The Vacuum Use of Molecular Sieves and other Desiccants

EVELYN

M. ROBSON

Edwards High Vacuum Ltd., Manor Royal, Crawley, Sussex

The use of molecular sieve as a desiccant in vacuum systems has been investigated.
A
comparison of the water adsorption efficiency of the sieve, under static equilibrium conditions,
with two other regenerable desiccants (silica gel and activated alumina) shows the sieve to be
far superior.
It is also shown that the non-regenerable desiccant phosphorus pentoxide, has
a greater adsorption capacity than the molecular sieve. A comparison of the adsorption
capacity of two column lengths of sieve under dynamic conditions is made, and the pressure
diflerentials across the columns are investigated.
The increase in the adsorption e$iciency
when the system is backed by an air ballast rotary pump is shown. Data are also given for
moisture breakthrough, heat of adsorption, and regeneration times. As the sieve readily
adsorbs air at atmospheric pressure, the time required for initial degassing is shown.
An
appendix describes a subsidiary water absorption test, made on a sample of phosphorus
pentoxide, to show the progressive effect of surface deliquescence and relative vapour pressures.

Static tests

quantities of water on to the desiccant sample in a closed

system, maintained at an air pressure of 0.02 Torr. At each
stage, when equilibrium conditions were reached, the vapour
pressure was measured by means of an oil manometer.

Experimental apparatus and procedure

The water adsorption efficiency of molecular sieve (type 4A,

& in. pellets) under static equilibrium conditions was determined by measuring its equilibrium vapour pressure at
varying moisture contents.
Similar information on samples
f3\

,,Extro
bulb

Results

Values of water vapour pressure up to 2.9 Torr and moisture

contents up to 17.9 per cent by weight, were obtained, and a
curve was plotted (see Graph 1). Similar curves for three

volume

Pironi
head

Water evaporated
On
to sample from heIe --

Desiccant
sample

test

II

III

u
Cold traps
6

Oil
manometer

% odsorbed

;-u

IO

12

14

moisture content

16

18

20

2.2

24

26

[by dry weigh4

GRAPH 1

FIG. 1. Experimentalapparatusfor staticadsorptiontests.

of silica gel, activated alumina and phosphorus

other desiccants : silica gel, activated alumina and phosphorus

pentoxide, are shown as previously published using this
methodl.
A comparison of these curves shows that under the static
adsorption conditions, the percentage moisture contents of

pentoxide

has been obtained as previously publishedl.

The general
arrangement of the apparatus is shown in Fig. 1.
The test procedure was to evaporate successive known
10

The Vacuum Use of Molecular Sieves and other Desiccants

the desiccants, exerting a water vapour pressure of 0.1 Torr
are as follows :
TABLE I

I
desiccants

Moisture content
at 0.1 Torr
(%)

Silica gel (granules)

Activated alumina (4/8 mesh)
Molecular sieve (a in. Delletsj
Non-regenerable~~~iccnnt
.
Phosphorus pentoxide

/
Fi
/
14:5
I
( >25 at 0.01 Torr

Regenerable

Bulk
density
(lb/f@)

1:
43

I
)

33

Dynamic tests
Experimental

apparatus

The apparatus to find the water adsorption efficiency of

the sieve under dynamic conditions was made completely
Pironi
head

gouge

Tao 2

Thermo-couple

Constcnt

temperature
Oil

FIG. 2.

Experimental

apparatus

for dynamic

manometer

adsorption

tests.

of glass, and was set up as shown in Fig. 2. A vessel containing water to be evaporated on to the sieve was connected
to the sieve column inlet via tap 1. The column was wound
with heating tape to enable the sieve to be regenerated in situ.
The wall of the column was dimpled in two places, each
dimple to house a thermocouple for temperature measurement
during regeneration, and to measure the heat of adsorption
of the sieve during the adsorption tests.
An oil manometer was placed across the column for
absolute pressure drop measurements, and a Pirani gauge
head was situated above the column to indicate the outlet
pressure.
The column was connected via tap 2 to a demountable cold
trap (-79 C) which condensed moisture at breakthrough.
The system was pumped via tap 3 by a two stage air ballasted
mechanical rotary pump with a free air displacement of
20 l/min.

11

Experimental procedure
The column was first filled with molecular

sieve (type 4A,

& in. dia. pellets) to the required height, the apparatus
(see Fig. 2) was then evacuated with tap 1 closed and taps 2
and 3 open. A known quantity of water was placed in the
vessel A, and the top flange was greased and replaced to
form a vacuum seal. When the pressure in the apparatus
reached 0.01 Torr, tap 1 was opened and the water in the
vessel was progressively evaporated on to the sieve. The
rate of evaporation (or water throughput rate) was dependent
on the constant temperature bath, and was changed for each
test. At the end of the test the quantity of water evaporated
was determined by subtraction.
The moisture that broke through the sieve column was
trapped out before reaching the pump. The trap was
isolated by closing taps 2 and 3, and was removed at intervals
without interrupting the test to enable the quantity of breakthrough water to be measured.
To ensure the quick and
complete removal of water from the warmed cold trap the
inner surfaces were coated with a silicone fluid (i.e. Repelcote
-a
dilute solution of chiorisilane dissolved in carbon
tetrachloride).
Throughout the tests record was made of moisture breakthrough quantities and change in sieve temperature.
Also,
readings were taken of pressure difference across the sieve
using the oil manometer, and sieve column outlet pressures
were noted by use of the Pirani gauge. Corrections made to
the Pirani gauge readings due to the pressure of water vapour
in the outlet pressure were found to be negligible in this
instance.
From these pressure readings, the sieve column
inlet pressures were calculated.
Regeneration of the sieve was carried out in situ under
vacuum. Tap 1 was closed and the column temperature was
raised to 300C. The apparatus was simultaneously pumped
and the liberated water vapour was condensed in the cold
trap. The quantity of condensed water was measured at
regular intervals as before, and as the amount of water
adsorbed by the sieve in the previous adsorption run was
known, the progressive state of wetness was determined.
Note : An approximate check on the dryness of the sieve
during regeneration could be made by observing the outlet
pressure on the Pirani, since a curve showing a rough relation
between pressure and moisture content under these dynamic
regeneration conditions had previously been obtained.
Results
Adsorption

efficiency

Dynamic water adsorption

tests were carried out as
described above. The dimensions of the molecular sieve
column were : length-13
in., diameter-l+
in. The sieve
weighed 220 g and was in type 4A, _ik in. diameter pellet
form. Water adsorption efficiency was determined by the
quantity of water adsorbed by the sieve at the time of breakthrough for a particular throughput rate. Breakthrough was
indicated by a pressure rise on the outlet side of the column,
and in all the tests, the time of breakthrough was recorded
when this pressure (as indicated by the Pirani gauge) increased
to 0.1 Torr. The results obtained are shown below in
Table II.

12

EVELYN M. ROBSON
investigated

TABLE II

length
Water throughput
rate
@+/hr)
(Lrlsec)

Nil
2.63
8.5
14.4
16.2
18.6

oA Moisture
adsorbed before
breakthrough

Time breakthrough
occurred.
(Pressure 0.1 Torr at
column outlet)

0.69
2.22
3.78
4.25
4.88

10 hr
2 hr
1 hr

30
27
8
50
29.5

14.5*
12.5
9.5
7.4
6.2
4.2

min
min
min
min
min

linear relationship

water adsorption,

was found

by weight,

and is shown in Graph

between

For this comparison

calculate

an average

pressure,

for

outlet
residual

I,, 6
Water

throughput,

find the effect

adsorption

sieve column

air pressure,

average

I, 12

IO

throughput

14

rote,

16

I8

20

of the column

efficiency,

1 22

the sieve.

: length-63

pressure

The latter was the

and the vapour pressure of the lower

This

as the moisture

vapour
content

pressure

was

increased

It will be seen from the results plotted in Graph

the test
3 that the

IO

246

throughput

12

14

16

18

20

22

24

cm3/hr

rate,

GRAPH 3
pressure drop-throughput
only

on the dynamic

surface of

during

of the sieve increased.

24

cm3/hf

length

to

inlet, and the

whilst the inlet pressure

rate,

Eflect of column length on dynamic adsorption efficiency

To

This

the average

2.

GRAPH 2

water

test.

between

pressures of the sieve.

Water
4

to be made it was first necessary

each adsorption

constant

throughput

pressure drop, up to the breakthrough

was the difference

existing

I
h

the effect of

at various

the resultant of this air pressure, the pressure of the water

the percentage

and the water throughput

to compare

pressure drop

rates of water.

drop

*Equilibrium value with sieve vapour pressure at 0.1 Torr, obtained

from static vapour pressure test (see Graph 1).
A

it was still interesting

on the column

slightly

greater

i.e. the difference

rate across the 13 in. column was

than that across the 6.5 in. column,

in pressure drop for the throughput

rates

l-6 cm3/hr was 9 per cent.

tests were made with a second

in., diameter-l+

in.

The results

TABLE III

Water throughput
rate
(cms/hr)
(L&ec)
0.635
3.12
3.48
5.6
5.86
8.0
8.2
obtained

0.166
0.818
0.912
1.46
1.53
2.1
2.15

Time breakthrough
occurred.
(Pressure 0.1 Torr at
column outlet)
23 hr
2 hr
2 hr
lhr

50min
17min
48 min
23 min
27 min

are shown below as Table III,

on Graph

A Moisture
adsorbed before
breakthrough

13.3
8.05
6.7
6.54
4.26
2.79
3.36
and are also plotted

The effect of column length on the pressure drop across it

Although

only two column

1
0

2 for comparison.
lengths (13 and 6.5 in.) were

Water

IO

throughput

12

14
rote,

GRAPH 4

16
cm3/hr

18

,
20

,
22

24

The Vacuum Use of Molecular Sieves and other Desiccants

Temperature

increase of sieve during water adsorption

During the adsorption tests, the temperature of the sieve

rose until breakthrough
occurred.
It then decreased by
2040C until the end of the tests. The highest temperature
reached by the sieve in each test has been plotted with
corresponding
water throughput
rate (Graph 4). The
temperatrires reached by the sieve in the 6+ in. column were
higher than those in the 13 in. column for corresponding
throughput rates.
Breakthrough

quantities

In each of the adsorption tests on the @ in. column,

20 cm3 of water were evaporated on to the sieve column at the
different throughput rates. The total breakthrough quantity
in each case has been plotted with water throughput rate
(Graph 5) and a linear relationship was found.

Increased adsorption
rotary pump

eficiency

dater

j
throughput

when backed by an air ballast

When the vacuum system is evacuated by an air ballast

pump (i.e. a pump capable of pumping, efficiently, small
quantities of water vapour) small breakthrough quantities
can be tolerated during water adsorption by the sieve. Since
this allows for a longer adsorption period, an increase in the
sieve adsorption capacity results, the degree of which is
related to the capacity of the air ballast pump.
Using a 110 g sieve sample, and assuming that a water
breakthrough
of 0.11 cm3 is permissible, the adsorption
capacity of the sieve at all water throughput rates is increased
by 3.7 per cent by weight (see Graph 7).
20

I
Curve A values taken after breakthrough
of 0,11cm3
Curve B values token at breakthrough
point

18

Total

13

1
I
-8

in
/ each test,=20cm3+-1
1
x
12
I

T
)

-1..

IO

I
0

I
2

I
4

Water
0

Water

throughput

rate,

IO

II

IO

throughput

12

14

rote,

16

1
22

24

cm3/hr

GRAPH 7

12

The limit to the adsorption capacity may then be the

vapour pressure exerted by the sieve at a high moisture
content, e.g. in equilibrium conditions a sieve sample with
an adsorption of 14.5 per cent, by weight, exerts a vapour
pressure of 0.1 Torr (see Graph 1).

sample

odsorption

20

cm3/ hr

GRAPH 5

Water

18

time,

exposed

to o dry

atmosphere

Time,

hr

hr

GRAPH 6

The relationship between the breakthrough quantities and

adsorption times obtained during several tests is also shown
(Graph 6).

GRAPH 8

for

I hr

14

EVELYN

Initial removal of air from molecular sieve

When the sieve is placed in a vacuum system the air it

contains is given off. From a typical pump down curve of a
system containing a fresh 2%g sieve sample, and using a
rotary vacuum pump with a displacement of 20 l/min it
may be seen that a pressure of 0.1 Torr is reached in a few
minutes, and the required 0.02 Torr in 13 hr. This is shown
in Graph 8. Thus when using a fresh sample of molecular
sieve, or a sample which has been exposed to the atmosphere,
this pre-pumping time must be taken into consideration.
It has been shown that small amounts of water drastically
reduce adsorption of oxygenz. If, therefore, exposure of the
sieve to the atmosphere after regeneration is unavoidable
it may be desirable to terminate the regeneration leaving a
small amount of residual moisture in the sieve.
Regeneration
The times required to regenerate the sieve at 300C under

the vacuum conditions previously described were found to

vary with the quantity of sieve, and are tabulated below.
TABLE IV

Approximate regeneration times

(Original sieve moisture content > 10 %
by weight)

Weight of sieve (g)

---

1 hr
2 hr
4 hr

36
110
220
A typical moisture

removal

10 min

curve is shown in Graph

9.

20~

,
-

M. ROBSON
temperature of the sieve was exactly 300C the pressure
(Pirani) at the outlet of the column was below 0.4 Torr.
Thus regeneration was stopped when this pressure was
reached.
r
It should be noted that at the start of the regeneration
under vacuum, water vapour evaporated from the sieve at
a very fast rate, and caused the sieve pellets to become
violently agitated.
Discussion

The static adsorption efficiency of molecular sieve at a

vapour pressure of 0.1 Torr has been shown to be much
greater than silica gel and activated alumina.
However,
it has also been shown (Table I) that phosphorus pentoxide
will absorb almost twice the amount of water than that
adsorbed by molecular sieve, and still maintain a vapour
pressure much lower than 0.1 Torr.
It may appear from this that phosphorus pentoxide is a
superior desiccant for use under vacuum. This is true
providing during absorption the exposed surface does not
deliquesce sufficiently to form a covering layer of phosphoric
acid. If the depth of the phosphorus pentoxide is greater
than approximately 2.5 cm, the exposed surface becomes
covered with phosphoric acid at around 25 per cent water
absorption, by weight, trapping unused phosphorus pentoxide
beneath it (this is shown by a subsidiary experiment as
detailed in the Appendix).
The phosphoric acid surface continues to adsorb water,
becomes diluted and very soon exerts a vapour pressure
much above 0.1 Torr. Thus it is necessary for phosphorus
pentoxide to have a large surface area with little depth if
it is to attain a 25 per cent water absorption, by weight,
whilst exerting a vapour pressure below 0.1 Torr.
It may also be noted that the phosphoric acid vapour
formed during water adsorption by phosphorus pentoxide
contaminates the desiccant chamber in which it is contained.
Conclusions

Regeneration

time,

hr

GRAPH 9

The moisture content of the sieve as obtained in the

previous adsorption run appeared
to have no definite effect
on the subsequent regeneration times.
For most practical purposes it is sufficient to reduce the
moisture content of the sieve to 0.3 per cent, by weight, for
in removing this final moisture content the regeneration
time is increased by one third of the total time.
It was also found that when the moisture content of the
sieve had been reduced to 0.3 per cent, by weight, and the

The molecular sieve was found to be a very efficient

regenerable desiccant for use in a vacuum system. In static
equilibrium conditions it can adsorb water up to 14.5 per
cent, by weight, whilst exerting a vapour pressure of 0.1 Torr.
Comparison with other desiccants per gramme weight, under
static equilibrium conditions, showed that molecular sieve
adsorbs sixteen times more water than silica gel and nearly
five times more than activated alumina, whilst exerting a
vapour pressure of approximately 0.05 Torr.
Under dynamic conditions the total water adsorption at
breakthrough was found to be linearly dependent on the
water throughput rate. The quantity of water adsorbed by
220 g of sieve (column length-13
in., dia.-l$
in.) was
increased from 4 per cent at 19 cm3/hr water throughput, to
14 per cent at 1 cmJ/hr.
An increase in the column length from 6t_ in. to 13 in.
produced an increase in the adsorption capacity per gramme
weight of the sieve. The increase varied with water throughput rate. Thus at very slow throughput rates, the effect
was negligible, but as the rate increased the adsorption

15

The Vacuum Use of Molecular Sieves and other Desiccants

Although it is usual to regenerate desiccants by passing

hot air through the column, it was found more convenient
in these tests to regenerate at 300C under vacuum, and so
keep the sieve consistently free of adsorbed air.
Regeneration time was dependent upon the weight of the
sieve and not upon the moisture content, and was found to
follow the usual expotential drying law. Thus the 13 in.
column containing 220 g of sieve required approximately
4 hr to regenerate whilst the 6+ in. column containing 110 g
required about 2 hr.
The violent agitation of the sieve which occurred at the
start of each regeneration run, should be taken into account
when a drying column is designed.

capacity of the 13 in. column to the 63 in. column increased,

until at a rate of 8 cms/hr it was five times greater. However
the pressure drop was found to be slightly greater across the
longer column (about 9 per cent) and therefore a maximum
length should be determined, in order that the back vapour
pressure exerted by the inlet surface of the sieve can be
tolerated.
The quantity of water adsorbed by the sieve can be increased
still further when the rotary pump used in the system is air
ballasted to deal with small quantities of water vapour.
The adsorption time would be lengthened until the rate of
water breakthrough equalled the vapour pumping rate of the
pump at normal pump temperature.
The heat of adsorption of the sieve was found to be quite
considerable during the tests. At a water throughput rate
of 8 cmJ/hr the temperature of the sieve rose from 20C to
73C in the 13 in. column, and from 20C to 100C in the
6+ in. column.
When molecular sieve has been exposed to the atmosphere
it is necessary to pump for approximately l+ hr before a
suitable vacuum is reached. This could be a serious limitation in a system where the moisture trap is open to the
L
atmosphere for a period of time after regeneration.

Acknowledgements

My thanks are due to Mr. T. W. G. Rowe, head of the

Drying & Distillation Research Div., for encouragement
in this work, and to the directors of Edwards High Vacuum
Ltd., for permission to publish this paper.
References

1 E. M. Robson ; Vacuum 4, 65 (1954).

2 D. W. Breck et al. ; J. Amer. Chem. Sm. 78, 7 (1956).

APPENDIX
IN

order to determine the efficiency of phosphorus pentoxide

in an air pressure of 0.02 Torr with relation to the state
of the exposed surface, the following subsidiary test was
made. A sample of dry phosphorus pentoxide (see Fig. A l(a))
was allowed to absorb controlled amounts of water in stages.
The vapour pressure and structure of the sample was noted
at each stage.
These are shown diagrammatically
in Figs. A l&-f).
(a)

It will be seen from Fig. A l(f) that the vapour pressure

decreased when a fresh surface of phosphorus pentoxide was
exposed.
Changes in sample structure during test

Diagrams showing percentage moisture absorbed, by

weight, after the number of days the test was continued.
(Time includes periods allowed for equilibrium conditions
to be reached at 24C.)
(b)

(c)

Dry

phosphorus

6% in 2days

20%

Vopour

0.001

in 5 days

Tom

, (e)

(d)
/

pressure<

\
Phosphoric

pad

Phosphoric

30%
Vapour

pressure

in 7doys

0.109 Tom.

acid

34%
Vapour

in 6 days

pressure

0.256

FIG. A 1

34%
TOW.

Vopour

in 9 days

pressure

0.0044

Tow