the measurement of the pressure and throughput. Provided that systematic errors due to vacuum-gauge calibration and flowmeter reading are avoided,
results of this quality are satis-
chamber is read on the with each of the apertures C set in position over the central hole. The conductance C through any of the apertures from Eq. (2-93) is known accurately from its diameter. From Eqs. (2-5) and (2-7) the total pumping speed of the diffusion pump and the aperture is given by
ionization gauge
P
Q
S
The pressure
in the test
S^
C
that
it is
(7-24)
in
in the region of
P =
where
ki+
(7-25)
of a vacuum gauge. The problems of calibration of ionization gauges and the errors inherent in the use of a McLeod gauge have been sufficiently emintegrity
Thus
1j-
k \o
(7-26)
so that the gauge reading is a linear function of the quantity 1/C and the intercept of the curve at 1/C = is reciprocal of the pumping speed Sj, of the pump alone. The pumping speed of the pump is thus
or the calibration
vacuum
in
gauge.
With
this
problem
mind,
constant k of the ionization gauge. The procedure of Oatley described above contains the elements of what is needed to meet the final recommendation of the Committee on
method
of measuring pumping speeds independent of the accuracy of vacuum-gauge calibration. The method consists essentially of comparing the pumping speed of a diffusion pump with the conductances of a series of apertures placed in series with the pump. The experimental arrangement as shown in Fig. 7-18 consists of a short tube A of diameter large as compared with that of the pump barrel to which is connected the test
which
is
B. The opening between B and A is obstructed by a thin plate D, in which there is a series of apertures C of various diameters, aijy one of which can be brought in line with the opening into A by rotating the plate. The plate D is mounted snugly on the base plate F so that leakage of gas from the test dome around the selected aperture into the region A is negligible. Gas enters the system through the maincapillaries L and between which there is a small chamber tained at a chosen value of the pressure by means of a mechanical can vacuum pump and throttling valve. The leak rate through
dome
by
this
means be
set
With a
Standards of the American Vacuum Society, i" that of calibrating the gauge by inserting a standard conductance between the pump and the test dome. By measuring the gas flow Q in the above procedure and inserting its value in (7-26), the gauge constant k is also determined from the slope of the line. The accuracy of the method described above depends upon the constancy of the flow rate Q throughout a series of readings for all values of the conductance C. The selection of aperture diameters must also be chosen so that the conductances are in reasonable proportion to the pumping speed to be measured. A pumpdown method of measuring the pumping speed at very low pressure in a situation in which Eq. (7-10) is applicable is described by Milleron.i^ Equation (7-10) is valid when the base pressure Pq The method of the system is small as compared with Pj and Pg. described by Milleron consists of baking the system and pumping down to a low base pressure P,,, increasing the pressure by admitting helium or neon gas, and then observing the pressure as a function of the time from Pj to Pj, where P^ > lOOP^ and P^ > lOPo- Since neither helium nor neon is adsorbed appreciably, the pumping speed with
302
or outgassing effects.
cancels out. If the measured pumping speeds for helium and neon are accurately proportional to the square roots of the molecular weights, then this law can be assumed to hold in general and the pumping speed for air and other gases computed accordingly. However, if
CHAPTER 8
the
for these
this
it is not clear how to extrapolate these results to other In that event a similar measurement using dry nitrogen would probably yield reasonably accurate results in spite of the somewhat greater problem due to sorption processes.
relationship,
gases.
REFERENCES
1.
consists of a vessel to be evacuated together with gauges, vapor jet and mechanical pumps, vapor traps, valves, connecting manifolds, and pipes. The vacuum vessel may be
A vacuum system
vacuum
the
2.
Standards for Steam Jet Ejectors, 3rd ed. (Heat Exchange Institute, New York, 1956). J. H. Leek, Pressure Measurement in Vacuum, Systems, 2nd ed. (Published for the Institute of Physics and the Physical Society by Chapman and Hall,
3.
4.
5.
London, 1964), pp. 173ff. B. B. Dayton, Ind. and Eng. Chem. 40, 795 (1948). J. F. Cleveland, Kinney Vacuum Division, The Xew York Air Brake Company, private communication, Sept. 19, 1949. D. L. Stevenson, in 1961 Vacuum, Symposium Transactions (Pergamon
Ltd.,
a processing tank, a space simulator, the chamber of a particle accelerator, or any enclosure within which the gas pressure must be reduced to a value substantially less than atmospheric. Designing the system consists of deciding upon the features of the vessel, selecting specific commercial vacuum components, and arranging these components together with interconnecting piping and accessories to achieve the
specified
vacuum
6.
Press,
London, 1957),
7.
p. 74.
Vessel. Vacuum vessels are built to meet such a wide variety of requirements that a detailed description applicable to all circumstances is not feasible. However, a design philosophy can be suggested, and certain generally desirable features and fabri8-1.
The Vacuum
8.
9.
10.
(Pergamon Press, London, 1959), p. 22. H. G. Noller, G. Reich, and W. Bachler, in 1959 Vacuum Symposium Transactions (Pergamon Press, London, 1960), p. 72. B. B. Dayton, in 195G Vacuum Symposium Transactions (Pergamon Press, London, 1957), p. 5. "Report of Committee on Standards and Nomenclature" (Chairman, B. B. Dayton), in 1955 Vacuum Symposium Transactions (Committee on Vacuum
Techniques, Boston, 1956), p. 91. J. D. Pinson and A. W. Peck, in 1962 Vacuum Symposium Transactions (The Macmillan Company, New York, 1962), p. 406. C. W. Oatley, Brit. J. of Appl. Phys. 5, 358 (1954). Norman Milleron, in 1958 Vacuum Symposium Transactions (Pergamon
Press,
cating methods will be described. Since a vacuum vessel is subjected to external atmospheric pressure, the usual precautions for the design of tanks subject to external pressure should be taken to ensure that the vessel can safely withstand an external pressure of somewhat more than 1 ton/ft^. Aside from the possible danger of buckling or collapsing, the walls of a vacuum vessel
will deflect
11
12 13
London, 1959),
p. 140.
during evacuation and return to normal when air is redeflections must be evaluated and any excessive deflections corrected by increasing the wall thickness or by providing structure to aid in sustaining the load imposed by the atmosphere. The materials of construction of vacuum vessels are predominantly mild steel, stainless steel, and aluminum, although brass and copper
admitted.
Such
porosity
common. Castings are generally troublesome owing to and are to be avoided whenever possible. Sections of glass the form of plates or of tubes are frequently used for viewing ports
or for electrical insulation. High-quality ceramics are frequently used as insulators for high-voltage leads. The vacuum vessels of large proton synchrotrons are probably the most complex in design and
303
304
305
exotic in the choice of materials because of the unusual combination Certain of the epoxy of electrical and mechanical properties required.
were first used extensively in this application and have since been used to advantage in other vacuum applications involving unusual electrical, magnetic, and mechanical requirements.
resins
Table
8-1.
large vacuum vessel can be a rather difficult undertaking since health hazards and the danger of an explosion are serious factors. The use of detergents with water is not recommended since detergents leave a thin, highly tenacious film. In the case of an accelerator cavity 9 ft in diameter and 90 ft in length the most satisfactory answer to the problem of cleaning has been to scour the surface with fine emery and then wash with distilled water using carefully
selected cloths.
Flash
point.
Toxicity,
vacuum
vessels
Solvent
Formula
pressure
at 20C, torr
M.A.C.
in
is
ppm
(_/
{j-HOlq
Carbon tetrachloride
Chloroform Aromatic hydrocarbons: Benzene Toluene Xylene Petroleum hydrocarbons: Stoddard solvent
Ethers:
CCI4
CHCI3
t_/rt_tle
60 88 180 80 23
5
87 77 61
200 25
100
C,Hg
80 110 140
155
35
30 40
25 200 200
such that only gas-shielded arc welding is recommended. Atomic hydrogen welding produces excellent results and was for several years the best commercially available process. The molten metal at the weld is protected by a blanket of atomic hydrogen, reducing any oxides present and preventing further oxidizing of the metal. More recently, arc welding with helium or argon has become generally available, helium being more common and inexpensive in the United States and argon in Europe. Using these methods a skillful welder can produce leak-proof welds with far greater dependability than by oxyacetylene or by ordinary electric arc welding in air. These latter methods
require the use of flux, which frequently results in occlusions which can cause leaks to develop after a weld has been tested and found to be
free of leaks.
25
500
C,H,0 ^4^^10^
440
180
71
-30 -20
400
1,000
C,H0 C.HgO
56 80
65 78 82
in a
^2
15
250 200
1,000
CH4O
C2H5O CoHoO 3 8
98 46 38
18 15
In order to present a smooth interior surface free of crevices the welds vacuum vessel should either be prepared for welding from inside or, if to be done from outside the vessel, should be prepared for full penetration. In either case no crack or crevice remains on the interior surface. The exposed weld metal can then be machined or ground to a
400
smooth
finish if desired.
seals,
Freon-MF(CCl3F)
Trichl orotrifluoroethane
700
284
57
24 48 93
None
1,000 1,000
Each
Freon-TF(CCl2FCClF2) ....
Tetrachlorodifluoroethane
None
None
by E.
I.
Froon-BF(CCl2FCCl2F)
*
1,000
From
du Pont de
by permission.
vessel should be as smooth and free of After fabrication the interior surfaces should be cleaned by vapor-phase degreasing or by the use of a highly volatile solvent to remove oil and other contaminants which would otherwise
interior of a
The
vacuum
crevices as possible.
such feature requires cutting through the wall of the vacuum vessel and welding in a tube or a flange and should be planned with care to facilitate the welding operation and thereby minimize the probability of a leak. Figure 8-1 illustrates acceptable designs for (A) a tube in the side of a cylindrical tank, {B) a flange at the end of a tube, (C) a side seam in a cylindrical tank and (D) the joining of an end bell onto the cylindrical section of a tank. If internal welded joints need reenforcement by additional welding on the outside, as illustrated in (B) above, the outside weld should not be continuous, but should consist of a
maintain a high organic vapor pressure long after the vessel is evacuated. A list of useful solvents and their properties is given in Table 8-1.
segments with gaps between. If the outside as well as the weld were continuous, then a volume of gas would be trapped between the two welds. A small hole in the inner weld would slowly
series of
inside
306
307
is
leak gas into the chamber but could not be located by a helium leak detector or any other method. Such leaks are referred to as virtual leaks and can arise whenever pockets of trapped gas are formed during
fabrication.
pumping speed
good design
0.2C
2.3^
liters/sec
(8-1)
a virtual leak are so difficult to diagnose that the designer should conscientiously avoid any possibility of pockets of trapped gas.
about 400
liters/sec.
'Bell
Tank
rK'
S
/,
Tank
Tube
Tube
1
^^Tcnk
(D) Bell-to-tonk Weld
(Where inside
welding possible
Where only
outside welding
possible
A)
ical
Tube-to-tank Weld
All
roughing pump with the leak-detector unit in the pumping line. welded joints should be inspected for leaks using a fine jet of probe
Outside
W^WVxVM
Vocuum
Where
inside
(
^th>Tank
side
Flange
'
gas in brief, carefully directed bursts in order to facilitate localizing any leak which may be present. With the superior leak-detection (see Chap. 4) and welding methods now available it is no longer necessary to accept as unavoidable a leakage rate which is great enough to constitute a serious limitation on the performance of a vacuum system. In the assembling, operating, and 8-2. Demountable Seals.
servicing of
welding possible)
(C)
welding possible
B)
Seam Weld
Flange-to-tube Weld
vacuum
(_B)
vessels
flange
and compononts. {A) Tube into on the end of a tube; (C) side seam in a
absolutely essential.
vacuum systems demountable seals are convenient if not The major components of a vacuum system are
end
bell
on a cylindrical tank.
Of major importance in the design of a vacuum vessel is the area pumping port. In the pressure range below 10"* torr the flow is molecular so that the conductance at room temperature through an aperture in a thin wall is that given by Eq. (2-93),
allocated to the
generally assembled using demountable seals to facilitate subsequent disassembly for servicing or modification. Access port covers, viewing windows, gauges, and other accessories are typically connected to the
Many types of seals have been vessel with demountable seals. used for this purpose, but with the development during World War II of rings for seals in aircraft hydraulic systems, the adoption of various rings has become an almost universal flange configurations sealed by
vacuum
practice.
c
which
for air at
room temperature
is
G =
where
11.6^
liters/s
a molded, ring-shaped gasket with circular cross section made in a wide range of dimensions and of any of several elastomers, depending upon its intended use. For vacuum service the material
An
ring
is
is
Levenson,
and Davis^* have shown that the transmission probability achievable in the design of the combined vapor baffle and valve for a diff"usion pump is about 0.35; that is, the conductance of the baffle and valve can be as great as 0.35 that of the aperture alone. ComMilleron,
Ho
chapter.
should have a Shore hardness of about 60. Buna N is commonly used because it does not deteriorate when exposed to oil and has excellent mechanical properties for room-temperature seals. Figure 8-2 illusFor sealing flanges an O ring trates a number of typical 0-ring seals. flanges, the is inserted in a groove cut in one member of the mating shapes of Several other member having a smooth, flat surface. proved to be have 8-2a, grooves, some of which are illustrated in Fig. depth should its groove, useful. However, whatever the shape of the not exceed two-thirds the cord diameter of the O ring, and the area
308
of
309
its cross section should exceed that of the ring. When the mating flanges are then forced together metal-to-metal, as shown in Fig. 8-26, the ring is sufficiently distorted to provide a good seal but does not fill completely the volume available in the groove. This
ring
is
not over-
If the
aAJ:ij
A=
B
=
~ ~
13/16
1
O-ring diameter
dimensions are such as to ensure that the O rings are not overstressed. Another advantage of the metal-to-metal contact is that the conductance from the O ring to the vacuum space is then very small so that outgassing of the elastomer is somewhat inhibited. rings are commercially available over a wide range of sizes, from about an eighth of an inch in major diameter to about eighteen inches. Morerings can be made from extruded buna over, much larger cord cut to the desired length and vulcanized, preferably with a 45 splice. The use of a double O-ring seal with a guard-ring groove between the rings is illustrated in Fig. 8-2c. The guard ring is normally vented Leakage at the flange can be to atmosphere through a small fitting. checked by evacuating the guard ring, which reduces the pressure on ring substantially and therefore also the leakage rate if a the inner Alternatively, a probe gas may be introduced into the leak is present. guard ring to provide a very sensitive test for leaks through the inner O-ring seal. The guard ring can also be filled with a low-vaporpressure liquid, such as diff'usion-pump oil, as a means of decreasing
Gland nut
Bell jar
Body
^Glond
^'' <'!<-/M/A
O-ring
(c)
Fig. 8-2. O-ring demountable seals, (a) Types of O-ring grooves; (6) flange with single O ring; (c) flange with double O ring and guard ring; {d) O-ring couplings; (e) double O-ring seal for a metal bell jar.
Farkass and Barry ^ report that by circulating cold through the guard ring of such a seal "significant reductions can be made in the total gas load of the system by lowering the temperature of the sealant material." For seals which are to be opened and closed frequently, as in the case of the seal at the base of a metal bell jar (see Fig. 8-2e), the use of a double O ring with the guard ring routinely pumped out by an auxiliary vacuum pump reduces substantially the leakage problem. rings can be used in convenient couplings for connecting lengths of pipe together or for connecting a glass tubing to a metal tubing. A typical coupling of this type is shown in Fig. S-2d and is merely
illustrative of
many
such couplings
now
available.
Vacuum
couplings
may
fail
to maintain a
sories
has been found necessary to compress the its original thickness to ensure a leakproof seal at low temperature. The rule of providing sufficient volume for the ring to expand laterally still applies even when the groove is as shallow as that required for this type of application. The groove design for an O-ring seal assumes that the bolting or clamping means is adequate to bring the faces of the mating flanges into contact (metal-tolittle
cryogenic applications
as one-third
to vacuum systems and for systems for bench use. Guthrie and Wakerling^ have described the use of square-cross-section rings synthetic rubber (buna N) gaskets for the purposes for which and convenience are now commonly used. Because of the greater is now material availability of rings the square-cross-section gasket seldom used. However, the double grooveless gasket described by Guthrie and WakerUng (see Fig. 8-3a) also developed during the Man-
metal).
In most cases the need for rigidity of the system ensures that the bolts or clamps provide much greater total compressional force than is needed to achieve metal-to-metal contact, and the groove
hatten Project period is now available commercially to fit several sizes of standard flanges. The grooveless gasket assembly consists of two rings of rubber vulcanized to both sides of a metal ring on which are mounted an array of metal spacers to limit the spacing between the flanges as they are bolted together to prevent overstressing the rubber
310
311
gaskets. The space between the two rubber rings may be used as a guard ring by drilling a hole through one flange and installing a fitting. Since no grooves are required, all flanges are ground flat. This type of gasket is particularly convenient and eff^ective for assembling forevacuum pipe lines. The Gask-0-Seal (Parker Seal Company) illustrated in Fig. 8-36 is a commercially available example of a grooveless
Farkass and Barry^ have carried out tests on the base pressure
vacuum chamber
using
rings
made
of various elas-
The O ring under test tomers. was placed in the inner groove of a double seal, between the grooves
of which was a guard -ring channel, as shown in Fig. 8-4. In these tests
A^
Gaskets
the
used
Atmospheric
pressure
Ultrahigh
and
flanges.
The
are
results
vacuum
-25C
shown
in
Spacers spot-welded
to
each face
(a)
Section
Table 8-2. In another set of tests the base pressure of the empty chamber was observed at room rings of temperature with inner Then either butyl or neoprene. ring of the a 143^-in. -diameter material to be tested was placed loosely in the chamber and the base
pressure again observed. The increase in base pressure multiplied
Fig.
8-4.
Double O-ring
seal
with
A-A
channel. cooling (1) Inner O ring; (2) cooling channel; [Taken with per(3) outer O ring. mission from I. Farkass and K. J.
guard
ring
of
London,
1961).]
the system was taken as a measure of the room-temperature outgassing rate. In Table 8-3 the results of these tests are shown together with the measured outgassing rates in micron liters per second per square centimeter of 0-ring surface area.
(6)
Fig. 8-3. (a) Grooveless gasket for use on standard pipe flanges. [Taken with permission from A. Guthrie and R. K. Wakorling (ods.), Vacuum Equipment
Table 8-2. Base Pressures with O Rings Made of Various Elastomers WITH Flange Temperature Maintained at 6C and at 25C*
Flange temperature 6C Flange temperature
and Techniques (McGraw-Hill Book Company, New York, 1949).] (6) Parker Gask-0-Soal. [Taken with permission from the Parker Seal Company, Hayward,Calif.]
25C
Ranges of
values,
torr
Elastomoter
Lowest
pressure attained,
torr
Lowest
Number
of runs
Ranges of
values,
torr
pressure attained,
torr
Number
of runs
gasket suitable for sealing between two flat flanges. A metal ring has grooves cut in both sides and filled with molded elastomer gaskets
which have protruding ridges. The shape of the gasket material is such that when the mating flanges are pulled metal-to-metal against both sides of the metal ring, there remains some free volume when the two ridges on each gasket are fully compressed. Thus the elastomer
is
1.0
4.5
2.1
X 10-" X 10-"
5
2 6
Buna
N
.
Silicone (red).
Silicone (green)
Viton A Teflon
, *
10-" 10-" 2.2 X 10-' 3.2 X 10-' 1.3 X 10-" 4.2 X 10-"
3.8
X X
X X
X
X X X X X
4
2
2
3 4
1.75
1.2 2.1
X 10-" X 10-"
2
2 2
X 10-"
X 10-" X 10-" X 10-"
1960 in jjt>i/ ni
4.8
5.6
1.0
2
2
5.5-5.7 0.9-1.1
X 10-"
X
10-"
never overstressed. The figure shows the gaskets before and after compression by the flanges.
Taken
,
T T7 I . ,! h,. J. T I. Farkass
and V
Rnrrv uarry,
TransVacuum Symposium a i-
actions
(Pergamon
Press,
London,
1961), p. 35.
312
313
As an example of one of the tests performed, the authors state that when the flange and 0-ring seal were at room temperature, the base pressure attained was 2 x 10~^ torr. However, when the temperature of the flange and O ring was raised to 40C, the pressure increased to 1.5 X 10~*torr; and when the temperature was 25C, the pressure decreased to 1.5 x 10^^" torr. These tests indicate that butyl is a preferred 0-ring material for room temperature or below, having both
Table
8-3.
When the buna N was replaced by Viton x 10-^ was attained. The mass spectromgases in the chamber with buna N and with
torr.
Room
Temperature*
Pressure with
Material of
Pressure with
Outgassing rate
of test O ring with entire surface exposed,
^fl/sec
vacuum
system
empty chamber
after 24 hr
Material of
tost
test
ring
ring
pumping,
torr
1.0
.
. .
24 hr pumping,
cm^
10-5 10-5 10-5 10-5 10-5 10-5 10-5
10-"
Neoprene
silicone (red) Silicone (green)
4.6 5.8
10-8
2.0 2,0
1.0
1.0 1.0
X 10- X lO-"
5.8
2.2
1.0
X 10-" X 10-"
2.0
X X X X
io-
io-i>
Teflon
10-" 10-
X X
X X X X X
2.04 2.16
X
X
actions
Taken with permission from I. Farkass and E. (Pergamon Press, London, 1961), p. 35.
Barry, in 1960
its
gas permeability
Neoprene and Viton A also rate very close to butyl according to these tests. It is to be noted that although the silicones tested have the lowest outgassing rates, the base pressure at 6C is very
must
also be low.
Addis, Pensak, and Scott* have compared Viton A and B with buna rings using a mass spectrometer and a vacuum chamber which could
rings shows that, aside from the total gas present being much Viton less with the Viton than with the buna N O rings, hydrocarbons are essentially absent with Viton, whereas the entire series of hydrocarbons rings. The authors recommend the use of are present with buna N Viton A or B O rings with baking the system at a temperature of about 250C for several hours whenever low base pressure and absence of hydrocarbons are important. For many years various forms of metal gaskets have been developed Because metal gaskets play to avoid the use of elastomers as gaskets. an important role in ultrahigh-vacuum technology, this topic is reserved Two distinct advantages may be gained by the use of metal for Chap. 9. gaskets: (1) the hydrocarbons and other gases associated with elastomers are eliminated from the system, and (2) metal gaskets permit baking the system to much higher temperatures than can be tolerated by elastomers, permitting a more thorough outgassing of the metal parts of the system. Manipulation of objects inside a vacuum 8-3. Motion Seals. chamber while the system is at high vacuum is a common requirement. In some cases the motion can be accomplished by mounting a soft steel member on a shaft inside the vacuum chamber and a rotatable permanent magnet just outside the chamber wall. The shaft then responds to the motion of the magnet as long as the torque required is not too great. Continuous rotation can be achieved by mounting the rotor of a small induction motor on a shaft inside the vacuum chamber and the stator outside with a thin metal casing between to provide the vacuum barrier. The rotating magnetic field from the stator penetrates through the thin wall of the casing and drives the rotor in the usual way.
in Sec. 5-13
That the gases given off by Viton as it is baked out to temperatures up to about 300C is predominantly water vapor, CO, and CO2', and no detectable hydrocarbons, whereas buna N rubber gives off water vapor and a whole series of hydrocarbons from C2H3 to C^Hjg and presumably beyond 2. That Viton does not decompose until the temperature exceeds 300C, whereas buna N cannot safely be baked at temperatures in
rings in vacuum at 200C for several days and then storing them in a desiccator greatly reduces the evolution of water, CO, and COj when they are finally put into use
is
made
possible
shown
in Fig. 8-5a.
by The
in the
vacuum chamber
using buna
a cylindrical fitting with a smoothly machined opening, the base of which has a conical contour around the shaft hole. Two elastomer washers, two metal spacers, and a packing nut make up the seal. The outside diameter of the elastomer washers is cut to fit the inside diameter of the casing closely, and the central hole diameter is about two-thirds that of the shaft. The first metal spacer has a central conical portion which matches the conical shape at the base of the casing. The second spacer is a plain circular collar which transmits pressure from the compression nut to the outer rim of the two
314
gaskets, causing
315
them
become
them from being pushed inward by the external pressure. In order to seal properly, the elastomer gaskets must be cut smoothly and lubricated with a low- vapor-pressure grease or with diffusion-pump oil.
The elastomer material should have a Shore hardness of 50 to 60. The region between the two washers may be pumped out to reduce
Pumpouf connection
In Fig. 8-6 is shown the rotating vacuum seal by Roberts,' which utilizes two Garlock Klozures (the Garlock Packing Company, Palmyra, New York) of silicone elastomer lubricated with Dow-Corning 704 Silicone diffusion-pump liquid.
available.
described
rings are used to seal against leakage between the Garlock Silicone Klozure units and the wall of the seal housing. The shaft is supported on two ball bearings, one of which
To mechonicol t
pump
Pumpout port
high vacuum. Although this problem can be solved by the use of low- vapor-pressure grease or a dry lubricant such as molybdenum disulfide, a better solution for some applications would be to locate both bearings outside the double seal. As used by Roberts, the space between the two seals is evacuated to
tion at
Silicone rubber
Gorlock Klozures
Fig. 8-5.
(4)
(a) The Wilson seal. (1) Body; compression ring; (5) compression nut;
(2) seal
washer;
gasket;
(3)
pumpout
ring;
a pressure of about 10"* torr by a mechanical pump. With a shaft diameter of 0.406 in. he reports a base pressure of 1 X 10"' torr with the shaft stationary and 7 x 10"' torr with the shaft rotating at 1,000
Scale
(6) shaft.
(3)
(6)
Chevron
ring;
seal.
(5)
(1)
Body;
(2) (6)
seat ring;
pumpout
(8) shaft.
pressure ring;
compression spring;
(7)
vacuum
seal con-
rpm.
from R. W.
a 300-liter/sec diffusion
leakage through the inner seal, may be filled with a low-vapor-pressure oil, or may be used for testing the leakage through the inner seal. When properly assembled, the leakage rate for a Wilson seal for a
pump
with
a liquid-nitrogen-cooled trap.
seals
according to Dawton,* not more than 10~' torr liter/sec, most of which is condensable at liquid nitrogen temperature. The leakage rate during rotation of the shaft is about three times the stationary leak and that due to translational motion inward is considerably greater. For repeated in-and-out motion the shaft must be kept
M-in. shaft
is,
8-7. In Fig. 8-7a is been machined in the wall of the clearance hole for the shaft. A third groove placed between the 0-ring grooves is connected by a small In assembling this type of seal drilled hole to a pump-out fitting. the rings are well lubricated with vacuum grease and placed in the grooves. The shaft with its leading end tapered is then thrust down
rings as shown in Fig. can be made using shown a seal in which two 0-ring grooves have
The surface of the shaft must be very smooth and, free of scratches in order for the washers to seal properly. Since atmospheric pressure tends to push the shaft into the vacuum chamber,
lubricated.
provision should be made to limit the motion of the shaft, particularly for shaft diameters in excess of 2 in. Very similar to the Wilson seal
is
The chevron
seal permits
through the seal. In Fig. 8-76 the shaft hole is drilled oversize down to the depth of the seal, the total diameter being about equal to the shaft diameter ring, so that the 0-ring plus 1.5 times the cord diameter of the of its initial three-quarters circular section will be compressed to about and a rings, the between diameter. A metal collar serves as a spacer
greater range of adjustment of the compressional loading with the result that materials such as Teflon may be used for the washers. number of commercial seals similar to the Wilson seal have
The region rings. nut permits adjustment of the pressure on the oil or may low-vapor-pressure between the rings may be filled with be used as a vacuum guard ring.
316
317
type of shaft seal frequently used in mechanical booster-pump is shown in Fig. 8-8. The heart of the seal consists of a highly polished graphite surface which is held in contact with a similarly
construction
Compression
nut
Metal bellows seals of this type have the great advantage over Wilson and similar seals in that they can be completely Their disadvantage is the relatively more difficult free of leakage. The stroke permitted by a job of effecting a repair in case of failure. designer must be strictly guided by the that bellows seal is limited so bellows manufacturer. Even the by provided the recommendations
Welded or brazed closures
Welded
or brazed closures
Air
Vacuum
^
Fig. 8-10. Metal bellows seal for transmitting rotary motion. (1) Guide bushing; (2) collar; (3) drive shaft; (4) bellows; (5) drive rotor; (6) driven rotor; (7) bushing; (8) driven shaft; [Taken with permission (9) housing.
Vacuum chamber
wall
Vacuum chamber
wall
Fig. 8-9. Metal bellows seal for linear motion. (1) Guide bushing; (2) stop
Elastomer O rings between one seal member and and gaskets are used to prevent leakage well shaft as as between the the
by means
of a spring.
[Taken (3) bellows; (4) shaft. with permission from Lawrence Radiation Lab. Rep. UCRL-7830, Apr. 15,
collar;
1964,
by T. H. Batzer.]
other seal
ing
member and
the wall
of the housing.
Ordinary lubricat-
UCRL-7830, Apr.
Batzer.]
15,
1964,
by T. H.
oil or a low-vapor-pressure oil such as diffusion-pump fluid may be used for lubricating the rotary seal, depending upon requirements. The principal advantage of this type of seal is the high rotational speed 1,200-3,600 rpm and more
By far then, metal bellows crack, usually near one of the end welds. the greatest use of bellows seals is in the construction of vacuum valves,
some examples of which are given
in the next section. Metal bellows can also be used for transmitting rotary motion into a vacuum vessel. An example of one of the many designs for this purpose is shown schematically in Fig. 8-10. Although several designs of rotary bellows seals have been developed, they are not used extensively, probably because of frequent failure and the complexity In all such devices some torque is imposed on the of construction.
of which
is
it is
capable.
When the
seal
Mechanical rotary shaft Sealing washer; (2) floating seat; (3) bellows; (4) retainer shell; (5) driving band; (6) disk; (7) spring; (8) spring holder; (9) seat sealing
Fig.
8-8.
(1)
seal.
age rate for permanent gas can Hbe very small, the principal difficulty being the tendency to weep lubricant along the shaft into the vacuum
region.
bellows during rotation of the shaft. Whenever bearings in which the shaft should turn freely tend to seize or gall, the bellows receives a twist which can cause it to buckle and crack. unique design of dual-motion feedthrough has been described by
For many years metal bellows have been used as schematically illustrated in Fig. 8-9 to transmit translational motion of a shaft into a vacuum chamber. Bellows of bronze and stainless steel are most
ring.
generally useful for this purpose, the bronze bellows normally being soldered and the stainless steel bellows generally being arc-welded for
Gerber .8 The construction of this device is shown in Fig. 8- 1 1 a detail of the weld joint at one end of the type 316 stainless steel bellows All joints are Heliaro welded and are designed is shown in Fig. 8-12. so that relatively thin mating edges can be fused without the use of any For a travel of 2 in. the bellows used has an extended filling material.
;
length of
3M
in.
in.
The rotary
318
319
are
motion is accomplished by means of an inner and an outer magnet assembly of which the inner is supported on the linear motion rod by
in
vacuum systems
means of a
Handwheel
unit
is
Leod -screw
key
Wheel -retoining
ring
Wheel bearing
plate
temperature bakeout, the inner magnet is replaced by a Kovar block machined in a quadrupole
configuration and the outer magnet assembly is removable during baking. For situations not permitting the use of a lubricant, the
Near the inlet of the mechanical roughing pump to permit checking 1 the performance of the mechanical pump isolated from the system and to permit bringing the mechanical pump up to operating speed before opening the valve for roughing down the system.
2. A small side valve and gauge connection in the short length of pipe between the main valve and the pump inlet facilitates the checking of mechanical pump perform-
Bellows-seoling
flonge
Heliarc weld
\^'
Vacuum chamber
Bellows
moving
Heliarc weld
may
ance.
3.
^
Liguid-nitrogen baffle
At the
to
4
5
Diffusion
Ball-bearing
pump
Valves.
be used in conjunction
^^<1-
mounting screw
8-4.
Vacuum
The
Outer
Ball beoring
Inner
magnets^'
Magnetic
locoting pin
magnets
convenience in operation of a vacuum system depends greatly upon the choice and location of
valves for isolating various portions of the system. Very early in the development of all-metal
with item 4 to isolate the diffusion pump from the system when the pressure exceeds the stalling limit, as during the pumpdown
period.
^
vacuum pump
pump
y Vacuum
reservoir
y^
Mechanical
'nner
magnet
A
this
Heliarc weld
capsule
tween
Vacuum spoce
pump pump
4.
outlet
provide
the
Gasket groove
vacuum systems
that
Fig.
view of dual motion feedthrough. [Taken with permission from J. F. Gerber, Rev.
8-11. Sectional
Sci. Instr. 34,
it was discovered which are entirely satisfactory for use on steam and
valves
1111 (1963).]
compressed-air
satisfactory for
systems
are
un-
operation before the pressure will exceed the stalling point. inlet to the diffusion pump, a large conductance valve to permit isolation of the diffusion pump from the system for testing
At the
vacuum apphca-
tions because of excessive leakage through the packing around the valve stem and at the joint between the bonnet, which carries the stem
packing, and the valve body. Many different methods for modifying standard steam globe valves to reduce or eliminate leakage may be found in the literature. Steam valves in
Fusion weld
which the bonnet is remachined to Welding relief accommodate a Wilson or chevron seal and an adequate gasket, such as an Fig. 8-12. Typical Heliarc weld ring at the joint between the bonnet joint. [Taken with permission and valve body, are satisfactory for from J. F. Gerber, Rev. Sci. many rough vacuum applications. Instr. 34, 1111 (1963).] However, there are so many different varieties of diaphragm and metallic bellows-sealed valves designed specifically for vacuum application now available commercially that improvising adaptations of steam valves is no longer necessary or expedient.
purposes and protection. the 5. A small side valve and gauge connection between item 4 and the on tests inlet to the diffusion pump to facilitate performance closed. diffusion pump when isolated from the system with valve 4 a bypass 6. At the vacuum vessel to provide a connection for atmospheric roughing line to permit pumping down the system from permits pressure with valves 3 and 4 closed. The roughing bypass pump diffusion batch operation of the system without subjecting the operating kept be that the diffusion pump may
to high-pressure gas, so
high. continuously isolated from the system when the pressure is too and systems large most This arrangement proves to be a convenience in batch repetitive a on is particularly advantageous in systems operating chamber. process involving frequent reloading of the vacuum connection for use gauge 7. On the vacuum vessel, a small valve and Actually, on whole. a as in testing the performance of the system will be vessel vacuum the most systems several such connections to
needed.
320
321
The valves listed above are generally of three fairly distinct types. Valves 1,3, and 6 may be referred to as fore vacuum or roughing valves
for
large.
For
this type of
D Closed
K-pipe
tap inlet
C Open
(a)
(&)
D^CIosed
(6)
(a)
Manual operation;
pneumatic
K-pipe tap
K-pipe tap
inlet
K-pipe tap
outlet
outlet
steam globe valves with modifications described above are suitable as are certain diaphragm-sealed valves. In Fig. 8-14 is illustrated the Kinney diaphragm-sealed valve which is available in sizes from 1 to 6 in. and for either manual or pneumatic operation. A cross-sectional view of the manually operated diaphragm valve is shown in Fig. 8-15. Bronze bellows-sealed valves derived from modified steam valves are
service,
2- and 3-
in.
valves
1-
and
l/2-in
valves
90 stem
45 stem
shown
in Fig.
view showing the metal bellows seal between the bonnet and the valve disk and the vacuum-tight gasket joint between the bonnet and the valve body. Outline drawings show the threaded and
flange connections as well as the 90 and 45 stem configurations. An impro^^ed
Fig.
8-15.
Cross
section
of
diaphragm-sealed valve.
is
and
h in
manually
In this design the bellows assembly, cover plate, stem, and valve disk can all be conveniently removed in one assembly from the valve body. This feature facilitates soldering or brazing the valve body into the
piping system.
(a)
(6)
Fig.
8-17.
valves,
(a)
Manually
operated valvo;
322
323
may
be a
PIN
STEM
"O" RING (BUNA-N)
COVER
(BRONZE)'
SCREW
THRUST WASHER^ 5.
tf.
A fabricated steel bellows-sealed valve is shown with its cover and bellows assembly removed in Fig. 8-19 and in cross-sectional view in Fig. 8-20. Since silicone rubber gaskets and valve disk are used the valve can be operated over the temperature range from about 90 to -fl75C. The stainless steel bellows, made of a series of thin
washers
BODY
(BRONZE)-
welded
alternately
on
BELLOWS (BRASS)
the inner and outer edges, provides greater flexibility and longer
life
than
bellows.
available in
4-in. sizes.
ij^-,
2-,
3-,
and
DISC SEAL
BOTTOM
NUT (BRASS)-
Bellows-sealed globe valves are convenient and relatively inexpensive for a limited range of sizes.
DISC (BUNA-N)
-WASHER
(BRASS)
However, even
unduly heavy
and cumbersome.
The passage
through a globe valve moreover is rather tortuous with the result that the conductance at high
vacuum is small. Use of the globe valve should therefore be limited to the pressure range above about 10-^ torr in which
the conductance
is
low conductance is not of serious concern. For many years gate valves of the type generally used in water systems have been used with modifications similar to those applied to steam valves to reduce leakage sufficiently for vacuum application
its
to auxiliary applications in
which
relatively
as
is illustrated in Fig. 8-21.9 ^^g advantages of the gate valve are the straight-through gas flow and the relatively short flange-to-flange dimension. The gate valve also provides a convenient basis for a
vacuum
lock through which items may be inserted into and withdrawn from a vacuum chamber without letting a significant amount of air into the chamber. The main difficulty encountered in attempting a
Fig. 8-19. Fabricated steel valve with cover and bellows assembly removed.
completely satisfactory vacuum conversion of the conventional gate valve is the long travel of the stem from the closed to the fully open position. Although compound bellows designs may be found in the
324
325
much more complicated than that for a globe valve. The modern vacuum gate valve seems to have descended from
design by Wahl, Forbes, Nyer, and Little" shown in Fig. 8-22. The sliding-plate type of gate valve has appeared in a variety of commercial designs adapted for hand wheel, toggle, or pneumatic operation as illustrated in Fig. Valves of this description are available 8-23.
over a wide range of sizes (2 to 32 in.). The flange-to-fiange dimension is much smaller than the for any other type of valve yet devised, and
conductance is correspondingly high. In most rings commercial valves of this type buna N temperthe case which in sealing, for used are ature range is limited. However, with Viton O rings the sliding-plate type of valve can be
it lies in the median plane of the valve body. When turned 90 from the closed position the valve has relatively good conductance, although not quite as good as a gate or sliding-plate valve of the same aperture. One end of the shaft on which the disk is mounted passes through an 0-ring seal on one side of the valve body and is fitted with a handle for setting the valve position. The butterfly valve has two distinct advantages over the gate and sliding-plate types of valves. One is that there is much less surface exposed to the vacuum space so that there is less difficulty with contaminants. The other advantage is that the butterfly valve is more compact in every dimension for a given aperture.
the disk
is
=o
As
less
is
clear
from Fig.
8-24
is
the
flange-to-flange dimension
even
The
can be thoroughly cleaned so that outgassing problems are a minimum and are determined primarily rings used in the assembly. by the type of Because of the wide range of sizes and high
is
good in
this respect.
class of valves
generally re-
ferred to as disk valves are particularly suitable for use in the low-pressure range in which high conductance is of primary concern.
Fig. 8-21. A commercial gate valve modified for vacuum application. The gate seats (A) are machined to a blunt edge, and rubber annulae {B) are inserted
conductance of the sliding-plate gate valve, it suitable not only for use in forevacuum lines but also on the high-vacuum side of diffusion pumps. This type of valve has in fact become
in recent years the
Fig. 8-22. Gate valve with sliding plate [Taken with permission from
J.
S.
Xyor,
the gates. The into jimction between the body and the bonnet is
most generally useful because of its compact design, straight-through opening, high conductance, and relatively low
cost.
made vacuum
tight
by
a groove cut in the The stem is body. sealed with a Wilson seal {D). [Taken with permission fromF.^v D. Kurie, Rev. Sci. Instr. 19, 485 (1948).]
.
Another type of valve which has distinct advantages for some installations is the butterfly valve, one type of which is illustrated in The butterfly valve is somewhat Fig. 8-24. similar to a damper of the type used in stove It consists of a relatively thin, ringhke pipes.
valve
bod}^ the inner surface of which is machined to conform to a spherical shape and a flat disk mounted on a shaft across its 0-ring groove disk diameter. The is thick enough to accommodate an which runs completely around the edge of the disk. An ring mounted in the groove makes an excellent seal when the disk is turned to a
diameter to the diffusion pump to which it is to be connected. The vertical travel of the disk in opening is chosen to be sufficient that the open conductance of the valve imposes a minimum restriction on the resultant pumping speed. Disk valves are most conveniently built as right-angle valves such as the commercial models shown in Fig. 8-25 in the manually and pneumatically operated forms. Valves of this type are commercially available in sizes ranging from 2 to 32 in. in aperture diameter, corresponding to the range of diffusion-pump diameters. Modern versions of the disk valve utilize O rings for sealing between the disk and the seal face, as
illustrated in the figure.
Both for testing vacuum-pump performance and for carrying out on vacuum systems a means of admitting air or other gases at a controllable, steady rate is needed. For many situations an ordinary gas needle valve is satisfactory. However, for control of a small gas
tests
326
327
-3
'^i>r<^
Fig. 8-23. Commercial sliding plate gate valves, (6) toggle operation; (c) hand wheel operation.
(a)
Pneumatic operation;
"
ur>|oo
2.250",
m
s
cS
>>>>
>>>/!?>
/7-r
pump
side
^W^
yj^j^^/zf ////// ///r9n^
vacuum
valve.
328
329
flow with reasonably steady throughput at a given setting, a needle valve of special design is required. An early but quite successful type of needle valve is that described by Bush" and illustrated in Fig. 8-26. The principal feature of the design is the slowly tapering needle fitting
snugly into a carefully reamed conical seat. For some applications a bellows seal may be substituted for the packing shown in the figure. Such valves are now commercially available. As Bush" mentions,
-Coated
lightly with glyptol
in their use at the inlets of primarily to condense vaporized pump fluid and products of decomposition of the pump fluid. The backstreaming of diff'usion pumps is discussed in some detail in Chap. 6, and the use of water-cooled baffles and refrigerated traps is also briefly described.
pumps
The
direct blast of
vapor backinlet
streaming from
diff'usion
the
is
to
pump
so
great that
in
without some precautions to condense and return the vaporized fluid to the diffusion-pump boiler the entire charge of working fluid would be lost and the pump would fail after a few days of continuous operation. Placing a water-cooled cap over the first-stage nozzle and cooling the upper portion of the pump barrel or using a set of water-cooled baffles, such as that
shown
of the
Fig.
8-26.
in Fig.
pump
Fig.
8-27.
Water-
or
Freon-cooled
suffice to eliminate
Vacuum
needle valve.
[Taken with permission from A. Guthrie Vacuum Equipment and Techniques (McGraw-Hill
baiHe unit.
the more virulent components of backstreaming vapor. These measures also greatly reduce the rate of loss of pump fluid so that the operating lifetime of a boiler filling
very
adjustment of the gas flow can be obtained by the use of two valves in series with a small volume in between. The first valve is adjusted to take most of the pressure drop and the second valve used Johnson and Good^^ also describe for fine adjustment of the flow rate. gas-flow rates (2 to 3 cm^/hr) using this a double needle valve for small
fine
not determined by loss of fluid by backstreaming but by eventual deterioration due to decomposition (except in the special case of mercury as the working fluid). To make these measures even more
is
same
principle.
8-5.
Vapor
:
Baffles
and Traps.
trap
is
applied
may be replaced by the use of refrigeration and other cooled surfaces are cooled to the lowest temperature at which the condensed fluid will still flow back into the diffusion pump. Both for mercury and for many of the organic pump
effective, the
water cooling
baffles
by which the
to a device
which
may
vacuum
system (a) to prevent the back migration of the vaporized pump fluid from a vacuum pump into the vacuum chamber and (6) to condense from the volume of the vacuum chamber any condensable vapor which may be present. In many cases a refrigerated vapor trap placed
a temperature of about 35C is optimum for this purpose, although this temperature is too low for those pump fluids which have
fluids
For many applications of off diffusion pumps the is sufficient to maintain the base pressure
and surface
between a diffusion pump and the vacuum chamber performs both these functions. However, depending upon circumstances, one function may be much more important than the other, in which case the choice of type and location of the refrigerated trap should be made with its principal function in mind. Cryogenic pumps which are
discussed in Chap. 9 are, in fact, vapor traps of category (b) operated at such a low temperature that many gases which are normally regarded as permanent gases condense on the cryogenically cooled surface.
cleanliness required. lonization-gauge readings of the order of 10-" torr are typically maintained in systems so equipped. Refrigerating the baffle at -35C will generally result in a reduction of the base pressure to about 10"' torr.
The backstreaming that still persists after elimination of the direct blast of high-temperature vapor from the hot jet region is the volume
surfaces.
migration of vapor at the approximate temperature of the condensing In the case of organic pump fluids which wet metal surfaces, the surface film of the fluid spreads out from the area of direct
330
331
condensation and by surface migration may also spread into the region beyond the baffle structure. The volume and surface migration results in the contamination of the high-vacuum chamber with approxi-
If the conductance for water vapor of the manifold up to the location of the trap is large as compared with the condensing speed given in
(8-3), then the pumping speed of the trap for water vapor can easily exceed that of the diffusion pump by a large factor. A thimble trap 2 in. in diameter and with a cold length of 4 in. will have a condensing speed of nearly 5,000 liters/sec for water vapor. Since the trap will
mately the room-temperature vapor pressure of the pumping fluid and its decomposition products. In a system
Thin-wo
sfainless-steel
tubing
Chomber wa
Liquid nitrogen
or dry
ice in
evacuated by an oil diffusion pump protected only by water or Freon-cooled baffles a base pressure reading on an ionization gauge less than IQ- torr (air calibration) can seldom be realized for an extended period of To realize base pressures significantly time.
Wellpolished
Table
8-4.
Temperature
C
less
than
K
373.1 423.1 273.1 233.1
Water
760 93
4.6
0.1
NH3
CO2
2.7
1.3
Hg
X X 2 X 1 X 3 X
10-1 10-2
surface
and
100 50
low -freezing-
of special design.
point liquid
vapor trap for many thimble trap, as shown in Fig. 8-28, cooled either by dry ice (solid CO2) at -78.5C in Fig. 8-28. Thimble trap. a low-freezing-point liquid such as trichlorethylene to provide good thermal contact or by liquid nitrogen at about 195C. A trap of this type has frequently been installed in the manifold between the disk valve at the inlet of the diffusion pump and the vacuum chamber. If the manifold has large enough dimensions, the conductance for flow of gas into the diffusion pump is not seriously impaired and the thimble trap is in a strategic position to condense backstreaming vapor from the diffusion pump and also to pump by condensation any water or other condensable vapors which may be present in the chamber. If the temperature of a thimble trap is very low as compared with that at which the vapor in the system would be at saturation, then the sticking probability for condensation conis very high and the trap acts like a nearly perfect pump for the only not thus densable vapors. The trap in the position indicated reduces significantly the pressure due to backstreaming from the diffusion pump, but also in most cases provides a much higher pumping speed for water vapor and other condensables from the vacuum chamber than the diffusion pump alone would provide. If the area of the cold section of the thimble trap is B cm2, then by reference to Eq. (2-93) the condensing speed of the trap for vapor is
Sr
3,220
-40
~78.5t
-120 -150
-195.81
194.6 153.1
123.1
77.3
540 42
0.2
760
10
6
10-13
6 X 10-4 10-11
X 10-2
10-8
* Source: Handbook of Phijsics and Chemistry (Chemical Rubber Publishing Company, Cleveland, Ohio). t I
Sublimation temperature of dry ice at pressure of 760 torr. Boiling point of liquid nitrogen at pressure of 760 torr.
J 3.64^
MI
Sr
(liters/sec)
(8-2)
result
is
14:.1B
liters/sec
(8-3)
accumulate a surface layer of condensate, the temperature must be low enough so that the equilibrium vapor pressure is low as compared with the base pressure required in the vacuum chamber. In Table 8-4 are given the values of the vapor pressure of water, NII3, COj, and mercury at a few significant values of the temperature. Note that for a base pressure of 10-^ torr the temperature of dry ice is not low enough for water vapor, so that the next convenient temperature is that of liquid nitrogen, for which the extrapolated value of the vapor pressure is negligible for most practical purposes. Whether anything approximating the equilibrium vapor pressure corresponding to the temperature of the trap is realized in practice depends upon the amount of material co^^densed on the trap. In the case of water vapor, the accumulation of ice, which has a very low thermal conductivity, may cause the surface exposed to the vacuum space to attain a temperature well above that of the metal or glass surface of the trap with the result that the limiting vapor pressure may be considerably above that expected. This particular difficulty applies to a thimble trap, the cold surfaces of which are usually exposed to surrounding walls of the system
332
at
333
encounter with a cold surface before it can enter the high-vacuum Surface migration is greatly inhibited and essentially stopped space.
Because thimble traps do provide very high pumping speed for condensing vapor under favorable circumstances, they are frequently installed directly in the vacuum chamber independently of the dilFusionpump manifold to ensure essentially infinite conductance for vapor A vacuum chamber thus equipped in reaching the condensing surface. is normally pumped down to the pressure at which the diffusion pumps can be put into operation (e.g., by closing the bypass roughing valve and opening the fore vacuum and disk valves) and the thimble trap then immediately filled with liquid nitrogen to ensure very rapid reduction of the pressure. In the case of large accelerators, which are usually evacuated for long periods after each pumpdown cycle, the thimble traps may later be allowed slowly to run out of liquid nitrogen and warm up to room temperature. The slowly evaporating vapors from the thimble trap in this case are pumped out of the system by the diffusion pumps over a period of m.any hours. The thimble trap
thus provides for a rapid reduction of total pressure to an acceptable operating level and may save several hours of machine time. In systems involving frequent recycling from atmospheric pressure to high vacuum, a thimble trap may decrease by a large factor the time In this type of application the thimble trap should be for each cycle. removed and cleaned each time the system is brought to atmospheric
pressure to ensure reasonable efficiency. The thimble trap placed in the diffusion-pump manifold
effective in systems using
is
by
so designing the trap that oil films migrating along surfaces will encounter a cold barrier before entering the high-vacuum space. Figure 8-30 illustrates a refrigerated trap of the chevron type with an oil creep barrier together with a water-cooled baffle to prevent an excessive backstreaming load on the liquid-nitrogen-cooled baffle
Clean vacuum
Cleon vacuum
Oil
-creep
Volume
migration
Surface
migration
barrier
Heat
roll
shield continuous
of polished stainless
_A_
Air
vent"
Diffusion
Conductance
-100
I
pump
iters /sec
by
HjO
then by
Liquid nitrogen
in
radiotion loss to
liquid nitrogen
pump
quite
fluid.
permission
mercury as a working fluid. Mercury does not wet surfaces and migrate along surfaces to any significant degree, nor does mercury decompose into products of widely differing vapor Any mercury molecules which happen to escape the pressures. condensing surface and enter the vacuum chamber will eventually return to the trap and be condensed. If the diffusion pump is prevented from blasting a hot stream of vapor back into the chamber by an effective combination of Freon-cooled baffles or jet cap and cooled pump barrel, the partial pressure of mercury in the system with a properly proportioned thimble trap at liquid-nitrogen temperature will normally be negligible as compared with that of the other elements
present.
Milleron, in 1958
Vacuum
Symposium
1959).]
Transactions
8-30. Schematic drawing of chevron type of vapor trap with surface creep barrier. [Taken with permission from Norman Milleron, in 1958 Vacuum Sym,posium, Transactions (Pergamon Press, London,
Fig.
19,59).]
system.
is
to R. F. Post.
that illusLiquid-
nitrogen-cooled vapor traps with surface creep barriers such as those illustrated in Fig. 8-30 and Fig. 8-31 have proved to be effective in reducing by a very large factor the backstreaming of condensable
When
much
organic
pump
is
pump back into the high-vacuum space. Traps of this tj^e, combined with other techniques to be described in Chap. 9, have in many systems contributed to the attainment of operating pressures below 10"^ torr in the region referred to as ultrahigh
materials from the diffusion
less
effective.
As discussed by
fluid
and
illustrated in
vacuum.
of vapor traps such as those shown in Figs. 8-30 and that the conductance is so low that the resulting pumping speed for permanent gases is impaired. This is a serious handicap since the operating pressure of systems using liquid-nitrogen-cooled traps is very
8-31
is
from the diffusion pump may invade the high- vacuum space either by volume migration or by surface migration. Volume migration can be prevented by arranging refrigerated surfaces in the vapor trap so that an oil molecule must have at least one
Fig. 8-29, the
working
A disadvantage
334
335
discussed
low SO that the highest possible pumping speed is needed to accommodate any appreciable throughput. The problem is to achieve the blocking effect of a well-designed vapor trap without introducing such a low conductance for the pumping of permanent gases. A commercial
by Levenson, Milleron, and Davis.i'^ Among vapor-trap configurations recommended by these authors is the "bulged-elbow" model, several versions of which are shown schematically in Fig. 8-32, together with
measured values of the transmission probability p, defined as
conductance of the trap configuration conductance of an aperture of diameter
Liquid-nitrogen
trap
r^^ v^ii Q
Bulged elbo*
W/D
2
Expenmentol
0.44
0.39 0.32
opening into the trap from the vacuum chamber and also of the opening out of the trap into the diffusion pump. p is thus a measure of transmission efficiency of the trap compared with that of a simple aperture utilizing the same wall area of the vacuum chamber. Figure 8-33 is a cutaway drawing of a commercial adaptation of the bulged-elbow trap with a chevron baffle combined
where
00
1.33
t***)
1,00
Chevron-valve-baffle seal
Cooling-medium
inlet
Vocuum-chomber
mating flonge
Volve-plate gasket
Water-cooled
baffle
Bulged elbow on d
G G
2,00
33
9
jet
1.33
0.30
Lifting lugs
cop
200
1
Valve plate
33
32 27
Copper chevron
ffusic n pufnp
200
1.66
1
0.38
0.35
0.31
C
Valve-plate actuator
33
KVB-32
A/B-5
Fig. 8-31. Water-cooled baffle and liquid-nitrogen-cooled trap with surcreep barrier. [Taken with permission from J. R. Ullman, in
face
1957
actions
1958).]
Vacuum Symposium
Trans-
Fig. 8-32. Several bulged-elbow vaportrap designs together with the experimentally determined value of the transmission probability p. [Taken with permission from L. L. Levenson, N. Milleron, and D. H. Davis, in 1960 V acuum, Symposium, Transactions
Cooling -woter
connections
1961).]
vapor trap meeting these objectives is that shown in Fig. 6-19. The pumping speed of the 6-in. diffusion pump alone, as indicated by the performance curve in Fig. 6-18, is about 1,400 liters /sec. The combined pumping speed with the vapor trap is about 600 liters/sec, as shown in Fig. 6-20, so that the diffusion pump and vapor trap combination delivers about 43 per cent of the pumping speed of the pump alone. The optimization of the vapor trap and diffusion pump for maximum
Oil drain
valve
Heaters
Fig. 8-33. Bulged-elbow trap with chevron baffle combined with disk valv
mounted on
32-in. diffusion
pump.
336
337
pump. The "pumping with a disk valve as used on a 32-in. diffusion pump as a function of the speed" at the inlet of the 32-in. diffusion speed of the pump and trap inlet pressure and also the pumping the performcombination are shown graphically in Fig. 8-34. From flange is S = 30,400 ance curves the pumping speed at the pump
~~'
The pumping-speed curve for the baffled 32-in. diffusion pump shown in Fig. 8-34 and that for the baffled 6-in. diffusion pump shown in Fig. 6-20 both show a peak in pumping speed in the pressure range
from about 10^* torr to about 10"^ torr. In this pressure range the molecular mean free path is equal to or less than the openings through the baffle so that in this pressure range the conductance of the baffle system is greater than its value at lower pressure. The performance curve finally falls off with increasing pressure even though the conductance of the baffle system continues to increase because the pumping
speed of the
In this pressure region the charis becoming viscous in character so that molecules no longer are assured of colliding with surfaces but may experience significant collisions with other molecules. Thus molecules which should condense on baffle surfaces can pass entirely through the trap without encountering the cold surfaces. In systems which use trapped oil diffusion pumps, but which must be kept free of organic contaminants, this awkward pressure region must be avoided as much as possible by arranging to pass through the pressure range
is
35,000
70P00
^
E 60P00
V
I
30P00
pump
falling rapidly.
o
"S
50POO
Un
)alt ed /
1
1
/
/
/
1
25,000
'
20,000
iping
^ y]
Q-
5 30300
2Q0O0
/
\ \
V-
15,000
10,000
lOPOO
,-4
1(
^
ic
'i.non
0-'
1
0-^
)-5
1(
10-'
-^
possible.
is filled
Pr ;ssu re, to rr
When
diffusion
a liquid-nitrogen-cooled trap
of Kinney Model KDP-32 Fig. 8-34. Typical puraping-speed curve baffle. chevron bulged-elbow pump with and without
liters/sec
between
fillings,
trap surface
warming up
slightly.
The
a sequence of abrupt
inlet to
the bulged
elbow
is
15,000 liters/sec.
From Eq.
(6-28)
on an ionization drops and available, such commercially number of systems gauge. There are a in the trap Kquid level maintain the 8-35, to Fig. as that illustrated in
slow increases in the pressure recorded
G B
W
Solenoid cord
S,,
is
SL-l-308
(15,100)(30,400)
30,400
15,100
Rubber stopper
The
about 5,190
G^
11.6^4
60,200 liters/sec
0.50, Thus for this case the transmission probability is p = C,lC'!v Levenson, by reported values the than higher which is somewhat of the pump Miheron, and Davis for a similar trap. Overall efficiericy as stated be trap combination may
S,
15,100
0.25
60,200
is
which
is
in
practice.
{a)
General arrangement;
338
339
nearly constant, so that only an occasional change of the large liquidnitrogen dewar is necessary to keep the trap continuously filled.
Another solution to the problem of automatic liquid-nitrogen supply and also an alternative effective form of vapor baffle with a surface migration barrier are described by Taylor^^ and illustrated in Fig. 8-36. In the reference is shown a
schematic diagram for the automatic control of the liquid-nitrogen flow through the cooling coil. The
liquid-nitrogen-cooled
baffle
is
This is, for example, true in the case of a the traps over the diffusion pumps are primarily intended to maintain a sufficiently low vapor pressure of mercury in the chamber. Liquid-nitrogen-cooled thimble traps may
mercury-pumped system
if
pressure materials.
Water
be used for rapid pumping of water vapor and other high-vapor Mechanical refrigerators of cascade or compound
fWrr
shielded from the direct blast of vapor from the pump by the 45
shield built into the water-cooled
Boffle
The baffle is a simple array with a minimum of two bounces for molecules to penetrate the trap. The pressures attainelbow.
able with this trap proved to be
about
Fig. 8-36. Liquid-nitrogen circulating
3x10-'
out
and about
vapor
lO-^"
tonReceiver
baffle.
(1)
Baffle
structure
(tough pitch copper); liquid(2) nitrogen coil, inlet, and outlet (copper) oil-migration barrier (stainless (3) steel); (4) heat shields; (5) watercooled elbow; (6) line-of-sight heat shield with drain hole; (7) dam to stop oil condensed on the elbow from draining into the baffle housing; (8) baffle housing with low heat con-
system at 250C
it
trap
is
such that
shown
tical position.
The high-conduc-
with permission from H. R. Smith and P. B. Kennedy, in 1959 posium, Transactions (Pergamon Press, London, I960).]
Vacuum Sym-
ductance inlet and outlet tubes. [Taken with permission from A. R. Taylor, in 1961 Vacuum, Symposium
Transactions
tance water-cooled elbow with the 45 shield eliminates the need for other water-cooled baffling over the oil diffusion pump, increasing
100
(Pergamon
Press,
London,
baffle is
1962).]
800
liters/sec
and
4,000 liters/sec.
The
measured pumping speed of the 4-in. pumping system was 367 liters/sec for air at 7 x 10"' torr and 1,024 liters/sec for hydrogen at 8.5 x 10"'
torr.
Vapor traps in large systems are sometimes preferably cooled by low-temperature mechanical refrigerators rather than by dry ice or liquid nitrogen. From the data in Table 8-4 it is evident that for many purposes a temperature intermediate between that of dry ice and liquid nitrogen would be entirely satisfactory from the point of view of the
with an adequate heat capacity for most vapor-trap applications. A number of such systems are discussed by Smith and Kennedy, 1' among them the double-cascade system illustrated in Fig. 8-37. The installation and operating costs of several types of lowtemperature refrigerators are compared with those of liquid nitrogen in Table 8-5 from the paper of Smith and Kennedy. The binary system mentioned in the table was developed for the heavy-ion linear accelerator at Berkeley and is still in service after many years of operation. Whereas the savings in liquid-nitrogen costs have been very large, the refrigerator units have been a major source of down time on the accelerator because of excessive servicing requirements. Experience to date is that the specialized servicing required and the frequency of failure largely negate the savings on liquid nitrogen by loss of operating time.
150C
340
341
can be lost from the boiler by migration into the forevacuum portion of the system as well as by backstreaming into the high-vacuum region. Also the working fluid of a diffusion pump can be contaminated by oil vapor from the backing pump. These two problems can be partially solved by the use of forevacuum condensers or traps the design of which is rather different
The working
pump
from the inlet to the mechanical pump and preferably upstream from a bend in the pipe to avoid accumulation of oil due to the direct spitting of oil droplets into the pumping line which most mechanical pumps Eliminating completely the back migration exhibit to some degree. sealing oil into the high-vacuum portion of the mechanical pump of system and the consequent hydrocarbon contamination from this source by the precautions described above is probably not feasible. However,
Table
8-5.
Quick-disconnect couplings
Cost
Compound
cascade
Cascade
Binary
Liquid
nitrogen
Flow
Operating:
Maintenance
Refrigerant
$330
315
270<*
$330
315-^
$330
315<^
$ 1,200"
....
6,100*
Thin-wGll
tubing
Liquid-nitrogen
Dry-ice "slush"
Copper tubing
Water Power
Initial installation.
stainless-steel
270*
6,050'^
270*
7,300'^
6,800
500^
7,800 7,350
Dewar'
flosk
Total,
first
year
Yearly^
*
7,715 1,595
6,965 1,520
8,215 1,645
Vacuum Symposium
"
"
Taken with permission from H. R. Smith and P. B. Kennedy, in 1959 Transactions (Pergamon Press, London, 1960), p. 271. Servicing trap or baffle. % 0.0085 kW/hr. " Does not include baffle. $ 0.107/liter 25% loss.
<*
Fig. 8-38. Two types of forevacuum traps which trichlorethylene or by liquid nitrogen.
may
be cooled by dry
ice in
$ 0.0003/gal.
Operating cost
+ 10%
depreciation.
The pressure in the forerange of viscous flow, well above the region of molecular flow, so that molecular encounters with the walls of the forevacuum system are normally rather rare events. In order to ensure reasonable efficiency of condensing vapor is it is necessary to provide a fairly long and tortuous flow path through a maze of condensing surfaces so arranged that the condensate will flow in the desired direction. Many commercial diffusion pumps include a vapor condenser at the outlet with provisions for cooling either by water or by a refrigerator unit. The need for an effective exhaust baffle system is particularly vital when mercury is used as the working fluid. The refrigerator coil used for cooling the inlet baffle and the top of the diffusion-pump barrel can conveniently be extended to cool the exhaust baffle as well. To prevent excessive backstreaming of mechanical pump oil into the diffusion pump, a rather large trap similar to a thimble trap built into an expanded section of the forevacuum line or a bend cooled by a slush of dry ice in trichlorethylene as shown in the sketches in Fig. 8-38 is fairly effective. Such a trap should be placed far enough
vacuum
measures described above A more superfluous. even some cases and in adequate are entirely is the use oil vapor of migration back preventing of effective method as an abalumina or zeolite artificial utilizing trap of an absorption
as described in the following section. the features of liquid-nitrogen-cooled highincorporating A system vacuum trap, water-cooled inlet, and exhaust baffle for the diffusion pump and a thimble trap near the inlet to the mechanical pump is The valve and manifold arrangement together illustrated in Fig. 8-39.
sorbing
medium
with the forevacuum ballast tank permits rapid cycling of the system without shutting down the diffusion pump. 8-6. Absorption Traps. In the vacuum lore of hand-assembled practice of using a bulb partly filled with glass vacuum systems the water vapor from atmospheric absorbing phosphorous pentoxide for air periodically admitted to the system and an absorption tube loosely filled with charcoal placed in the forevacuum line to prevent backstreaming of mechanical pump oil into the high-vacuum system was well known." Chemical and physical absorption have been used in
many forms
in
vacuum
practice.
342
343
advances have been made in absorption traps, both for forevacuum and high-vacuum apphcations. A few examples will suffice to give an impression of what can now be done with absorbing traps and what components can be obtained commercially. The development of modern absorption traps appears to have
started with the report
the period of time a room-temperature copper-foil trap remains effective In a small, glass vacuum system using a two- or after bakeout.
three-stage glass diffusion pump with Octoil-S as the working fluid a copper-foil trap 6 in. in length and 2 in. in diameter was installed
by Alpert"
when he
Main vocuum
isolation valve
lonizotion-
Haas
chamber
Thermocouplegauge tube
gauge tube
Baffle Flexible
connector
between the diffusion pump and the test chamber. The base pressure attainable after bakeout was typically in the range lO-i" torr. The graph in Fig. 8-41 shows the history of the pressure recorded by the ionization gauge on the test chamber over a period of several days. For a period of nearly twenty days the pressure rose slowly but remained below 10-9 ^orr. However, at about twenty days after initiation of the test the pressure then rose rapidly to about 10^' torr. Mass
spectrometer analysis of the gases after saturation was reached indicated that carbon monoxide was by far the major gas present. Carmichael and Lange then made
up a number of combinations of
small
traps
(2-in. -diameter)
copper-foil
time"
each
Fig.
shown
in
8-42.
combination, as Additional
Fig. 8-40. Copper-foil trap. [Taken with permission from D. Alport, Rev.
Fig. 8-39. Diffusion-pump system incorporating traps and baffles to prevent oil migration both in th high-vacuum and forevacuum portions of the system.
pump and the copper- Sci. Instr. 24, 1004 (1953).] trap with the result that the smaller the conductance the longer the stay-down time of the trap became. This very simple type of trap is useful on small systems in which performance of a week or less is all that is required. However, efforts to make a high-conductance trap of the copper-foil type for use
diffusion
added a roll of corrugated copper foil to a glass liquid-nitrogen-cooled vapor trap to reduce fluctuations in the gas pressure due to change in the liquid-nitrogen level on the outside wall of the trap, as shown in Fig. 8-40. The procedure had been to bake out the trap with its spiral of copper foil and then cool it down to liquid-nitrogen temperature to make observations at very low pressure. During the course of the experiments, however, it was found that after high-temperature bakeout the copper-foil trap became fully effective at room temperature and did not have to be cooled down to low temperature. Carmichael and Lange^" have conducted a series of tests to determine
on large systems have thus far not been successful because of the very short stay-down time or period of effectiveness which results. A much more promising type of absorption trap is that described by A Biondi^i utilizing artificial zeolite* or activated aluminaf pellets. m shown small glass trap of this type and the arrangement for testing is 8-14 Fig. 8-43. Either with 3^-in.-diameter pellets of zeolite or with of 8 hr mesh chips of alumina in the trap the system after bakeout
Linde porous alkaU metal aluminosilicate, type 13X, manufactured by Union Carbide Corp. in the form of 8-14 mesh t A highly porous material provided for these tests chips by Aluminum Company of America.
*
division of
344
345
Type of trap
10-^
T
1
Vacuum system
1
~2
-*^\"\^
days
Diffusion
Copper
trap ,
X~
g^^^^
2
IO-8[_
pump
k2"^
'
6 days
The principal disadvantage encountered in the use of zeolite and alumina in vapor traps is the enormous amount of gas and vapor given off by the material in the bakeout period. Milleron and Levenson22 measured an output of 28 g of water from 250 g of zeolite pellets baked at 450C for 48 hr. This measurement was made in the course of developing a high-conductance room-temperature trap as illustrated
in Fig. 8-46.
fc
^= ^
1"
^ i"k2"*^
6 days
k-
4
10 15
~15days
U-2"-*l
20
Time, days
Fig. 8-41. Typical pressure dependence with time over a period of several days for a small system in which a room-temperature copperfoil trap 6 in. in length is used between the oil diffusion pump and the test chamber. [Taken with permission
8-42. Stay-down times for Fig. various copper-foil trap arrangements. [Taken with permission from J. H.
The trap was made by lining a right-angle elbow with about two layers of zeolite pellets, the pellets being about in. in diameter and length. A No. 10 mesh stainless steel screen was used to retain the zeolite pellets as shown in the figure. After bakeout the pressure history observed was similar to that reported by Biondi,^! but in this case the test was continued for a much longer period. The pressure after bakeout was 3 x lO"!" torr initially but rose to 1 x 10"^ torr where it remained for a period of about one year. Biondi^i
Carmichael and
W.
J.
Lange, in 1958
Transactions
1959).]
To ultrahigh
Vacuum
(Pergamon
Symposium
Press,
London,
H. Carmichael and W. J. Lange, in 1958 Vacuum Symposium Transactions (Pergamon Press, London, 1959).]
from
J.
vocuum system
at 450C remained in the lO-^" torr range for about three months.
The
diffusion-pump fluid used for these tests was Octoil-S. Biondi^i also describes the performance of the large room-temperature zeolite trap illustrated in Fig. 8-44. The trap provides a series of
trays in which artificial zeolite pellets are placed. The trays are so arranged that molecules in passing through the trap must encounter
Before each test the trap and test chamber were baked out and the system typically reached an ionization-gauge reading of slightly over lO"!" torr after cooling to room temperature. Using Octoil-S in the diffusion pump resulted in a stay-down time (below 10^^ torr) of only about twenty days. However, when Convoil-20 was substituted for the Octoil-S the base pressure remained at 10~"i" 1 X for more than one-hundred days without any evidence of rising. The results of these tests are given in Fig. 8-45 showing three different tests with Octoil-S and one test run with Convoil-20. The striking difference in performance of the room-temperature zeolite trap depending upon the choice of fluid used in the diffusion pump is apparently not understood. Room-temperature traps with zeolite or alumina apparently provide no protection whatever against mercury
zeolite several times.
Flow poth
Flow path
]
1 r I
Bokeout oven
(b)
Scale-
in.
Fig.
8-43. (a)
.Scale
drawing of
backstreaming.
[Taken with permission from M. A. Biondi, in 1960 Vacuum Symposium Transactions (Pergamon Press, London, 1961).]
346
347
occurred within about a half hour after the diffusion pump was put back In a subsequent experiment Levenson and Milleron^* into operation. determined that the evaporation of less than 1 cm^ of Convoil-20 into a room-temperature trap containing about 436 g of zeolite resulted in
the eifective saturation of the zeolite. steel housing of the zeolite trap was cooled with liquid nitrogen with the
result that the pressure typically
"O
10
15
20 25 30 35 40
Timej days
45 50
dropped from 2 x 10~* torr at room temperature to 6 X lO^^" torr reached in about one hour after
application of liquid-nitrogen cooling.
Fig.
8-45. History
of
test
ionization
gauge
in
oil
chamber
by
zeolite at
room
diffusion
pump
by
room-temperature trap over a period of many days with Ootoil-S and Convoil 20 as working fluids. [Taken with permission from M. A. Biondi, in 1960 Vacuum
zeolite
Symposium
Press,
Transactions (Pergamon
1961).]
London,
temperature and found no evidence of absorption of hydrogen, helium, nitrogen, argon, carbon monoxide, and methane, but did observe some absorption for carbon dioxide and n-butane. In the tests performed by Milleron and Levenson, power failures occurred and in one into nearly the original pressure
The
that of the more effectively cooled stainless steel walls in lowering the
pressure on application of liquid-
nitrogen cooling
is
walls.
Fig.
8-47.
stance the forevacuum valve was closed and the power to the diffusion-
Liquid-nitrogen-cooled
pump
heater turned
off.
Recovery
ultrahigh-vacuum adsorption pump. (2) copper wool; (3) (1) Zeolite; copper retaining screen; (4) stainless
steel heat shield;
(5)
two
liquid-nitrogen
gauge
bakable pressure reservoir; (6) In Fig. 8-47 is illustrated release; (7) copper pinch gasket. adsorbing an ultrahigh-vacuum [Taken with permission from T. H. pump developed by Batzer and Batzer and R. H. McFarland, Rev. McFarland^* to provide a hydroSci. Instr. 36, 328 (1965).] carbon free vacuum for electron beam excitation and ionization experiments. The base pressure indicated on a Bayard- Alpert gauge is typically of the order of 10~* torr. The second application of liquid-nitrogen-cooled zeolite is in the adsorption pumping of a clean vacuum system from atmospheric
service.
down to a pressure of about 10"^ torr, at which point getter-ion pumps can take over. The obvious advantage of this combination of roughing by adsorption pumping and finishing by getter-ion pumping
pressure
'Water-cooled baffle
diffusion
pump
the complete elimination of hydrocarbon comtaminants. In this pumping unit is first outgassed by baking at a temperature of about 200C with a valve open to the atmosphere so that water vapor and other adsorbed gases can escape. The vent valve
is
isolation
zeolite.
[Taken with permission from N. Milleron and L. L. Levenson, in 1960 Vacuum Symposium Transactions (Pergamon Press, London, 1961).]
then closed and the unit allowed to cool down with both valves, the vent valve and the valve to the system, closed. The unit is then immersed in liquid nitrogen and the valve to the system opened. By
is
348
349
having a number of units of this type connected to a manifold, large vacuum chambers can be quickly pumped down from atmospheric pressure to the point where a getter-ion pump can be put into operation. This topic will be discussed further in Chap. 9. 8-7. The Pumpdown Time. The simplest situation one can imagine for computing the pumpdown time of a vacuum system is one in which there is no leakage from the outside, no outgassing from the Inwalls, and the pumping speed is independent of the pressure. tegration of Eq. (7-9) then yields the pressure as a function of the time
P^
pump and
S
^
pumping speed
is
= ^
eS^ and
2 3,
becomes
Z(l^gilZl/^ + i^gi^Zi^
Sd\
Cj
2
+ l^iliZi/f +
^3,4
etc
^2,3
(8-8)
Experience at the Kinney Vacuum Laboratory has shown that the volumetric efficiencies of pumps of different displacement speeds but of the same general design are very nearly the same at various values Thus for such a series of pumps one may write of the pressure.
Pie-<-^"'>*
(8-4)
where P^ P,
= initial pressure = pressure after pumping S = pumping speed V = volume of the system
for such
t
F=
for the time
t
2.30 P"g\
^-^^-
'g- ^-^^^2,3
'''- ^-^"^^
^3,4
etc.)
'
(8-9)
^1,2
which
is
is
all
pumps
a simple case
size.
2.30-log,^
Pi
_F
Sr
In the operating pressure range of mechanical roughing pumps the assumptions made above are frequently all nearly valid except that the pumping speed decreases with the pressure, as shown in the performance curves of Figs. 5-5 and 5-6. However, even in this case one can apply Eq. (8-5) to successive intervals of the pressure and obtain a set of time
intervals as follows
<i.2
^1\'P,
(8-10)
where P, is the last pressure value in the series in Eq. (8-9) and is the Generally pressure for which the pumpdown time is to be determined. 760 torr, but the F curve may be used in calculations of interest P^ in calculating the pumpdown time from any other initial value of the
pressure as desired.
The quantity
is
referred to as the
is is
pumpdown
= 2.30^^1ogio^
'S'1,2
particular type of
pump and
In Fig. 8-48
formance curves.
logio
F
2.30
2,3
P.
P.,
2.30
-logio-^ etc.
(8-6)
function of the pressure for single-stage and compound Kinney mechObviously F is just the anical pumps as a function of the pressure. time required for a pump to reach the pressure P( from atmospheric pressure when evacuating a chamber of volume equal to the displace-
in
which S^^^
S^^s,
speed in the pressure intervals Pj to Pj, P2 to P3, P3 to P4, the total pumpdown time is
tl.2
consisting of a mechanical
+
/lo^
<3,4
etc.
backing a mechanical booster pump can be computed from the performance curves such as those shown in Fig. 5-14 and that from in Fig. 5-5, choosing an appropriate pressure for the changeover a Such pump. bypassing to pumping through the mechanical booster
pump
2.30F
PilP:
^
logioP,/P3
etc.
(8-7)
pumpdown
factor
Si.
So
S3
The quantity within the brackets of the above equation can be obtained
directly
Sec. 5-3,
we
pump in question. If, as in designate Sjj the theoretical displacement speed of the
pump (KMBfor the combination of a 1,200-cfm mechanical booster 1200) and a 130-cfm backing pump (KDH-130). From the curves of the type shown in Figs. 8-48 and 8-49 the pumpdown time from atmospheric to any desired pressure can be calculated by multiplying the value of F on the graph by the ratio F/;S^ for the
is
shown
in Fig. 8-49
351
\
^^i
T
\
1-
14
22
20
18
\ \
\ -^
\,
t1
system, provided only that there is no significant leakage or outgassing. With the leak-hunting and welding techniques available today actual leaks of any significance need not be tolerated so that it is reasonable Out0. to insist upon a sound system in which the throughput Q avoiding and matter. By porosity gassing, however, is another thorough the by cleaning, and crevices inside the vacuum system
II
\
III
pumps
16
14
12
\
1 1
^
/I
1
^,
\<
/
/
Table
8-6.
\tmosphere - -t1
10 8 6
"
1^
Observed Pumpdown Compared with That Computed by Pumpdown Factor F Volume of System: V = 11,800 ft^ Displacement Speed of Roughing Pump: So = 500 cfm
t
^^
'
F =
t
23.6
i*'
min
V V.
^^ '^
computed,
observed,
^
-^
2
> 1
P, torr
1
min
1-
mm
35.4 52.0
68.7
System factor
1.0
0()00
001
0. 01
00
100
1,0
Pr 'ssure ,tor r
Pumpdown
factor
Kinney
.single-
1.5
30 28
26
6.2
7.1
8.1
0.5 0.3
8.8
24
0.14
10.0
0.96
1.00
1.02 1.02 1.06
1.04
1.04
1.06
1.06
1.11
1.30
3.66
22
20
\
V
\
s
t O
i o
18
14
\\
\
1
KMBl?On
\
s
1-
i" 12
10
s l/ <J ^,
\<
1/
KDH 130 /
^^F
t
ansition pressure
T
n
KMB1200
i
"v
pressure at which outgassing becomes important can be made quite low. In Table 8-6 is shown the pressure at various times during the evacuation from atmospheric pressure of a vacuum tank of volume equal 11,800 ft^ by a Kinney single-stage pump of displacement speed The 8-50. Fig. in plotted to 500 cfm. The pumpdown curve is not are times pumpdown errors between the computed and observed
large until the system pressure
\,
^ ^v
0.1
^
^
1.0
becomes
less
than about
torr.
For
^..^^
^\
-4.
"'^
increasmg dispressure values below 1 torr experience indicates an times. pumpdown observed and crepancy between the calculated the with associated mostly effects to outgassing
This discrepancy
is
N
100
^_
1,000
0.0001
0.001
0.01
10 15
Pressure, torr
Fig. 8-49.
and a
Pumpdown factor for the combination of a mechanical booster single-stage mechanical backing pump (KMB-1200 and KDH-130).
350
chamber. When presence of water vapor on the walls of the vacuum to atmospheric down" "let is vacuum a svstem which has been under evacuation subsequent on then nitrogen, pressure by admitting dry factor pumpdown the from computed that the pumpdown curve follows
due
down
to
much
352
353
The discrepancy between the actual pumpdown curve and that computed by use of the pumpdown factor (provided that the pump is in good condition) is due to the extent to which the surfaces of the
This discrepancy is therefore gases. not an indication of the failure of the pump to do its job, but is an indication of the extent to which the internal surfaces of the system
1,000
s.
0.005 to 0.010 torr on a tight, dry system, but it seems doubtful whether a continuation of the pumpdown test illustrated would ever reach a pressure less than 0.1 torr. In the above discussion the conductance of the connecting piping has not been taken into account. However, the criterion used for
selecting the pipe size for connecting a roughing
is
\
100
\ \
in determining the pumpdown time. As discussed in Sec. 2-4, the pipe size is selected to ensure an acceptably small pressure drop when the system has reached its normal operating range (i.e., the lowest pressure of practical interest). The criterion frequently applied is that the pressure drop in the line (up to the inlet of the mechanical pumps)
V
\
a
10
\ \
1
\
\
r ompu ^pH
-
should not exceed 10 per cent of the pressure. However, during atmospheric pressure this pressure drop is negligible as compared with the pressure itself. The conductance of the connecting piping is proportional to the pressure in this pressure range so that over most of the pumpdown cycle the conductance is very large indeed as compared with the pumping speed of the pump and is there-
pumpdown from
pumpdown
J<
Min.
/ Obse rved
\.
N
\
10
Otor
at 8 35
~^
n
:>
1
30
60
90
120
150
210
time within the range of mechanical pump operation. Many pumpdown experiments have been carried out under both favorable and unfavorable conditions. When there are sizable leaks present in the system or when puddles of water have accumulated at some low point in the plumbing, then the pumpdown process becomes stalled and the system factor approaches infinity. However, when
there are no leaks present, when the interior of the system has been cleaned section by section before assembly, and when no unforeseen
Fig. 8-50.
Pumpdown
down
factor F.
curve compared with that computed by use of the pumpData of Table 8-6.
adsorbed water, and possibly other between the observed and calculated pumpdown times may conveniently be called the system factor, values of which for the pumpdown data in Table 8-6 are given in the last column. Because of minor errors in pressure readings, changes in temperature during the period of the pumpdown, minor discrepahcy in the actual as compared with the assumed rotational speed of the pump, etc., a system factor in the range 0.95 to 1.05 may be considered not to be significantly different from 1.00. It will be noted, however, that the computed system factor in this case rapidly exceeds 1.05 when the pressure drops below 1 torr. The very large increase in the system factor for pumpdown to 0.14 torr is not typical and was most probably caused by the presence of a leak of the order of 40 torr cfm. The pump is capable of reaching an ultimate pressure of the order of
oil
are contaminated
by
films,
condensable materials.
The
ratio
event has created puddles of water somewhere in the system, then experience shows that rather definite values of the system factor apply to the pumpdown time, depending upon the pressure limit involved, such that
^(actual)
(calc)
V -
i^^^^^
(8-11)
The recommended system factor makes allowances for the normal outgassing of surfaces exposed to atmospheric air and provides a basis for judging whether the system is pumping down normally or whether some problem exists which must be corrected. On the basis of experience, therefore, recommended system factors are given in Table 8-7 not only for single-stage mechanical pumps, but also for compound
booster pumps.
It should be
such as letting
down
354
pressure
355
by admitting only dry nitrogen, be reahzed. those computed using the recommended system factors can apsystematic and Naundorf25 has attempted a rather complete the into extending time proach to the determination of the pumpdown solution a to leads approach lange of diffusion-pump operation. His typify based upon the graphical representation of two quantities which
Table
8-7.
pump, Naundorf was able to demonstrate good agreement between the predicted and actual pumpdown schedule. The outgassing rate was determined experimentally by closing the valve into
diffusion
Qi
Oi
a calculation of the gas load as a function of the time for most practical situations would be a formidable task. However, in the case of the relatively simple case of a stainless steel chamber 4 ft in diameter and 6 ft in length, evacuated by a
32-in
and ceramics.
The
difficulty is that
\
I
Pressure range,
torr
Single-stage
Compound
mechanical
1.0
1.1
Mechanical booster
10^
(J
mechanical
pump
pump
pump*
1.15
10
.
V
3
10
">
10
760-20 20-1
1-0.5
0.5-0.1
1.0
1.1
1.25
1.5
f Net pumping
\
\
Q2
T,
^^Gos
load
/
/
/
/
Pz
capacity
5-0^
o
0.1-0.02
^10'
P,,
0.02-0.001
*
Based upon bypass operation until the booster pump is put into operation. the Larger system factors apply if rough pumping flow must pass through gettmg and valves operating for needed time Any idling mechanical booster. the mechanical booster pump up to speed must also be added.
the throughput as a function of the pressure as represented in Fig. 8-51 and (2) the gas load as a function of the time as represented in Fig. 8-52. At every instant of time, in order for the pressure to be the observed value P, the throughput of the system Q must equal the gas load L existing at the time of the observation. the system:
(1)
T2
1
^^^
'
1
,0
10"^
10"^
10"
10"
10"'
10"'
10"
10'
10^
Pressure P.torr
Time
t,hr
Fig. 8-51.
pressure.
pump system
from
C.
Fig. 8-52. Gas load as a function of the time. [Taken with permission from C. H. Naundorf, in 1960
Vacuum
Si/m,posium,
Transactions
1961).]
Vacuum
Symposium
Transactions
1961).]
will Unless there is a dominating leak in the system, the gas load There 8-52. Fig. in decrease with the time more or less as illustrated the throughput as a is no difficulty about determining the form of function of the pressure. This curve can be quite accurately predicted from the pumping speeds of the pumps used and the conductances of traps and other components introduced. At each value of the pressure the throughput Q = PS, where S is the resultant pumping speed of the
combined system. The gas load L as a function of the time is somewhat more difficult to construct. In order to predict the form of the gas-load curve as a function of the time one must know a great deal in detail about the processes of the adsorption, chemisorption, diffusion of gases through materials, and are topics These solubilities of gases in materials of construction. are data of tables discussed at length by Dayton^" in a paper in which
given for
all
pump once each hour and measuring the rate of pressure which multiplied by the volume of the tank gave the gas load due to outgassing. The data thus obtained were plotted as shown in Fig. 8-53, and prove to be in excellent agreement with the data on outgassing of stainless steel contained in Dayton's paper. The result of combining the gas-load curve with the throughput-capacity curve of the pumping system is shown in Fig. 8-54. A horizontal line drawn through any value of the throughput and gas load intersects the throughput-vs. -pressure curve and the load- vs. -time curve. Dropping vertical lines down from each intersection yields the pumpdown time
the diffusion
rise,
This procedure
pumpdown
many
357
Inte-
shown
mcf-15,000 system
used
.''Key:
\.
Gas load
(by rate-of-rise
method)
Predicted
.
vs.
Pressure read at
V S P
for the
+C
Pj^.
(8-13)
valve inlet
gas lood
time
^y*
Measured values
and McLeod gauge
data
f
I ^
Log time.hr
10
10'' 10"'
10"'
10
10"'
10'
10^ Time,hr
10"' Pressure,
torr
10"^
10"^
Fig. 8-53. Experimental gas load as a function of the time for a stainless steel tank of total surface
Fig. 8-54. Gas load as a function of the time combined with throughput as a
function
line
area of 165 sq ft. [Taken with permission from C. H. Naundorf, in 1960 Vacuum Symposium Transactions
1961).]
A vertical of the pressure. at any value of the pressure to the throughput curve, then a horizontal line drawn to the intersection with
drawn
the gas-load-vs.-time curve, and finally a
vertical line drawn down from this intersection gives the pumpdown time for
Here P^ is a paramwhere a is the same as that in Eq. (8-12) and V is the volume of the system. The wide range over which the above pumpdown time relationship holds in practice is demonstrated in Fig. 8-57. When organic materials such as elastomers and plastics dominate the outgassing properties of a system, however, the pumpdown relationship is more complicated. In this case (P Pe)~^ is more nearly a linear function of the pumpdown time, and the equation cannot be integrated to anything approximating Eq. (8-13). From this discussion of pumpdown time in the range of diffusioneter of the system being defined as
aj V,
pump
the chosen value of the pressure. [Taken with permission from C. H. Naundorf, in 1960 Vacuum Symposium Transactions
evident that firm predictions are much more in the pressure range for mechanical pumps. Factors not taken into consideration are the use of refrigerated traps and the application of mild heating to the vacuum chamber. The
operation
it is
difficult to
make than
1961).]
vacuum designer make reasonable choices in pump sizes to make possible the attainment of the desired pressures in the specified time. However, the
the gas-load curve must be corrected and in general will have the form
shown
in Fig. 8-55.
MO"'
1
Further understanding of the problem of predicting pumpdown times that for a metal is provided in a paper by Kraus" in which it is stated by the expressed is time the of function a apparatus the pressure as equation following differential
dt
/
Sio^ Adsorbed
and absorbed
d{P
S
P^;)-
const
(8-12)
gas load
vsJimeX^
S^
2'10"'
/
/
y
^t
6
10
time
5-10"
the pressure attained after pumping for the time t, P^ is the ultimate pressure attainable after pumping for a long time, and S is the pumping speed of the system. The prediction of this equation is that P^)"! is a linear function of the pumping time the quantity (P
Inleakage and
10'
P is
permeation
10"
/A ,'V
,;>
4
10
10'
10^
Time.hr
Min
*S is
a constant.
That
this is indeed
shown by the graph in Fig. 8-56. The two curves in the figure were obtained from the same system, the steeper curve after the system had been exposed to atmosphere for only 2 mm and the less steep curve after an exposure to atmosphere of 2 hr. The value of the constant a in the above equation depends upon the initial
Fig. 8-55. Gas load as a function of the time corrected for the presence of significant leakage and
[Taken with permission t. from Th. Kraus, in 1968 Vacuum Symposium, Transactions (Pergamon Press, London, 1959).]
time
358
359
many cases affect the performance much more than minor changes
in in
10"
little
the
8-8.
original
choice
of
design
roughing pumps required for this function can be computed with very uncertainty from Eq. (8-11) using the system factors given in Table 8-7. Whether this battery of roughing pumps consists of
single-stage single-stage
Nude
ion
\,
\>
parameters.
pumps alone or mechanical booster pumps backed by pumps is an economic question which can only be answered
gouge
Cold trap
/
10"
y' Switchover to
diffusion
pump
^:)
\
10'
-3
Log(P-PE)
t
^-Log(t10"
1
K
V
1
ponents. The conventional vacuum system consists of mechanical pumps, diffusion pumps, valves, vapor traps, vacuum gauges, and interconnecting plumbing all assembled for the purpose of attaining and maintaining the specified environment in a vacuum chamber. Because the vacuum designer is faced with several alternative combinations of components which will meet the specified per100
Ion
Gote valve
Interconnecting
valve
Roughing
,0
wjvolve
10
Min
Fig. 8-57. Pressure-time curve for a vacuum annealing furnace demonstrating the wide pressure range
formance, the final choice involves judgment regarding the most convenient and economical combination of components which will serve the purpose. In this section the functions of each of the
-@)-^14
^Thermocouple
over which Eq. (8-13) is applicable. [Taken with permission from Th. Kraus, in 195& Vacuum Symposium Transactions (Pergamon Press,
components in
Diffusion-pump^
foreline trap
Freon^
compressor
Roughing pump
be briefly and some London, 1959).] combinathe specifying given for tions and capacities of components needed. Mechanical Pumps. The mechanical pumps of a conventional high-vacuum system have two rather separate functions: (1) to pump down the system to the level necessary for the diffusion pumps to be put into operation and (2) to maintain the backing pressure during regular operation at an acceptable pressure for optimum operation of the diffusion pumps. These two requirements frequently lead to very different values of the capacity for the mechanical pumps. In many large systems the time for roughing down the system is much longer than that required to reach operating pressure once the diffusion pumps and refrigerated traps can be put into operation, after which the mechanical-pump capacity required to maintain the needed backing pressure is very small. In such systems, as illustrated in Fig. 8-58, it is economical and convenient to install a battery of largecapacity mechanical pumps connected directly to the vacuum vessel by means of a bypass line to rough out the system to a pressure below that at which the diffusion pumps can operate. The capacity of the
conventional high-vacuum system. (1) Vacuum chamber; (2) internal liquid-nitrogen thimble trap; (3) liquid-nitrogon-cpoled diffusion pump baffle; (4) gate valve (in optimum design the connecting pipe would be as short as possible; (5) Freon-cooled baffle; (6) diffusion pump with Freon-cooled exhaust condenser; (7) forevacuum oil vapor trap; (8) forevacuum valve; (9) backing pump; (10) roughing pipe and valve with oil vapor trap; (II) roughing pump; (12) interconnection between roughing and backing lines with close-off valve; (13) ionization gauges nude and tubulated in vacuum
Fig.
8-58. Representative
chamber;
(14)
lines.
by computing the pumpdown time for various combinations of pumps and their associated plumbing and then comparing the resulting
performance with the cost of each combination. In principle the determination of the capacity of the backing pump is a simple matter. If the throughput Q of the system during normal operation is known, then the pumping speed for backing is Sj, = QjPi,, where Q is the throughput and P^ is the backing pressure required during operation. This determination is generally much more difficult to judge in advance than the pumpdown capacity required because of the uncertainty in the value of Q due to gas flow, outgassing, and
360
361
permeation. The system should capacity to override minor leaks sufficiently to get the diffusion pumps An interconnection between into operation and expedite leak hunting. in Fig. 8-58 can be invaluable the backing and roughing pumps as shown with this added flexibility even However, during periods of difficulty. backing pump above the the of capacity ill the system, a factor of 2 in
calculated from the anticipated throughput is recommended. Because of uncertainty in the knowledge of the value of the throughput, an even larger margin in capacity may be required. Even with a fairly generous factor applied to the throughput for determining the
have
some
excess
accidental leakage or a larger gas flow for whatever process is involved, this additional construction cost will in most cases be at least partly compensated for by the reduced pumping and processing time which usually result from excess pumping capacity. A vacuum system with
minimum
is too small to do the allotted job is much less economically sound than one which has a moderate excess capacity. Loss of time during operation can be very expensive and in a short time dissipate the initial savings one might make by installing insufficient
pumping
ventional
capacity.
Accessories.
capacity of the backing pump, however, in most cases that capacity is very much smaller than that required for the roughing pump. Operation of a system thus usually consists of pumping the system down through the bypass and roughing pumps to a pressure of perhaps 0.1 or 0.2 torr, then closing the bypass valve and opening the gate valve The diffusion pump has presumably already into the diffusion pump.
The
accessories
from those
shown
1. Multiplicity of ionization gauges. In many large systems it is convenient to install ionization gauges in pairs, one with and one without a glass liquid-nitrogen trap. The discrepancy between the gauges
been in operation with the gate valve closed and backed by the backing pump. The next step is then to cool down the liquid-nitrogen-cooled baffle over the diffusion pump and also to fill the thimble trap with Meanwhile the liquid nitrogen (assuming a thimble trap is used). system down the roughing used for pumps mechanical capacity large
due primarily to condensables (mostly water vapor) so that an experienced operator can readily ascertain the condition of the system and diagnose many troubles.
is
2. Thermocouple or Pirani gauges are indicated in Fig. 8-58, but the advantage of a multiplicity of such gauges in the roughing and backing sections of the system should be emphasized.
3.
pump and
Diffusion Pumps. The pumping speed required for the diffusion associated baffles and gate valves must also be considered from the point of view both of pumpdown time and of the required operating pressure. The pumpdown time can best be approached by
flexibility
introduced
is
by the
well
method ofNaundorf^^ outlined in the previous section. For various combinations of diffusion pumps, gate valves, and baffles one can estimate the throughput capacity of the system as a function of the pressure. From the outgassing data supplied in the paper by
the
Dayton^' and the exposed areas of various materials one can construct a gas-load curve as a function of the time. By combining these curves as in Fig. 8-54 the pumpdown time as a function of the pressure can be roughly predicted for any particular combination of diffusion pump, baffle system, and gate valve for which the overall pumping speed is known. The choice from this point of view must then be compared with the pumping speed required to maintain the desired operating pressure for the predicted gas load by applying S = QjPo, where P
is
worth the cost. Aside from the gate valve for isolating the diffusion pump from the vacuum chamber, valves should be installed at the following positions (a) at the vacuum chamber end of the roughing line, (b) in the forevacuum line near the outlet of each diffusion pump, (c) at the inlet of each mechanical pump, either roughing or forevacuum, and (d) in a line interconnecting the forevacuum and roughing lines. Also recommended are small, normally closed valves installed between each shutoff valve and its mechanical pump for testing and diagnosing the source of trouble in the system, and a small, normally closed valve on the vacuum chamber for letting down the chamber to atmospheric pressure. Provision should be made to admit commercial dry nitrogen or dry air through a drying unit.
:
is
baffle,
and
from
latter con-
Conventional vacuum systems of the type described above should give excellent service with base pressure (untrapped ionization-gauge reading) of 10^' torr and should perform well in the range of 10-^ torr. When operation at significantly lower pressure is desired, the
techniques of ultrahigh
results
to override
vacuum
are required.
This
is
next chapter.
362
REFERENCES
1.
2.
3.
L. L. Levenson, Xorman Milleron, and D. H. Davis, in 1960 Vacuum Symposium Transactions (Pergamon Press, London, 1961), p. 372. Transactions I. Farkass and E. J. Barry, in 1960 Vacuum Symposium (Pergamon Press, London, 1961), p. 35. A. Guthrie and R. K. Wakerling (eds.). Vacuum Equipment and Techniques
CHAPTER
4.
5. 6. 7. 8. 9.
10.
(McGraw-Hill Book Company, New York, 1949), pp. 148-158. R. R. Addis, Jr., L. Pensak, and Nancy J. Scott, in 1960 Vacuum Symposium Transactions (Pergamon Press, London, 1961), p. 39. R. R. Wilson, Rev. Sci. Instr. 12, 91 (1941). R. H. V. M. Dawton, Brit. J. Appl. Phys. 8, 414 (1957). R. W. Roberts, Rev. Sci. Instr. 32, 750 (1961). J. F. Gerber, Rev. Sci. Instr. 34, 1111 (1963). F. N. D. Kurie, Rev. Sci. Instr. 19, 485 (1948). J. S. Wahl, S. G. Forbes, W. E. Nyer, and R. N. Little, Rev. Sci. Instr. 23,
379 (1952). William E. Bush, A. Guthrie, and R. K. Wakerling (eds.). Vacuum Equipment and Techniques (McGraw-Hill Book Company, New York, 1949), Chap. 4, p. 179. J. W. Johnson and W. M. Good, Rev. Sci. Instr. 32, 219 (1961). Norman Milleron, in 1958 Vacuum Symposium Transactions (Pergamon Press, London, 1959), p. 140. J. R. Ullman, in 1957 Vacuum Symposium Transactions (Pergamon Press, London, 1958), p. 95. L. L. Levenson, Norman Milleron, and D. H. Davis, in 1960 Vacuum Symposium Transactions (Pergamon Press, London, 1961), p. 372. A. R. Taylor, in 1961 Vacuum Symposium Transactions (Pergamon Press, London, 1962), p. 1328. H. R. Smith and P. B. Kennedy, in 1959 Vacuum Symposium Transactions (Pergamon Press, London, 1960), p. 271. John Strong in collaboration with H. Victor Neher, Albert E. Whitford, C. Hawloy Cartwright, and Roger Hayward, Procedures in Experimental
Physics (Prentice-Hall, Inc., Englewood
Cliffs,
ULTRAHIGH VACUUM
11.
The term ultrahigh vacuum has come into use in recent years to designate the range of pressure below about IQ-' torr which cannot easily be attained by the conventional methods and techniques described in the previous chapter. In order to reach significantly lower pressure, additional or alternative techniques must be applied. The techniques thus far found to be useful in attaining operating pressures in the range IQ-s to lO"" torr or lower will be briefly described in this chapter.
9-1. The Dominance of Surface Phenomena. From the papers of Daytoni-3* on the outgassing of "clean" metal surfaces at room temperature it is evident that after exposure to normal atmospheric air for several hours the amount of gas readily available for desorption
12. 13.
14.
15.
16.
from the surface at room temperature amounts to many molecular layers. As an example, Dayton's tables^ show that after 10 hr of
outgassing rate for a stainless steel surface is and is decreasing very slowly. Therefore to maintain a base pressure of IQ-s torr in the presence of such an outgassing rate requires a pumping speed of at least 20 liters/sec for each square foot of internally exposed surface. Most large vacuum
17.
X 10-5
18.
19.
20.
21.
22.
23.
24. 25.
26.
27.
D. Alpert, Rev. Sci. Instr. 24, 1004 (1953). J. H. Carmichael and W. J. Lange, in 1958 Vacuum Symposium Transactions (Pergamon Press, London, 1959), p. 137. M. A. Biondi, in 1960 Vacuum Symposium Transactions (Pergamon PreSs, London, 1961), p. 24. N. Milleron and L. L. Levenson, in 1960 Vacuum Symposium Transactions (Pergamon Press, London, 1961), p. 213. L. L. Levenson and N. Milleron, in 1961 Vacuum Symposium Transactions (Pergamon Press, London, 1962), p. 91. T. H. Batzer and R. H. McFarland, Rev. Sci. Instr. 36, 328 (1965). C. H. Naundorf, in 1960 Vacuum Symposium Transactions (Pergamon Press, London, 1961), p. 60. B. B. Dayton, in 1959 Vacuum Symposium Transactions (Pergamon Press, London, 1960), p. 101. T. Kraus, in 1958 Vacuum Symposium Transactions (Pergamon Press, London,
1959), p. 38.
chambers consist of outer walls and a complex inner structure, the which must be considered. Also the wall area
vacuum pumping is usually limited by the many other demands of the system for access ports, high voltgage insulators, and a variety of accessories essential to the vacuum process. The result is
available for
that a design figure of 20 liters/sec for each square foot of internal surface can generally be realized or even somewhat exceeded in practice so that the base pressure is limited to about lO"" torr even after many hours or days of pumping.
In the previous chapter mention was made of speeding process of outgassing by increasing the temperature of the
up the
vacuum
363
364
ULTRAHIGH VACUUM
365
chamber, and this practice has been followed for many years. Howsystem is not ever, the gain in the ultimate pressure attainable by a The factor. large a by 100C) to (say baking improved by a mild typical the reach to required time the that principal advantage is To reduce significantly limit of the system may be greatly reduced. at temperatures baking requires pressure the attainable operating
experimental data of this type in the curves shown in Fig. 9-1 for steel and in Fig. 9-2 for aluminum. Based upon the macroscopic surface area, the total amount of gas given off by metal surfaces at room temperature over a period of 10 to 50 hr of vacuum pumping ranges from 20 to 100 molecular layers. What is perhaps more to the point is that untreated
much
This requirement introduces a number of complications into the design which are not encountered in conventional
greater than 100C.
vacuum-system design. Experience has shown that outgassing from metal surfaces in vacuum The character of the surface is predominantly due to water vapor. discussed by many investiis surfaces metal deposition of water on Hebling, Mongodin and and Lichtman gators: Kraus,* Hayashi,^ the surfaces of metals Because Noller.* and Prevot,' and Flecken consist of somewhat construction vacuum-chamber generally used in state of the physical exact the defining of problem porous oxides, the metal is complicated. the of surface the on contaminants water and other As Dushman* has explained, there are three mechanisms by which a the gas can be taken up by a solid material, all generally included under
general term sorption.
10-7
Chemisorption refers to the formation of a chemical compound by interaction of a gas with the wall material, as in the case of the formation of an oxide film. on a metal 2. Adsorption refers to the surface condensation of a gas of only film a in This process is generally believed to result surface.
1.
9-1. Outgassing rate versus time for steel at room temperature. Curves for rusty steel, sandblasted
Fig.
a very few molecular layers of gas. molecules penetrate 3. Absorption refers to a process by which the gas dissolved into the interior of the wall material and in a sense the gas is
in the solid.
steel, and stainless steel (curve 1) from the data of Blears et al.^" Stainless steel curve 2 from the data of Gellor,!' and curve 3 from that of Basalaova'^ on untreated stainless steel. [Taken with permission from B. B. Dayton, in 1961
Vacuum Symposium
Transactions
1962).]
J'iG. 9-2. Outgassing rate versus time at room temperature. Curve 1, untreated duraluminum, and curve 3, duraluminum scoured and washed with benzol and acetone, are based on the data of Basalaeva.i^ Curve 2, duraluminum, is based on the data of Geller.ii Curve 4, aluminum bright-rolled and cleaned in Stergene, and curve 5, anodized aluminum, are based on the data of Blears et al.^" [Taken with permission from B. B. Dayton, in 1961 Vacuum Stjmpo-
The outgassing history is expected on theoretical grounds to depend incritically upon which of these three mechanisms of sorption are is extensive, very been has which question, volved. Research on this The area. surface effective the of complicated by the definition roughness factor (i.e., the effective microscopic surface area compared with the gross, or macroscopic, surface area) inferred by various investigators ranges from about 20 to 100 depending upon the details applied of outgassing experiments and the assumptions of the theory lack the and theory the of complexity the Because of to the situation.
the of sufficient detailed knowledge of the microscopic character of experiupon based problem the to surfaces, an empirical approach mental outgassing results at various temperatures seems to be the only
practical course to follow at the present time
.
metal samples outgas at the rate of about I0-' torr liter/sec cm^ after 1 hr of vacuum pumping at room temperature, and this rate of outis about inversely proportional to the pumping time. These statements have to do with room-temperature outgassing for which water vapor is by far the dominating substance. At high temperature other factors, such as diffusion of absorbed gases through the metal, become important so that high-temperature outgassing cannot be inferred simply by integrating the room-temperature curves and compressing the resulting output over a shorter period of time. 9-2. High-tetnperature Bakeout. In the previous section it was stated that operating pressures much less than about 10^* torr
gassing
366
ULTRAHIGH VACUUM
conventional stainless steel diffusion
367
cannot easily be attained because of the long-persisting outgassing of The total amount internally exposed surfaces at room temperature. of gas available on metal surfaces is so great that even with the inverse dependence upon the pumping time of the outgassing rate the base pressure attainable is seldom as low as lO"' unless some action is taken
to change drastically the source of gas available for desorption.
pump
The liquid-nitrogen-cooled vapor trap was that 8-30, featuring an oil-creep barrier and a water- or
oil
radiation-
The
Mc X system
10-^
te
mp
400C
on first bake
/Sp
bokeout
/
/Max
10'* /
//I / /
'
\
\
\ \^
N.
Cut heat
system
\^
>\<
10"
:
Cut heot
\\ -^^ \|
1st
bake
System
cooling
1
^
1 1
X
1
Air
\
\
1
\l
f"s
1
\
Pressure pips occurred wherA.
\
Time
to heot
'
^^
1^
8hr
muffle
>C^y-K
A^i
1
\ aK
1 1
1
1x10-'
-5x10"'
1
vapor from the diffusion-pump jet. The volume of the vacuum chamber was 70 liters and was equipped with metal gaskets to permit bakeout temperatures up to The vacuum chamber had been used for many previous 450C. ultrahigh-vacuum experiments so that the stainless steel walls were far cleaner than those of a newly constructed chamber. As indicated in Fig. 9-3 the base pressure attained with the chamber at room temperature was about 2 x 10"' torr. During bakeout the temperature reached the value of 400C in about 1 hr and was maintained at that temperature for about 5 hr, the heater power having been on for The pressure in the chamber reached a maximum of 10~* torr 6 hr. at about the same time the temperature reached 400C and then slowly decreased until it had declined to a value of about 2 x 10"' torr after 6 hr of baking when the heater was turned off. A crude integration of the area under the outgassing curve yields a total gas output of about 30 torr liters during the bakeout, since the measured pumping speed of the system was about 40 liters/sec. The projected area of the interior walls of the vacuum chamber was about 7,500 cm^.
Since the number of molecules in 1 torr liter of gas is 4.5 x 10", the gas removed during bakeout was equivalent to 30 x 4.5 x 10"/7,500 2 X 10" molecules/cm^ of projected area. If one assumes a roughness factor of 50, which may be somewhat generous, the gas extracted from
10"
1
1
>|~2weeks
Bokeout
started
Time,lir
Fig. 9-3. Plot of the pressure versus the time for two typical bakoout cycles of a chamber evacuated by an oil diffusion pump through a liquid-nitrogen trap with and without anticreep barrier. [Taken with permission from N. Milleron,
in 1958
Vacuum Symposium
the metal during bakeout was equivalent to about 40 molecular layers. At the pressure existing before bakeout the time required to remove this quantity of gas would be about 100 hr. However, at the end of the bakeout when the chamber reached room temperature the pressure
most
change is to bake the and including the liquid-nitrogen-cooled baffle system at a high temperature for an extended period of time. The supply of gas sorbed in the metal is desorbed rapidly and pumped at a much higher pressure during the period the chamber is maintained
effective
means of
entire
vacuum chamber up
to
at high temperature, so that the total throughput of desorbed gas increased during this period by a large factor. When the system
is
had dropped to 5 x lO"" torr, where it remained for the period of the test with no indication of rising. A similar test run without the oil-creep barrier in place reached a base pressure of about 1 x lO"" torr after a shorter bakeout period (about two hours) but then began to rise rapidly after about two weeks operation and in a period of one day rose back up to a value of 1 X 10~' torr, where it then remained. For the conditions of this test, therefore,
the surface migration time for the oil film to bypass the liquid-nitrogencooled trap along the approximately room-temperature surface of the
is
then allowed to cool back down to room temperature, the great quantities of easily desorbed gas have been removed from the surfaces, the
outgassing rate is decreased by many factors of 10, and the resulting base pressure is correspondingly decreased. The results of two typical bakeout cycles reported by Milleron" The system was evacuated by a are shown graphically in Fig. 9-3.
housing surrounding the vapor trap was apparently about two weeks. From this result one may conclude that the oil-creep barrier is not a necessity for the attainment of ultrahigh vacuum with an oil diffusion pump as long as the period of interest is less than about two weeks, but for longer periods the barrier is an absolute necessity.
368
ULTRAHIGH VACUUM
369
gouge
results are therefore not indicative of the gas content of raw nickel generally, but illustrate a method of quantitative analysis which
w
Go-ln-Sn seol
should be of considerable help in disentangling the contributions of adsorption, absorption, chemisorption, and diffusion to the process of
To
oil
outgassing.
ditfusion
pump
Even though processes involved in sorption and outgassing are not yet understood quantitatively in the detail desired, experience has shown that a well-designed diffusion-pump and vapor-trap system
augmented by baking the vacuum chamber at temperatures from 250 to 450C can attain base pressures of 10^" torr or less. Experience further indicates that a newly fabricated and carefully cleaned system must generally be subjected to a baking temperature of at least 300C
for
Furncce
about 24 hours in order to reach the lO"!" torr range. Subsequent bakeout temperatures can be at lower temperatures and for shorter periods of time, depending upon the character of the gas exposure
Thermocouple
Fig. 9-4. Gas analytical apparatus for analysis and quantitative moasuroment of gases given out during bakeout. [Taken with permission from P. F. Varadi, in 1960 Vacuum Symposium Transactions (Pergamon Press, London, 1961).]
Outgassing experiments of the type just described are very informative from an empirical point of view and lead to concrete indications of the ultrahigh-vacuum capabilities of oil diffusion pumps coupled with appropriately designed vapor traps. However, the outgassing in the case just discussed was measured with an ionization gauge so that the details of the outgassing process in terms of the This gases being evolved as a function of time were not determined.
point
is covered by Varadi^* in a cleverly executed experiment in which the throughput of the evolved gases are analyzed and measured quantitatively during the bakeout cycle. Varadi's apparatus is
and incorporates a known conductance and r-f mass spectrometer to identify and measure the quantity of each component of gas evolved as a function of time as the sample is put through the bakeout cycle. In Fig. 9-5 are shown graphically the quantitative
results of outgassing a nickel cathode
The data shown in the figure are taken from three separate samples prepared in the same way, showing the excellent reproducibility of the results. The predominance of hydrogen during the early phase of the outgassing period and its subsequent rapid decrease to a minor constituent is an interesting feature of the
for a period of several minutes.
process.
Since the materials being tested in Varadi's experiment are cathodes already prepared for use in vacuum tubes, the material had been treated for an hour in a hydrogen atmosphere at 1150C. The
Fig. 9-5. Graphical presentation of the thermal degassing properties of a cathode nickel material. [Taken with permission from P. V. Varadi, in 1960 Vacuum
Press,
London,
1961).]
370
ULTRAHIGH VACUUM
371
during the intervening period. Best results are obtained by admitting nitrogen from evaporating liquid nitrogen to the system and maintaining a flow of dry nitrogen while the system is open so that a pressure slightly By in excess of atmospheric pressure is maintained in the system. this procedure water vapor can be almost completely excluded from
the system for the subsequent pumpdown cycle. The necessity of baking the vacuum chamber and all its accessories at high temperature introduces a number of
Flange
Cylinder
wall
temperatures used for baking out the system are not stated. Considering that these techniques were in use in the mid-thirties emphasizes that the eff'ort to develop satisfactory metal gaskets has a long history.
tomers are restricted to a relatively limited temperature range, the sealing of flanges and valve seats presents serious problems to which reasonably satisfactory solutions have
experience" was in the design and construction of the accelerating tube for the 2.5-million-volt Van de Graaff generator built at MIT in 1937. The vacuum system was evacuated by mercury diffusion pumps and was constructed of porcelain and steel with fuse-wire gaskets at all metal-to-porcelain
lOmils
been found.
9-3.
20mils
Metal
Gaskets.
The
fact
that
vacuum
Fig. 9-6.
flange
Metal gasketed
design described by John Strong. [Taken with permission from John Strong, Procedures of Experimental Physics
(Prentice-Hall,
Englowood
systems because of outgassing and permeation has been recognized for many years. In Chap. 8 the properties of a number of elastomers are discussed and the advantage particularly of Viton A and B for use in systems with bakeout temperatures up to 300C is mentioned. Another technique discussed is that of circulating a coolant rings through a guard ring between two with the result that the outgassing and per-
about 250C and that the fuse room temperature. If the mechanical design of the flange is such that the compressional loading is mainwire cold-flows even at tained as the fuse-wire material
creeps, the gasket eventually be-
20
mils
meation of the 0-ring material is substantially reduced. Alternatively, a copper tubing for circulating the coolant may be installed in the guard ring thus avoiding a rather messy situation when the flange is disassembled. These alternatives are available, and in those instances in which the resulting base pressure is satisfactory for the process in hand more exotic solutions to the problem may not be justified. The fact that the best of elastomers, even with the improved techniques for their use, are not satisfactory for many applications still remains. The use of various softer metals such as gold, indium, and
lead fuse wire has been mentioned in the
years.
comes very thin and usually deIn spite of this feature fuse-wire gaskets have been used to advantage in a great many situations in which the available elastomer gaskets could not be used. Another metal wire material which has proved to be useful is an indium- tin alloy which is, however, limited to bakeout temperatures of about 100C. This material has the advantage that fusing joints is not necessary. The wire can be overlapped at the joint and will seal
velops a leak.
tightly when the flange surfaces are pulled up snugly compressing the gasket joint. On large or irregularly shaped flanges this feature of the indium-tin wire may be a distinct advantage. In recent years with the greater demand for ultrahigh-vacuum systems requiring high-temperature (450C) bakeout there has been
[Taken with permission from Lawrence Radiation Lab. Eng. Note EXA-122, Mar. 16, 1961, by T. H. Batzer and J. R. Ullman.]
Fig. 9-7. Copper shear gasket.
a greatly enhanced effort on the development of metal gaskets. Copper gaskets have been developed in a number of different forms
In Procedures in
described.
It is
emphasized that lead wire gaskets joined by butt welding to form a ring can be used to high bakeout temperature and that aluminum wire remains vacuum tight for even higher bakeout temperatures. The
and have been used with some degree of success. In Fig. 9-7 is illustrated the copper shear gasket" developed for the Atomic Energy Commission by the Westinghouse Research Laboratories. The mating flanges are designed with an interleaving step form so that when the two flanges are forced together under loading from the bolt circle, the flat copper gasket 0.045 to 0.050 in. thick is sheared to a depth of
372
ULTRAHIGH VACUUM
373
about one-half
thickness.
was at
first
recommended
for this
on the Stanford
of copper ridge seal is described by Goerzi" 2 -mi electron accelerator. As shown in Fig.
developed during high-temperature bakeout. Details of the flange and gasket design were developed in some detail by Batzer and Ullman," who found that annealed cupronickel is a much more satisfactory material than copper for As developed this type of seal. <- 0.015 in. + 0.005 Ullman the shear and Batzer by 0.005 gasket is very reliable for use +0.002_i_ with stainless steel flanges and bakeout temperatures up to 450C. t The proper machining of the flanges requires adherence to close
in.
r-
the rather thick (3/16-in.) copper gasket has a narrow ridge of in. and width at its base of 0.018 in. When the gasket is compressed between the flanges, the ridges are distorted as shown in the figure. The complete assembly of the waveguide flange
height equal to about 0.010
joint
is
shown
in Fig. 9-10.
This case
is
wave-
2A
16
Clomping bar
tolerances,
.
and the
resulting prod-
uct
in.
is
easily subject to
damage.
A>' /i6
Fig.
in.
B>V32
in.
V8
9-8. Cross
section
of the
soft
copper bead gasket before compression between steel flanges with plane faces. [Taken with permission from N. Milleron, in 195S Vacuum Symposium Transactions (Pergamon Press, London,
1959).]
scratches.
gasket
must be replaced.
is
The
Before seolinq
After seoling
principal disadvantage
tively
high cost as compared particularly with the aluminum-foil gaskets to be discussed later. Milleron^^ describes a form of gasket which he calls the copper bead The gasket as gasket, the cross section of which is shown in Fig. 9-8. shown in the flgure is placed between polished flanges with plane surfaces and is compressed by a symmetric tightening of the bolts on the bolt circle. The total area of the protruding beads is small with the result that the concentrated loading of the flanges flattens the beads
until the steel flange surfaces
Fig.
Fig. 9-10. Waveguide flange assembly utilizing ridged copper gasket of Fig. 9-9. [Taken with permission
from
1961).]
D.
Vacuum
and therefore
is
come
much
larger
area of the main body of the gasket. Further tightening of the bolts results in no signiflcant further compression of the gasket but a stretching of the bolts, the stress due to which keeps the seal tight during the bakeout and cooling cycle. This gasket configuration proved to be
There have been a number of designs of knife-edge seals reported, one example of which is that reported by Lichtman and Hebling^' and
illustrated in Fig.
9-11.
much
superior to copper or
aluminum wire
OFHC
copper gasket
is
area and therefore develop flat, there are no difficult tolerances to hold or expensive operations to carry out in the fabrication of the flanges. Making satisfactory copper bead gaskets, however, has proved to be somewhat more difficult, but the problem has been solved by the use of chemical milling using concentrated
eff'ective
silver-plated. Both stainless steel flanges have machined on them circular ridges, referred to as knife edges, which deform the copper gaskets as the flanges are drawn together. The mechanical process involved would appear to be somewhat similar to that of the copper ridge gasket. In any case the authors report excellent reliability at bakeout temperatures up to 600C with a system utilizing many such
seals
nitric acid.
up to a flange diameter of 12
in.
374
ULTBAHIGH VACUUM
Stellarator devices at Princeton
375
and
is
described
by Grove. ^^
Grove
O.OIOR
attributes the origin of this type of seal to Hickam,^^ but the idea seems to have occurred independently to others as well. According to
0.001 or 0.002
A-7o-A
Stainless -steel
Grove the important factors in the design are a radial clearance of in. between the inner and outer members of flanges which slide together, a surface finish of 16 microin. and a snug fit for the
^>^swivel flange
stainless steel
gold ring in the corner of the flange. The material is 24-karat gold. The initial diameter of the wire is 0.020 in., and it is compressed to 0.010 in. when the flanges are pulled tightly together closing the
Stoinless-steel
Silver-plated
oxygen -free
copper washer
in
[Taken with permission from D. Lichtman 1960 Vacuum Symposium Transactions (Pergamon Press,
between them on the inner surface. These seals have repeatedly withstood bakeout to 450C without failure, and were in use in sizes in. to 8 in. diameter at the time the report was written. from Power and Robson^^ summarize a series of experiences using wire One difficulty encountered gaskets between flat stainless steel flanges. during their tests, and indeed also encountered by others, is the tendency of an aluminum-5 per cent silicon wire gasket to form a strong bond to the surface of the flange, resulting in tearing the gasket and
crevice
Lichtman and Hebling^" also report on an all-metal flare seal shown which provides a compact and reliable means of connecting a small-diameter copper tubing to a fitting on a manifold or chamber. The authors state that the seal is assembled by slipping the copper tubing through the stainless steel flange and then flaring the tubing with a standard tool. The seal is said to be reliable even after several cycles of disassembling and reassembling and to withstand the high
in Fig. 9-12
leaving portions thereof tightly stuck to the flange. Their practice was to compress the 0.040-in.-diameter wire to 0.011 in. between the flanges, controlling the limit of compression by using shims placed
between the flange faces. The adherence of the gasket material to the flange is a major nuisance since cleaning up the flange for subsequent use is then a tedious chore. Power and Robson report that this
completely eliminated by applying a surface coating of indium to the aluminum wire gasket. An unexpected advantage is that the compressive force required is reduced by a factor of 2 or 3 when the indium coating is applied. The authors report that the difficulties in making successful seals in the manner described increase with the dimensions of the flange, and that they have not been consistently successful in sealing flanges for 400C bakeout for diameters
difficulty is
seal
vacuum
designs.
in Fig.
Copper tubing
Compression
of 12 in. or greater.
Fitting
Braze seal
^Manifold
tubing
0.020 diam
gold wire
Fig. 9-12. All-metal flare seal. [Taken with permission from D. Lichtman and A. Hebling, in 1960 Vacuum Symposium Transactions (Pergamon Press, London,
1961).]
Fig. 9-13. Simple gold seal. [Taken with permission from D. J. Grove, in 1958 Vacuum' Symposium Transactions
1959).]
metal gasket development of particular promise is that of Batzer and Ryan, 2* who have devised a bakable aluminum gasketed flange which appears to be more reliable than any previously described up to The large diameters and is far less expensive than most other designs. flanges, mating 9-14. The aluminum-foil flange is illustrated in Fig. which may be either stainless steel or aluminum, are machined with mating faces which are slightly conical at the angle Q relative to the plane of the flange. A gasket is cut from commercial aluminum foil 0.015 in. thick (using tin snips, for example) and placed between the flanges, which are then bolted together. Contact will initially occur at the inner diameter of the contacting conical surfaces, but the point of contact wifl move to larger and larger radii as the bolts are tightened and the flanges
>
376
ULTRAHIGH VACUUM
377
as a function of the
modulus of elasticity
of the flange material, the dimensions, and the applied loading. The aluminum-foil seal is loaded at 2,000 to 3,000 Ib/lin in. of gasket.
The
Pipe
0D + 2t
gasket is under compression for a wide range of temperature. Thermal cycling of aluminum-foil seals
thoroughly reliable is that illustrated in Fig. 9-15. This seal has been used extensively in ultrahigh-vacuum applications by the Compagnie General de Telegraphic sans Fil. The flanges are stainless steel and are machined with a ridge on one flange and a mating groove on the Since the half-angle of the ridge is 30 and that of the groove other. is 45, the aluminum foil, which is compressed into the groove by the The flange distortion ridge, is under a tapering compressional load. stored energy necessary to the provides and stretching of the bolts at all times during sustained will be ensure that a compressional load
the bakeout.
0.0015 Alfoil
type 1145-0
7/16diam(approx, 2.00"
spacing)
All dimensions
in
mm
Fig. 9-14.
flange.
aluminum
foil
over-
cr(max)
Maximum = MRjZ.
in
flange:
Rotational deflec-
tion in flange:
= MR^jEI.
moment due
reaction.
R =
to bolt load and gasket radius of bolt circle. flange section modulus.
*nnii
ntO.l
r^r^
0*0
From
r+u.i
this
experience
'.-to
-0.3mm
Aluminumrto.l
foil
attempted using the aluminutti-foil technique with overlapping secVacuum Society, Inc.] tions in order to meet an, actual need in the controlled-fusion program. Presumably because of the relatively low local pressures developed between the flanges and the foil gasket no serious difficulty of bonding of the gasket material to the flanges has been experienced. From the performance to date the aluminum-foil seal appears to be the most economical and successful demountable, bakable all-metal seal for ultrahigh-vacuum application. Another form of aluminum-foil seal which has been found to be
Symposium Transactions. Copyright 1963 by The American
Vacuum
iJliil
p tO.2
Aluminum-foil gasket with wedge seal flange of the Compagnie per Generale de Telographie sans Fil. The circular gasket is cut from 99.95 [Taken with permission from thickness. cent pure aluminum foil of 0.3 Boetie, Pans.] Societe de Reoherches Techniques et Industrielles, HI, rue la
Fig.
9-15.
mm
378
ULTRAHIGH VACUUM
a differential
379
method
gasket of any kind has been described by Farkass and Vanderschmidt.^^ The method as illustrated in Fig. 9-16 consists of welding onto each flange a thin, stainless steel skirt which protrudes out beyond the The machining of the flange to provide a thin lip of metal flange.
onto which to weld the skirt is a typical example of the proper preparation for a Heliarc weld of a thin section of metal to a heavy
member.
The two
flanges
are
mounted
in
position with the outer peripheries of the skirts in contact, and the edge is then welded shut all
the way around. The resulting joint is then When the flange completely sealed by welding around the edge weld the disassembled, is to be rewelded can be a seal Such off. is ground
.
replaced.
several times before the old skirtings must be The advantage of absolute welded
tightness
is
obvious.
In
many
have to be ground off and the seal rewelded might become a major handicap. In the previous 9-4. Bakable Valves.
Fig. 9-16.
seal connection
Reweldable between
two heavy
to
flanges,
up
30
in.
in
vacuum valves of several varieties and illustrated. Many of the described were have metal bellows to transdiscussed designs
chapter,
mit the motion for opening and closing the valve port. However, in all the examples given the sealing of the valve port is accomplished by
so that the
is
maximum
limited
by the choice
is
bakeout to
The and
steel-fabricated valve
Fig. 8-20
capable of
gaskets.
and
The same
somewhat higher-temperature
For many ultrahightemperature must be much higher vacuum applications the bakeout than any available elastomer will tolerate. The first major step toward the solution to the bakable valve problem is that reported by Alpert^' and illustrated in Fig. 9-17. The valve consists of two main components: (1) a vacuum-tight capsule with a heavy OFHC copper base sealed over one side by a corrugated Kovar diaphragm which carries a Kovar plunger which can be driven snugly into one of two 0.25-in. holes drilled through the base and
bakeout with Viton
or
screw thread by which the (2) a driver assembly with plunger is driven into or withdrawn from the sealing hole. Kovar tubes are soldered into the two holes in the copper base plate and are connected to the system either by glass or metal tubing. The Kovar plunger has a 60 tapered tip and forms its own seat by deforming the copper about the entrance to the sealing The conductance of the valve when hole. closed is reported to be in the range 10"" to 10-11 liter/sec which is almost beyond the detectable limit. Valves of the Alpert type are available commercially and permit operation in the pressure range IQ-i" torr or less. The virtue of this type of valve is that it may be thoroughly outgassed by high-temperature bakeout either in the open or closed U'-'-l--i Y'l position. Its principal disadvantage is its Fig. 9-17. All-metal bakrelatively low conductance even when fully able vacuum valve. The open. valve body consists of valve having many larger A considerably a 1^ -in. -diameter (1) of the features of that just described is that copper cup with l/4-in.reported by Grove^^ and shown in cross sec- diameter holes for the The open conductance of valve seat and openings, tion in Fig. 9-18. (2) a flexible Kovar diathe valve shown is 80 liters/sec and the closed brazed to the phragm conductance is typically 10"^ liter/sec. In copper cup and to (3) a this design the nose piece with the 45 face Kovar member with a is of copper and the seat is stainless steel highly polished conical surwith a 1%-in. opening. A stainless steel face. The driver mechanism consists of (1) an bellows allows adequate motion to provide outside screw and stainless Valves the open conductance stated above. steel housing, (2) a siliconof this design have been used extensively in bronze differential screw, the construction of Stellarators and have (3) a drive screw, (4) a plate to withstand backing position been baked at 450C in the closed the large forces exerted their in repeatedly without deterioration upon the valve body.
ability
to
seal
properly.
The
differential
[Taken
with
permission
screw method of applying final loading on from D. Alpert, Rev. Sci. Instr. 22, 536 (1951).] the valve is provided by a separate assembly which is removed when the valve is baked. A new design principle has been introduced in the gasketless valve described by Wishart and Bancroft." When a conical spring washer is flattened by compression between two flat surfaces, the washer
increases in diameter.
ULTEAHIGH VACUUM
381
Flattened position
Relaxed
position
the aiaerture of a valve in the manner illustrated in Fig. 9-19. A H-in. valve utilizing this principle is shown in cross section in Fig. 9-20. The clearance between the disk and the bore within which it travels is only a few thousandths of an inch, so that as the disk is thrust against the circular stop, the disk expands radially and seals against the It is found that in applying this principle of sealing cylindrical wall. The disk is it is not necessary to use a soft metal such as copper.
made
The
difference in hardness
and the body of type 304 stainless steel. between these two materials is sufficient
The
% to 10
seating principle involved in this design has in. in diameter with no indication of
Valves of the
2-in.
size
D.
J.
Grove, in 1958
Vacuum Sym(Pergamon
permission from J. Wishart and G. H. Bancroft, in 1960 Vacuum Symposium, Transactions (Pergamon
Press,
London,
1961).]
have been baked out repeatedly to 250C without impairment in operation, although it would appear from the lack of mention of successful bakeout at high temperature that 250C may be the practical limit of the design as presently developed. A 3^-in. valve was tested to more than 1,000 openings and closings, and a 2-in. valve to 300 without any
evidence of significant leakage rate through the closed seat. A bakable gate valve as described by Conner et al.^^ utilizes a stainless steel knife-edge seal against an OFHC copper disk and pneumatic pressure between a pair of concentric metal bellows to provide the necessary seal loading. A schematic drawing of a 4-in. valve of
Kinematical alignment
of gate
shown in Fig. 9-21. The copper valve disk is raised and lowered by a toggle mechanism sealed by a stainless steel bellows. The valve has been tested in repeated opening and closing operations, rate less is bakable to 450C, and is reported to have a closed leakage than 10-12 liter/sec atmospheric pressure. No leakage was detected by a mass spectrometer with helium at atmospheric pressure maintained on one side of the valve for 3^ hr. The authors state that the design successfully is not limited to the simple circular aperture but has been
this design is
Pressure
9-20. One-half inch bakable valve utilizing the sealing principle illustrated in Fig. 9-19. (1) Pressure ring; (2) expanding disk; (3) valve body; (4) shoulder to limit travel of valve-disk assembly; (5) end piece of valve body; (6) operating nut for opening and closing valve; (7) pinned color for retracting valve-disk assembly; disk carrying assembly. (8) [Taken with permission from J. Wishart and G. H. Bancroft, in 1960 Vacuum, Sym,posium Transactions
Fig.
chamber
OFHC
gate
copper
Stpinless
knife edge
applied to a rectangular aperture of 2 by 11 in. A large bakable gate valve has been described by Batzer^^ and is The valve incorporates an inflatable metal illustrated in Fig. 9-22. bladder the expansion of which under the pressure applied by hydraulic The valve disk is put into position by a rotary fluid affects the seal. motion seal with a metallic bellows and an internal mechanical linkage
is
and
Fig. 9-21. Schematic drawing of 4-in. bakable gate valve. [Taken with
fluid in the
raised to a position of close contact with the bladder seal. bladder chamber is then pressurized as shown most clearly
Hydraulic
permission from R. J. Conner, R. S. Buritz, and T. von Zwock, in 1961 Vacuum Symposium Transactions
in Fig. 9-22a.
utilizes
for
1961).]
1962).]
assembly of the demountable parts. annular diaphragm 0.032 in. thick clamped vacuum tight between The hydraulic pressure flanges utilizing the shear seal technique. required for sealing is typically 6,000 psi.
382
ULTRAHIGH VACUUM
383
is
oil
pump.
The inner
evacuated by a 6-in. diffusion pump through a liquidnitrogen-cooled chevron baffle. The 6-in. diffusion pump is backed by a 2-in. diffusion pump and a compound mechanical vacuum pump with a cold trap (usually dry-ice-cooled) between the 2-in. diffusion pump and the mechanical pump. The inner chamber is a stainless steel bell jar, the bottom seal for which is a 13-in. -diameter flat flange on a plane face plate with a simple copper wire gasket. The flange loading is provided by 12 drop-
chamber
forged
(a)
C clamps. Two 3-in. sight ports are mounted on the inner chamber wall, secured by flanges with metal wire seals. The outer chamber is fitted with heater elements on the inner surface and water-cooling coils on the outer surface. The system is normally pumped down with the outer chamber removed. After leak checks are completed, the outer chamber reaches a peak temperature of about 490C. The inner chamber pressure as a function of the time during bakeout and subsequent cooling to room temperature is shown graphAt the end of the baking cycle compressed air is ically in Fig. 9-24. first blown through the cooling coils on the inner chamber to facilitate rapid cooling to room temperature. Liquid nitrogen is then circulated
Radiation shielding
Inner bell jar
Outer
Sight ports
,
bell jar
l-|ar
Outer bel
water cooling
(6)
Representative
Heater windings
Gouge
Chevron trap
High -vacuum valve
Fig. 9-22. (a) Valve seat and gate for bakable gate valve. [Taken with permission from T. H. Batzer, in 1959 Vacuum Symposium Transactions (Pergamon
Cross-sectional view of bakablo gate valve. [Taken with permission from T. H. Batzer, in 1959 Vacuum Symposium Transactions (Pergamon Press, London, I960).]
Press,
I960).]
(6)
High-vacuum
valve
London,
Roughing valve
9-5. Two-region Vacuum Systems. An approach to the design of an ultrahigh- vacuum system which in some cases greatly simphfies the problems of motion seals and demountable joints is that of sur-
rounding the ultrahigh-vacuum chamber by a conventional vacuum chamber so that the pressure difference determining the leakage into the inner chamber is of the order of 10~^ torr instead of atmospheric
pressure.
5-cfm
mechanical
pump
A commercial vacuum
shown schematically
in
system designed on the two-region principle is Fig. 9-23 from a paper by Rivera and Le
in 1959
384
ULTRAHIGH VACUUM
385
X
t:
Average curve
-L
31
P,=f(t)
through the trap and the pressure drops rapidly, usually into the 10~i* torr range. The complete bakeout
Bakeout terminated
+
Bakeout begon_
Another example of a two-region system is described by Metcalfe and Trabert^^ ^^d illustrated schematically in Fig. 9-26. The outer vacuum chamber is 48 in. in diameter and is evacuated by a conventional diffusion-pump system consisting of a 12-in. diffusion pump with
a water-cooled baffle typically producing a pressure of 2 x 10-* torr bakeout cycle is completed. The inner chamber is 32 in. in diameter and is fabricated of Inconel, in. thick, the total working volume of which is 16 ft^. The inner chamber is evacuated through a
after the
iguid-nitrogen
A
cally
somewhat
is
similar two-region
vacuum system
and Molpi
3
described
9-25.
by Ehlers
illustrated schemati-
12
15
18
21
in
Time.hr
chamber
able to
Fig. 9-24. Pressure vs. time for the inner chamber of a two-region ultra-
thin-walled
20-in. -diameter pumping manifold within which there are two chevron baffles separated from each other in the manifold and isolated
Radiation shields
high-vacuum system. [Taken with permission from M. Rivera and R. LeRiche, in 1959 Vacuum Symposium Transactions (Pergamon Press, London, I960).]
the thin metal wall material. Because of the low heat capacity the heating period is very short and the total bakeout period at 400 to
450C is about iH hr. The inner chamber and an internal coil are then cooled by circulating compressed air, and finally liquid-nitrogen cooling is applied not only to the diffusion-pump trap but also the internal cooling coil. The result is a relatively short bakeout and cool-down cycle with a total about 6 hr to operating pressures in the range 3 x ID"' to 5 x lO"" torr as read on a Bayard-Alpert gauge, depending upon the detailed features of the system and the operating cycle.
Window system
High vacuum 10"* to 10"* torr
containing the innermost chevron are heated by radiation heaters in one zone of the outer
baffle
Heaters
vacuum
space,
\^^
Ttiin-wall bakoble
contoiner
5r^
higti
^,Alpert gouge
similarly Vacuum Symposium Transactions heated by heaters in the second (Pergamon Press, London, 1962).] zone of the outer vacuum space. The elbow leading from the pumping manifold to the diffusion pump is separately heated by heater elements on its outer surface. There is a third chevron baffle just over the diffusion pump. According to the authors the net pumping speed out of the inner chamber is calculated to be 750 liters/sec.
most chevron
Schematic drawing of two-region ultrahigh-vacuum system. [Taken with permission from R. A. Metcalfe and F. W. Trabert, in 1961
9-26.
Fig.
(Heated by
o-c
(j)!^
('i
=
=>^'^~\
Bokable tube
Liquid -air baffle
-Cooling -water baffle
After bakeout of the inner chevron trap and the inner chamber to 400C for 6 to 8 hr, cooling for several hours, and finally cooling the inner chevron trap with liquid nitrogen, Bayard-Alpert gauge readings
"corrected for X-ray limit" reached base pressures in the range 2 to 3 X 10-" torr. The pressure-versus-time curve for the bakeout and
pfT~^ \
Alpert gauge o
Cooling -water boffle
^Oil
I
pumpdown
diffusion
pump
Oil diffusion
pump
Quick cooling
is shown in Fig. 9-27. An interesting feature of the curve the series of ionization-gauge peaks which are observed as the liquidnitrogen supply to the chevron baffles is shut off and the baffles slowly
is
warm
up.
distillation of the
Rotary pump
These peaks are interpreted as being due to the fractional condensed gases off the baffle surfaces.
Fig. 9-25. Two-region ultrahigh-vacuum system with thin-wall inner chamber. [Taken with permission from H. Ehlers and J. Moll, in 1959 Vacuum Symposium Transactions (Pergamon Press, London, I960).]
Pumping. Thus far in this chapter we have ways in which the performance of the conventional diffusion pump and vapor-trap combination, together with various modifications in the methods of sealing and the technique of baking to high temperature, could be improved for the purpose of attaining
considered various
9-6. Getter-ion
386
ULTRAHIGH VACUUM
387
much lower pressures than are typically obtained in conventional vacuum systems. The diffusion pump and its mechanical backing pump are sources of contaminants
which must be prevented from backstreaming into the highvacuum portion of the system and
limiting the base pressure to
the ions and electrons formed by ionization are constrained to move in more or less tight spirals along the lines of force. A neutral molecule
which wanders into a discharge column in a magnetic field quickly becomes ionized, trapped in a spiral path in the magnetic field, and
forced to leave the region where it entered the discharge by spir'aling along the magnetic field. The pumping action of discharges in magnetic fields has therefore been well known for a long time. However, the first serious attempt to develop a vacuum pump utilizing
this
some
this
much
desired.
problem of backstreaming, engineers in recent years have made an intensified effort to exploit methods of pumping which do not generate hydrocarbon contaminants and give promise of providing "clean" vacuum spaces with
relatively simple combinations of
appears to be that of Foster, Lawrence, and Lofgren. ^s The device took the form shown in Fig. 9-28, which illustrates a discharge
eff'ect
capable of pumping at the rate of about 5,000 liters/sec. was produced by an array of coils mounted on the long, cylindrical body of the pump, except for the central region where the pump housing is enlarged to provide high entrance conductance. Across this enlarged section of the pump the coil was in the form of a rather open spiral conductor carrying a large current to maintain the magnetic field strength and still permit gas molecules to diffuse
ion
pump
The
Fig. 9-27. Typical bakeout and pumpdown cycle of the two-region vacuum
degree of success was discussed toward the end of Chap. 5, where recent progress in the development of molecular pumps and axial-flow molecular turbine pumps was de-
The distribution of currents in the coils determines the shape of the magnetic field which is optimum when the lines of force bulge slightly in the central section of the pump and converge somewhat symmetrically toward both ends.
The discharge is a large PIG discharge first investigated by Penning. There are two cathodes, one on each end of the device. Experience with a variety of hot- and cold-cathode designs resulted in the final
on one end and a cold, hollow cathode The anode for the discharge is the long cylindrical pump body reaching from the enlarged pumpinlet section to the anode constriction on each end. The discharge column is limited in diameter by the anode constrictions. The shape of the magnetic field and the diameter of the cylindrical anode from the constriction to the pumping section of the device are sensitive parameters. In order to maintain the proper discharge conditions in the central region of the pump it was found to be necessary to maintain a neutral gas density in the cathode chambers not less than about 5 X 10-* torr, which is the forevacuum against which the ion pump works. The gas from the cathode chambers fiows in from both ends of the pump, is ionized, and as positive ions is carried back to the cathodes, where the ions are neutralized. Molecules fiowing into the discharge column in the central section are ionized and also are carried out the ends as ions and are neutralized at the cathodes. Many of the on the other, as shown in the
figure.
and F. W. Trabert, in 1961 Vacuum Symposium Transactions (Pergamon Press, London, 1962).]
We
pressure range.
It is well
degree.
This
discharges have the ability to pump to some particularly true of discharges in magnetic fields, since
Pressure
~10'^
Exit-gas leak
torr
Side magnet
Hollow cold
ttiode
-300
volts
Anode constriction'
Pressure~5xlO"''
torr
"To forevocuum
End view
Fig. 9-28.
a magnetic field. [Taken with permission from .J. S. Foster, and E. J. Lofgren, Rev. Sci. Instr. 24, 388 (1953).]
E. O. Lawrence,
ions striking the cathodes interact chemically. This process proceeds at such a rate that the forevacuum valves at the two ends of the
pump
388
into the cathode
ultrahigh vacuum
389
chambers in order to maintain the minimum operating pressure of 5 x 10"* torr required to maintain the discharge. The operating characteristics of the pump are shown in Table 9-1. From these characteristics it is evident that the ion pump of Foster,
Lawrence, and Lofgren cannot be classed as an ultrahigh-vacuum pump since the typical base pressure was about 1 X 10"* torr. However, the pump did have the specific advantage of not producing hydrocarbon impurities. The feature of continuing to pump even
about 12 in. in diameter. (2) An electron-emitting filament and double grid system which accelerates electrons radially outward, ionizes residual gas molecules, and drives the ions with energies up to 1,000 eV into the walls of the pump body which are coated with the evaporated
titanium.
is
pumped
in the Evapor-ion
Table
9-1.
pump, driven into the wall coating, and covered up by subsequent layers of evaporated metal.
Pumping speed
Base pressure Arc voltage Arc current Cathode
Magnet power
400-300 V 20-10 A Radiantly heated tungsten cathode Heating power, 4.5 kW Side magnets, 20 kW Center helix, 12 kW
Pumping speeds measured for various gases by Swartz^* were as given in Table 9-2 when the rate
of titanium evaporation was 5.3
Filament F
Inner grid G|
* Reproduced by permission from J. S. Foster, Jr., E. O. Lawrence, and E. J. Lofgren, Rev. Sci. Instr. 24, 388 (1953).
mg/min. The most extensive use of the Evapor-ion type of pump is on the 30 X lOi'-eV AGS proton synchroton at Brookhaven, where over 50 units have been in use for
several
briefly
culties
Outer grid Gj
of the
valves were closed contributed further to the system relative to hydrocarbon contaminants.
years.
Gould^^ reported
diffi-
on experiences and
Unfortunately, the pressure in the cathode region had to be maintained at least at 5 x 10~* torr, and the compression ratio which the pump could maintain against the pressure in the cathode chambers was never better than about 10^, so that pumping at pressures much lower than
10~* torr with this particular type of ion
pump
promising.
During studies of the performance of closed-off systems with a Bayard- Alpert gauge in operation Bayard and Alpert^^ observed a very definite pumping action of the gauge involving chemisorption and ion Herb and his collafcoraburial in metal coatings in the gauge tube. ^Qj.g36,37 have reported on the operation of a large device designed The specifically to exploit these mechanisms for vacuum pumping. device, known as the Evapor-ion pump, is illustrated schematically in Fig. 9-29, and involves two principal features: (1) A feed mechanism by which titanium wire is fed in a sequence of discrete steps from an
'
encountered in the early use of these pumps in such great multiplicity. The method of evaporation of titanium has been changed to one of sublimation from a heated titanium rod as described by Gould and Mandel*" and also by Herb, Pauly, Welton, and Fisher.*! The Evapor-ion pumps which have been changed over to the new continuous sublimation technique are operated as
described by Gould and Mandel" under the control of an automatic
Bleeder system
Fig. 9-29. The Evapor-ion pump. [Taken with permission from R. H. Davis and A. S. Divatia, Rev. Sci.
Instr. 25, 1193 (1954).]
internal spool
down a guide so that the tip of the wire periodically touches a post of tantalum-tungsten alloy which is heated by electron bombardment to such a high temperature that a short length of the wire is evaporated each time the tip of the wire touches the post. The evaporated titanium coats the walls of the pump housing, which is
pressure detector in the pressure range 2 x 10"' to 2 x 10"* torr. The ultimate pressure thus far attainable using the new technique is 2 x 10^' torr, which the authors believe is determined by the impurities present in the commercial (non-vacuum-processed) titanium which is used.
The pumps equipped with three sections of titanium rod for sublimation deposit, as shown in the photograph in Fig. 9-30, are expected to
390
ULTRAHIGH VACUUM
391
Table
operate satisfactorily for a period of about two years before replacement of the titanium rods will be required.
Pumping
speed,
liters/sec
Air
1 1
X 10-5
370
1,000
Oxygen
Nitrogen
X 10-5
20
5
5 X 10-5
* Reproduced by permission from J. C. Swartz, in 1955 Vacuum Symposium Transactions (Committee on Vacuum Techniques, Boston, 1956), p. 38.
Although as used on the Brookhaven AGS the operating pressure is not very low, it is an acceptable range for the present needs. What is important is that the system appears to be essentially free of hydrocarbon contaminants. The system is initially pumped down to a pressure of about 10-* torr by a group of compound mechanical booster pumps backed by single-stage mechanical roughing pumps. The system is then isolated from the mechanical pumps and the Evapor-ion pumps are put into operation. The relative simplicity of the Penning Penning gauge gauge system in other respects seems to have fully justified the considerable expenditure of effort in perfecting the Evapor-ion pump to the point of high reliability. A getter-ion pump in which the
ionization
are
Tungsten filaments
completely separated than in the Evapor-ion pump has been described by Gale.*^ The pumping unit together with the test reservoir for admitting various
gases under controlled conditions
is
more
pump
of Gale*^
for
reservoir
admitting various gases under controlled conditions. [Taken with permission from A. J. Gale, in 1956 Vacuum Symposium Transactions
shown
in Fig. 9-31.
1957).]
within one titanium metal is evaporated from heated tungsten filaments wound with titanium wire, and within the other is a Penning type of ionization gauge. One interesting feature of this arrangement is that
Fig. 9-30. Inner structure of Evapor-ion pump with three sections of titanivim for sublimation coating. [Reprinted with permission from The Macmillan Co.,
from
C. L.
Copyright
Gould and P. Mandel, in 1962 Vacuum, Symposium, Transactions. 1962 by The American Vacuum Society, Inc.]
formed in the Penning discharge unit cannot strike directly the walls on which the titanium metal is deposited. Even so, there is a marked difference in the pumping characteristics, depending upon whether the Penning discharge is in operation Particularly in the pumping of argon and helium the pumping speed is greatly enhanced. To determine the pumping effectiveness of the combined unit, the procedure followed was first to pump out the entire system with a diffusion pump and outgas the pump and structure by heating the tungsten filaments below the evaporating temperature. The valve to the diffusion pump was then closed and the filaments were raised arbitrarily to a temperature at which titanium was evaporated onto the walls of the chamber and then the filament current was turned off. The valve between the reservoir and the getter-ion pump was then closed and gas admitted to the reservoir to a predetermined pressure. The gas sample was shared with the
ions
.
392
ULTRAHIGH VACUUM
Pumping Speed op the Getter-ion System of Various Gases*
^'^
Air
Fig. 9-31
393
opening the interconnecting valve, and finally Since the pressure in the reservoir was observed as a function of time. the volume of the reservoir, which was about 700 cm', was about equal to that of the pumping unit including the Penning discharge chamber,
pump chamber by
fob
Pum,ping speed,
16 28
ig
cmfijsec
the operation of opening the valve to the reservoir and sharing the gas sample between the two volumes accounted for an immediate drop in pressure to one-half that initially Thereafter, the in the reservoir.
fall
Oxygen
Nitrogen
24
9
in pressure with time provided a measurement of the pumping speed. In the test runs reported, the initial pressure in the reservoir
Reproduced by permission from A. J. Gale, in 1956 Vacuum Symposium Transactions (Pergamon Press, London, 1957), p. 12.
pumping unit
argon.
is
pumping speed
for
helium and
was
S
10"
X 10^2
torr
or
3.5
x lO-^
its
When an
pumping
Absorption choVocteristic of
getter-ion
pump
for
dry oir
The results of nine ion pump. runs in sequence are shown graphiThree regions cally in Fig. 9-32. of performance can be distinguished: (1) the pressure range from 3.5 X 10-2 cio^n to about 3 x 10-* torr in which the slope of the pumpdown curves, and therefore the
evaporated metal coating has reached saturation and loses efficiency, the surface can be restored to its original
Anode
Titoniumatom
Electron
t3
pumping
5
10
speed,
is
generally less
15
20 25 30 35 40 45 50 55 60
Elapsed time.min
Fig.
9-32. Pressure-vs.-time
curves
pump shown
for
in Fig. 9-31
and furthermore defrom run to run as the getter surface appears to become
sure range,
creases
was determined
1956
saturated;
(2)
Vacuum Symposium
Transactions
1957).]
from 3 X 10-* to about 5 x 10- torr, over which the pumping speed is the same from run to run
Cathode
number of
absorbed by the getter surface; (3) the pressure range from about 5 X 10-* to 6 or 7 X 10-' torr, over which the pumping speed decreases toward zero at an ultimate pressure at which the absorption and desorption rates of the getter surface appear to reach equilibrium. The pumping speed in the pressure region (2), in which it is constant
insensitive to the gas absorbed by the getter surface, is shown in Table 9-3 for several gases. According to Gale*^ the pumping speed for hydrogen is much greater than that for other gases, with the result that the slopes of the curves were too steep to permit a measurement of its value. The most interesting feature of the performance of the
Vac Ion pump. [Taken with permission from L. D. Hall, in 1958 Vacuum Symposium Transactions (Pergamon Press, London, 1959).]
9-34. Assumed pumping mechanism of the Vac Ion getter-ion pump. [Taken with permission from L. D. Hall, in 1958 Vacuum, Symposium Transactions (Pergamon Press, London, 1959).]
Fig.
and
performance by heating the filaments for a few minutes and depositing a new coating of evaporated titanium. This process can be repeated until the titanium wound on the tungsten filaments has been essentially completely consumed by evaporation. A major advance in the development of getter-ion pumping was initiated by Hall in the development of the Vac Ion pump, which is
illustrated in Fig. 9-33.
The device
consists of a rectangular
which
is
mounted an
electrically insulated
394
ULTRAHIGH VACUUM
reached.
395
of thin metal plates (usually titanium) arranged to produce an array On each of the inner flat surfaces of of cells with square cross section. the boxlike stainless steel casing a plate of titanium or other active
secured with a small clearance between the surface of the two The rather flat fiat electrodes and the insulated cell structure. assembly is put between the poles of a magnet so that the field lines pass through the square cells of the insulated electrode and are perpendicular to the surfaces of the two titanium plates on either side. When a positive electric potential is applied to the insulated electrode,
metal
is
In a test of this point a small Vac Ion pump was heated to 400C while it was in its magnet with the voltgage applied to the anode. The pump continued to operate at this temperature and reduced the pressure at this temperature to about 2 X 10-* torr at the end of a 3-hr bakeout. After the system was
allowed to cool
5
down
to
10-10 torr.
each cell of the device acts like a separate PIG or Penning discharge. Because of the large cathode area involved the pressure at which the discharge will start and continue to pass current is very low when a potential difference of the order of 5,000 V is applied to the central
electrode.
diQ^: I
>
The mechanism of pumping by the Vac Ion pump as visualized by Hall" is illustrated in Fig. 9-34. As in any PIG discharge any electrons which are present oscillate in the electric field between the cathodes and are restricted from moving to the side and striking an anode plate by the magnetic field. The electrons are thus very efficiently used for producing ionization, i.e., not only is a positive ion produced and drawn
by the
cathode but in addition in each ionizing event another electron is produced which carries on the process of producing more ions. The ions are propelled into the cathode plate with energies of several kiloelectron volts and sputter cathode
electric field into the surface of the
material (such as titanium), some of which settles on the surfaces of the anode plate structure. The freshly deposited active metal has strong chemical affinity for most gases with the result that gas atoms are
accumulated and held by chemisorption on the anode plates. The cathode plates are slowly eroded by the sputtering process. Figure 9-35 is a photograph of a cathode plate after long service showing the deep holes eroded opposite each cell of the anode. Hall" reports that hydrocarbon contamination of the Vac Ion type of pump can easily prevent the pump from starting to operate. Several hours of exposure to the pumping action of a mechanical vacuum pump Baking the Vac Ion will make the Vac Ion pump difficult to start. pump to 400C for 2 hr in air may restore the pump to normal operation. However, repeated contamination by hydrocarbons eventually results in the pump no longer responding to air baking and the cathodes must be replaced to put the pump back into operation. The gas which is absorbed on the anode surfaces is partly very tenaciously held and is partly rather weakly bound. A pump will therefore both pump and release gas during operation, and the question is what limit of base pressure can one hope to realize when the balance
2i
'
'3'
after long service showing pattern [Taken with permission from L. D. Hall, in 1958 Vacuum Symposium Transactions (Pergamon Press, London, 1959), and through the courtesy of Varian Associates, Palo Alto, Calif.]
The earliest pumps of the Vac Ion type were quite small and typically had pumping speeds for air of the order of 5 liters/sec. Zaphiropoulos and Lloyd*^ have discussed some of the design considerations which arise in scaling up the Vac Ion tjrpe of pump to much larger sizes. Figure 9-36 shows schematically a quadrupole and an octupole configuration for very large pumps of this type. Figure 9-37 is a photograph of a 5,000-liters/sec Vac Ion pump, showing one example of a satisfactory scaling up of the Vac Ion concept. For the 5,000-liters/sec
396
ULTEAHIGH VACUUM
397
the power supply is current-limited beyond a specified value and the In some applications of large pumps, potential drops to about 500 V. the sections are brought into operation one at a time in order to avoid Inner bore Cathode -anode
sections
excessive
power drain, and all units are turned on only when the pressure has decreased to 10"^ torr or
less.
W=W
the
pumping of
argon.
is
insignificant,
is
pumped
2L
apparently by being rather deeply buried in the cathode material. With argon the
quite well,
situation
is
Fig. 9-36. Illustration of configuration for large Vac Ion pump designs
quite different,
and
Fig. 9-37. Photograph of 5,000 liters/sec
(magnets not shown). [Taken with permission from R. Zaphiropoulos and W. A. Lloyd, in 1959 Vacuum Sympo-
the problem is discussed in some A typical detail by Jepsen et al.* pattern of the pressure versus time for a getter-ion pump exhibiting
Vac Ion pump. [Taken with permission from R. Zaphiropoulos and W. A. Lloyd, in 1959 Vacuum Symposium Transactions (Pergamon Press, London, I960).]
cathodes appears to be to provide an optimum solution. (1) Pumping and the pumping speed slightly higher than with the plane cathodes. (2) The pumping speed for argon is about 10 per cent of that for air and is stable for all values of the pressure
for air is completely stable,
argon instability
9-38.
is
shown
j
in Fig.
The
periodic
umps in pres-
One sure by a factor of 10 or more are characteristic of this difficulty. incorto is Brubaker*' solution to this problem which was proposed by the porate a third electrode in the form of a grid between the anode and cathode true the outer plate electrode, such that the new grid becomes 9-39. and the side plates become auxihary electrodes as illustrated in Fig. suita By pump. This arrangement is referred to as the triode getter-ion pump triode the for able choice of design and operating parameters the Brubaker was able to show a generally improved pumping speed for However, air. and noble gases and completely stable pumping of argon principal one of there are several disadvantages to the triode design, the cathodes. the of which seems to be the very much shortened life
In order to avoid the complexities and et al.* resulting from the triode getter-ion pump design, Jepsen type of diode the of investigated the effect of slotting the cathodes
loss of cathode lifetime
below 10-^
torr.
(3)
The cathode
life
XlO"''
MO"'
xlO"^
xlO"''
Time
pump
in the
manner shown
in Fig. 9-40.
The
Fig. 9-38. Typical pattern of pressure vs. time for a getter-ion pump exhibiting the argon instability. [Taken with permission from R. L. Jepsen, A. B. Francis, S. L. Rutherford, and B. E. Kietzmann, in 1960 Vacuum Symposium Transactions
1961).]
398
ULTRAHIGH VACUUM
399
impaired by the slotting nor do the benefits of the slotting disappear with aging of the cathode. Aside from the problem of rather extreme sensitivity to hydrocarbon contamination of the cathodes, the getter-ion pump of the Vac Ion type has undergone steady improvement and certainly must be regarded as one of the most effective available means for ultrahigh-vacuum pumping. The question of hydrocarbon contamination can be completely
Auxiliary
V=V| "-*
electrode
^^^^^^^^^^'-^^^^^^^^^^ ^^^^v^^^'^
D D
D
^=~-Hl
V=V,
ID
Cathode
v=oSlotted
-j^Clouds
of tropped
cathode
Positive
Sputtered
Fig. 9-39. Cross section of the triode getter-ion pump showing the open
Fig.
9-40. Cross
section
of slotted
cathode structure and the side plate as an auxihary electrode. [Taken with permission from R. L. Jepsen, A. B. Francis, S. L. Rutherford, and B. E. Kietzmann, in 1960 Vacuum Sympo-
cathode configuration of the diode getter-ion pump. [Taken with permission from R. L. Jepsen, A. B. Francis, S. L. Rutherford, and B. E. Kietzmann, in 1960 Vacuum Sym,posium, Transactions
(Pergamon Press,
L(760mmHg)
lOOg
zeolite
London,
1961).]
(b)
avoided in systems in which the roughing-down operation is carried out by a mechanical pump with an artificial zeolite trap in the pumping line or by the use of absorption pumping starting from atmospheric
pressure.
Fig. 9-41. (a) Pumping speed of a molecular sieve pump at 0.1 torr as a function of amount of gas already pumped. (6) Final base pressure as a function of the amount of gas pumped. [Taken with permission from P. F. Varadi and K. Ettre,
in 1960
Vacuum Symposium
9-7.
Absorption Pumping.
Artificial zeolite* as a
vapor-trap
The recent more zeolite absorption pumps for roughing out vacuum systems of large volume in order to avoid the possibility of hydrocarbon contamination from the sealing oil of a mechanical
material has already been discussed at length in Sec. 8-6.
roughing
pump
is
* 1 3X Zeolite is an alkali methal aluminosilicate of unusually porous structure manufactured by the Linde Division of the Union Carbide Cornpany.
Varadi and Ettre''* have carried out a series of tests on a 13X Zeolite type of "molecular sieve" absorption pump to determine (1) the pumping speed of the absorption pump for various gases as a function of the amount already absorbed, and (2) the final pressure attainable as a function of the amount of each gas absorbed. The pumping-speed measurements were all made at a pressure of 0.1 torr. The results of these two types of tests are shown graphically in Fig. 9-41. The absorption pump contained 100 g of 13X Zeolite which was cooled by
liquid nitrogen.
400
ULTRAHIGH VACUUM
liquid-helium cryogenic
pressure.
401
pumping speeds for different species of be noted by the very small line near the origin of the figure, the ability of 13X Zeolite to absorb hydrogen is almost nil. Another interesting feature of the tests is the apparent fatigue effect due to repeated absorption and expulsion of air. It will be noted that there are two curves for air in the figure: (1) for the first run on a new sample of zeolite and (2) for zeolite which had been recycled several times through the absorption and degassing routine.
gas
is
The rather
pumping
very striking.
As
will
Getter-ion evaporation or sputtering of active metals, usually titanium, together with ionization of the gas by electron collisions
9-8.
pumps
utilize the
to enhance the gettering effectiveness of the newly deposited reactive metal. This combination provides very satisfactory vacuum pumping
However,
for
many
applications the evaporative deposition of an active metal without ionization of the residual gas has also been found to be useful and
Liquid helium
800
trap
liters/sec
Main chamber
Bellows valve
Auxiliary
\
Liquid
helium trap
Omegatron
Work spoce
Support
ion-getter pump.
lonizotion
gouge 8 liters/sec
Bellows
Pironi
5W
25
I
pump
iters /sec
For many years barium particularly has been used as getter In the mass production of receiving tubes, vacuum pumping is normally accomplished by mechanical pumps arranged in groups on a rotating and indexing machine having a large number of ports to which the tubes being evacuated are connected. As the machine rotates, each tube is connected in turn to a rough pumping section followed by stages
effective.
Front View
Sorption
gouge
pump
Simultaneously the tube is subjected to induction heating of the internal metallic parts and oven heating of the The tubulation is then glass envelope to ensure thorough outgassing. sealed off and a getter capsule is flashed dispensing barium metal as a
operating at lower pressures.
getter material to clean
process.
Fig. 9-42. Schematic diagram of ultrahigh-vaouum system incorporating sorption, ion-getter, and liquid helium cryogenic pumping. [Taken with permission
in
1961
Vacuum Symposium
From
it
The use of a getter for completing the evacuation process by a large factor the cost of initial evacuation of electronic tubes and in addition provides a means of continued chemisorption of any gases which might be produced during the operation of the tube
reduces
in service.
upon the amount of zeolite contained. The use of a sorption pump under conditions
which avoidance of hydrocarbon contaminants is necessary is represented in the system shown in Fig. 9-42, which illustrates the arrangement developed by Honig.*' The system is one of fairly small volume and is roughed out to 10~^ torr by a sorption pump cooled to liquid-nitrogen temperature. The system is then further pumped by the getter-ion pump and finally liquid nitrogen is introduced to cool thoroughly the surroundings of the liquid-helium thimble trap before introducing the liquid helium. When the pumpdown cycle was preceded by a bakeout at 350C for 15 hr, the base pressure realized in this system was judged to be less than IQ-i" torr on the most favorable run. Typical pressure readings were between 10""' and 10"!" torr. This system represents most of the more advanced techniques of fully bakable systems with freedom from hydrocarbon contamination and the introduction of
titanium and molybdenum, deposited on surfaces within a vacuum chamber in pumping out residual gas by chemisorption. This technique has proved to be effective in two specific situations: (1) as a separate pumping unit acting as a getter pump which can be attached
to a
vacuum chamber,
the evaporated metal being confined in this case pump itself, and (2) within the vacuum
chamber itself, in which case the evaporated metal is deposited directly on the walls of the vacuum chamber. In case 1 ) the getter pump must be connected to the vacuum chamber and is therefore limited in pumping speed by the conductance of the connecting tube or manifold. In case (2) the walls of the vacuum chamber itself become absorbing with the result that a much larger effective pumping speed can be
(
achieved.
402
ULTRAHIGH VACUUM
403
In contrast with the performance of a getter-ion pump (see Sec. 9-6) which is capable of pumping the noble gases to some degree, a getter device without ionization has no such capability. Only those gases which interact chemically with the evaporated metal are effectively pumped. Getter pumps are therefore usually supplemented by the
formation of a molten drop of metal, which is supported on the tip of the wire by surface tension forces. For metals such as titanium and molybthere is no serious problem in controlling the electron bombardment heating so that the molten ball at the end of the wire is stable and the rate of evaporation steady. Typical evaporating rates are given as 0.05 g/min, which could be maintained for several hours by
feeding the wire at the proper rate from the reel in through the capillary channel.
denum
Negatively biased
cylindrical grid
5'
pumping action of a well-baffled diffusion pump. Since the noble gases make up a small fraction
(about
1
Nickel supports
-Molten
5-10 kv/^"
ball
0.030-1
0.1
amp
'^
Capillary tube
High
surrounded by water
2
vacuum
Feeder rolls
Wire degassed
byl^R heating
metal pumping of high pumping speed for reactive gases supplemented by a trapped diffusion pump of relatively small pumping speed is an effective combination for many applications. Such a getter pump may, for example, consist of a chamber with one or more units for evaporating a reactive metal on the interior surfaces connected to a well-baffled diffusion pump backed by a mechanical vacuum
air at sea level, active
Using the above method of evaporating molybdenum, Milleron and Popp^i have measured the pumping of hydrogen gas admitted to the chamber in pulses. The conditions of the test were as follows
Volume
7,500 cm^ Microscopic area as measured by low-temperature adsorption of argon on molybdenum surface more than 20 times the projected area
Base pressure less than 10~i torr Quantity of hydrogen per pulse 10~^ torr liter Time duration of gas pulse less than 1 sec Maximum pressure rise per gas pulse with newly coated walls
x 10~'
torr
Etficient
_liquid-nitrogen,
pump.
Several methods of evaporating
trapped 4
oil
Equivalent pumping speed greater than 10* liters/sec Quantity of hydrogen to saturate coating and raise pressure to 10~^ torr approximately 1 torr liter
in.
diffusion
pump
Insulated
Fig. 9-43. Device for evaporating active metals by electron bombardment. [Taken with permission from N. Milleron, in 1957 Vacuum Symposium Transactions (Pergamon Press, London,
1958).]
metals such as titanium, zirconium, and molybdenum, all of which have been used as getter materials, have been developed. Milleron^" has described a method of evaporation from the end of a wire, the tip of which was heated by electron bombardment in the device
is
Molybdenum, zirconium, and titanium are all effective in the above type of gettering device, but molybdenum is found to be more effective in pumping hydrogen (the gas of major concern in the Milleron and Popp development) in the pressure range below 10"^ torr, whereas titanium was found to be capable of absorbing more hydrogen gas. The reason for this difference would appear to be that molybdenum may be heated to a higher temperature without the danger of melting than titanium or zirconium, resulting in a more thorough outgassing of the metal before it is evaporated into the pump chamber, whereas the latter two metals have a greater total capacity for reacting with
hydrogen. Other methods of evaporating the most commonly used reactive metals, titanium and molybdenum, than that described above are mentioned in Sec. 9-6, particularly as an adjunct to the Evapor-ion pump. The simplest and most convenient method thus far developed is similar to that described by Gale,*^ in which a wire of the metal to be evaporated is wound on a somewhat larger-diameter tungsten wire, which acts as a heater element. In one of the large thermonuclear research machines at the University of California Lawrence Radiation
by being heated to as high a temperature as the material will stand by an electric current. The wire is then fed by a system of rolls through a
water-cooled copper capillary tube into the evaporation chamber. The small conductance around the wire through the capillary tube prevents any appreciable flow of gas desorbed by the wire from entering the evaporation chamber. The end of the wire protruding from the
capillary tube is bombarded and heated by electrons from a circular filament at 5 to 10 kV negative potential relative to the wire. The end of the wire is heated beyond the melting point, resulting in the
404
ULTRAHIGH VACUUM
405
Laboratory (Livermore) known as ALICE, both titanium and molybdenum units are used in a beam tube in which the pressure must be reduced from about 10~^ torr of hydrogen and water vapor at one end to an operating pressure of about 10~' torr over a distance of about ten feet. Titanium evaporation combined with well-baffled diffusion pumps is used in pumping units as shown schematically in Fig. 9-44 in portions of the system in which the pressure Metollic is comparatively high and the main evaporotor requirement is that of relatively Liquid-nitrogen Flap valve high throughput. In the main cooled liner " " n nn vacuum chamber where the pressure must be maintained at 10~' or less with the beam on, molybdenum is evaporated directly on a liquidLiquid -nitrogen-
may require many hours before the system can be cooled the evaporative deposition process started under conditions which will ensure effective pumping at pressures less than 10"* torr. The practice of coating a liquid-nitrogen-cooled inner liner of the vacuum chamber has become accepted for certain classes of controlleddown and
fusion research devices.
is
One of the most famous such installation the Ogra machine of the Kurchatov Institute for Atomic Energy in Moscow. The central chamber of the machine is 1.6 in diameter
and 19
Table
volume
is
about 38,500
liters.
The
9-4.
Temperature,
cooled trap
Diffusion
4 x IQ-^ 1 X 10-2
in diameter
tungsten
a
filament
* t
Fig. 9-44.
Pumping
wound with
titanium
0.020-in. -diameter
an
with a liquidnitrogen-cooled baffle. [This drawing was kindly provided by Mr. William S. Neef, Jr., Lawrence Radiation Laboratory, Livermore, California.]
oil
diffusion
pump
or molybdenum wire. The heating current for the titanium evaporators is about 125 A, whereas that for the molybdenum
is
Quoted with permission from the RCA Review, June, 1957. These values of vapor pressure are considerably lower than those quoted in the Smithsonian Physical Tables (Smithsonian Institution, 1954) from the results of Brewer, The Thermodynamic and Physical Properties of the Elements, Report This suggests that many of the older data on for the Manhattan Project, 1946.
vapor pressures of metals are suspect.
steel
about 190 A, consistent with the high-temperature characteristics of the molybdenum. Herb et al.*^ quote the data given in Table 9-4 from the RCA Review of June 1957, from which it is evident that vapor pressures for titanium of the order of 10~' torr are attainable for sublimation,
i.e.,
thin-walled stainless
is
reached.
Herb
et
al.
estimate that
at a temperature of
1715K sublimation from a titanium surface of 10 cm^ should produce a pumping speed for active diatomic gases of about
1,000 liters/sec
evaporated periodically from several evaporating units placed along the length of the chamber. An electron beam of current up to 1 A at 3 keV bombards the end of a 2-cmdiameter titanium rod on each evaporator, resulting in a maximum evaporation rate of about 50 g/hr of titanium. According to Simonov,
steel liner
when
the pressure
is
liter/sec at
increased
by a
factor of 10
the temperature of the titanium is raised to 1850K. Whether performance of this order is realized in practice will depend greatly upon the extent to which the titanium metal is outgassed before evaporation. The process of evaporative deposition of active metals for getter pumping includes a preliminary baking of the system and thorough outgassing of the getter metal at a temperature which is high enough not to evaporate any significant portion of the metal but high enough to drive out absorbed gases. In some cases the baking and
Kleimenov, Mileshkin, and Kochnev^^ the combination of the active metal coating on the cooled liner backed by an array of well-baffled mercury diffusion pumps provides a pumping speed for hydrogen of 2 X 10" liters/sec, a base pressure of 1 x lO-i" torr, and an operating pressure during injection of a powerful molecular ion beam of about
5 X 10-" torr. Effective pumping speeds of millions of liters per second could not possibly be achieved unless the walls of the chamber themselves are highly absorbing. Only limited portions of the chamber walls, primarily the ends, are available for openings into pumping manifolds, so that no matter how high the pumping speed of the pumps connected to these manifolds the net pumping speed limited by the conductance of the openings alone would not exceed 100,000 liters/sec.
406
ULTRAHIGH VACUUM
407
If in addition one allows for the conductance of the chamber itself toward both ends from the middle, the net conductance from the median plane (where most of the gas originates in this machine) to both ends, assuming that both ends of the tank are completely open, is only about 50,000 liters/sec. From this it is clear than the only possibility of
the probability of absorption per surface encounter is stated to be about 0.1 that is, about 10 collisions with the coated wall are necessary on the average for a hydrogen molecule to stick under the conditions
;
of the experiments.
As gases
are
Dz
achieving
machine
to
make
the walls as
Fig. 9-45. Apparatus for the investigation of the sorption of gases at low pressures on renewable surfaces
completely absorbing as possible. Simonov et al.^^ investigated the sorption of hydrogen gas by renewable surfaces of chemically active metals as a function of the temperature of the surface using the apparatus shown schematically in Fig. 9-45. The metal to be investigated was evaporated either continuously or periodically and condensed upon a substrate, the temperature of which could be controlled and measured over the range
sorbed by the surface, the probability of absorption of hydrogen decreases markedly. The process of dissociative chemisorption of pure hydro-
(a)
/
H2
gen on a pure metal surface is said by the authors to be independent of the temperature over a wide range.
/<
HD
100
i-200
-100
TemperQture,C
which are weakly bound to the metal and are thus desorbed. The authors have shown, however, that
when
face
Vacuum
chamber;
tion
of
+100C. Two time-ofmass spectrometers were used, one for determining the comto
flight
-195
pronounced
-100
Temperature, C
specimen under test and the other for determining the composition of the gas generally throughout the chamber. The authors state that one of the most important characteristics of baffles; (8) getter -ion pump. the sorption of hydrogen by a reactive metal is that the sorbed hydrogen molecules are dissociated into atoms which migrate over the surface of the metal and readily react with other sorbed atoms producing volatile compounds which may be desorbed. This process at least partly defeats the purpose of the original sorption process, resulting in a portion of the sorbed hydrogen and other gases being reemitted from the surface in a form which is no longer effectively sorbed by the surface but must be pumped out through the diffusion pumps or otherwise disposed of. Gases whose presence greatly affects the rate of permanent sorption of hydrogen are oxygen and nitrogen. For a typical freshly deposited surface of titanium exposed only to pure hydrogen
metal under study; (4) sample of structural material under investigation; mass spectrometer for (5) analysis of the gas desorbed from the surface of the specimen; (6) mass spectrometer for analysis of the gas in the chamber; (7) mercury vapor pump with liquid-nitrogen-cooled
by the
+100
and a
re-
duction in the rate of desorption of gas from the surface. In Fig. 9-46 are shown experimental curves of the partial pressure of hydrogen (Hg),
Fig. 9-46. Dependence of the rate of sorption of hydrogen, the rate of deuterium exchange, and the rate
of
continually
renewed
by
deuterium (D^), deuterium-hydrogen ambient pressures of these gases in (HD), methane and deuterio-me th- the chamber. The evaporated ane [IiC(H, 0)4] resulting from the metals are {a) nickel and (6) admission of a constant flow of a titanium. The rates of admission hydrogen-deuterium mixture mixture of hydrogen and deuterium of wore (a) 0.0,5 /d/sec and (6) 0.06 filj into the vacuum chamber in which The conductance of the input sec. the active metal surface was conpipe was 4,000 liters/sec. tinuously renewed by evaporation onto a substrate, the temperature of which was varied, Titanium and nickel were studied in this manner. It is evident from the curves that the sticking coefficient remained unchanged from 100 to 170C
408
for nickel (a)
ULTRAHIGH VACUUM
409
(b) and the catalytic exchange and the formation of methane do not decrease significantly with the temperature over these temperature ranges. Further cooling below 170C for nickel and 130C for titanium results in a marked increase in the sticking coefficient for hydrogen and a very marked decrease in the rate of catalytic reactions. Thus the major problem in the use of evaporated metal coatings for gettering hydrogen which arises from the catalytic processes which occur at a high rate at high temperatures appears to be completely solved by cooling the surface below a critical temperature, depending upon the metal being evaporated, such as 170C for nickel and 130C for titanium. The sorption of oxygen, nitrogen, water vapor, carbon monoxide, and carbon dioxide by evaporated metal surfaces at low temperatures was also investigated by Simonov et al.^^ somewhat less quantitatively. However, they report that all these gases are sorbed with considerable effectiveness. Argon is also sorbed but with rather poor sticking probability on titanium at temperatures from 170 to 195C. Lower temperatures were not investigated. For practical applications Simonov et al. recommend titanium as the best metal for evaporative coating of surfaces as a gettering agent because of its relatively wide temperature range for nondissociative sorption, ease of evaporation, and availability in sufficiently pure form. The findings of Simonov et al.*^ are generally consistent with those of Milleron and Popp.^^ However, as a practical matter both titanium and molybdenum are found to be useful in ultrahigh-vacuum pumping in the ALICE machine, with molybdenum being the most effective in the region of pressure below about 10"^* torr as was previously mentioned. Evaporative deposition of active metals on low-temperature surfaces is particularly effective for providing very high pumping speeds for For long short periods of time, e.g., for periods of a few minutes. periods of continuous operation the accumulated deposit of evaporated metal flakes off the cold surface. If a fairly large throughput is to be pumped in this manner, the rate of evaporation of the active jnetal must also be large so that the problem of supplying metal at the high
reactions
of hydrogen-deuterium
condensing surfaces. Liquid nitrogen and refrigerated traps generally for example, be regarded as cryogenic pumps for the hmited number of gases, such as water vapor, for which the resulting vapor pressure is sufficiently low for the purpose at hand. Cryogenic pumping utilizing condensing temperatures well below that of liquid nitrogen However, with the need for high pumping speeds is not a new concept.
may,
both in controlled fusion research and more recently in space research the development of large-scale cryopumping systems has undergone a considerable acceleration. This acceleration has been greatly aided by the spectacular growth of the cryogenics industry to meet the needs of the missile program for huge quantities of liquid oxygen, nitrogen, hydrogen, and helium. As is well known, the vapor pressure of a solidified gas as a function of the absolute temperature can conveniently be represented by an equation of the form
at low pressure
logio
P =A
B
(9-1)
Thus
line.
for
any
Such a graph
is
shown
common
From these curves it is evident that neon and especially hygases. drogen stand out as gases which may present special difficulties because of their relatively high vapor pressures at low temperatures. Helium Since the vapor pressure of helium is not even shown on the graph. is equal to atmospheric pressure at a temperature of only 4.2K,
not feasible at temperatures which are For all other gases the vapor pressure at present of practical interest. is well into the ultrahigh-vacuum range at a temperature of 20.4K, the highest vapor pressure of these then being about 5 x lO'^" torr
cryogenic
pumping of helium
is
pumping
is
convenient manner.
9-9.
Cryogenic Pumping.
gases in a
vacuum chamber
surfaces as long as the partial pressures of the component gases are above the equilibrium vapor pressure at the temperature of the
due to nitrogen. The vapor pressure of hydrogen is much higher so that even at liquid-helium temperature (4.2K) it is about 3 x 10"' Neon, however, torr and rising very rapidly with the temperature. range by ultrahigh-vacuum can be satisfactorily cryopumped in the pressure vapor the surfaces maintained at 6K, at which temperature 10~ii of neon is about torr. Temperatures which can be conveniently maintained and are frequently used in cryogenic pumping are those corresponding to the These boiling points of common gases at normal atmospheric pressure. neon, for 20.4K; hydrogen, temperatures are for helium, 4.2K; for and helium Hquid practice 27.3K; and for nitrogen, 77.4K. In liquid hydrogen are widely used for maintaining surfaces at the desired condensing temperature, and liquid nitrogen is used extensively to
410
ulteahigh vacuum
411
provide an intermediate temperature for a protective shield between the cryogenic pumping surface and the room-temperature walls of the chamber to reduce the evaporation-loss rate of the liquid helium or
Gas
Nitrogen (Ng)
Per cent by
volume
78.08 20.95 0.934 0.03 18 X 10-4 5.3 X 10-4 2 X 10-4 1.1 X 10-4 0.5 X 10-4 0.08 X 10-4
hydrogen by radiative heat transfer. In the description above, emphasis has been placed upon cryopumping in the ultrahigh-vacuum range. Bailey and Chuan^^ have applied cryopumping on a large scale to a low-density wind-tunnel
+3
+2
+
1
\s
s
\
\.
\V
>s s
\
V
Xenon
(Xe)
593 159
7.2 1.4
4.0
1.5
8.4 3.8
6.1
\
-1
H2
-2
^
'\\\^
\
N2
\ s.
\^
^e
t -3
CO
NS
* See U.S. Standard Atmosphere, 1962 (National Aeronautics and Space Administration, U.S. Air Force and U.S. Weather Bureau, Washington, D.C.,
I. 2.7, p. 9.
5-5
-6
-7
CO
1 w \m
K
f k I
s
\
\
\, V
s.
\N
\N
\S
gas throughput would be prohibitively expensive. Bailey and Chuan therefore adopted a combination of a mechanical vacuum pump and a cryopump, reducing the pressure to
N
l/Temperature
-10
m\\
0.02
I,
\,
\
\,
\
:^
0.16
0.04
I
Q06
I
008
I
O.IO
i_
0.12
I
0.14
I
018
O20
0.22
Q24
1^
15
12
10
i-
_L
Tennperature,K
about 1 torr by use of mechanical roughing pumps and then applying cryopumping at a total pressure such that the remaining pressure due to the uncondensed gases is reduced to a value of about 2.4 x 10-5 and Chuan torr. Bailey
illustrate the
cryopump
Fig. 9-47.
The logarithm of the vapor pressure plotted as a function of the number of the common gases.
pumpdown from
of
at0.1
mospheric
pressure
200-ft3
0.5
1.0
5.0
10
50 100
500 ipOO
Pumping time,min
chamber
installation to provide very high
pumping
Fig. 9-48.
pumping speeds
pressure range from about 1 torr down to about lO"* torr. A limitation in the cryopumping of atmospheric air arises from the mixture of
a 40-hp cryopump alone, (2) a 4-hp mechanical vacuum pump alone, (3) a 40-hp
combinations:
Pumpdown
of a 200-ft3
applying cryo-
component gases present, as listed in Table 9-5. The components neon, helium, and hydrogen add up to a total of 23.8 x 10-4 per cent of the total atmosphere or a total pressure of about 1.84 x 10-^ torr. If a chamber were to be evacuated simply by condensing all the air from atmospheric pressure on surfaces at a temperature of 20K these three gases would not be condensed by the cooled surface, but would continue to make collisions with the outer wall of the chamber at room temperature and maintain a pressure of the order of 2 x 10-^ torr in the
chamber. To operate a large condensing surface at lower temperature to reduce this base pressure significantly under conditions of large
pumping after pumping down to mechanical vacuum pump alone, pumps. mechanical 1 torr by and (4) the combination of the [Taken with permission from B. M. 40-hp mechanical vacuum pump Bailey and R. L. Chuan, in 19S8 roughing the chamber down to a Vacuum Sym,posium Transactions pressure of 1 torr and then a cryo(Pergamon Press, London, 1959).] pump of low horsepower condensing the remaining gas for a rapid reduction of the pressure to a value
well
below 10-*
torr.
The
pumpdown methods
are illustrated in Fig. 9-48, the dotted line indicating the pressure decrease when cryopumping is applied at a pressure of 1 torr in the
412
ULTRAHIGH VACUUM
plane surface condenser as shown in Fig. 9-49.
are:
413
In the combination system the heat load of condensing the gas is mostly eliminated so that a cryopump of "nominal" horsepower is all that is required to obtain very rapid pumpdown from 1 torr on down. For dynamic pumping of gas in a low-density wind tunnel, Bailey and Chuan developed a gas-cycle helium refrigerator with power input In Table of 37.5 kW and space requirement of 105 ft^ of floor space. 9-6 the performance of this cryopump is compared in terms of power
combined system.
free path is large as compared with the surfaces. the distance L between 2. The condensing surface is maintained at temperature T^ and has a deposit of condensed solid formed from gas from the source surface. 3. Of the mass flow rate w-^ of
1.
9-6. Dynamic Pumping Speeds of Cryopump op 37.5 kW Power Input and 300 Watts Capacity at 20K with Floor-area Requirements or 105 ft^ Compared with Power and Floor-area Requirements for Conventional Pumping Equipment of Comparable Performance*
Table
molecules from the source surface that strike the condensing surface, the fraction / stick and the rest
are diffusely reflected.
4.
Pumping
speed,
liters/sec
Diffusion
Mechanical
pumps
Space,
ft2
pumps
Space,
ft2
Totals
the condenser constitute a mass flow of (1 /)Wi and have a velocity distribution corresponding to the temperature T^ of the condens-
Source
Condenser
ing surface,
Space,
ft2
i.e.,
Power,
Power,
Power,
coefficient is unity.
5.
kW
300 300
kW
145 180 156 22
kW
445 480 156 22
The
6.5 X 10-
4.8
9.9 8.4
6 X 10-3
0.1 1.0
*
360 360
6.5 X 103
if it
were in equilibrium
9-49. Model for analysis of cryopumping between infinite parallel planes. [Taken with permission from
Fig.
Taken with permission from B. M. Bailey and R. L. Chuan, in 1958 Vacuum Symposium Transactions (Pergamon Press, London, 1959), p. 262. f These pumping speeds based upon data obtained at one-third full capacity (100- watt refrigeration) and multiplied by 3 to make a direct comparison with:
the other figures.
(Wj) consists of
R.
W.
Moore,
Jr.,
in 7.967
Vacuum.
(Tfj) and diffusely Press, London, 1962).] molecules {w^ which strike reflected the source surface, all leaving the source surface with a velocity dis-
by the source
The
requirements and floor area needed with conventional pumping systems consisting of diffusion and mechanical pumps or mechanical pumps According to the alone, depending upon the operating pressure. provides system the cryopumping results show that authors these of the for all values system over the conventional clear advantage the higher pressures 1 torr. At than about operating pressure less
conventional
preferable.
The gas flow between the source and condensing surfaces consists of two oppositely directed streams w-^ flowing from the source toward the
:
is
condensing surface with a velocity distribution corresponding to a temperature of Tj, and w^ flowing from the condensing surface toward the source with a velocity distribution corresponding to a temperature of T^. On the basis of the infinite plane model adopted and the
Cryopumping and experimentally by Moore. 5* The cryopumping process is analyzed in some detail for a simple model consisting of an infinite plane source of gas separated by a distance L from an infinite
theoretically
assumptions listed above, Moore^* finds that the mass flows for these two opposing streams are given by
W,
w^
+
f
w.
(9-2)
414
ULTRAHIGH VACUUM
at the probe pressures Pj, given in Fig. 9-50.
(9-3)
415
and
w.
-/)
f
W,,
may
(9-6)
W^ and W^^ by
The flow
may
and
(9-7) as follows:
be Case
o
Cose A
Pp="ikT,VT^ V [j
Case B
mn
(9-4)
A ^
W2
mn2V2.
Case B:
^ / 2nkTX m \ ^ / 2nkTX m j \
/
(9-8)
W^i(2
-/)
2Af
2 If.
(9-9)
/27TkT:'
(mkT\^
W, I:
\
27
Case C:
W,{1 -/) 4 W^
(9-10)
Af
(9-5)
where
7^1
A =
mass
flux
through a
unit area
where T^ is the temperature of the pressure probe. One important conclusion is that the contribution to the probe pressure indication, by the mass flow rate due to evaporation from the condenser deposit W^^, is always the same and equal to
m=
41V
^
On
^277fcT.,V-^W^ l2jrkTX'
\
(9-11)
/
Af
Pp
fn,kT,
M^.^T./I
Case C
Pp-"2kVl
Fig. 9-50. Pressure determined by open-ended probes variously oriented in a region between infinite plane source and condensing surfaces. [Taken with permission from R. W. Moore, Jr., in 1961
The molecular densities of the two streams can then be determined from the preceding four equations.
*'
_ / 277 V^ ~ [mkTj
/
W,
/
Tf.
where P^a is the vapor pressure of the condensed gas at the temperature T2 of the condenser. Thus the contribution due to reevaporation from the condensing surface to the probe pressure indication is always
(9-6)
~
that
is,
^"^
(9-13)
[tJ
Vacuum Symposium
Transactions
1962).]
f W.il [mkTj
27r
-/)
f
(9-7)
the vapor pressure of the deposition on the condenser corrected temperature between the pressure probe and the condenser. This quantity is thus independent of the sticking co-
efficient /.
Whereas the total molecular density between the source and condenser surfaces is the average of the above two expressions, the disnot isotropic. Thus the "pressure" in the usual If one were to measure sense of an isotropic pressure does not exist. the pressure in the region between the two surfaces with a probe device, the reading would depend upon the orientation of the opening to the
tribution in general
is
The contribution to the pressure probe indication by the molecular flow into the system from the source W^ is, however, dependent upon the value of the sticking coefficient /, as illustrated in Table 9-7, in which the probe pressures are listed on the assumption that the contribution due to reevaporation from the condensing surface P^^.
probe and upon the temperature of the probe, as illustrated in Fig. 9-50. Assuming that the pressure sensed by the probe oriented in either of the three directions shown in the figure will be determined by the molecular density in the probe space at which the influx of molecules is just equal to the efflux and by the probe temperature, Moore arrived
extreme values of the sticking coefficient / < 1 and / = 1. Obviously, since / < 1 results in no appreciable cryopumping by the condensing surface, the pressure is independent of the orientation of the probe, i.e., the pressure is isotropic. However, whenever there is significant cyropumping, the pressure is noniso tropic, as shown in the
for
416
ULTRAHIGH VACUUM
417
Moore^* proceeds from the above analysis to offer two alternative pumping speed of a cryogenic system such as that shown in Fig. 9-49. Because the pressure in the region between the source and the condensing surfaces is not isotropic, it is not immediately evident what value of pressure to use in an equation such a,s S = QjP for the pumping speed. For some purposes a pumping speed per unit of area Sg^jA based upon the molecular density %i sensed by the source of gas is most significant, e.g., in the case of space simulation. A
definitions for the
P^^iTiJT^f'^, the value of the pumping speed apparent to the source approaches zero. From this analysis it is clear that in order to ensure maximum pumping speed at the source, the temperature of the condensing surface Tg must be low enough such that Pj,2 < Ps\\ then the pumping speed sensed per unit area of the source is
8^SI
1-/4
which
is
(9-17)
Table 9-7. Probe Pressure Indications for Various Orientations of the Probe and Extreme Values of the Sticking Coefficient / when Reevaporation trom the condensing surface is unimportant {w g2 =0)*
Case
<
A
and when /
~'-'
(9-18)
f<
\
S.SI
00
A B
c
\
(9-19)
m m m
Af Af Af
m m
A
2A
because in this latter case the gas flux returning to the source is essentially zero, as one would like to achieve in simulating the space environment. An alternative definition of pumping speed 8^, for the cryopumping system which is also useful is based upon the mass flow input W^ and the pressure sensed by a probe in the position corresponding to Case B of Fig. 9-50. This definition corresponds more directly to that normally used for diff"usion pumps. In this case the pumping speed
per unit of area
is
(2nkTj^^ W^
\
j
\
* Reproduced by permission from R. W. Moore, Jr., in 1961 Vacuum Symposium Transactions (Pergamon Press, London, 1962), p. 426.
cavity in the source surface would receive the same molecular flux C of Fig. 9-50 and the corresponding
C with T^
Thus
A
From
Amrij,
AmPj,
(9-20)
from
(9-10)
SI
and
(9-13)
P,(T,
T,)
l 27TkT^
Y
/
-^
Af
SI
m
K
Psi\T2
(9-14)
8^
(9-15)
2/
P IT
1
so that since
/
As
before,
kT,
the
m
is
/kj\f
\277m/
(9-21)
maximum pumping
speed
realized
when
P^j
in
is
given by
which case
SSI
Q,SI
W,
Amric ^""'si
^
AP. SI
/ 4
(9-22)
As long as the vapor pressure P^a ^^ the condenser temperature T^ is very small as compared with the pressure P^i sensed by the gas source (P2 ^ -Psi); then the pumping speed apparent to the gas source is independent of the pressure and the temperature of the condenser and has the value ,
/
Vl
sufficiently small.
1-/4
temperature is that of the pressure probe, which in conventional pumping systems does not normally present a problem since the temperatures of all components of the system (with minor exceptions) are ambient. However, in a cryogenic system hightemperature gradients exist and the pressure probe temperature may
The
significant
^
418
ULTEAHIGH VACUUM
419
not be the ambient value, and this effect must then be taken into account. From Eq. (9-22) it follows that for/ < 1 the pumping speed
per unit area becomes
Zlr-.f'Hl
(9-23)
A
and when /
1
(9-24)
Of these two
differs
results (9-23)
is
similar to that given in (9-18), but (9-24) in (9-19) because the pressure probe
determining the pumping speed of the device, because a fraction of the molecules incident on the radiation baffle is reflected back toward the source and fails to reach the condensing surface. For such an array Moore^* introduces an overall capture probability G, which is the fraction of the total number of molecules incident upon the baffle system which is finally captured on the cryogenic surface. The impedance to gas flow interposed between the source and the condensing surface is not quite as disadvantageous as would first appear because that same
impedance ensures that those molecules which do penetrate the baffle but are not condensed on the first encounter with the cryogenic surface, do not necessarily escape back toward the source. This same impedance results in many such molecules being reflected back toward
the cryogenic surface for a second opportunity to be condensed. In Fig. 9-51 is shown schematically a typical cryogenic pumping
array.
Surface
Chevrons
Temperature
Emissivity
Condenser
Bock shield
Tank wall
0.90
In addition to the condensing surface and the front shield consisting of a chevron baffle system there is a back shield at an inter-
0.50 0.20
0.50
mediate temperature to reduce radiation loss to the wall of the vacuum chamber, which is normally maintained near the normal room temperaVacuum
chamberwoll
Back shield
,,,,,,-,,,,,,,,/,,,,,/,/,/,,/,/,,,,/,,//////
condenser
>>>>>>>>>>>>>>>>>Chevrons
Inlet
Heat
rodiotion
The temperatures and emissivities of the various surfaces typical of such an array are also given in the figure. The condensing surface, as shown in the figure, generally consists of an array of shields with gaps between so that some of the molecules which penetrate the
ture.
0.2
r?'^"^
y^
/r
0.4
0.6
0.8
1.0
f
chevron
fExperiment]
back
is
Fig.
9-51. Portion
of
cryogenic
pumping
1961
Fig. 9-52. Dependence of the overall capture probability of a cryogenic pumping array on the sticking coefficient at the condensing surface. [Taken with permission from B. W.
and have a second chance to condense on the back side of the condensing panels. The pumping effectiveness of the entire array
shield,
will
analyzed in terms of the overall capture probability G for gases which be cryopumped on a surface maintained at 20K.
HQs
baffle,
is
Moore,
Jr., in
1961
sium, Transactions
is
the probability that a molecule wiU pass through the chevron the sticking coefficient for molecules which strike the con-
London,
in the orientation of Case
1962).]
still
receives
some molecular
surface/ =
flux
even
denser surface, and g^ is the probability that a molecule incident on the plane of the condenser will pass through one of the gaps (equal to the fractional open area) then Moore shows that the overall capture probability is given by
when the
1.
Cryopumping surfaces cannot generally be exposed directly to^ a source of gas at normal room temperature because the heat load due to radiation would then exceed by far that due to the condensation of gas molecules. Therefore, the cryogenic surface is usually protected on the side facing the gas source by an optically opaque baffle system at an intermediate temperature to act as a radiation shield. Gas from the source must penetrate the baffle to be condensed on the cold surface, but is impeded by the limited conductance of the baffle system. The sticking coefficient / of the cryogenic surface is then not the only factor
G =
1
1
2(1
sr,)(l
-/)
(1
yJ2(i
-(1
-g,){l -f){2~g,)
+(1
-grj[(l
^fY g, -g^)^l - ff ~
g,^]
(9-25)
Assuming that for the chevron shield gr, 0.23 and that for penetration through the gaps of the condenser surface g^ 0.25, Moore
has computed the value of the capture probability (? as a function of the sticking coefficient / on the condenser. The results of this calculation are
shown graphically
in Fig. 9-52.
420
ULTRAHIGH VACUUM
421
of Fig. 9-50, is the same as that given in Eq. (9-22) with / replaced by G, so that
relatively
abnormally high pumping speed for water vapor and other gases of low vapor pressure and a reduction in the heat load on the low-temperature cryogenic surface by eliminating the heat of con-
S^
^
From
^
2
^^
-G4
'"_E
(9-26)
evident that the overall capture probability is not a sensitive function of the sticking coefficient / provided that / is greater than about 0.4. This feature arises from the
it is
densation of these gases. Moore^* has determined experimentally the pumping speed of the cryogenic pumping system shown in Fig. 9-53 incorporating the The results of these measurements are features discussed above.
05D0FC,P4
600
Black
Noiicondensobl eJT]_
with N? deposit
man's
with Njdeposit
J,
flow
500
man's
man's
p400
r300
Sapphire
f=0,7
-t = 0,4
window
200
100
10-'
10"
I0-'
lO'
P,,torr
10-"
10"^
Fig. 9-54.
Liquid-nitrogen
Pumping speed
in Fig. 9-53.
pumping
Jr., in
system shown
Liquid-nitrogen-cooled
outer shields (77K)
[Taken with permission from R. W. Moore, 1961 Vacuum Symposium, Transactions (Pergamon Press, London, 1962).]
in Fig. 9-54 in
is
Fig. 9-53. Cross section of cryogenic pumping system and vacuum chamber. [Taken with permission from R. W. Moore, Jr., in 1961 Vacuum Symposium
shown
fact previously
mentioned that once a molecule penetrates the chevron baffle it is to some degree trapped between the baffle and the condenser and has considerably more than a single encounter with the condensing
surface.
which the pumping speed for nitrogen as defined measured as a function of the pressure in the vacuum chamber. Note that the units of pumping speed are cubic feet per second (cfs). An interesting feature of the pumping-speed curve is the
in (9-26)
increase in pumping speed with increasing pressure for values of the pressure in excess of about 10~^ torr. By comparison of the results below 10~5 torr with the theory, Moore concludes that the value of the
is between 0.4 and 0.7, the value of the between 0.177 and 0.209, and the pumping speed per unit area of the array for nitrogen is between 71.4 and 84.9
The chevron baffle at a temperature of 77 to 100K serves an additional role as an unshielded cryopump for gases such as water vapor,
the vapor pressure of which at the chevron temperature is sufficiently Because molecules incident upon a chevron baffle system bounce low. several times on the average against the baffle surfaces, the capture probability of such an array as a cryopump exceeds the sticking The advantages in a system involving coefficient / by a large factor. the
cfs/ft2.
pumping of appreciable
would be an
The large-scale cryopumping of hydrogen at modest temperatures with the aid of a catalytic surface has been demonstrated by Grobman.^^ As has been previously pointed out, the cryopumping of hydrogen by a pure condensation process requires that the condensing surface
422
ULTRAHIGH VACUUM
423
is much lower than is practical apphcations. Grobman introduces a catalytic surface consisting of palladium-coated alumina pellets to convert hydrogen gas to water vapor, which is then readily condensed at liquid-nitrogen (77K) or lower temperatures. Two arrangements of catalytic beds
were tested in a chamber as shown in Fig. 9-55a and b. A 6-in. diffusion pump with a cold trap and a 15-cfm mechanical pump evacuated
maintained at the intended operating temperature before each series of measurements. A representative set of data is shown graphically in Fig. 9-56 for which the diffusion-pump heaters were turned off A marked decrease is observed for the chamber pressure when the oxveen flow is turned on and set at an optimum value depending upon the hydrogen flow. The optimum setting for the oxygen flow proved not
ConfigInstrument port
Stainless-Steel
cover plote
Cotolyst,
lb
3.1
Diffusion
uration
a
pumps
Oft
Hydrogen -i-|
H-x-Oxygen
A
A A A B
B
b
c
3.1
On
Off
Instrument port
6,000
Radiation
5.8 5.8
Hydrogen -^n
^
/
d
e
On
Off
r**- Oxygen
1,400
heater
6.5
5,000 Annular
cotolyst
6.5
On
1
-So
1,200
Cotolyst bed
Gloss cylinder
50-mesh screen/
l"
bed
2"
I'S 1,000
Without oxygen
4,000
oddition
S 800
-?
With oxygen
^
/
3,000
-
f* I2"0D^
qi" 21
oddition
600
2,000
-I7JID
/
i
1
1
(| copper
tubing)
J.
I 400
1,000
-
Stoinless- steel
;.
^ 200
//
manitold^^.
.--ti-nn'
I
n
100
of
LJ^Ill /_fc1-3TtTlll
10
Pressure upstream
Instrument ports
I
cotolyst bed,
Instrument port
microns
Hg
microns Hg
lO^
6 pipe
JUl
(b)
Fig. 9-55. Schematic drawings of test chamber with catalytic bed arrangements for converting hydrogen to water vapor and condensing the water vapor on a cryogenic surface, (a) Flat cyhndrical arrangement of the catalytic bed (configuration A); (b) annular arrangement of the catalytic bed (configuration B).
[Taken with permission from Jack Grobman, Transactions (Pergamon Press, London, 1962).]
in
1961
Vacuum Symposium
Fig. 9-56. Variation in tost chamber pressure as a function of total hydrogen throughput with and without oxygen addition for configuration of Fig. 9-55 with 3.1 lb of catalyst. Diffusion-pump heaters turned off. [Taken with permission from Jack Grobman, in 1961 Vacuum Sym,posium, Transactions (Pergamon Press,
Fig.
9-57. Variation
in
the
net
with
test
London,
1962).]
London,
1962).]
instrument port and means of introducing 9ontrolled flows of hydrogen and oxygen were provided at the top of the chamber. The performance of configuration A was tested with 3.1 and 5.8 lb of catalyst and that of configuration B with 6.5 lb of catalyst. In each case performance was measured (1) with the oil diffusion pump in operation and backed by the mechanical pump and (2) with the
the chamber.
An
any
case
stoichiometric value.
The performance of the catalytic-cryogenic pumping system was studied as a function of the temperature of the catalytic bed and found not to depend noticeably on the temperature. Heaters are essential,
however, to outgas the catalytic bed before each period of use. The performance of each of the two configurations was measured for differ-
diffusion-pump heater off and only the mechanical pump in operation. The temperature of the catalyst bed was varied from to 600F. The test chamber was outgassed for a period of 24 hr with the catalyst
shown in Fig. 9-57. The net hydrogen throughput in this figure represents the additional hydrogen throughput
ent quantities of catalyst as
424
resulting
ULTRAHIGH VACUUM
425
from turning on and optimizing the oxygen flow. the demonstration of the cryopumping of hydrogen by catalysis as described by Grobman^^ was carried out at relatively high temperatures and pressures, the results are suggestive of a method which might be extended to low condensing temperatures and pressures by a careful study of various catalysts and alternative geometries. Of much greater promise for ultrahigh-vacuum systems is the process of cryotrapping, which has been investigated by Brackman
have studied the trapping of normally noncondensible gases, nitrogen and argon, by condensed water vapor deposited on a stainless steel surface maintained at about 77K by liquidnitrogen cooling. From combined measurements of adsorption and trapping the authors conclude that water vapor condensed at 77K forms a porous deposit with an effective area of about six hundred
square meters per gram of water. The quantity of nitrogen and argon required to saturate the surface
deposit of water is proportional to the quantity of water deposited,
-ii
10"
/^
^
/".Qt. R m
10'
/
5
^P
fiP
Theory:
in
Fig.
9-59.
This
Liquid he!
S.
10"
Emperi col
10"*
number appears
to
3
M,
^
Fig. 9-59.
10"'
10"2
10"'
stant value of about 10-2 for partial pressures of nitrogen above about
0.1 torr
Slit width
grams of HjO
Pressure p.torr
0.3
mm
^A
\-^\
Fig. 9-58. Quantity of nitrogen and argon required to saturate an ice coating deposited by condensation as a function of the accumvilated mass of condensate (Temperature 77K.) [Reprinted with permission
.
from
F.
The
Macmillan
Co.,
from
W.
Schmidlin, L. O. Hefiinger,
and E. L. Garwin, in 1962 Vacuum Symposium Transactions. Copyright 1962 by American Vacuuin
Society.]
of molecules of nitrogen trapped per molecule of water condensed on a surface at 77K as a function of the partial pressure of nitrogen. [Reprinted with permission from The Macmillan Co., from F. W. Schmidlin, L. O. Hefiinger, and E. L. Garwin, in 1962 Vacuum Symposium, Transactions. Copyright 1962 by American Vacuum
Number
value of about 5
torr.
8cmM
PTh Surface at
W
liquid-air
temperature
(900
cmM
Society.]
Hengevoss and Trendelenburg^' have investigated the cryotrapping of hydrogen and helium by condensed argon at a temperature of 4.2K in a pressure range of
much
high-vacuum
The
and
vacuum system used for measuring the cryotrapping of hydrogen and helium by a deposit of condensed argon on a silver surface at 4.2K. [Reprinted with permission from The Macmillan Co., from J. Hengevoss and E. A. Trendelenburg, in 1963 Vacuum, Sym^posium Transactions. Copyright 1963 by
Fig. 9-60. Cryostatic
ments
entire
tially
is
Brackman and Fite reported that gases on cooled surfaces on which water vapor has been con-
American Vacuum
Society.]
densed with the result that the partial pressures attainable for a number of the common gases using cryopumping techniques may be significantly lower than the equilibrium vapor pressures at the temperature of the cooled surface. This process is referred to as cryotrapping and offers the possibility of cryopumping of gases such as nitrogen, hydrogen, and argon much more effectively in the presence of a "contaminating" agent such as water vapor than in a system from which all such agents have been carefully removed and subsequently excluded.
baked at a temperature of 450C and was evacuated by two oil pumps in series through a low-conductance tube. The pumping speed of the evacuating system is largely determined by the conductance of the connecting tube and was carefully calibrated for the gases used in the experiments. After bakeout the pressure in the vacuum chamber before refrigeration was less than 10-' torr (nitrogen equivalent) as read on a Bayard- Alpert gauge. The conductance of the gauge connection tubing for hydrogen was 19 liters/sec, which by preliminary tests was proved to be adequate to prevent the measured pumping speed
diffusion
426
ULTEAHIGH VACUUM
427
of the gauge from falsifying the pressure reading by the gauge or the mass spectrometer also connected to the same tubing for measurement Since the ionization gauge and mass spectromof partial pressures. eter were both at room temperature (300K), whereas the gas in the
chamber was 83K, the readings were multiplied by (83/300)'-^ to determine the correct value of the pressure. Liquid helium was used to cool the isolated inner cylinder to 4.2K. At the bottom of this cylinder a section of silver protrudes through an aperture between the upper and lower sections of the vacuum chamber and exposes a surface area of 8 cm^ to the lower section of the chamber. The conductance between the upper and lower sections of the chamber through the clearance around this protruding surface is very small so that the cryostatic pumping speed exposed to the lower chamber was that due to the protruding surface only and not to any significant degree due to flow through the annular clearance. With the chamber immersed in liquid nitrogen and with liquid helium in the inner cylinder, pure hydrogen was first admitted at a steady flow rate and the pressure resulting from the dynamic balance between the throughput and pumping action of the system was measured. At small flow rates for which the hydrogen pressure was below the saturated value corresponding to the temperature of 4.2K the pumping speed was entirely due to the diffusion-pump arrangement. At sufficiently high flow rates such that the resulting hydrogen pressure
exceeded the saturated value the pumping action was the sum of the pumping speed of the diffusion-pump arrangement and the condensing speed of the cryogenic surface. Under the conditions of the experiment corresponding to the chamber temperature of 83K the saturation hydrogen pressure was 1.3 x 10^ torr. The next step consisted of adding argon initially at a very small flow rate and then progressively increasing the argon-flow rate with the hydrogen flow held constant. As is evident from Fig. 9-47, the saturation vapor pressure of argon
is unmeasurably small so that even a very small addition of argon results in condensation on the cryogenic surface, for which process a sticking probability of 0.7 was measured in preliminary
that no condensation occurs and the pumping action is only that of the diffusion pumps. In each of these cases, when the argon partial pressure reaches about 0.1 that of the hydrogen, the hydrogen partial pressure drops suddenly by a factor of 10 or more and then remains at this lower value as the argon-flow rate and partial pressure are increased further. The decrease in hydrogen partial pressure is due to additional pumping resulting from the trapping of hydrogen by the condensed argon. From the trapping rates and the area of the cryoImpinging flrgon, atoms/cm^ sec static surface the sticking coefficient 10' '"'5 10'2 lo" 10' lo" in'" ml-r for hydrogen on the argon deposit
I
III
III
il
was determined to be 0.4, so that 4 out of 10 hydrogen molecules striking the surface are trapped. straight line is drawn at 45 in the
figure
10
~~
10"
-io
H
N ,u...
i
T-'>
V > A
/
X
J
and
is
"y
1-2= U.I
'
the lower inflection points of curves (a) and (6). Auxiliary scales showing flow rates in terms of molecules
A
10'
10"'
/
A
10
'
Iio'Ve
6lO'"
n-9 10 '
10
IlO
per square centimeter per second are included. The 45 line corre-
at 4.2K
experiments.
The
partial pressures of
measured for various values of the argon-flow rate holding the hydrogenThis procedure was repeated for several different flow rate constant. values of the hydrogen-flow rate. The results are shown graphically in Fig. 9-61, in which the hydrogen partial pressure is plotted against the argon partial pressure for three different values of the hydrogenflow rate.
the hydrogen-flow rates are both so low that the equilibrium pressure is below that for saturation at 4.2K so
(a)
and
(b)
sponds to the case in which one hydrogen molecule is trapped by one condensed argon molecule. Curve (c) is taken at a hydrogenflow rate which is great enough that Vacuum Society.] the hydrogen partial pressure exceeds the saturated value at 4.2K so that condensation on the cryostatic surface occurs even in the absence of any argon. Thus in curve (c) both condensation and trapping occur at the same time so that the break in the curve corresponding to the onset of cryotrapping by argon occurs at an appreciably lower argonflow rate than that corresponding to the intercept with the 45 line drawn through the inflection points of (a) and (6), indicating that about 10 times as many hydrogen molecules are deposited by the combination of condensation and trapping as are argon molecules. In a subsequent experiment the connection to the diffusion pump was sealed off and a large amount of hydrogen admitted to the chamber. After cutting off the hydrogen flow an equilibrium hydrogen pressure of 1.3 X 10-', corresponding to the saturation value at 4.2K, was reached. A continuous flow of argon resulting in an argon partial pressure of 4 x 10~' torr was then introduced, and the partial pressure of hydrogen slowly fell to 2 x 10^* torr. The argon flow was then
9-61. Hydrogen cryotrapping by argon for different values of hydrogen and argon flow rates. [Reprinted with permission from The Maomillan Co., from J. Hengevoss and E. A. Trendelenburg, in 2.963 Vacuum Symposium Transactions. Copyright 1963 by American
Fig.
428
off
ULTRAHIGH VACUUM
9-62
429
an unmeasurable turned value, whereas the partial pressure of hydrogen remained at 2 x 10~* torr, indicating that the hydrogen was permanently trapped by the condensed argon deposit. Similar experiments were carried out to determine whether helium could also be cryotrapped by condensing argon. The results showed that the sticking probability of helium on the argon deposit is about
to
Figs. and 9-63. Because hydrogen cannot be condensed at liquid-hydrogen temperature, there is great advantage in adding materials which are effective in adsorbing hydrogen, particularly in
situations in
which hydrogen
controlled-fusion research.
typical of a series primarily for the purpose of meeting the needs of the Soviet controlledfusion research program. As is evident from Figs. 9-62 and 9-63
is the major gas component present, as in The two adsorption pumps illustrated are of such devices developed by Lazerev and Fedorova
together with their captions, the cryogenic adsorption pumps consist of a central, double-walled cylinder, open at one end as the pumping
aperture and lined with small chunks of graphite held in place by a wire mesh. The liquid-nitrogen-cooled shield is designated as component 1 and not only serves to reduce the radiation heat load on the inner
liquid-hydrogen-cooled component 2, but also acts as an auxiliary adsorption pump for nitrogen, oxygen, and argon. This feature is said to be important because it permits evacuation of the system to 10~* torr or less of other common gases before pouring in the liquid hydrogen
central component 2 to 20.4K. The inner adsorbing thus preserved for pumping hydrogen without appreciable contamination due to the adsorption of the other more easily adsorbed
Fig. 9-62. Model of liquid-hydrogencooled charcoal adsorption pump with liquid-nitrogen-cooled shield having an adsorbing section upstream from the liquid-hydrogencooled section and of the same diameter. [Taken with permission from the American Institute of Physics, from B. G. Lazarev and M. F. Fedorova, Soviet Phys.-Tech.
charcoal adsorption pump with liquid-nitrogen-cooled shield having an adsorbing layer completely surrounding a similar liquid-hydrogen-cooled adsorbing unit except for the pumping aperture. [Taken with permission from the American Institute of Physics, from B. G. Lazarev and M. F. Fedorova, Soviet Phys.Tech. Phys. 6, 624 (1962)[.]
cooled
0.03
and that about 30 argon molecules are required to trap one mole-
cule of helium.
Although the process of cryotrapping has been only partially investigated and the mechanism of the process is not understood, the results described above are most encouraging for the enhancement of the normal cryopumping process by the trapping of otherwise noncondensable gases on low-temperature surfaces. The process of adsorption pumping discussed in Sees. 8-6 and 9-7 has been extended into the cryogenic region by Lazarev and Fedorova*" particularly for the purpose of pumping hydrogen with high pumping speeds at low pressure. Several designs of liquid-hydrogen-cooled adsorption pumps were developed, two of which are illustrated in
connecting it connecting it to a mechanical vacuum pump for roughing out the system. The sequence of operation is first to rough out the system with both valves open to a pressure of about 10"^ torr, then close valve 5. Liquid nitrogen is then poured into the reservoir of component 1 after which the pressure in the system quickly drops to a value of 10~^ or 10^* torr. Liquid hydrogen is then introduced into the reservoir of component 2 and the adsorption pump is then ready to pump hydrogen, which can then be admitted to the vacuum chamber as needed. The pumping speed of the adsorption pump illustrated in Fig. 9-62 is shown as a function of the pressure (on a logu scale) in Fig. 9-64, in which the pressure indicated is that measured at the inlet to the pump. Over the pressure range tested, the pumping speed for hydrogen increased from about 400 liters/sec at 8 x 10~* torr to about 900 liters/sec at 10~^ torr.
is
(4) for
and a valve
(5) for
it is clear that even at the higher not choked by the conductance of the inlet, which for hydrogen must be at least a factor of 5 greater than the measured pumping speeds. The lower curve 2 is the pumping speed as a function of inlet pressure which would normally be realized in the second step of the evacuating procedure when the valve 5 is turned off and component 1 of the pump is cooled with liquid nitrogen. Curve
From
pressure the
pump
is
430
3
is
ULTRAHIGH VACUUM
1
431
component
and component
Substitution of liquid-hydrogen
cooling (20.4K) for liquid-nitrogen cooling (77K) appears to increase the pumping speed of the adsorption
a factor of 3 or
800
more and reduces the attainable base pressure by nearly a factor of 10. If is worth noting that the liquid-nitrogen filling lasts for a period of 20 to 40 hr and that of liquid hydrogen for 24 hr or more depending on the
details of design.
baffles not only serve as heat shields but also condense gases such as nitrogen with very high effective pumping speed. The pumping speed for hydrogen achieved was about 2,000 liters/sec and that for nitrogen was about 3,000 liters/sec. The consumption rate of helium was about 0.5 liter/hr. According to measurements of Bachler et al. and those of Borovik, Grishin, and Grishina,^^ the equilibrium vapor pressure of hydrogen is approximately lO-^ torr at about 3.3K, so that at
2.5K
Pressure gauge
600
Vacuum chamber
400
200
Preliminary tests have been made on a similar design of adsorption pump by Lazarev and Fedorova using liquid-helium (4.2K) cooling on the inner component for the cryogenic
10'"
10"'
10
Auxiliary
pump
10
pumping
Pressure, torr
Fig. 9-64. Porformance of the adsorption pump illustrated in Fig. 9-62. Curve 1: Pumping speed for hydrogen as a function of the pressure, liquid-hydrogen cooled. Curve 2: Pumping speed of the adsorption pump with liquidnitrogen cooling in component 1 only as a function of the pressure. Curve 3: Pumping speed of the adsorption pump with liquidnitrogen cooling in both compo-
of helium in the pressure range 10~^ to 10~* torr. Bachler, Klipping, and Mascher*^ have made a study of cryopumping in the temperature range from 4.2 to 2.5K by controlling the pressure over liquid helium. Because of the very steep dependence of the equilib-
Liquefied gas
reservair
rium vapor pressure of hydrogen on temperature in this range, this procedure provides a possible solution to the cryogenic pumping of hydrogen
Fig. 9-65. Arrangement for controlling pressure of cryogenic gas and therefore the temperature of the condenser. By this system temperatures as low as 2.5K are achieved in condensing hydrogen. [Reprinted with permission from
nent 1 and component 2. [Taken with permission from the American Institute of Physics, from B. G. Lazarev and M. F. Fedorova, Soviet Phys.-Tech. Phys. 6, 624
(1962).]
by condensation.
The Macmillan Co., from W. Bachler, G. Klipping, and W. Mascher, in 1963 Vacuum Symposium Transactions. Copyright 1963 by American Vacuum
Society.]
maintained at any desired temperature, either higher or lower than 4.2K, is shown in Fig. 9-65. Valve
the vapor pressure of hydrogen should be well below IQ-i" torr, ensuring that the sticking coefficient and pumping speed for hydrogen due to
is
and valve
is
adjusted to provide the needed pumping speed to attain any desired temperature in the condenser. Bachler et al. report that control of the temperature to within 0.01K is achievable by this system. In Fig. 9-66 is shown a schematic drawing of a condenser unit to be operated at temperatures below 4.2K. The low-temperature coil is shielded above and below by chevron brffles which are cooled by the
cold exhaust gas evaporated from the low-temperature
coil.
condensation on a surface maintained at 2.5K will be independent of the pressure well into the ultrahigh-vacuum range. Bachler et al. note that whereas in their device nitrogen is pumped with about the maximum speed anticipated from theoretical calculations, hydrogen is pumped with about half the theoretical rate. However, since for hydrogen the theoretical rate is nearly four times that of nitrogen, the
is still very favorable for pumping hydrogen. The use of a liquid-helium-cooled condensing surface under conditions in which the heat load from the process going on within the vacuum chamber (in this case a controlled-fusion plasma) may be a problem
result
These
432
ULTRAHIGH VACUUM
433
In a series has been investigated by Borovik, Busol, and Kovalenko."' geometries Borovik of experiments with trapping surfaces of various tolerable for the proper et al. determined the maximum thermal load of liquid maintenance of the temperature near the value 4.2K typical atmospheric helium at normal Since the heat of evappressure. oration of liquid helium is small (about twenty calories per mole)
made
The liquid-helium-cooled condenser was thick, and was also highly polished. minimum radiation heat load on the liquid-
mm
much
^^www;^
UooooocoooppociooooocB
the conditions of heat transfer become characterized by eruptive boiling at the metal surface if the heat load g (watts per square
centimeter) reaches some critical
value
g,r-
generally such as to determine the temperature of the condensing surface as a function of the heat
load.
From
these experiments
it
was observed that the temperature of the surface was relatively independent of the heat load (increasing very slowly with the heat
Fig. 9-66. Condenser unit with prochevron radiation baffles tective cooled by the exhaust gas from the These low-temperature condenser. baffles serve to condense gases such
as nitrogen,
i=LJ# ^^^^^^^
Fig. 9-67. Vertical cross-sectional
Fig.
radiation
wall;
shields.
(2)
(1)
sectional
Chamber outer
liquid-
shown in Fig.
watt/cm^ above which value the surface temperature increased abruptly due to the onset of erup10^*
tive boiling.
[Taken with
liquid-nitrogen-cooled louver;
and the low-temperature condenser condenses hydrogen. [Reprinted with permission from The
Macmillan
Co.,
Borovik, F.
I.
from
W.
Bachler,
permission from the American Institute of Physics, from E. S. Borovik, F. I. Busol, and V. A. Kovalenko, Soviet Phys.-Tech. Phys. 8, 68 (1963).]
G. Klipping, and
W.
Mascher, in 1963
Transactions.
Vacuum
Copyright
Symposium
1963
by
American
pumping.
Vacuum
and
Society.]
Borovik
The liquid-helium-cooled, double-walled cylindrical surface chamber wall by a liquid nitrogen-cooled (3) is protected from the unit connected cylinder with skirts above and below the condenser
9-68.
louver-type baffles. to the reservoir for a set of liquid-nitrogen-cooled and within this assembly is another cylindrical
1. With a heater of 10.5 output inside the inner shield the rate of evaporation of liquid helium due to the heat input was only 0.04 liter/hr, corresponding to a thermal load on the heat transfer surface of only 7.5 x 10^^ watt/cm^, which is about a factor of 5 below the
kW
critical value.
2. The pumping speed of the condensation pump was found to be about II per cent of that of a perfectly condensing surface or about 1.25
Concentric with arrangement of louver-type baffles, in this case water-cooled. The brightly outer surface of the liquid-nitrogen-cooled outer shield was water-cooled polished (copper), whereas the inner liquid-nitrogen- and
liters/sec
cm^ of the
The authors
^
434
ULTRAHIGH VACUUM
435
assume that the pumping speed for hydrogen would be 4.68 Hters/sec cm^, but did not make the measurement.
Borovik et al. conclude that cryogenic pumping is an effective means pumping hydrogen in controlled-fusion research devices. Private communication with Professor Borovik reveals that a magnetic mirror machine utilizing liquid-helium-cooled surfaces for cryogenic pumping has been constructed and is now in operation.
of
and trap combination to provide much higher overall system pumping speeds at low pressure than had been previously attempted. As the requirements of the controlled-fusion program became more
demanding, auxiliary techniques were added, such as the evaporation of active metals, first on the walls of the vacuum chamber at room temperature and then on liquid-nitrogen-cooled inner liners. Elimination of hydrocarbon contaminants by the use of either room-temperature or liquid-nitrogen-cooled absorption pumps combined with getter-ion pumps provides another solution to the problem of relatively high-speed pumping in the ultrahigh-vacuum range. The more recent advent of space research and simulation has expanded much further the demand for large ultrahigh-vacuum chambers with extreme requirements of high pumping speed at very low pressure. For this service the combination of very large diffusion pumps with
Freon- and liquid-nitrogen-cooled traps, augmented by the extensive use of cryopumping at liquid-hydrogen or liquid-helium temperatures to achieve pumping speeds in the multimillion liters per second range, has been most commonly adopted. Each ultrahigh-vacuum system is itself a special design problem which must be solved by a careful appraisal of the requirements to be met and the selection of the most effective combination of the techniques described in the preceding sections capable of meeting the requirements. Because of the requirements of metal gaskets, bakable
pressure measurement
9-10. Ultrahigh-vacuum Systems. In this chapter and to some extent in the preceding chapter the techniques of ultrahigh vacuum have been discussed. The problem of the vacuum engineer is to utilize
these techniques in the design of ultrahigh-vacuum systems to achieve the required performance as economically and effectively as possible.
of the techniques described have been applied under rather and are not necessarily applicable to a wide range of vacuum problems. However, the most important parameters to be considered are
specialized circumstances
1.
Some
com-
ponent gases.
2. The operating pressure desired for the process to be carried out, either in terms of total pressure or in terms of the partial pressure of a
particular
component
gas.
Ultrahigh-vacuum systems tjrpically utilize some combination of the techniques discussed in this and the preceding chapters. For systems involving essentially no throughput of gas other than the outgassing of the surfaces, the achievement of very low pressures can be accomplished with low pumping speed and the thorough outgassing of the
surfaces
Oil or
by baking the system at temperatures up to 450C or higher. mercury diffusion pumps with a combination of Freon- and
liquid-nitrogen-cooled baffles can provide the modest pumping speeds required for such systems. Indeed, as was shown by Alpert,^^ a thoroughly outgassed system can be maintained in the ultrahigh-
components, and extreme freedom from leaks, and the difficulties in and the like, an improper choice of techniques to be applied can result in excessive costs of construction and operation and seriously jeopardize the chances of achieving the required performance. However, the means are now available for achieving almost any desired base pressure and enormous pumping speeds at low pressure. If properly applied, the methods already developed are capable of achieving spectacular goals. What is perhaps more important, the development of new techniques proceeds at such a pace
that what seems spectacular today will most certainly be commonplace within the near future in the rapidly expanding field of ultrahigh
vacuum range by the pumping action of the ionization gauge alone when closed off from the vacuum pump by a sufficiently tight bakable valve.
Systems of this type may be regarded as static systems in which ultrahigh-vacuum conditions are attained on a small scale with essentially zero
vacuum.
REFERENCES
1.
throughput. Beginning with the requirements of controlled-fusion research, static systems could no longer be relied upon to maintain the desired low
pressure because the experimental equipment was relatively large in volume and an appreciable gas throughput required high pumping speeds. This need was initially met by optimizing the diffusion pump,
2.
3.
Vacuum Symposium Transactions (Pergamon Press, London, 1960), p. 101. B. B. Dayton, in 1961 Vacuum Symposium Transactions (Pergamon Press, London, 1962), p. 42. B. B. Dayton, in 1962 Vacuum Symposium Transactions (The Macmillan Company, New York, 1962), p. 293.
B. B. Dayton, in 1959
T. Kraus,
4.
Vakuum-Technik
8, 39 (1959).
r
436
5.
ULTRAHIGH VACUUM
Press,
31.
437
10.
11.
12. 13.
London, 1958), p. 13. D. Lichtman and A. Hebling, in 1960 Vacuum Symposium Transactions (Pergamon Press, London, 1961), p. 187. G. Mongodin and F. Prevot, Le Vide 11, 3 (1956). F. A. Flecken and H. G. Noller, in 1961 Vacuum Symposium Transactions (Pergamon Press, London, 1962), p. 58. 5. Dushman, Scientific Foundations oj Vacuum Technique (John Wiley & Sons, Inc., New York, 1949), pp. 387, 388. J. Blears, E. J. Greer, and J. Nightingale, Advances in Vacuum Science and Technology (Pergamon Press, Oxford, 1960), Vol. II, p. 473. R. Geller, Le Vide 13, 71 (1958). N. Basalaeva, Soviet Phys.-Tech. Phys. (New York) 3, 1027 (1958). N. Milleron, in 1958 Vacuum, Symposium, Transactions (Pergamon Press, London, 1959), p. 140.
1960 Vacuum Symposium Transactions (Pergamon Press,
32.
33.
H. Ehlers and J. Moll, in 1959 Vacuum Symposium Transactions (Pergamon Press, London, 1960), p. 261. R. A. Metcalfe and F. W. Trabert, in 1961 Vacuum Symposium Transactions (Pergamon Press, London, 1962), p. 1211. J. S. Foster, Jr., E. O. Lawrence, and E. J. Lofgren, Rev. Sci Instr 24 388
'
(1953).
34. F. 35.
'
'
36.
37.
M. Penning, Physica 4, 71 (1937). R. T. Bayard and D. Alpert, Rev. Sci. Instr. 21, 571 (1950). R. G. Herb, R. H. Davis, A. S. Divatia, and D. Saxon, Rev.
331 (1948).
Sci. Instr.
19
38. J.
39.
14. P. F. Varadi, in
40. 41.
London, 1961),
15.
p. 149.
16.
John Strong, Procedures in Experimental Physics (Prentice-Hall, Inc., Englewood Cliffs, N.J., 1938), p. 128. C. M. Van Atta, R. J. Van de Graaff, and L. C. Van Atta, Phys. Rev. 51, 1013(A) (1937); C. M. Van Atta, R. J. Van de Graaff, L. C. Van Atta, and D. L. Northrop, Phys. Rev. 57, 536(A) (1940); and L. C. Van Atta, D. L. Northrop, R. J. Van de Graaff, and C. M. Van Atta, Rev. Sci. Instr. 12, 534
(1941).
FF 1054-Rl, Dec. 31, 1956. 18. T. H. Batzer and R. Ullman, University of California, Lawrence Radiation Laboratory (Livermore) Engineering Note ENA-122, Mar. 1, 1961, rev. Oct.
17.
6,
46.
1961.
47.
48. 49.
R. H. Davis and A. S. Divatia, Rev. Sci. Instr. 25, 1193 (1954). C. Swartz, in 1955 Vacuum Symposium Transactions (Committee on Vacuum Techniques, Boston, 1956), p. 83. C. L. Gould, in 1956 Vacuum Symposium Transactions (Pergamon Press, London, 1957), p. 39. C. L. Gould and P. Mandel, in 1962 Vacuum Symposium Transactions (The Macmillan Company, New York, 1962), p. 360. R. G. Herb, T. Pauly, R. D. Welton, and K. J. Fisher, Rev. Sci. Instr. 35, 573 (1964). A. J. Gale, in 1956 Vacuum Symposium Transactions (Pergamon Press, London, 1957), p. 12. L. D. Hall, Rev. Sci. Instr. 29, 367 (1958). L. D. Hall, in 1958 Vacuum Symposium Transactions (Pergamon Press, London, 1959), p. 158. R. Zaphiropoulos and W. A. Lloyd, in 1959 Vacuum Symposium Transactions (Pergamon Press, London, 1960), p. 307. R. L. Jepsen, A. B. Francis, S. L. Rutherford, and B. E. Kietzmann, in 1960 Vacuum Symposium Transactions (Pergamon Press, London, 1961),
p. 45.
D. J. Goerz, Jr., in 1960 Vacuum Symposium Transactions (Pergamon Press, London, 1961), p. 16. 20. D. Lichtman and A. Hebling, in 1960 Vacuum Symposium Transactions (Pergamon Press, London, 1961), p. 187; and D. Lichtman, J. Appl. Phys.
19.
London, 1959),
22.
p. 9.
Sci. Instr. 20,
472 (1949). 23. B. D. Power and F. C. Robson, in 1961 Vacuum Symposium Transactions
24.
W. M. Hickam, Rev.
50.
51.
28.
29. 30.
(Pergamon Press, London, 1962), p. 1175. T. H. Batzer and J. F. Ryan, in 1963 Vacuum Symposium Transactions (The Macmillan Company, New York, 1963), p. 166. I. Farkass and G. F. Vanderschmidt, in 1959 Vacuum Symposium: Transactions (Pergamon Press, London, 1960), p. 42. D. Alpert, Rev. Sci. Instr. 22, 536 (1951). J. Wishart and G. H. Bancroft, in 1960 Vacuum Symposium Transactions (Pergamon Press, London, 1961), p. 13. R. J. Conner, R. S. Buritz, and T. von Zweck, in 1961 Vacuum Symposium Transactions (Pergamon Press, London, 1962), p. 1151. T. H. Batzer, in 1959 Vacuum Symposium Transactions (Pergamon Press, London, 1960), p. 265. M. Rivera and R. LeRiche, in 1959 Vacuum Symposium Transactions (Pergamon Press, London, 1960), p. 55.
52.
Brubaker, in 1959 Vacuum Symposium Transactions (Pergamon London, 1960), p. 302. P. F. Varadi and K. Ettre, in 1960 Vacuum Symposium, Transactions (Pergamon Press, London, 1961), p. 248. R. H. Honig, in 1961 Vacuum Symposium Transactions (Pergamon Press, London, 1962), p. 1166. Norman Milleron, in 1957 Vacuum Symposium Transactions (Pergamon Press, London, 1958), p. 148. Norman Milleron and E. C. Popp, in 1958 Vacuum Symposium Transactions (Pergamon Press, London, 1959), p. 163. V. A. Simonov, G. F. Kleimenov, A. G. Mileshkin, and V. A. Kochnev, Paper No. 255, Conference on Plasma Physics and Controlled Nuclear Fusion
Press,
W. M.
Research, Conference Proceedings, Salzburg, Sept. 4-9, 1961, AEC-tr-5589, Book 1 (U.S. Atomic Energy Commission, Division of Technical Information,
February 1963),
53. B.
54.
p. 168.
in 1958
p. 262.
Vacuum Symposium
Transactions
55.
56.
57.
Moore, Jr., in 1961 Vacuum Symposium Transactions (Pergamon London, 1962), p. 426. Jack Grobman, in 1961 Vacuum Symposium Transactions (Pergamon Press, London, 1962), p. 421. R. T. Brackman and W. L. Fite, J. Chem. Phys. 34, 1572 (1961). D. A. Degras, Second European Vacuum Symposium, Frankfurt am Main,
R.
W.
Press,
438
VACUUM
SCIE^
AND ENGINEERING
58.
59.
June 5-7, 1963 [Proceedings published for the Deutschen Arbeitsgemeinschaft Vakuum by Rudolf A. Lang, Esch (Taunus), 1963], pp. 54ff. F. W. Schmidlin, L. O. Heflinger, and E. L. Garwin, in 1962 Vacuum Symposium Transactions (The Macmillan Company, New York, 1962), p. 197. J. Hengevoss and E. A. Trendelenburg, in 1963 Vacuum Symposium Transactions (The Macmillan Company, New York, 1963), p. 101; and A. Naturforsch. 18a, 481 (1963).
Lazarev and M. F. Fedorova, Soviet Phys.-Teoh. Phys. 6, 624 (1962). Bachler, G. Klipping, and W. Mascher, in 1963 Vacuum Symposium Transactions (The Macmillan Company, New York, 1963), p. 216. 62. E. S. Borovik, S. F. Grishin, and E. I. Grishina, Zhur. Tech. Phys. 30, 539
60. B. G.
61.
APPENDIX
W.
(1960).
63. E. S. Borovik, F. I. Busol, 8, 68 (1963).
Formula
Helium
Argon
He Ne
Ar
Kr Xe H2 N2
O2
CI2
Nitrooren
Oxvoren Chlorine
Air
28.98 (mean)
chloride
sulfide.
Hydrogen Hydrogen
Sulfur dioxide
Nitric oxide
36.47 34.08
64.06
30.01 44.02 17.03
28.01 44.01
NO
N2O
Nitrous oxide
NH3
Carbon monoxide .... Carbon dioxide
CO
CO2
CH4
C2-H.2
16.04
Ethylene
C2H4
26.04 28.05
439
APPENDIX APPENDIX
Critical Constants,
II
III
Van der Waals' Constants, Molecular Diameters, AND Mean Free Paths Computed from the Constant 6 According to EqS. (1-40) AND (1-38)
A*
Gas
Viscosity op Gases at 0C and 760 Torr together with Computed Values OP Molecular Diameters and Mean Free Paths in Accordance with Eqs. (1-62) and (1-38)
A
cm
torr
Formula
P*
atm
(cm^/mole)'
atm
X
io-
cm^/mole
cm
X
10-
P=
T=
X
cm
1
Gas
Formula
Viscosity, fi*
torr
micropoises
cm
X 10-8
P = T =
0C
O^C
10-3
X 10-3
13.2
Helium
Neon
Argon Krypton
He Ne
Ar
Xenon
Hydrogen
Nitrogen
Kr Xe
Ha Na
O2
Clj
0.03412 0.2107
1.345 2.318
4.194
0.245 1.390 1.360 6.493 8.093
3.667 5.464 4.431 6.714 1.340 3.782 4.170
1.485 3.592 2.253 4.390 4.471 11.62
3.43
Helium
He
Ne
Ar
Neon
Kr Xe
Hvdrosren
2.20 2.55
3.69
9.82
4.27
4.87
216.5
84.7 166.6
Oxygen
Chlorine
33.5 49.7
76.1
Mercury
Hg
HCl
>1500
51.4 374.0 100.4 157.2 94.0 36.5 132.4
>200
81.6
H2 N2
O2
CI2
2.68
3.78 3.65
5.51
4.45
4.77
7.61
5.08 11.2
Chlorine
H2O
HjS
SO2
217.72 88.9
77.7 65.0 71.7
NO
.
. .
N2O
Ammonia
Carbon monoxide Carbon dioxide Methane
.
NH3
CO
.
111.5
35.0 73.0
6.27
7.61 6.07
Air
3.76
3.24 3.55
2.81
Hydrogen Hydrogen
chloride.
sulfide
. .
HCl HjS
SO2
4.53
4.73 5.55
3.71
Sulfur dioxide
Nitric oxide
NO
N2O
117 179.0
136.1
4.62 2.90
3.05
139.0
31.1
Nitrous oxide
4.69
4.57 3.79
C02
NHg
Carbon monoxide Carbon dioxide
. . .
88.9
CH,
CgHj C2H4
.
>1550
36.0
9.7
>200
62.0 50.9 76.0
Acetylene Ethylene
Carbon
disulfide
cs.
273.0
CO
CO2 CH4 C2H2
C2H4
165.8
4.42
2.93
137.6 103.2
93.5
4.66
4.18
3.64
2.59
AmericMn Institute of Physics Handbook (McGraw-Hill Book Company, New^York, 1963), 2nd ed.; Handbook of Physics and Chemistry (Chemical Rubber Publishing Co., Cleveland,
* Sources:
4.96 5.05
Ethylene
93.6
2.50
440
441
APPENDIX IV
1
443
= = =
1
1.667 X 10-2 cfs 471.95 cm3/sec = 0.47195 liter/sec 28.32 liters/min = 1,699 liters/hr
1.699 mS/hr
0.0283 m^/min
APPENDIX
IV
= = = = = = =
fx
liter/sec
(fi
l/sec)
10-3 torr liter /sec 2.119 X 10-3 torr cfm 2.119 / cfm 1.316 X 10-3 atm cm3/sec
1,000 n liters/sec (fi I/sec) 1,000 torr cm3/sec 2.119 torr cfm
torr liter/sec
= = = = = = = = = = = = =
760
mm Hg of density
=
where g
bar
1.0133 X 10dynes/cm2 14.696 psi = 2,116.2 Ib/ft^ 29.921 in. Hg at 32F 33.899 ft of water at 39.1 F 10* dynes/cm^ = 10*/ibar
hter/sec
{fi
1/seo)
= = =
=
750.06
mm Hg of density
=
where g
atm cm3/sec
torr
micron
{/i)
0.98692 atm 14.504 psi 1 Hg of density 13.595 g/cm^ where g = 980.665 cm/sec^ 1,333 fih&T (dynes/cm^) 10^^ torr = 1 millitorr
760 ^
liter/sec
mm
torr
cfm
= = = =
= 6 x 10^^ liter/min = 10~' m^/sec 6 X lO-^mS/min = 3.6 x lO-^m^/hr 3.531 X 10-5 cfs ^ 2.119 x 10-3 cfm
10-^ liter/sec
3.6 liters/hr
1 /i
cfm
1 liter
per second
(liter/sec)
=10^
cm'/sec
=6x10* cm^/min
1
= = =
contimetor (cm)
0.4719 / Hter/sec 0.4719 torr cm3/sec 4.719 X 10-* torr hter/sec 6.210 X 10-* atm cm3/sec
1 liter
per hour
(liter/hr)
= 3.6 X 106cm3/hr = IQ-Sm^/sec = 6 X 10-2ni3/jnin = 3.6 m^/hr = 3.532 X 10-2 cfs = 2.119 cfm = 2.778 x 10"* liter/sec = 1.667 X 10-2 liter/min = IQ-^m^/hr = lO^cm^/hr = 1.667 x lO-Sm^/min' = 16.67 cmS/min = 5.886 x 10-* ofm = 9.72 X lO-Scfs = 2.778 x 10-* m^/sec = 1.667 x 10-2m3/min = 10^ Hters/hr = 10* cm3/hr = 0.2778 liter/sec = 277.8 cm3/sec = 0.583 cfm = 9.72 X 10-3 gfg
1
Length
meter (m)
= = = = =
mm mm
mm mm
inch
1
(in.)
foot
(ft)
30.48
cm =
304.8
Area
square centimeter (cm2)
Source: Handbook of Chemistry and Physics (Chemical Co., Cleveland, 1963), 44th ed.
*
Rubber Publishing
= = =
442
444
1
square inch
(in.^)
square foot
(ft^)
= = = = = = = =
1,550.0 in.2
10.764 ft2 6.944 x lO-^ ft^ 6.452 cm2 = 645.2 6.452 X 10-* m2
mm^ mm^
144
in.^
929.0
cm2 =
9.29 x 10*
APPENDIX V
Values op Some Physical Constants*
Symbol
9.290 X 10-2
m2
10'
Volume
1
1 liter
We
1 liter 1
1.000027 x 10-' 1,000.027 use hereafter the approximation = 10' cm' = 10-' m' = 10* mm' = 3.531 X 10-2 ft' = 61.025 in.'
= = = =
= lO^^m' =
mm'
Name
Losohmidt's number (molecular density of a gas at 760 ton- and 0C) Avogadro's number (number of molecules in 1 g mole) Volume of 1 g mole of gas at 760 torr and 0C Gas constant Boltzmann's constant Atomic mass unit (chemical scale in which
32 mi exactly) O2 Electronic rest mass Electronic charge 2.687
Value
cm' =
m'
X lO'Vcm'
cubic inch
1
(in.')
cubic foot
(ft')
= = = = = = =
10' liters
10*
cm'
6.1023 X 10* in.' 1.639 x 10-2 liter 5.787 X 10-* ft' 28.316 liters 2.8317 X 10* cm'
1,728 in.'
R
1.639 x 10*
"K
mm'
^oln^ol
1.3805
1.660 9.109 4.803 1.602 2.893 9.649
X 10-" erg/K
10-2*
X
X X
2.8317 x 10'
mm'
= =
P=
9
mol
Faraday
Acceleration due to gravity: Approximate range over U.S.A. Assumed "standard" value
X X 10"
X
10*
Temperature
K =
"R
F
Rankine to centigrade
= = =
F
32
+ + +
Cleveland
1.8C
Item
Boiling point of water* Melting point of ice* Sublimation point of dry ice* .... Boiling point of liquid Nj* Boiling point of liquid H2* Boiling point of liquid He* Absolute zero
K
373.16 273.16
195.0
op
R
671 69 491.69 351.0
139.2
36.5
7.59
100
212 32
At standard atmospheric
pressure.
445
AUTHOR INDEX
Abbott, T. A., 140, 168 Ackley, J. W., 165-167, 1(>8 Addis, R. R., Jr., 312, 362 Alexander, P., 234, 27.3
Alpert, D., 99, 103, 104, 131, 342, 343, 362, 378, 388, 434, 436, 437
Burch, R. C, 268, 273 Buritz, R. S., 380, 381,436 Bush, William E., 328, 362 Busol, F. I., 432-434, 438
Biichler,
430-432, 438 Backus, J., 133, 135 Bailey, B. M., 410-412, 437 Bancroft, G. H., 379, 380, 436 Barrington, A. E., 165-167, 168 Barry, E. J., 309, 311, 312, 362 Basalaeva, N., 365, 436
Batzer, T. H., 243, 272, 273, 347, 362, 371, 372, 375, 376, 381, 382, 436
-^
Bayard, R. T., 99, 103, 104, 131, 388, 437 Beams, J. W., 212, 213, 218, 313 Beck, A. H., 118, 132
Becker, J. A., 89, 131 Becker, Wilh, 214, 215, 218 Bennett, W. H., 155, 168 Benson, J. M., 85, 86, 131 Bills, D. G., 107, 132 Biondi, M. A., 343-346, 362
Blears, J., 100, 131, 242, 273, 365, 436
Carmichaol, J. H., 342-344, 362 Cartwright, C. Hawley, 341, 362 Chapman, S., 21, 2^ Charles, D., 145, 147-149, 168 Charpentier, D. E., 139, 142, 143, 167, 168 Chuan, R. L., 410-412, 437 Chupp, Warren W., 247, 273 Clausing, P., 51, 53, 54, 62 Cleveland, J. F. 285, 286, 302 Conner, R. J., 380, 381, 436 Crawley, D. .T., 230, 24.5-247, 258, 260, 264, 273
335, 362
Davis, R. H., 388, 389, 437 Davis, W. D., 140, 141, 168 Dawton, R. H. V. M., 314, 362 Dayton, B. B., 99, 285, 294-296, 302, 354, 355, 360, 362, 363-365, 435
Borovik, E. S., 431-434, 438 Bowie, R. M., 99-101, 131 Brackman, R. T., 424, 437 Briggs, W. F., 162-165, 168 Brisbane, A. D., 118, 132 Brubaker, W. M., 396, 437 Buckley, O. E., 95, 131
Degras, D. A., 424, 437 Dempster, A. J., 133, 167 Denison, D. R., 107, 132 Dennis, N. T. M., 241, 244-247, 271, 273 Divatia, A. S., 388, 389, 437 Dobrowolski, Z. C, 200, 218 Doughty, E. G., 140, 168
447
448
Downing,
J. R., 122, 132 Dubrovin, J., 68, 131 Dunlop, G. C, 83, 131 Dunoyer, L., 272
AUTHOR INDEX
Hablanian, M. H., 261-264, 273, 289, 302 Haefer, R. A., 102, 118, 131, 132 Hall, L. D., 393-395, 437 Hall, L. G., 142, 168 Hamilton, A. R., 87, 89, 131 Hasse, T., 83, 131 Hayashi, C., 364, 436 Hay ward, Roger, 341, 362 Hobling, A., 150, 168, 364, 373, 374, 436 Heflinger, L. O., 424, 438 Hengevoss, J., 102, 131, 424, 425, 427, 438 Herb, R. G., 388, 389, 404, 437 Hickam, W. M., 375, 436 Hickey, J. S., 159, 168
AUTHOR INDEX
Kochnev, V. A., 405-408, 437 Kovalenko, V. A., 432-434, 438
Kraus, Th., 356-358, 362, 364, 435 Kruger, Charles H., 215-217, 218 Kuhn, H. J., 65, 131 Kurie, F. N. D., 323, 324, 362
Moll, J., 384, 437
449
Dushman,
Mongodin, G., 364, 436 Moody, R. E., 153-155, 168 Moore, R. W., Jr., 412-421, 437 Morse, R. S., 99-101, 131
Nakayama, K.,
Lafferty, J. M., 107, 108, 110, 111, 132
Naundorf,
Neef,
362
Ehlbeck, H. W., 154, 156, 157, 168 Ehlers, H., 384, 437 Eklund, S., 212, 218 Enskog, T>., 21, 22 Ettre, K., 399, 437
W., 98, 131 Landfors, A. A., 289, 302 Lane, C. T., 83, 131 Lange, W. J., 342-344, 362 Langmuir, I., 101, 131, 159, 168 Latham, D., 241, 244, 271, 273
C.
Lampson,
W.
S., Jr.,
404
Farkass, I., 309, 311, 312, 362, 378, 436 Feaks, F., 129, 132 Fedorova, M. F., 428-430, 438 Fisher, K. J., 389, 437 Fite, W. L., 424, 437 Flecken, F. A., 364, 436 Florescu, N. A., 230, 233, 234, 261, 262,
Hickman, K.
Hippie, J. A., 139, 143, 167, 168 Ho, T. L., 240, 253, 273 Honig, R. H., 400, 437 Hustrulid, A., 140, 168
273 Fondrk, V. V., 220, 223, 224, 272 Forbes, S. G., 324, 325, 362
Forsythe,
Ishii,
Lauer, E. J., 106, 132 Lavender, R., 87, 88, 131 Lawrence, E. O., 386-388, 437 Lazarev, B. G., 428-430, 438 LeBlanc, M., 219, 272 Lech, J., 145-147, 151, 152, 168 Leek, J. H., vi, 67, 77, 87, 92, 94, 97-99, 113, 115, 123, 132, 283, 302 LeRiche, R., 382-384, 436 Levenson, L. L., 52, 54-57, 61, 62, 265, 273, 306, 334, 335, 345-347, 362
Noher, H. Victor, 341, 362 Nicollian, E. H., 149, 150, 168 Nienhuis, K., 114, 132 Nier, A. O. C, 140, 167, 168 Nightingale, J., 365, 436 Noeller, H. G. (see Noller, H. G.)
Nollcr,
H. G., 202, 205, 218, 230, 237, 250, 252, 273, 289, 302, 364, 437
Normand,
C. E., 253, 273 Northrop, D. L., 371, 436 Nottingham, W. B., 95, 96, 102, 107, 131, 132 Nyer, W. E., 324, 325, 362
W.
E., 4
Foster, J. S., Jr., 386-388, 437 Francis, A. B., 165-167, 168, 396-398,
16,
22
Jacobs, Robert B., 209, 218 Jaokel, R., 230, 237, 273 Jaycox, E. K., 99 Jepsen, R. L., 165-167, 168, 396-398, 437 Johnson, J. W., 328, 362 Jones, A. C., 162-165. 168
Lichtman, D., 150, 168, 364, 373, 374, 436 Little, R. N., 324, 325, 362
Lloyd, W. A., 395-397, 437 Loeb, L. B., 12, 14, 22 Loecherer, K. H., 154, 156, 157, 168 Loevinger, R., 138, 167 Lofgren, E. J., 386-388, 437 Lothrop, C. F., 165-167, 168
Gaede, W., 179, 205, 211, 218, 227, 230, 273 Gale, A. J., 391, 392, 403, 437 Garwin, E. L., 424, 438 Geller, R., 365, 436 Gerber, J. F., 317, 318, 362 Goerz, D. J., Jr., 373, 436 Good, W. M., 328, 362 Gould, C. L., 389, 390, 437 Green, C. B., 89, 131 Greer, E. J., 365, 436 Grishin, S. F., 431, 438 Grishina, E. I., 431, 438 Grobman, Jack, 421-424, 437 Grove, D. J., 374, 375, 379, 380, 436
Guthrie, A., 133, 136, 138, 162, 167, 168, 309, 310, 362
Kenna, R. A., 180, 218 Kennard, E. H., 78, 131, 231, 235, 273 Kennedy, P. B., 339, 340, 362 Kietzmann, B. E., 396-398, 437 Kingdon, K. H., 159, 168
Kinsella, J. J., 269, 273
McLeod, H.,
69, 131
Klages, G., 83, 131 Kleimenov, G. F., 405-408, 437 Klipping, G., 430-432, 438 Klopfer, A., HI, 112, 132 Klumb, H., 83, 131
Mandel, P., 389, 390, 437 Mandoli, H., 165-167, 168 Marks, Lionel S., 4 Mascher, W., 430-432, 438 Mellen, G., 122, 132 Menshikov, M. I., 247, 259, 260, 273 Metcalfe, R. A., 385, 437 Mileshkin, A- G., 405-408, 437
Milleron,
Pauly, T., 389, 437 Pearson, G. L., 89, 131 Peck, A. W., 298, 302 Penning, F. M., 113-115, 132, 387, 437 Pensak, L., 312, 362 Perkins, G. D., 142, 143, 168 Peters, J. L., 141, 142, 168 Pinson, J. D., 298, 302 Pirani, M., 86, 131 Popp, E. C., 403, 408, 437 Post, R. F., 333 Power, B. D., 180, 218, 230, 241, 244247, 258, 259, 264, 271, 273, 375, 436 Pressey, D. C, 67, 131 Prevot, F., 364, 436
Rabinovich,
I. S.,
Redhead, P. A.,
Norman,
265, 273, 301, 302, 306, 332-335, 34.5-347, 362, 366, 372, 402, 403, 408, 436, 437
129, 132, 154, 168 Reich, G., 101, 131, 157-159, 168, 250,
450
Riddiford, L., 101, 131
AUTHOR INDEX
Torney, F.
L., Jr.,
95, 96,
129,
131,
Riddoch, A., 115, 132 Ridenour, L. N., 98, 131 Rivera, M., 382-384, 436
Roberts, J. A., 162-165, 168 Roberts, R. W., 315, 362 Robinson, C. F., 142, 168 Robson, F. C, 375, 436 Roehrig, J. R., 126, 127, 132 Romann, M. P., 77, 131 Rovner, L. H., 121, 132 Ruf, J., 154, 156, 157, 168 Rufer, C. E., 288, 302 Rutherford, S. L., 396-398, 437 Ryan, J. F., 375, 376, 436
168
Trabert, F. W., 385, 437 Trendelenburg, E. A., 424, 425, 427, 438 Trump, H. G., 83, 131
SUBJECT INDEX
Ullman,
J. R., 333, 334, 362, 371,
436
Santeler, D. J., 106, 129, 130, 132 Saxon, D., 388, 437
Schuemann, W. C, 104, 105, 73,2 Schuetze, H. J., 104, 131, 154, 156,
168 Schwartz, C. M., 87, 88, 131 Schwarz, Helmut, 118
Scott,
157,
Vacca, R. H., 123, 132 Atta, C. M., 186, 195-198, 218, 371, 436 Van Atta, L. C, 371, 436 Van de Graaff, R. J., 371, 436 Vanderschmidt, G. F., 378, 436 Vandershce, T. A., 140, 141, 168 van Oostrom, A., 103, 131 Varadi, P. F., 368, 369, 399, 436, 437 Vekshinsky, S. A., 247, 259, 260, 273 Voego, W., 83, 131 Von Friesen, S., 211, 218 von Zweck, T., 380, 381, 436
Van
Absorption, definition of, 364 pumping by ahimina, copper and zeolite, 343-348, 398-400
Accommodation
of,
coefficient, definition
79
role, in
Nancy
J.,
312, 362
C,
16,
22
Smith, H. R., 260, 273, 339, 340, 362 Smith, J. H., 272 Smith, P. T., 93, 131 Sommer, H., 143, 168 Stoinherz, H. A., 261-263, 273 Stevens, C. M., 140, 168 Stevenson, D. L., 250, 253, 254, 265267, 273, 287, 302 Stork, F., 104, 131 Strong, John, 341, 362, 370, 436 Swartz, J. C, 389, 390, 437 Sylvester, R. L., 186, 218
Wahl, J. S., 324, 325, 362 Wakerling, R. K., 133, 136, 138, 162, 167, 168, 309, 310, 362 Wallace, R. A., 145-147, 151, 152, 168 Warnecke, R. J., Jr., 145, 147-149, 168 Watson, W. R., 145-147, 151, 152, 168 Webber, R. J., 83, 131 Weinhart, H. W., 99 Welton, R. D., 389, 437 White, W. H., 159, 168 Whitford, Albert E., 341, 362 Williams, C. E., 212, 213, 218, 313 Williams, T. W., 139, 167 Wilson, R. R., 313, 314, 362 Winters, H. F., 107, 132 Winzenburger, E. A., 186, 189, 190, 218 Wishart, J., 379, 380, 436 Worcester, W. G., 140, 168
124 thermal conductivity gauge, 79-81 Adsorption, definition of, 364 pumping by graphite at cryogenic temperatures, 428-430 Air, effective molecular weight of, 2 normal composition of, 411 Alphatron gauge, 122 Argon, instability in Vac Ion pumps,
in
pumping
effects,
105
gauges; Penning discharge gauge; Pressure gauges) Bellows seals (see Metal bellows) Blears's effect, 100-102, 242, 297 Boiling points of common gases, 409
ionization
Backstreaming
in
diffusion
pumps,
257-268, 329
catalytic effect of materials of nozzle
Calibration of vacuum gauges, 124-128 aperture method for ionization gauges, 126-128
assembly, 267 dependence, on pressure, 263 on shape of first-stage nozzle, 265-267 effect on ultimate pressure, 257-268 measurement of, 258, 261-264 reduction by water-cooled cap over first-stage nozzle, 259, 263
role
McLeod gauge,
as absolute standard,
Tate, J. T., 93, 131 Taylor, A. R., 338, 362 Thees, R., 203, 218
259-264 Bayard- Alport ionization gauge, 103107 comparison, with conventional ionization gauge, 103 with hot cathode magnetron gauge, 109
Thomas, H.
on viscosity, 21 Compression ratio, for diffusion pumps, 230-240 for mechanical booster (blower) pumps, 185, 188, 199-201, 204 for mechanical oil-sealed pumps, 171 for molecular drag pumps, 205-214
451
462
Compression
ratio, for
SUBJECT INDEX
molecular turaccelerated
re-
SUBJECT INDEX
Displacement
speed
453
moval by high-temperature bakeout, 363-370 backstreaming, from diffusion pumps, 258-268, 329 from oil-sealed mechanical pumps, 174, 340 from steam ejectors, 227 dominance of water vapor following pumpdown, 137, 330-332 effect, on performance of mechanical oil-sealed pumps, 177-179, 351 on reading of McLeod gauge, 71
Conductance, general formulas, viscous flow pressure range, formula for long tube, 29, 58 pipe -size formula based on
pressure drop, 30, 59
vacuum pumps,
Dubrovin gauge, 68
of 172
mechanical
Conductance
factors,
Knudsen and
Clausing, 52-57
Cryo-adsorption, 421-424, 428-430 Cryogenic pumping, 408-434 boiling points of common gases, 409
elimination from oil-sealed mechanical pumps, 179-185 vapor compression action of mechanical booster pumps, 204 vapor traps, absorption type, 341348 refrigerated, 328-341 Conductance, definition, 24, 58 general formulas, combined with
combined, with catalytic process, 421-424 with mechanical pumping, 410-412 cryo-adsorption, 421-424 cryotrapping, 424-428 liquid-helium-cooled thimble trap, 400 pumping speed of cryogenic pumps, 411-421
shielded
Elastomers, 307-316 Electron volt, unit of energy, 91 Electronic charge, definition and value, 4 Emissivity, definition, 81 Evapor-ion pump, 388-391
Gases, universal gas constant, 3 vapor pressures at low temperatures, 409 velocity of sownd in, 11 Gaskets, elastomer, 307-316 metal, 370-378
con-
pumping speed, 25, 58 conductances, in parallel, 26, 58 in series, 25, 58 molecular flow pressure range, 43-51, 60-62 annulus between two concentric tubes, 51, 62 aperture in thin wall, 47-49, 60 channel of rectangular cross
section, 50, 61
theory
Faraday, definition and value, 4 Farvitron mass spectrometer, 157-159 Forepressure limit of diffusion pumps, 254-257 dependence on design and operating parameters, 255 design compromises, 255 limitation due to decomposition of working fluid, 256 process of jet breakdown, 254 throughput dependence on, 256 booster diff'usion pumps, 257 oil vapor ejector pumps, 257
391-393
leak detection application of, 165167
discharges,
long
section,
cross
Monte Carlo
57
narrow
slot
51, 61
36-43, 59
dependence on pressure, 37-43 formula for long tube, 37, 41, 59 limits of, 40-42
viscous flow pressure range, 26-30, 34-36, 58 change in character due to surface slip, 34
pump working fluids, 240-249 decomposition of organic, 242-245, 264 mercury and organic, comparative advantages, 244-249 vapor pressures of, 240-244 Diffusion pumps, 227-272 backstreaming, 257-268, 329 Blears's effect, 100-102, 242, 297 booster, 257 compression ratio, 230-240 ejector, oil vapor, 257 forepressure, limiting valvie of, 254257 fractionating, 268-271 Ho coefficient, 248-250, 253 modern types of, 227-229 principles of operation, 227-240 pumping speeds of, 240, 249-254, 293-302 with vapor trap, 240, 252
Diffusion
Gas ballast, 179-183 Gas flow, 14, 23-62, 277-291 through a hole, 12, 47-49 low pressure range, 43-57 methods of measurement of, 277-291
molecular flow, 14, 23, 43-57, 60 Poiseuille's law, 26-30, 34-36
argon instability, 396 hydrocarbon contamination, 394 mechanism of operation, 394 slotted cathode construction, 396398 triode getter-ion pump, 396-398 Getter pumping, 401-408 deposition of reactive metals, 401408 molybdenum, 402-404, 408 nickel, 407 titanium, 401-408 zirconium, 402
Reynolds number, 31
pressure range, 15, 23, 36-43, 59 turbulent flow, 31-34 viscous flow, 14, 23, 26-30, 34-36, 58 Gas law, general, 1-4 Boyle's law, 1 Charles's and Gay-Lussac's law, 1 Gases, boiling points of, 409 general gas law, 1-4
ideal gas, deflnition, 2
pumping
effectiveness as function of
transition
Halogen leak detectors, 159-161, 162 Helium leak detectors {see Leak
detectors)
purging, 271
molecular constitution of, 4 molecular weights of, 2 molecules per unit volume, 5 nature and behavior of, 1-22
ratio of specific heats, y, 11, 220 specific heats at constant pressure
speed factor, 253, 272 ultimate pressure, 257-268 working fluids, 240-249
and
at constant volume, 11
Ideal gas, deflnition, 2 Ionization of gases, 90-98 cross section for electrons, 93 ionization potential, 92 ionization probability, 93 Ionization gauge, conventional cathode type, 90-102
hot-
454
Ionization
SUBJECT INDEX
gauge, conventional hotcathode type, alternative methods of operation, 92 Blears's effect, 100-102, 242, 297 cahbration of, 97, 125-128
cross section for ionization, 93 design of gauge tubes, 99-101
SUBJECT INDEX
Mass
flow, definition, 24
of,
455
Mass spectrometer leak detectors (see Leak detectors) Mass spectrometer vacuum analyzers,
133-159 magnetic deflection types, 133-143
cycloidal focusing, 142
Dempster magnetic focusing, 133140 Nier 60 degree deflection, 140 Vanderslice 90 degree deflection, 140 resonance types, Farvitron, 157-159 linear accelerator, 152-157 omegatron, 143-152
Bayard-Alpert ionization
gauge; Magnetron ionization gauges; Penning discharge gauge) Isentropic flow, 219-221, 278-282
McLeod
gauge, 69-78
of condensable reading, 71
vapor
on
accommodation
coefficient, effect
on
due to connecting tube, 73-75, 125 due to liquid-nitrogen-cooled trap, 72, 125, 127 methods of controlling mercury
error,
level, 75-77 primary standard for pressure measurement, 72 response formula, 70-72 scales, linear and quadratic, 71
Metal bellows, 316 for rotary motion seals, 317-318 for translational motion seals, 316 for valve-stem seals, 320-323 Metal gaskets, 370-378 aluminum foil, 375-377 copper bead, 372 copper ridge, 373 copper shear, 371 flare seal, 374 knife-edge seals, 373 soft metal ring, 370, 374 reweldable flanges, 378 Molecular drag pumps, 205-214 analysis of performance, 205-210 performance of various designs, 211214 Molecular mean free path, 5, 13-15, 21,
23 derived in terms of molecular diameter, 13 relationship to viscosity of a gas, 21
multiple, 77
sensitivity, 72
Magnetron
ionization
gauges,
cold
helium
(see
of, below)
cathode inverted magnetron, 118120 cold-cathode magnetron, 120-122 hot-cathode magnetron, 107-111 {See also Bayard-Alpert ionization gauge; Ionization gauge, conventional type; hot-cathode Penr'ing discharge gauge; Pressure gauges)
determining character of gas ''2, 23 Molecular tu.bine pump, 214-218 Molecular weights, 2, 10
role in
flow,
ionization of,
Maxwell-Boltzmann velocity
bution
of,
distri-
Dempster 180
deflection,
8-11
of, 5,
purification
and
circulation,
mean
184
13-15, 21, 23
456
Molecules, polyatomic, 11 velocities of, 7-1
SUBJECT INDEX
Pressure, gas, definition of, 1 dependence, on kinetic energy of
SUBJECT INDEX
457
Pumpdown
Motion
seals,
313-318
Seals, elastomer, O-ring gaskets, couplings, quick connect, 309 groove designs for, 307-309 guard ring with double seal, 309 properties of various elastomers, 311-313
354-358
Nozzles,
converging-diverging
219 220 diffusion pump, 227-230, 265-267 isontropic flow, 219-221 mass flow through, 221 velocity of gas flow through, 220
critical pressure,
of nonisotropic
system
factors,
352-354
Pumping
square-cross-section gaskets, 309 Solvents, properties of, 304 Sorption processes, absorption, 364
434
315 Oil ejector pumps, 257 Omegatron, 143-152 argon vs. helium leak
rings, 307-313,
permanent, 18, 63 vapor pressure, 17, 71, 331, 409 Pressure drop formula for viscous flow,
detection
bution, 416-418 methods of measurement, 291-302 performance, adsorption pumps, 347 cryogenic pumps, 411-421 diffusion pumps, 249-254 Evapor-ion pumps, 390 getter-ion pumps, 391-393
absorption pumping, 398-400 adsorption, 364 adsorption pumping, 428-430 chemisorption, 364
cryosorption, 421-424 cryotrapping, 424-428 desorption, 365-370 Specific heats of gases, 1 Standard conditions of temperature
mechanical booster (blower) pumps, 189, 193-202 mechanical oil-sealed pumps, 171177 molecular drag pumps, 212 molecular turbine pumps, 218
278-282
mass 280-282
standardized dimensions of, 280 subcritical gas flow through, 280, 282 Outgassing, bakeout procedures and
effectiveness,
effect
101
effect
363-369
at
rate
room
Alphatron, 122 cold-cathode types, 113-122 magnetron, inverted Haefer 118-120 Penning discharge, 113-118 Redhead magnetron, 120-122 hot-cathode types, 90-113 Klopfer magnetically collimated electron beam gauge, 111-113 Lafferty hot-cathode magnetron, 107-111 {See also Bayard-Alpert ionization gauge; Ionization gauge, conventionalhot-cathode type) Knudsen radiometer type, 123 McLeod, 69-78 manometer, diaphragm, 65-68 liquid, 63-65 partial {see Mass spectrometer vacanalyzers) thermal conductivity, 78-90 Pirani type {see Thermal conductivity pressure gauges)
vacuum systems, 26, 277, 360 resultant for pump combined with a
conductance, 25, 58
units of, 24
Pumping speed
factor
for
diffusion
227-272
Temperature, absolute
Evapor-ion, 388-391
Partial
pressure
gauges
(see
Mass
uum
scales, 2, 4 absolute zero of, 2 centigrade scale, 2, 4 dependence, of gas pressure on, 1-3 of vapor pressure at low tempera-
spectrometer
118
vacuum
analyzers)
Penning discharge gauge (PIG), 113114-116 principles of operation, 113 useful pressure range, 114-118 Pipe sizes, selection of, for viscous flow, 30 Pirani pressure gauge {see Thermal conductivity pressure gauges) Poiseuille's law, 26-30, 34-36
erratic behavior of,
thermocouple type
(see
Thermal
Pumpdown Pumpdown
factor
for
mechanical
185 molecular drag, 205-214 molecular turbine, 214-218 oil vapor ejector, 257 steam ejector, 219-227 Vac Ion, 393-398
molecul'
pressure,, 78
pumDS, 349-351
time, 348-361 effect of outgassing, 351-358 for mechanical pumps, factor
78-90
basic principles, 78-83
307-318
accommodation
coefficient,
79-81
349-351
458
SUBJECT INDEX
Traps, vapor, creep barrier for organic fluids, 332, 367
SUBJECT INDEX
459
9, li
mechanical, 169-218
vapor-jet, 219-272
average,
Stefan-Boltzmann law, 81 thermal conduction loss along filament, 81-83 thermal conductivity of rarefied
gases, 78
exhaust baffles for diffusion pumps, 340 forevacuum, 340-342 functions of, 328-330
inlet baffles for diffusion
303-307
pumps,
Pirani gauge, 86-90 control circuits for, alternative, 87 pressure range of, 87 principles of operation of, 86 response curve vs. pressure, 89-91 thermistor type of, 87-91
thermocouple gauge, 83-86 compensation for ambient temperature, 85 matched tubes, 84 multi-station control circuit, 85 principles of operation, 83
240, 252, 329-339 mechanically refrigerated, 338340 performance of diffusion pump with, 240, 252, 335-337 surface migration of organic fluids, 332 temperatures for various applications, 244, 246, 329-331 thimble traps, 330-332 Turbulent flow, 31-34 occurrence in vacuum systems, 3234
cleaning of interior surfaces, 304 external pressure requirement, 303 finish of interior surfaces, 304 leak hunting, 161-167, 307 materials of construction, 303
Maxwell-Boltzmann distribution
of,
most probable,
root
9,
10
7, 9,
mean
square,
n
to, 11
pumping
Thermocouple
gauge
(see
Thermal
Ultrahigh vacuum techniques, 363-435 absorption pumping, 398-400 bakeable valves, 378-382 bakeout procedures, 364-370 cryogenic pumping, 408-434
getter-ion
welding specifications, 305 Valves, vacuum, 318-328, 378-382 bakeable, 378-382 functions of, in vacuum systems, 319 gate, 323-328 butterfly type, 324-326 disk typos, 325, 327 modified plumbing types, 323 sliding plate typos, 324-326 globe, 318-323 bellows sealed types, 318, 320323 diaphragm sealed type, 320 elastomer sealed types, 318
needle type, for control of gas
flow, 325, 328
34
of 172
Volumetric
efficiency
mechanical
vacuum pumps,
Water
\'apor,
sealod mechanical
of
thimble
trap
in
Mechanical
relation to
pumps, 385-398
change of phase
pressure,
vapor and
steam ejectors, attainable water vapor pressure, 223-225 backstreaming in, 227 vapor pressure as function of temperature, 331
Titanium,
pumping
by
de-
thimble trap, 400 metal gaskets, 370-378 reactive metal deposition, 401408
liquid-helium-cooled
surface
temperature and
Zeolite, absorbent material for
vapor
definition of, 18
traps, at
phenomena, dominance
of,
347
refrigerated, 347,
traps, 328-341
absorption materials, 341-343 bakeout cycle, 342-345, 347 capacity for gases and vapors,
Vapor Vapor
baffles
and
pressure,
of
gases
at
low
344-348 copper foil type, 342-344 stay-down times for, 342 effectiveness as a pump, 347, 398-400 liquid-nitrogen-cooled, 347 tray design, 344 ultimate pressure, 344-347 vapor, 328-341 automatic liquid nitrogen level control, 337 conductance of baffle systems, 333 337
absorbing various gases, 344-347, 399 quantity of vapor and gas evolved during bakeout, 345
effectiveness
in
Vacuum components,
tion,
358-361
valves, 361
Vacuum
gauges
(see
Pressure gauges)
Other
special
features
include
methods
Here
of
of
vapor
'
authoritative coverage of
vacuum
sys-
problems associated
every aspect of
vacuum technology.
GERHARD LEWIN
illustrations
knowledge
of
vacuum technology,
this
unique reference
theory equations
.
been involved with physical apparatus requiring larger than normal vacuum systems. Since 1937, he has acted as
consultant
in
"
gas flow
. .
surface effects
.
. .
...
punnping processes
.measurements
all
com-
ponents
...
vacuum technology
to in-
dustrial firms
gaged
in
The book
systems
evaluates
basic
vacuum
effort of the
equipment for your specif ic purposes and guides you in the actual design of special equipment.
Filled with
and
vacuum
'
as a working tech-
"
/
y \
'
'
"
.
University
of
California,