The Units of Pressure Measurement

All vacuum problems started with the Italians. There was one obstacle to
mining in Italy in the early 1600s ... Water. It was everywhere. Before the
miners could dig, the water had to be pumped to the surface. It irked them
that their 'suction' pumps could only 'suck' water up to a height of 32 feet
above flood level (or rather, the contemporary equivalent of 32 feet). At
that point, the pump effluent had to be spilled into vats and another pump
used to suck the next 32 feet. Why couldn't they use one pump for the
whole distance? What was magical about this height of 32 feet?

They posed the problem to Galileo but he did little with it until three
months before his death when he tossed it to a mathematician named
Torricelli who came to study under him. Torricelli had been kicking
around ideas about 'oceans of air' surrounding us and concluded that he
could bring this pump problem to a manageable size by using a fluid
denser than water. Mercury seemed a good choice. He had his assistant
fill a glass tube (closed at one end) with mercury, placed the open end in
a dish with more mercury, and raised the closed end. The mercury
reached a level equivalent to 32 feet times the ratio of the densities of
water to mercury. What would be above the mercury if the glass tube was
long enough? A vacuum, of course!

Ah, there's the rub, as Shakespeare said about the same time. Galileo
believed that vacuums (vacua?) could not exist and he had already been
put under house arrest by the Church for saying things like the earth went
around the sun. There was no way Torricelli was going to broadcast the
results of the 'vacuum experiment' himself. But he did write to a friend in
Rome who copied the letter as sent it to a Father Mersenne in Paris.
Mersenne, a minorite friar, acted as sort of a medieval computer bulletin
board. He promptly copied the letter again (where was Xerox when it was
needed most?) for his friend Blaise Pascal who lived close by and
therefore, at a sufficient distance from Rome to ignore the Church's word
- to a certain extent.

Pascal, being a literal kind of man, set up the experiment in full scale
using water and 'mirabile dictu' confirmed the existence of a vacuum. It
followed from Torricelli's ideas that if the weight of air pushed mercury
so far up the tube, then the mercury level would be reduced if the test
were done at higher elevations. Blaise Pascal's brother-in-law lived in
central France in an area surrounded by mountains and was apparently
adventurous (and strong) enough to march a complete mercury barometer
to the top of the nearest mountain. The rest (to use a very bad pun) was
downhill from there.

In my opinion, Torricelli, a mathematician, was a premature software

guy. Blaise and his brother-in-law got the job of proving the prediction
worked on a grand scale and up a mountain, less because of the Church's
decree and more because Torricelli was reluctant to get his hands dirty.
Whatever the reason, Torricelli was honored later by someone naming the
pressure measurement unit of 1 millimeter of mercury, the torr.

Nice story, but the Battle of the Units had only just begun. Life with torr
would have been lovely, except the Brits had to tinker with it. Liking
everything to be in Imperial Units, they converted the 760 mm Hg
pressure of the standard day to 29.92 inches Hg. Well, if you like inches,
that's ok. But notice how the weather forecasters on American TV long
ago forgot that it was the height of mercury they quote every night. Every
one of them says, "The barometer is 29 inches and rising." What's this ...
psycho-kinesis?

The inch thing really got out of hand when manufacturers of rough pumps
came on the scene. Rough pumps are arbitrarily defined as those used for:
in-house vacuum systems; meat packing; impregnating lumber and
transformer coils; making freeze dried coffee, tea or foods (got ya!). That
is, any pump that hauls great loads of gas and vapor day-after-day to a
modest vacuum level. These manufacturers noted that if atmospheric
pressure was 29.92 inches Hg, they would be shooting for 0 inches. That
would look bad in their brochures. So, they calmly inverted the scale.
Atmospheric pressure is 0 inches Hg and the best possible vacuum is
29.92 inches Hg, they said. Which left the rest of us struggling with
converting inches Hg to torr. (First, subtract the given inch pressure from
29.92 inches, then multiply the answer by 25.4.)

Since the standard meter and kilogram are kept in Paris, I blame the
French for the metric system. Not that I really object to it. After all, I
accept 760 torr is really 760 mm Hg without too much argument. But
someone, somewhere, noted that 760 mm Hg could not be related to any
basic measurement units. The column's height depended on the mercury's
density and that wasn't basic in anyone's scheme. "Let's make the pressure
unit conform to the cgs (centimeter/gram/second) system" they said with
glee, knowing how much they would confuse the rest of us.

And how is this done? First, we must understand that "pressure" is force
per unit area. What's the unit of force in the cgs system? Remember way
back at school, your science teacher smacked your knuckles for not
remembering that the dyne is that force which gives 1 gram an
acceleration of 1 centimeter per second per second? You really should
have listened because pressure in the cgs system is measured in dynes per
square centimeter or, to give its proper name, microbar. Why micro
(millionth) bar? Well, try this for an explanation. One million microbars
(or 1 bar) is 750.06 torr. That is, the atmospheric pressure of the standard
day is 1.0133 bar which is easier to remember than a number which must
be multiplied by ten to the 'minus oh-dear-I've-forgotten'. The cgs
advocates rounded off this part of the story by declaring pressures will be
measured in millibar (because it's close in value to the torr?). Indeed, to
this day, millibar is the unit used for recording both vacuum and weather
pressures in Europe.

If I can blame the French for the cgs system, what defence can they offer
for the next leap? Some august body set up an international standard for
measurements called Systéme International d'Unités (or SI). What could
be more French than that? They threw out the cgs system's claim to fame
and installed the MKS (meter/kilogram/second) system. The unit of force
in SI units, that is the force to accelerate 1 kilogram at 1 meter per second
per second, is called the newton. Don't you get the flavor of collusion
between Brits and French here? And what do you think the MKS pressure
unit (1 newton per square meter) is called? You've got it - the pascal! I
rest my case.
Of course, since the units of length and mass in cgs and MKS are all
related by powers of ten, millibar (mbar) and pascal (Pa) have the simple
correspondence:
100 Pa = 1 mbar ( = 0.75 torr)
But does that really make you sleep easier at night? If you want
nightmares, try this one. In pressure measurement, when does a milli
equal a micro? Answer: consider the humble torr. If one torr equals 1
millimeter of mercury, then 1 millitorr must be equivalent to 1
micrometer of mercury, right? And what do we call a micrometer? - a
micron. All of which says that:
1 millitorr = 1 micron Hg
But how often have you heard someone quote a pressure of "250 microns
of mercury"? We all think of the unit as "micron" leaving the neophyte
vacuum person with the impression that a milli (unit) equals a micro
(unit). Just remember, like the weather forecasters we have simply
forgotten to add "mercury" to the microunits.
Pressure Measurement By Manometer

The relationship between pressure and head is used to measure pressure with a
manometer (also know as a liquid gauge).

Objective:

• To demonstrate the analysis and use of various types of manometers for

pressure measurement.

1. The Piezometer Tube Manometer

The simplest manometer is a tube, open at the top, which is attached to the top of
a vessel containing liquid at a pressure (higher than atmospheric) to be measured.
An example can be seen in the figure below. This simple device is known as a
Piezometer tube. As the tube is open to the atmosphere the pressure measured is
relative to atmospheric so is gauge pressure.

A simple piezometer tube manometer

This method can only be used for liquids (i.e. not for gases) and only when the
liquid height is convenient to measure. It must not be too small or too large and
pressure changes must be detectable.
2. The "U"-Tube Manometer

Using a "U"-Tube enables the pressure of both liquids and gases to be measured
with the same instrument. The "U" is connected as in the figure below and filled
with a fluid called the manometric fluid. The fluid whose pressure is being
measured should have a mass density less than that of the manometric fluid and
the two fluids should not be able to mix readily - that is, they must be immiscible.

A "U"-Tube manometer

Pressure in a continuous static fluid is the same at any horizontal level so,

For the left hand arm

For the right hand arm

As we are measuring gauge pressure we can subtract giving

If the fluid being measured is a gas, the density will probably be very low in
comparison to the density of the manometric fluid i.e. rman >> r. In this case the
term can be neglected, and the gauge pressure give by

3. Measurement Of Pressure Difference Using a "U"-Tube Manometer.

If the "U"-tube manometer is connected to a pressurised vessel at two points the

pressure difference between these two points can be measured.

Pressure difference measurement by the "U"-Tube manometer

If the manometer is arranged as in the figure above, then

Giving the pressure difference

Again, if the fluid whose pressure difference is being measured is a gas and
, then the terms involving can be neglected, so

4. Advances to the "U" tube manometer.

The "U"-tube manometer has the disadvantage that the change in height of the
liquid in both sides must be read. This can be avoided by making the diameter of
one side very large compared to the other. In this case the side with the large area
moves very little when the small area side move considerably more.

Assume the manometer is arranged as above to measure the pressure difference

of a gas of (negligible density) and that pressure difference is . If the
datum line indicates the level of the manometric fluid when the pressure
difference is zero and the height differences when pressure is applied is as shown,
the volume of liquid transferred from the left side to the right

And the fall in level of the left side is

We know from the theory of the "U" tube manometer that the height different in
the two columns gives the pressure difference so

Clearly if D is very much larger than d then (d/D)2 is very small so

So only one reading need be taken to measure the pressure difference.

If the pressure to be measured is very small then tilting the arm provides a
convenient way of obtaining a larger (more easily read) movement of the
manometer. The above arrangement with a tilted arm is shown in the figure
below.

Tilted manometer.

The pressure difference is still given by the height change of the manometric
fluid but by placing the scale along the line of the tilted arm and taking this
reading large movements will be observed. The pressure difference is then given
by

The sensitivity to pressure change can be increased further by a greater

inclination of the manometer arm, alternatively the density of the manometric
fluid may be changed.

5. Choice Of Manometer

Care must be taken when attaching the manometer to vessel, no burrs must be
present around this joint. Burrs would alter the flow causing local pressure
variations to affect the measurement.

Some disadvantages of manometers:

• Slow response - only really useful for very slowly varying pressures - no
use at all for fluctuating pressures;
• For the "U" tube manometer two measurements must be taken
simultaneously to get the h value. This may be avoided by using a tube
with a much larger cross-sectional area on one side of the manometer than
the other;
• It is often difficult to measure small variations in pressure - a different
manometric fluid may be required - alternatively a sloping manometer may
be employed; It cannot be used for very large pressures unless several
manometers are connected in series;
• For very accurate work the temperature and relationship between
temperature and r must be known;

Some advantages of manometers:

• They are very simple.

• No calibration is required - the pressure can be calculated from first
principles.
Pump Types and theory – by Varian

Diffusion Pump

Historical Notes
Producing low pressures is the function of vacuum pumps, one type of
which is the diffusion pump. Diffusion pumps were first conceived and
constructed by W. Gaede (1915-Germany) and I. Langmuir (1916-
U.S.A). They operate on the principle of transferring momentum from
high velocity vapor molecules to the gas molecules that are to be moved
out of the system. The vapor molecules are formed by heating a suitable
condensable fluid. The early pumps used mercury for this purpose.

In the late 1920s, C.R. Burch (England) and K.C. Hickman (U.S.A.)
found that certain high molecular weight oils having high boiling points
and low vapor pressures could be used as pumping fluids. These oils, not
generally synthetic hydrocarbons, were useful because they remained in
the pump indefinitely and allowed lower pressures to be attained without
the use of a cold trap (see section on Baffles and Traps). Today, with the
exception of a few isolated applications like some analytical instruments,
all diffusion pumps utilize some form of oil. For additional information in
this area, see the discussion on pumping fluids, below.

As industrial and scientific requirements for rarefied atmospheres

increased, research and development into the nature and production of
high vacuum increased. By the early 1940s, a well-developed vacuum
technology existed and was intensified both during World War ll and by
the space effort of the 1960s. Engineering has continued in the vacuum
field and in 1965, Varian's M.H. Hablanian, et al. made a significant
contribution to diffusion pump design that markedly increased pumping
speeds.

Applications
Due to its simplicity, high performance, and low initial cost, the diffusion
pump remains as the primary industrial high vacuum pumping
mechanism. Applications for this type of pump are found in such diverse
areas as:
· Analytical instruments
· Coating, functional
· Coating, ornamental
· Electron tube manufacture
· Metallurgy
· Optics
· Outer space simulation
· Particle accelerators
· Petrochemicals
· Pharmaceuticals
· R&D laboratories
· Semiconductor manufacture
By the proper choice of motive fluids, traps, baffles, and valves, diffusion
pumps can be used in a wide variety of applications and over pressure
ranges from 1 x 10-3 Torr to 2 x 10-11 Torr.

Basic Performance Factors

1. Pumping speed is volume per unit time. It is generally
specified in liters/second and is an important parameter in determining
the ultimate pressure of a system. This is expressed by the relationship
Q = PS
Where:
o Q is the system gas load in Torr-liters/second
o P is the attainable pressure in Torr
o S is the effective pump speed at the system
o 'Q' is the total leakage of the system which includes
vapors given off by dirt and outgassing of internal surfaces as well as
holes to the outside world. Ultimate pressure is also affected by such
factors as the compression ratio for light gases and the nature of the
pumping fluid.
2. Maximum throughput is the pump's maximum mass gas
transfer capability pressure x volume per unit time. It is generally
specified in Torr-liters/second.
3. Tolerable forepressure is the maximum allowable pressure
in the foreline. It is maintained at or below this value by a suitably-sized
mechanical foreline (backing) pump. If this pressure increases above that
specified for a given pump, gas will diffuse back through the pump and
pumping will stop. It should be noted that the size of this mechanical
pump can affect the maximum throughput value.
4. Backstreaming rate is the rate at which the pumping fluid
vapor leaves the inlet opening of the pump, moving back in the direction
of the system being pumped. It is measured in milligrams per cm2 per
unit time and will vary with the type of motive fluid employed.

Operation
Diffusion pumps are vapor jet pumps that work on the basis of
momentum transfer from a heavy high speed vapor molecule to a gas
molecule. This results in the gas molecules being moved through the
pump. With reference to Figure 1, the bottom of the pump contains an
electric heater which is used to produce the vapor by heating the pumping
(motive) fluid to its boiling point at reduced pressure.

This means that before the pump is started, it must be 'rough pumped'
down to and held at an acceptable pressure, typically 10-1 Torr. (For
information on rough pumping, see section on Primary Pumps.) To do
otherwise will result in no pumping actionand possible damage to the
pumping fluids. Once boiling of the fluid has begun, the vapor is forced
up the central columns of the jet assembly. It then exits at each
downward-directed jet in the form of a molecular curtain that impacts the
water-cooled pump body. Here, the vapor condenses and runs back down
to the boiler. This refluxing action continues as long as proper heat and
forepressure are maintained.
As gas molecules from the system randomly enter the pump (molecular
flow conditions), they encounter the top jet. Some of them are correctly
impacted and driven on to the next jet. Subsequently, they reach the
foreline where they are exhausted to the atmosphere by the mechanical
backing pump.

The diffusion pump is similar in character to other compression pumps in

that it develops a relatively high exhaust pressure compared to the inlet
pressure. This compression ratio for an inlet pressure of 2 x 10-7 Torr and
a foreline pressure of 2.0 x 10-1 would be ten million to one for most
gases. Figure 2 shows how the pumping speed varies with pressure. Note
that the speed remains constant from the 10-3 Torr scale to the X-10 Torr
scale and then falls off as a result of the compression ratio for hydrogen
and helium plus the vapor pressure contribution of the pumping fluid.

Typical plot of diffusion pump performance. Four regions are evident: 1)

Effect of the pressure ratio limit; 2) Normal operating range with constant
speed; 3) Throughput limited condition; 4) Effect of backing pump.
In the same way that the pump must be rough pumped before starting, so
must the system to be evacuated by rough pumping prior to exposure to
the pump. Exposing a hot pump to a rush of air at atmospheric pressure
could be catastrophic for the equipment and possibly explosive,
depending upon the pump fluid being used. For further information in this
area, see the discussion on pumping fluids, below, and the section on
valves.

Design Features
Design features unique to Varian diffusion pumps provide positive
benefits to the customer, such as:
1. Varian oil diffusion pumps incorporate an ejector stage as well
as the full fractionation jets. This feature assures the user of constantly
purified pumping fluid and the capability of maintaining low pressures.
2. Varian oil diffusion pumps incorporate insulated jet drip
shields which prevent re-boiling of oil droplets outside the jet assembly.
This feature assures the user of the lowest backstreaming rates attainable.
3. Varian water cooling coils are attached by a proprietary
weld/brace technique. This special technique means excellent thermal
contact and no chance for coils to 'melt' away from the pump body in
cases of accidental overheating.
4. Varian pumps all incorporate a water-cooled cold cap which
reduces 98 percent of the backstreaming common to most diffusion
pumps. Thus, the user is assured of a cleaner system.
5. Varian water-cooled pumps incorporate the quick cool boiler
coils, allowing faster shutdown of the system with no damage to the oil.
6. Varian pumps utilize standard ASA flanges. This feature
permits wide flexibility for mating with systems and other hardware.
7. Varian (4-inch and larger) pumps have a thermal protection
switch as a standard feature. This device prevents damage to the pump
and surroundings due to overheating.

Pumping Fluids
In an oil diffusion pump, high speed heated oil vapor provides the kinetic
energy that moves gas molecules to the foreline and prevents their back-
migration. These oils may be derived from a petroleum base but more
typically are synthesized from phthalates, sebacates, phenyl groups, or
siloxanes.

To be an effective pumping fluid, the compound must have a relatively

high molecular weight and a low vapor pressure at elevated temperatures.
Other desirable properties are inertness and stability in order to resist
chemical reaction and disintegration into undesirable fractions.
Phenyl ethers such as Neovac-SY and Santovac-5 are fairly resistant to
oxidation and are used successfully around electronic devices. These oils
polymerize into a conducting film when bombarded with electrons and
thus do not promote static charge build-up. In addition, they are quite
soluble and 'clean up' easily. Neovac-SY has the advantage of economy
while Santovac-5 is more durable and has a lower vapor pressure.

For additional oxidation resistance, many applications lend themselves to

the use of silicone fluids. These are phenyl siloxane compounds that
polymerize as a non-conducting film that can allow static charge buildup
and are difficult to 'clean up'. Two common fluids of this type are DC-
704 and DC-705; the former has four phenyl groups and the latter has
five. The DC-705 is, therefore, a heavier molecule, and it has a lower
vapor pressure, so it is highly suitable for achieving very low pressures.
However, it is somewhat less effective under high throughput conditions
than DC-704, due to the fact that fewer molecules emerge from the top
jet.
Another extremely stable fluid under reactive conditions is the fluorinated
polyphenyl ether (Fomblin or Krytox). This oil is widely used in
mechanical oil-sealed pumps where large amounts of oxygen are pumped.
It is also suitable as a diffusion pump fluid where large quantities of
oxygen or other reactive gases may be encountered.

Varian Low-Profile Water-Cooled Baffles combine 100% optical density

with high conductance and unusually low overall height. They are
especially useful in applications where clean operation down to 10E-8
Torr is required but cryogenic traps are not. With M-series pumps, these
traps retain approximately 50% of the pumping speed. Mechanical
refrigeration can be used to reduce the re-evaporation of pump fluid and
attain a partial trapping effect.

The liquid nitrogen Cryotraps provide optimum performance for diffusion

pumps. These traps offer optical density intercepting 100% of primary
backstreaming while giving additional pumping speed for condensables.
Each trap has a large built-in reservoir that gives long, unattended service.
Cryogenic temperatures are maintained even as liquid nitrogen level
drops. High conductance internal geometry achieves the highest possible
pumping speed at the inlet taking full advantage of the diffusion pump's
speed. Varian's Halo Baffles are used instead of a standard cold cap and
are therefore integral to the pump, adding no height to the pump. They
reduce primary backstreaming by approximately 90% while only cutting
the pump speed by less than 80%, about half that of opaque chevron
baffles. Pumps can be ordered with halo baffles installed or can be
retrofitted in the field.

Extended Cold Caps are used in place of the standard cold caps in the
VHS-series diffusion pumps. They reduce primary backstreaming to
levels that cannot be measured by the American Vacuum Society
standard collection methods. They can be ordered installed in a new
pump or can be retrofitted in the field.

The accompanying table shows the speed of Varian diffusion pumps

when using the appropriate baffle or trap.

Ion Pumps
Thin films of reactive materials have been used for 'gas cleanup' or
'gettering' for over a century. The early electron tube makers were only
able to mechanically pump to about 1 x 10-4 mbar, but through the use of
'getters' flashed on the internal surfaces, pressures in the low 10-7 mbar
scale were attained. These getters were typically metals like barium,
titanium, zirconium, or thorium. Gettering materials are still used in tubes
today even though pressures of 1 x 10-8 mbar are readily attained by the
pumps in the manufacturing process.
Gettering was not employed extensively in vacuum systems until the
1960s, when it was found to be highly compatible with ion pumping.
Titanium was the metal commonly used because of its availability and its
ability to sublime readily over a moderate temperature range.
Applications
Due to cleanliness, bakeability, low power consumption, vibration-free
operation, long pumping life, and high pumping speed, Titanium
Sublimation Pumping (TSP) is the ideal cost-effective companion to ion
pumping in ultrahigh vacuum.
Applications for this pumping mode are found in many areas, such as:
1. Auger electron spectrometry
2. Electron spectroscopy for chemical analysis
3. Electron tube manufacturing
4. Mass spectrometers
5. Materials science conductor R&D
6. Nuclear physics
7. Outer space simulation
8. Particle accelerators
9. Secondary ions mass spectroscopy
10. Solid state semiconductors

Basic Performance Factors

1. Pumping Speed. The pumping speed of a Ti film is proportional to the
film area and to the sticking coefficient that is the probability that an
impinging gas molecule reacts with Ti forming a stable compound. The
pumping speeds per unit area of a fresh evaporated Titanium film are
reported in Table 1. Using these coefficients the intrinsic pumping speed
(Si) of a Ti film can be evaluated using the following equation:

Si[l/s] = Coefficient x Surface

As the gas molecules react with the surface Ti atoms, the number of
active sites decrease and, as a consequence, the pumping speed decreases.
A plot of the specific pumping speed vs time at different pressures is
reported in Figure 1. Using this plot it is possible to estimate how
frequently the Ti film has to be renewed.

It must be noted that the actual pumping speed S of a TSP depends on the
conductance C between the active surface and the vacuum vessel
according to the following equation:
1/S = 1/C + 1/Si
2. Throughput. When the impingement rate of the gas molecules on the
active film becomes higher than the Ti sublimation rate (excess of gas
molecules respect the available Ti atoms), the pumping speed does not
depend any more on the sticking coefficient. It is simply controlled by the
quantity of the available Ti atoms according to stoichiometric reaction.
If n Ti atoms need to pump a gas molecule
example: 2Ti + N2 -> 2TiN, n = 2
the gas throughput Q is given by:
Q [mbar l/s] = (0.13)/n R[gr/h]

where R is the Ti sublimation rate. In this condition the pumping speed is

not constant but it depends on the pressure and is directly proportional to
the sublimation rate (Figure 2).
 Typical Pumping Speeds Per Square Inch (Per Square Centimeter) of
Titanium Sublimation Surface for Various Gases (Liters Per Second)
Gas H2 N2 O2 CO CO2 H 20 CH4 Ar He
Surface 20 C 20 (3.1) 30 (4.7) 60 (9.3) 60 (9.3) 50 (7.8) 20 (3.1) 0 0 0
Temp -195 C 65 (10.1) 65 (10.1) 70 (10.9) 70 (10.9) 60 (9.3) 90 (13.9) 0 0 0

3. Other factors. The overall performance of a titanium sublimation

pump is a function of several variables, including gas species, pressure,
gas temperature, getter film temperature, get ter film area, the geometry
of the area, sublimation rate,sticking coefficient, and the conductance
from the film to the area being evacuated. For further information, write
for 'Predicting and Evaluating Titanium Sublimation Pump Performance'
by D.J. Harra, 1974 (Vacuum Report VR-88).

Operation
Titanium Sublimation Pumping is accomplished by coating the inner
surfaces of a vacuum system with sublimed titanium films. Since it
involves a chemical reaction, this kind of pumping is useful where mainly
active gases are present. The pumping speed of a unit area varies with
various reactive gas species as shown in the following table. It can also be
seen that cooling the substrate to liquid nitrogen temperature markedly
increases the speed for hydrogen and water.
The gases thus 'gettered' form stable compounds with titanium and are
stored in the system as such until they are removed by cleaning. Since
there is generally un-reacted pyrophoric titanium buried in the deposits,
caution should be used in cleaning. If the desired gas throughput is
known (Q = pumping speed x pressure) the maximum theoretical
operating time is given by:
Operating time [h] = 0.13 T [gr]
______ _______
n Q [mbar l/s]

where T is the usable Titanium.

For example, using our cryopanel at 1 x 10-8 mbar with a three filaments
cartridge Ti source, the theoretical operating time is given by:
0.13 3.6 [gr] = 46,800 hrs = about 5 years
___ _________________
2 500 [l/s] x 10-8 [mbar]

After this time the filament cartridge should be replaced.

Design Features
· All sources are mounted on 23 4" Varian ConFlat Flanges and
fit through 11 2" ports.
· The three-filament source contains 3.6 grams of useful
titanium.
· The Varian Mini Ti-Ball source contains 15 grams of useful
titanium.
· The Varian Standard Ti-Ball source contains 35 grams of
useful titanium.
· Control units for the Ti-Ball source configuration
automatically maintain the temperature dependent titanium phase change
and thereby insure the most efficient use of the available getter.

Turbo Pumps

Turbomolecular Pump Bearings

The Turbo-V pumps incorporate Varian's innovative ceramic bearing
design with a proprietary ultra-low vapor pressure solid lubricant, which
enables these pumps to provide a long service life and a high degree of
cleanliness under many operating conditions. The ceramic bearings utilize
balls made of silicon nitride, a polycristalline material with an amorphous
intergranular binder base that offers the following advantages.
Hardness This is a critical aspect of bearing design, and it closely relates
to bearing performance and reliability. The silicon nitride material used in
Varian's Turbo-V bearing system is twice as hard as conventional steel
providing dramatic improvement in wear resistance while minimizing the
effects of surface contact and stress.

Weight Silicon nitride is 40% less dense than conventional steels, which
helps to reduce centrifugal loading and stress levels at high rotational
speeds, especially in the bearing race area.

Friction Silicon nitride's low coefficient of friction enhances wear

resistance and adds to the bearing's operational life.

Thermal stability With its low thermal expansion coefficient, the silicon
nitride bearing material ensures that tight tolerances and mating
component fit will be maintained over an extremely wide temperature
range. In addition, silicon nitride has an outstanding resistance to fracture
by thermal shock.

Chemical Stability Silicon nitride is virtually inert.

Another feature of the Turbo-V bearing system is its proprietary lubricant
which has an extremely low vapor pressure and is virtually hydrocarbon-
free. The use of this lubricant in the permanently sealed bearing system
ensures clean, reliable operation without the need for any maintenance
whatsoever.

General Care
There are some simple steps you can take that will further enhance the
efficient operation of the Turbo-V pump. For example, sudden exposure
to atmospheric pressure should be avoided because this will produce an
overload condition. The bearing temperature (displayed on the control
panel) should be kept within proper limits to avoid operation in poor
lubrication conditions. This will be achieved if the cooling instructions
are followed.

A vibration monitor can also be useful for preventive maintenance, as an

increase in the bearing vibration levels may be an indication that
replacement of the bearings is required.
When the pump is initially installed, or after a long period of inactivity, it
is good practice to use a soft start sequence, which will gradually
accelerate the rotor to full speed, allowing proper redistribution of the
lubricant in the bearings.

The MacroTorr Concept

The Varian award-winning, patented MacroTorr design, which was
developed in 1991, is the result of the improvement of the original design
of the Gaede molecular pump.
· It is based on the idea of substituting (rather than adding)
molecular impeller disks to some turbo-bladed stages.
· The molecular impellers consist of a disk rotating in a
channel in which the inlet and outlet are divided by a stripper.
· The cross section of the channels decreases from the top to
the bottom of the pump (from high vacuum to low vacuum or from the
low pressure to the high pressure zone).
· Gas molecules gain momentum after each collision with the
moving surface of the impellers.
· The gas is then forced to pass through a hole to the next stage
due to the stripper.

How to Select a Turbo-V Pump

The right choice of a turbomolecular pump depends on the application.
As a general rule, we can reduce the choice to two types of use:
· HV and UHV operation (low gas load)
· gas handling operation
The first choice includes most cases in which the turbomolecular pump is
employed to create vacuum in systems where the gas load is produced
only by outgassing. In this application, the choice is typically based on
the desired base pressure within a desired time as a function of the
foreseen outgassing rate, i.e

where:
· p is the desired base pressure (mbar)
· Q is the total outgassing rate at the desired time (mbar l/s)
· Seff is the effective pumping speed
The second choice relates to all operations where process gases must be
used. The main parameters are, therefore, the desired operating pressure
and the process gas flow

where Q' is the total gas flow and p' is the operating pressure.

How to Select the Backing Pump

The selection of a backing pump should be based on the required
roughing time and the minimum size of the pump for properly backing
the turbopump.

Roughing
Once the desired roughing time is established, the size of the forepump
can be determined through the following formula:

where
· Sforeline is the pumping speed of the roughing pump (l/min)
· V is the volume of the chamber to be evacuated (l)
· t is the desired roughing time (min)
· p0 is the starting pressure (mbar)
· p1 is the end pressure (mbar)
When using a foreline pump much larger than the recommended size, a
by-pass line might be necessary to achieve the calculated roughing time.
Backing
The backing pump must be big enough to achieve an effective pumping
speed as close as possible to the nominal speed.

where

• Sforeline is the pumping speed of the foreline pump

• Q is the gas load
• p is the operating foreline pressure

It should be noted that Q is the total gas load on the pump and includes
process gases and turbo purge gases when used.
The size of the backing pump can be calculated according to the
following rule:

where

• S is the pumping speed of the turbopump

• Sforeline is the pumping speed of the backing pump
• K is the maximum compression ratio of the turbopump
for a given gas (i.e. process gas) at the operating foreline
pressure

The pumping speed of the backing pump should be the highest of the two
values calculated as above (roughing and backing).

It is possible to use dry diaphragm pumps for hydrocarbon-free operation

when MacroTorr-type pumps are used

Vacuum Gauges

Historical Notes
Early interest in pressure measurement was stimulated in the 17th century
by engineers who were concerned about the inability of suction pumps to
remove water from mines. The pumps were limited to about 30 feet. For
example, the Duke of Tuscany (Italy) commissioned Galileo to
investigate the problem.
Galileo, among others, devised a number of experiments to investigate
the properties of air.
Among these experiments were pistons for measuring the force of
vacuum' and a water barometer that stood about 34 feet tall.

After Galileo's death in 1642, the work was carried on by his associate,
Evangelista Torricelli. Torricelli invented the mercury barometer (Figure
1); he concluded that atmospheric air forced water up to a height of 33.6
feet.

Figure 1 Notes

1. Air has weight and mass

o 2 Ibs. per cubic yard, or
o 1.293 grams per liter
2. Pressure = force per unit area
3. The Barometer At 'standard conditions', the height of the
Hg column above the surface of the Hg In the dish will be
760 mm or 29.9". The density of mercury is 0.49 Ibs. per
cubic inch and, if the column is 29.9" high, it would then
exert a force per unit area of 0.49 x 29.9 = 14.7 pounds/In2.

The weight of the atmosphere exerts a force of 14.7 pounds per square
inch on the surface of the Hg in the dish. The height of the mercury
column is therefore a direct measure of the pressure and the unit of
pressure is 1/760 of an atmosphere, which is called a Torr. The
international pressure unit is Pascal, equal to one Newton per meter
square.

In 1644, the French mathematician, Blaise Pascal, sent a group of

mountaineers up into the Alps with a barometer and proved that air
pressure decreased with altitude. The average height of the mercury
column at sea level is 760mm, and this is defined as a standard
atmosphere. This also is 1.01 x 105 Pascals or 1.01 x 105 dynes cm2. The
1/760 of this value is called a Torr in honor of Torricelli.

An extension of the mercury barometer was the mercury U-tube

manometer (Figure 2). Varying atmospheric pressures causes the mercury
level to rise and fall in the 'Torricellian Void.' Likewise, if the pressure at
the other end of the tube is artificially reduced by a vacuum pump, the
mercury in the tube falls drastically.

With both the barometer and the manometer, it is the difference in heights
of the mercury levels that indicates the pressure, that is, the force (weight
of Hg) per unit area that the air pressure will support.

As the pressure on the system side is reduced, the height of the columns
on either side of the U-tube approaches the same, and any difference
becomes very difficult to measure (Figure 2).

Many schemes were tried to magnify the very small differences that
occurred at very low pressures. But the only one that really extended the
range of the manometer was invented by H. McLeod in 1872.

This gauge is an application of Boyle's Law and is still in use today as a

standard for calibrating secondary gauges (Figure 3).
P2 V2
(P1 + h) bh = P1V1
P1 bh + bh3 = P1V1
bh3 = P1V1 - P1 bh)
bh3 = P1 (V1 - bh)

V1 bh
V1 = Total volume, capillary plus bulb (cm3)
P1 = Pressure in system
b = Volume of capillary (in cubic cm)
mm length
h = Difference in height of mercury columns
V2 = bh (cm3) volume in capillary
P2 = Pressure in capillary = P1 + h

Applications
The vacuum gauges in use today mainly fall into three categories:
mechanical, manometric, and electronic. Which gauge is used in a
particular application generally depends on the pressure range it is
intended to measure. Figure 4 shows useful pressure ranges of some
typical gauges.
High pressures such as those found in the rough pumping of a vacuum
system are generally measured with a thermocouple gauge. This
instrument measures heat transfer rate from a heated wire. As gas is
removed from the system, less heat is removed. The changes in
temperature are measured by a thermocouple junction and its output is
displayed as changes in pressure. The most useful pressure range for this
gauge is from 5 Torr to 5 microns.

At lower pressures from 1 x 10-2 Torr to 1 x 10-7 Torr found in many

industrial applications, the cold cathode gauge is very useful.

This instrument is basically a gaseous electric discharge cell which

operates on the same principle as a diode-type ion pump. It is a rugged
gauge that does not use a hot filament.

The most commonly used measurement device for high vacuum is the hot
filament ionization gauge. This type of gauge can be designed to measure
pressures as high as 5 x 10-1 Torr, and as low as 5 x 10-12 Torr. Since it is
found in many industrial and scientific applications, it will be treated here
in more detail.
Basic Performance Factors of Ionization Gauges

1. Reproducibility. A gauge must be able to give reproducible

readings for identical physical situations. Drift due to electric or
geometric instabilities can have adverse effects on the
measurement process.
2. Accuracy. Gauges can be calibrated but, in general, their accuracy
is limited to about ± 20%.
3. Sensitivity. In the ion gauge, the positive ions that impact the
collector (and pick up electrons from ground) account for the
positive ion current (I+).

The electrons emitted from the filament (I-) produce the required
ionization. The fundamental relationship is:

I+ = (I-)PK (1)
Where: (P) is the number of molecules present (Pressure)

And: (K) is the gauge constant which depends on the geometry of the
device and the electrical parameters employed (K) is also referred to as
the sensitivity (S), and: S = I+
PI-
Where: Both (1+) and (1-) are measured in amperes and (P) is in Torr.
Hence: S = (I+) Amps = (a number)Torr (I-) Amps Torr. For instance, the
sensitivity of the Varian UHV-24 nude gauge is 25 per Torr.

Operation of Ionization Gauges

When an ionization gauge is used to measure pressure, two physically
observable parameters are of interest; namely, emission current and ion
current. These two currents must be observed simultaneously, and the
pressure can be calculated by the following rule:

P = (I+) 1
(I-) S
I+ = Observed ion current
I- = Observed emission current
S = Gauge sensitivity (constant for any particular gauge)

Although the McLeod gauge uses mercury in a way different from the
manometer, it still expresses pressure in terms of the height of a mercury
column. At the pressures attained by modern vacuum systems, gauges
that depend on the mechanical effects of pressure are ineffective. So,
other means had to be found that could take advantage of other properties
of atoms and molecules, such as heat conductivity or the ability to be
ionized.

There are many ways to express pressure and some of the more common
units are listed in Table 1.

The gauge sensitivity, S, is a function of the design and construction of

the gauge. For the Varian 563 Bayard-Alpert gauge, S has a nominal
value of 10 Torr. For the Varian 507 Triode gauge, S has a nominal value
of 17 Torr.

One could use an instrument that measured both the I+ and I- currents
with a high degree of accuracy. However, the absolute values of I+ and I-
are unimportant in determining P; only their ratio must be measured.
Therefore, entirely equivalent results can be obtained with an instrument
which measures I+ as a fraction of I- (a ratiometric instrument).

It is convenient to maintain a constant emission current at a preselected

value, rather than to observe it for each measurement of pressure. Thus, in
some gauge controls, the emission current is regulated at a nominal value
of 9 ma (Bayard-Alpert) or 6 ma (triode). This value of emission leads to
ion currents equal to 0.1 amp/Torr. During the calibration procedure,
emission current is sensed by the electrometer amplifier and displayed on
the panel meter.

When the ion current is measured, the same amplifier and meter are used.
Hence, the meter deflection observed during ion current measurement is
automatically interpreted as a fraction of emission current, providing a
true ratiometric measurement, even though the absolute value of emission
current may be 20% different from nominal.

In conventional gauge controls, emission current is measured by the panel

meter with suitable precision resistor shunts. The ion current is then
amplified by an electrometer and displayed on the meter. Overall
accuracy of this kind of system depends on the individual accuracies of
the shunt resistors, the meter movement, and the electrometer gain. Since
these items all function independently, the errors can add up.

However, in the Varian ionization gauge controls, the electrometer and

meter are always used together, whether measuring emission or ion
currents. Thus, some of the errors are cancelled in the calibration process.
As a result, the accuracy of the Varian controls is not critically dependent
on the tolerances of a large number of components, and long-term
accuracy and repeatability are assured.

Some additional conversion factors for pressure and flow units are shown
in Table 2.
Pressures

Standard atm = 1.01325 x 106 Dynes cm-2

= 760 mm Hg (at 0 C)
= 29.9213" of Hg (at 32 F)
Bar = 1 x 106 Dynes cm-2
75.0062 cm Hg (at 0 C)
0.986 atm
Torr = 1333 Dynes cm-2
= 1 mm Hg (at 0 C)
= (760)-1 Standard atm
Micron = (length) 1000 Angstroms
(1A = 10-10 meter)
= (pressure) 1.33 Dyne cm-2
= (pressure) 1 x 10-3 Torr
Barye = 1 Dyne cm-2
= 9.869 x 10-7 atm
= 1 x 10-6 Bar

Dyne = Force necessary to give a one-gram mass an acceleration of one

cm/sec/sec
Flows
Cubic Foot = 28.3 liters
= 2.83 x 104 cm3
CFM = 28.3 liters min-1
= 0.47 liter sec-1
Liter sec-1 = 2.12 cfm
= 3.53 x 10-2 cubic feet sec-1
 Vacuum Valves
Historical Notes
Vacuum valves of a crude type date back to the early 1600's, and were
originally adapted from the stopcocks used to tap barrels of various
fluids. Improvements in valve design by Robert Hooke in England and
Otto Von Gureicke in Germany reduced leakage into the vacuum
environment and thus permitted work at lower pressures and for longer
time periods. No improvements on these designs were made until the
1800's, when the requirements for machinery in the industrial revolution
became more sophisticated.

Applications
The field of vacuum can be roughly divided into three areas: Low
vacuum, which is relatively high pressure (10-1 to 10-4 Torr); medium
vacuum (10-4 to 10-7 Torr); and high vacuum (below 1 x 10-7 Torr). A
further subdivision for pressures below 1 x 10-9 Torr is often included
and called ultrahigh vacuum. The vacuum hardware for each area requires
special design features to match the type of application, and this is
naturally true of valves.

Valves used in low vacuum are assembled from castings of brass or

aluminum and have elastomer seals, such as Viton O-Rings. Valves used
in medium vacuum are usually brass or aluminum and will generally
incorporate a hydroformed bronze bellows as a shaft seal. High vacuum
valves are usually made from stainless steel (typically, series 304); use
welded stainless steel or inconel bellows as a shaft seal; and use either
Viton, Polyimide, or a metal gasket for the main seal. Valves used in
ultrahigh vacuum applications will always use metal gaskets (typically,
copper or gold) because of the baking requirement (250 to 400 C).

Common Applications

1. Low vacuum
o Freeze drying
o Food processing
o Metal ore refining
o Steam plant condensers
o Vacuum Distillation
2. Medium vacuum
o Decorative coatings
o Functional coatings
o Chemical processes
 o Electron microscopes
o Microscopy sample processing
3. High/Ultrahigh vacuum
o Physics research
ƒ Optics
ƒ High energy
o Semiconductor manufacturing
o Electron tube manufacturing
o Surface analysis (Auger Spectroscopy)
o Molecular beam epitaxy
o Outer space simulation

The last three high vacuum applications must be able to attain pressures
of 1 x 10-10 Torr or less.

Valve Choice Criteria

Once a chamber design is planned and the effective pumping speed
needed in order to attain the required pressure is determined, the
remaining plumbing configuration should be laid out according to the
pressure range, space limitations, and process requirements. Based on the
chamber pressure required, a valve and other components for the
plumbing configuration should be assumed and tested for suitability in
the next steps of the selection procedure.

Certain criteria have already been established for the choice of valve. The
pressure range and cleanliness level required for the process of interest
will determine the vacuum level, materials of construction, and
bakeability needed for the valve. Since delivered pumping speed (Sd) will
rely on conductance, and the conductance of the valve relies on pipe size,
a larger size (higher conductance) should be chosen for higher chamber
pumping speeds.

Choice of pump and other components will help to determine how the
valve will connect to the system (solder, weld, ConFlat flange, etc.), and
space limitations and conductance will influence the flow pattern that is
chosen. Contact a Varian sales engineer for assistance in selecting
materials of construction, bakeability, conductance, flow patterns, and
connections that are appropriate for your application.

Narrow Down Your Valve Choice

The next step in valve selection is to determine the means of actuation
needed for the valve. The intended type of valve may be available with a
choice of manual, air-operated, or electromagnetic actuation. In making
this choice, parameters to be considered are frequency of use, utilities
required, and whether operator attention is required during the stage of
the process at which the valve is to be actuated.

Remember, Varian stocks many types of valves. Just contact your Varian
sales engineer for immediate information on application, price, and
delivery of any of our valves.

Varian Leak Detectors

Historical Notes
The helium mass spectrometer leak detector was born of necessity during
the Manhattan Project in the early 1940s. Since then, it has become,
through successive technical innovations, an increasingly useful device
for quality control of a wide variety of systems and products. The mass
spectrometer dates back to 1912 when the British physicist, J.J.
Thompson, used a magnetic field to separate various gas ions by their
mass-to-charge ratios. See Figure 1.
Development of nuclear devices under the Manhattan Project required the
utmost integrity of gaseous diffusion processing systems, a tightness far
beyond all existing means of measurement or detection. As in other
phases of the project, the uniqueness of the requirement demanded a new
and different technique for identifying and measuring leakage paths
smaller than previously known to exist.

The technique and tracer gas selected were tailormade for this stringent
requirement. Helium was selected as the tracer primarily because, as the
lightest inert gas, it penetrated small leaks readily. Helium was also
known to be nontoxic, nonhazardous, nondestructive, plentiful, relatively
inexpensive, and present in the atmosphere only in minute quantities (5
ppm). These features enable it to go through small passages without
affecting parts or processes, and be detected easily, reliably, and
economically.

Applications
Some applications include testing for and measuring the vacuum integrity
of:

1. Components
a. Automotive parts
b. Bellows
c. Beverage and food containers

• Pressurized fluids
• Vacuum-packed solids

d. Electronic feedthroughs
e. Electron tubes
f. Glass-to-metal seals
g. Heat exchangers
h. Hermetic seals
i. Medical devices
j. Nuclear components
k. Quartz crystal packages
l. Refrigeration parts
m. Relays
n. Semiconductor packages
o. Transducers
p. Vacuum hardware
q. Vacuum switches
2. Systems
a. Electron microscopes
b. Evaporation coils
c. Freeze dryers
d. Ion implanters
e. Nuclear reactors
f. Plumbing
g. Space simulation
h. Sputtering
i. Vacuum evaporators
j. Vacuum furnaces

Basic Performance Factors

1. Sensitivity
Is the minimum detectable partial pressure of helium in the spectrometer
tube. This factor relates to the minimum detectable leak rate in std cc per
sec for helium molecules (1 std cc/sec = 0.76 Torr-liter/sec). This number
is sometimes specified in terms of air molecules, which results in a lower
number, and at first glance, appears better than the number for helium
molecules.
2. Cycle time is the minimum time constant for the machine between leak
check cycles. This time includes pumpdown, crossover, response time,
and cleanup. It basically relates to the design and efficiency of the
vacuum system.

3. Maximum detectable leak is the gross leak detection capability and is

limited by the maximum pressure that the system can tolerate.

4. Ease of operation and maintenance. This factor is basically the cost

of training, errors, and maintenance. It relates directly to uptime.

5. Fail-safe operation is the machine's immunity to accidents. This can

include utility failures and sudden exposure of the vacuum system to full
atmospheric pressure. It relates directly to uptime and maintenance cost.

Operation
A mass spectrometer is a machine that can produce, sort, and detect
gaseous ion species. As such, it requires a high vacuum to operate
effectively.

A helium leak detector basically consists of a simple mass spectrometer

tuned to helium and an associated high vacuum system.

Helium is extremely well-suited as a probe (test) gas because:

1. It is a small light atom that can penetrate small holes and move rapidly.
This latter quality reduces both response and clean-up time.
2. It is inert, therefore non-reactive and non-toxic.
3. It has an isolated peak in the spectrum that does not require the use of a
high resolution instrument.
4. It is readily available and inexpensive, and it also has a low
atmospheric partial pressure, so it presents a low background signal.

A helium mass spectrometer leak detector (HMSLD) is a complete

system for locating and/or measuring the size of leaks into or out of a
device or a container. In use, this method of leak detection is initiated
when the tracer gas, helium, is introduced to a test part that is connected
to the HMSLD system.

The helium from the test part leak travels through the system, its partial
pressure is measured, and results are displayed on a leak rate
measurement meter. The HMSLD operating principle consists of ionizing
gases in a vacuum and accelerating the various ions through an electric
and magnetic field. The helium ions are separated and collected, and the
resulting ion current is amplified and measured.

The heart of the leak detector is the mass spectrometer tube (spec tube)
together with its control circuitry. See Figure 2.

The spectrometer tube is typically designed and adjusted so that hydrogen

ions are deflected 135, helium ions 90, and all heavier species less than
90 , as shown in Figure 2. Consequently, only helium ions pass through
the field exit slits and arrive at the collector. The collector current is,
therefore, proportional to the partial pressure of helium in the
spectrometer tube and, within the normal operating pressure range of the
HMSLD, is not affected by the pressure of other residual gases.

The collector current is measured by an electrometer amplifier and

displayed on the meter as a leak rate rather than as a partial pressure.

The Varian spectrometer tube, illustrated in Figure 2, shows a cold

cathode ionization gauge as an integral part of the tube. This gauge serves
two functions. First, it monitors total pressure within the tube to initiate
protection of the filament in case of an excessive pressure rise. Second,
because of its location between the filament and the vacuum system, it
minimizes the quantity of gaseous hydrocarbons reaching and damaging
the filament.

There are basically two accepted approaches to helium leak detection the
conventional system shown in Figure 4, and the Contra-Flow method
shown in Figure 5.

With an HMSLD such as this, the part to be tested is simply coupled to

the test port. A switch is manually actuated, whereupon the part is
automatically evacuated to a suitable pressure and then valved off from
the rough vacuum system and onto the HMSLD spectrometer tube
vacuum system. A lamp indicates when this process is complete, and leak
testing can be started.

At the conclusion of leak testing, the switch is returned to its original

position, the test port is automatically isolated from the HMSLD and
vented to air, and the test piece is ready to be uncoupled from the leak
detector. A newer development is the Contra-Flow leak detector, shown
in Figure 5.

The Contra-Flow technique takes advantage of the differences in

compression ratios (outlet pressure divided by inlet pressure) produced by
the diffusion pump for gases of different molecular weights. For example,
the maximum compression ratio for helium may be 10 or 100, while for
oxygen, nitrogen, and other gases contained in air, the ratios are normally
far in excess of 1 million.

This principle is implemented in the leak detector by introducing helium

(and other inlet gases, such as those resulting from a leak in the test piece)
into the high vacuum pump outlet (foreline) rather than into the 'normal'
pump inlet, as in conventional leak detectors. Helium, having a much
lower maximum compression ratio than other gases contained in air,
diffuses backwards through the pump to reach the spectrometer tube,
where it is detected in the normal manner.

Although the mechanical pump is also attached to the high vacuum pump
foreline and removes all inlet gases, including helium, there is no
appreciable loss of sensitivity in the Contra-Flow leak detector. (The
sensitivity of the automatic conventional model is 10-11 std cc/sec, while
the sensitivity of the Contra-Flow is 10-10 std cc/sec.)

One advantage of the Contra-Flow is that no liquid nitrogen is required, a

feature that results in a substantial reduction in operating cost. Another
advantage is the ability to begin leak detection at a higher (rough
vacuum) pressure. This allows the presence of outgassing during leak
detection. Conventional leak detectors must rough pump to a much lower
pressure before they can be used to detect leaks or must use some sort of
throttling device that admits only a small portion of the gas stream, with
subsequent loss of sensitivity.