RELAXATION STUDIES OF ADSORPTION ON
KENNETH DARE LAWSON
A DISSERTATION PRESENTED TO THE GRADUATE COUNCIL OF
THE UNIVERSITY OF FLORIDA
IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE
DEGREE OF DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
To Dr. Wallace S. Brey, Jr., director of this project, the
author would like to express his sincere appreciation.
The author would also like to thank Dr. M. E. Fuller II,
unofficial co-director of the project, for both physical and
Thanks are also extended to the Atomic Energy Commission
for financial aid received during the course of the investiga-
TABLE OF CONTENTS
ACKNOWLEDGMENTS . . . .
LIST OF FIGURES . . . .
LIST OF TABLES . . . .
9 9 9...........9 9 9 9
. 9 9 9...........9 9
9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9
I INTRODUCTION . . . . . . . . . . ..
II LITERATURE REVIEW . . . . . . . . . .
A. Adsorption Studies by Magnetic Resonance
Techniques . . . . . . . . .
B. Adsorption Studies by Dielectric Measurements . .
III EXPERIMENTAL TECHNIQUES . . . . . . ... .
A. Samples . . . . . . . . . . .
B. Vacuum System and Techniques Used in the Nuclear
Resonance Experiments . . . . . .
C. Description and Procedures of the Magnetic
Resonance Studies . . . . . . . .
D. Determination of T and T2 . . . . . .
E. Measurement of Hf . . . . . . . . .
F. Systems and Techniques Used in the Dielectric
Experiments . . . . . . . . . .
IV PRESENTATION AND DISCUSSION OF RESULTS . . . . .
A. Magnetic Resonance Results . . . . . . .
B. Dielectric Results . . . . . . . . .
C. Comparison of Dielectric and Magnetic Resonance
Results . . . . . . . . . . .
V SUMMARY . . . . . . . . . . . . 105
BIBLIOGRAPHY . . . . . . . .. . . . 109
APPENDIX A: Tables of Numerical Data . . . . . . 113
APPENDIX B: Theory of Nuclear Magnetic Resonance . . . 159
APPENDIX C: Theory of Dielectric Measurements . . . 163
LIST OF FIGURES
1. A Typical B. E. T. Isotherm Used for Determination
of the Surface Area . e . . . . . . . 26
2. Vacuum System Used in the N.M.R. Experiments . . 27
3. Determination of the Magnitude of H1 for Various
Attenuations . . . . . . . . . . 34
4. Vacuum System Used in the Dielectric Experiments . 35
5. Electrodes and Switching Device for the Dielectric
Cell . . . . . . . . . . . . 38
6. Assembled Dielectric Cell . . . . . . .. 40
7. Line Width of Water on Thorium Oxide Prepared from
Thorium Hydroxide . . . . . . . . . 48
8. The Observed Relaxation Times of Water on Thorium
Oxide Prepared from Thorium Hydroxide . . . . 51
9. Line Width of Water on Thorium Oxide Prepared from
Thorium Oxalate . . . . . . . . . 53
10. The Observed Relaxation Times of Water on Thorium
Oxide Prepared from Thorium Oxalate . . . . . 54
11. Line Widths of Various Adsorbates on Thorium Oxide . 56
12. The Observed Relaxation Times of Methanol on
Thorium Oxide . . . . . . . . . . 58
13. The Temperature Dependence of the Line Width of Water
on Thorium Oxide . . . . . . . . . 61
14. The Temperature Dependence of the Observed Relaxation
Times of Water on Thorium Oxide . . . . . . 62
15. Log of the Line Width versus the Reciprocal Temperature
for Water on Thorium Oxide Prepared from Thorium
Hydroxide . . . . . . . . . . . 65
16. Log of the Line Width versus the Reciprocal Tempera-
ture for Water on Thorium Oxide Prepared from Thorium
Oxalate . . . . . . . . . . . .
17. The Temperature Dependence of the Line Width of
Methanol on Thorium Oxide . . . . . . . .
18. The Temperature Dependence of the Line Width of
Ethanol Adsorbed on Thorium Oxide . . . . . .
19. The Temperature Dependence of the Line Width of
Butylamine Adsorbed on Thorium Oxide . . . . .
20. Presentations of the Absorption Line of Butylamine
Adsorbed on Thorium Oxide . . . . . . . .
21. Fluorine Spectrum of Hexafluoropropene Adsorbed on
Aluminum Oxide . . . . . . . . . .
22. The Temperature Dependence of the Line Widths of
56.0 mg. C3F6/g. A1203 . . . . . . . .
23. The Temperature Dependence of the Line Widths of
93.6 mg. C3F6/g. A120 . . . . . . . .
24. Capacitance Changes with Frequency for the Cell, Dry
Sample, and Sample with Adsorbed Water . . . .
25. Conductance of Water on Thorium Oxide Prepared from
Thorium Oxalate . . . . . .
26. Conductance of Water on Thorium Oxide
Thorium Hydroxide . . . . .
27. Capacitance of Water on Thorium Oxide
Thorium Hydroxide . . . . .
28. Capacitance of Water on Thorium Oxide
Thorium Oxalate . . . . .
29. Conductance of Water on Thorium Oxide
of 2.0 Megacycles . . . . .
30. Tangent of the Loss Angle of Water on
Prepared from Thorium Oxalate . .
31. Tangent of the Loss Angle of Water on
Prepared from Thorium Hydroxide . .
at a Frequency
Thorium Oxide. . . .
. . . . .
32. Capacitance of Methanol on Thorium Oxide Prepared
from Thorium Hydroxide . . . . . . . . 99
33. Capacitance of Methanol on Thorium Oxide Prepared
from Thorium Oxalate ............... 100
34. Nuclear Relaxation Times as a Function of the
Correlation Time . . . . . . . . . 162
LIST OF TABLES
Summary of Sample Preparations and Properties . . .
2. Dependence of the Line
Times on the Amount of
3. Dependence of the Line
Times on the Amount of
4. Dependence of the Line
Times on the Amount of
5. Dependence of the Line
Times on the Amount of
6. Dependence of the Line
Times on the Amount of
7. Dependence of the Line
Times on the Amount of
Width and Observed Relaxation
for Sample VI-D .
for Sample VI-E .
for Sample IX-A .
for Sample IX-B .
for Sample OX-3 .
Adsorbate for Sample OX-4 . ..
8. Temperature Dependence of the Line Width and the
Observed Spin-spin Relaxation Time of Water Adsorbed
on Sample VI-D . . . . . . . . . .
9. Temperature Dependence of the Line Width and the
Observed Spin-spin Relaxation Time of Water Adsorbed
on Sample VI-E . . . . . . .. . .
10. Temperature Dependence of the Line Width and Observed
Relaxation Times of Water Adsorbed on Sample IX-A .
11. Temperature Dependence of the Line Width and Observed
Relaxation Times of Water Adsorbed on Sample IX-B .
12. Temperature Dependence of the Line Width and Observed
Relaxation Times of Water Adsorbed on Sample OX-3 .
13. Temperature Dependence of the Line Width and Observed
Relaxation Times of Methanol Adsorbed on Sample OX-3
14. Temperature Dependence of the Line Width and Observed
Relaxation Times of Methanol Adsorbed on Sample VI-D
15. Temperature Dependence of the Line Width and Observed
Relaxation Times of Ethanol Adsorbed on Sample VI-E . 131
16. Temperature Dependence of the Line Width and Observed
Relaxation Times of Butylamine Adsorbed on Sample VI-D 132
17. Apparent Activation Energies of Processes Contributing
to the Spin-spin Relaxation of Water on Thorium Oxide . 134
18. Line Widths of Thorium Hydroxide . . . . . . 135
19. Line Widths of Thorium Oxide in the Presence of Excess
Water . . . . . . . .. . . . . . 135
20. Line Width as a Function of the Coverage of CFaFb = CFxCF3
Adsorbed on Aluminum Oxide . . . . . . . 136
21. Observed Relaxation Times of CFaFb = CFxCF3 Adsorbed on
Aluminum Oxide . . . . . . . . . . 137
22. Chemical Shifts of CFaFb = CFxCF3 Adsorbed on Aluminum
Oxide . . . . . . . .. . . . 138
23. Temperature Dependence of the Chemical Shifts of
CFaFb = CFxCF3 Adsorbed on Aluminum Oxide . . . . 139
24. Temperature Dependence of the Line Widths of
CFaFb = CFxCF3 Adsorbed on Aluminum Oxide . . . . 140
25. Changes with Frequency in the Capacitance of the
Dielectric Cell and in the Capacitance of an RC Network 141
26. Dielectric Properties of Water Adsorbed on Sample X-B . 142
27. Dielectric Properties of Water Adsorbed on Sample OX-4 147
28. Dielectric Properties of Methanol Adsorbed on Sample
X-B . . . . . . . . . . . 152
29. Dielectric Properties of Methanol Adsorbed on Sample
OX-4 . . . . . . . . . . . . 154
30. Tangent of the Loss Angle of Water Adsorbed on Thorium
Oxide . . . . . . . . . . . . 156
31. Frequency Dependence of the Tangent of the Loss Angle
of Water and Methanol Adsorbed on Thorium Oxide . . 158
"And ye shall know the truth and the truth shall
make you free."
The investigation of the mechanism of solid-gas interactions in
systems such as sorbed gases and/or liquids on solid substrates has
long been of considerable interest in several areas of scientific
research. One of the major areas of interest has been that of hetero-
The ability of thorium oxide to catalyze a number of vapor phase
chemical reactions has long been known.1-5 However, up to the present
time the mechanisms of the various reactions, in terms of the behavior
of the oxide surface, are not completely understood. Several features
of the surface which must in some manner be related to the activity of
the catalyst toward a given reaction are: the surface area, the pore
structure, and the presence or absence of "active" sites on the surface.
Each of these features depends on the past history of the catalyst, and
thus, to obtain a catalyst having the desired properties, the history
must be controlled. In order to do this the processes which produce
the various properties must be understood.
This investigation is concerned with the interactions between the
surface of thorium oxide and adsorbed polar molecules. To determine
the nature of the interactions and the "degree of interaction" between
the surface and the adsorbates, several samples of thorium oxide,
differing in various properties, have been investigated by nuclear
magnetic resonance techniques and dielectric measurements.
In particular nuclear resonance techniques applied to nuclei in
sorbed liquids and/or gases give several kinds of information about
both the sorbed molecules and the adsorbing surface. As an example
the surface of the adsorbent may consist of sites on which the adsorbate
is adsorbed with different strengths. If this is true, different phases
may be distinguished by the presence of different sets of nuclear relax-
ation times. The accessibility of a catalyst and the influence of the
method of preparation upon its activity can also be investigated by
determination of the relaxation times of the sorbed molecules.
In the same manner dielectric measurements on sorbed systems can
also give pertinent information about the interactions between the
adsorbate and adsorbent through the determination of such quantities
as the capacitance and conductance changes.
In addition to the interest in thorium oxide arising from its use
as a catalyst there has been considerable interest in the relation of
its surface properties to its ability to cake. This interest rises
from its use as a breeder fuel in nuclear reactors.6
Yet another reason for investigating these particular systems is
simply the fact that there is no data of the type presented here
existing in the present scientific literature. Therefore, this study
improves and extends methods of studying solid-gas interactions in
relation to catalytic activity. This is especially true of nuclear
magnetic resonance techniques.7
In summary, the purpose of the present study is to investigate
solid-gas interactions by magnetic resonance and dielectric measure-
ments and to relate the results to the catalytic properties of the
solid. In addition, it is believed that the results will provide an
insight into the mechanism of the adsorption process itself.
"There is no new thing under the sun."
A. Adsorption Studies by Magnetic Resonance Techniques
Although nuclear magnetic resonance (n.m.r.) has become a powerful
tool for molecular structure determination its application to the study
of gas-solid interactions has been somewhat limited.
In the infancy of the technique n.m.r. was used to determine the
water content of various materials8 and to observe in a qualitative way
the amount of water adsorbed on starch and egg albumin.9 Two lines
were observed in the study of water adsorbed on the albumins. These
consisted of a very narrow line superimposed on a broader line. The
broad line is attributed to the water in the solid and water "sorbed"
on the surface of solid produces the narrow line. The width of the
narrow line decreased as the amount of adsorbed was increased while the
broad component remained essentially constant.
Similar behavior was later observed by Tanaka and Yamagata10 for
water adsorbed on carbon. In this system the wide line is attributed
to the water in the first layer and the narrow line is attributed to
the water in the upper layers. A study of water sorbed on fibrous
materials11 has shown that there are two states of mobile water with a
boundary at about twenty per cent moisture content. These two states
show a different dependence of the line width upon the amount of water
in the states. Dielectric measurements of the same systemsl2,13 indi-
cate that in addition to these two states another phase is also present.
This latter phase is attributed to localized adsorbed water. It is
believed by the authors that the sensitivity of the n.m.r. apparatus
used in the above study was such that a line from this localized water
could not be obtained.
The water-titanium dioxide system has been studied by two groups
of workers. Mays and Brady14 studied the temperature dependence of the
line width and found that motion persists down to 77 K. This motion
is interpreted as molecules either "flopping about" at an active site
or jumping between these sites. It was concluded that for very low
coverages the water molecules are held to the surface by forces suffi-
ciently strong that formation of ice clusters is prevented even at 770 K.
However, when the coverage is increased to the point where multilayer
formation begins an ice-like structure results.
Fuschillo and Aston15 have also studied water and methane adsorbed
on rutile. For water adsorbed to the amount of about 0.1 monolayers the
line width is such as to indicate that the molecules are sufficiently
far apart so that little or no interaction is present among them. As
the coverage increases the width of the line decreases and at a coverage
of about two monolayers the width is about one-half of that found for the
0.1 monolayer coverage.
The line width for methane adsorbed on rutile was found to be much
less than that for water adsorbed on rutile. It was concluded that the
reorientation processes occur much more readily in this system than in
the water-rutile system.
Carbon tetrafluoride has been adsorbed on rutile and studied by
Stottlemeyer and coworkers.16 Rotation is found to persist in the
outer layer at all temperatures down to 200 K. Self-diffusion persists
down to 55 K. and, at a coverage of one-half a monolayer, down to 300 K.
One component of the absorption line, for a coverage of 3.24 layers,
is believed to be due to rotation of the molecules in the outer layer.
For this coverage it is believed that the inner layers are rigid.
O'Reilly and coworkers17 studied the residual protons in silica
gel and in a silica-alumina catalyst and found that the line widths are
essentially independent of temperature from 280 C. to -210* C. The
spin-spin relaxation time (see Appendix B for a discussion of relaxation
times) is of the order of 10-4 seconds for both samples. It was con-
cluded on the basis of chemical shift measurements that the protons were
present as hydroxyl groups rather than water molecules.
Graham and Phillips18 observed chemical shifts for several organic
compounds adsorbed on various types of carbon. This study indicates
that the adsorption frequencies of the organic material in the more
mobile part of an adsorbed film are much lower than those of the free
compounds. For a material having more than one resonance line, all the
lines are shifted by essentially the same amount. The product of the
observed shift and the thickness of the adsorbed layer, excluding the
first layer, was used to calculate a number which represents the thick-
ness of the material through which the perturbations of the surface
can be felt. This number is essentially constant for all combinations
of carbons and organic and corresponds to about seven molecular layers.
The shifts of the organic materials are of such a magnitude that
they must be attributed to the effect of paramagnetism in the surface
of the solid. However, previous to the adsorption of the organic
materials, magnetic susceptibility tests indicated that the bulk solids
were diamagnetic. Paramagnetic resonance techniques also indicated no
unpaired electrons in the carbons.
Pulsed or "spin-echo" nuclear resonance techniques have been used
by several groups of workers to study adsorbed materials. Zimmerman
and coworkersl9,20*21 have made an extensive investigation of the water-
silica gel system by this technique. For a coverage of greater than
two layers two distinct spin-spin relaxation times were observed in
the preliminary study.19 One of these remains nearly constant with
increasing coverage while the other increases sharply. The constant
component is associated with the water adsorbed in the monolayer while
the second component, which increases with coverage, is associated with
the water adsorbed on the monolayer. In the region between one-half
and two monolayers only one relaxation time is observed. This is taken
as an indication that protons in this condition have the same average
interactions with the local magnetic fields. For a coverage of less than
one-half a monolayer two values are again observed. The shorter of
these two is associated with the protons bound to impurity sites in
the silica gel.
A qualitative correlation was made between these results and the
dielectric results obtained by Kurosaki22 on the same system. The
changes in the slope of the plots of the apparent dielectric constant
against the amount of coverage occur near the points where the changes
occur in the spin-spin relaxation time.
*This work will be discussed in more detail in a later section.
In a later investigation20 measurements of the spin-lattice
relaxation time (T1) produced only one value below a coverage of one
monolayer. This value increases with increasing coverage. From the
data collected in these two studies the average life-time of a water
molecule in an adsorbed phase on silica gel is calculated to be approxi-
mately three milliseconds. From this calculation, it is concluded that
the value of T1 found is the value which is the average of the two
times for the protons in the different phases.
In a more refined study21 Zimmerman and Lasater again found two
values of the spin-spin relaxation time (T2) at low coverages. These
components remain constant up to a coverage of about two layers. At
this point the longer time commences to increase. This increase is
attributed to the effect of capillary condensation which occurs at this
coverage. The shorter time is associated with water molecules adsorbed
on high energy sites on the solid. An abrupt change in T2 which occurred
at about one-half a monolayer in the previous study20 was not observed
in this investigation. This is not considered an inconsistency, however,
because the small values of T2 found here would obscure this effect.
In the region from one-half to three-fourths of a monolayer the
spin-lattice relaxation time (T1) also has two components. All other
coverages produce only one value. The appearance of two phases in this
particular region is attributed to strong molecular interactions which
produce ordering of the molecules in the adsorbed phase. The spin-
lattice relaxation time measured over the total range of coverage up
to two layers exhibits a minimum between one-half and one monolayer.
This minimum is in the region where the two values are observed. It
was concluded that a single phase T1 will be observed if the average
lifetime of a water molecule in either of the two phases is much less
than the T1 of that phase. It was also concluded that the two phases
observed in the spin-spin relaxation data are identical to the two
phases which produce the two values of the spin-lattice relaxation
Hickmott and Selwood23 have used spin-echo techniques to study
the proton reorientation process in several associated organic liquids
adsorbed on catalytic solids. For all liquids studied, methanol,
ethanol, water, and n-hexane, there is a linear increase in the spin-
lattice relaxation time as the amount of adsorbed material is increased.
Methanol adsorbed on gamma-alumina has a relaxation time slightly
greater than water but slightly less than ethanol. Hlexane has a relaxa-
tion time greater than any of the polar liquids. The hexane relaxation
time decreases when the alumina is completely dehydrated indicating a
greater interaction between the hexane and the dehydrated alumina surface.
Only one relaxation time was found in any particular measurement,
indicating only small variations in the environment of the individual
A qualitative explanation is offered to explain the behavior of
these systems. It is suggested that the first liquid adsorbed is
bound tightly to the solid surface and the protons exchange with lattice
vacancies and with protons incorporated in the alumina. As more liquid
is added a larger fraction of the total amount of liquid approaches
the liquid state and the relaxation times approach the values for the
bulk liquids. The smaller change noted for hexane indicates again that
hexane interacts only weakly with the surface of the solid. The fact
that dehydrated alumina produces a shorter relaxation time indicates
that hexane, being immiscible with water, can approach the solid sub-
strate closer after dehydration.
A study of the same liquids adsorbed on supported metal oxides
shows a similar behavior with the exception of water adsorbed on
supported copper oxide. This system shows a nearly constant change
in the relaxation time with the amount of water adsorbed regardless
of the initial concentration of paramagnetic ions. There is an indica-
tion that free cupric ions are present on the surface of the oxide which
are taken into the water and thus these ions rather than the solid
itself control the relaxation time. It was also found that those cata-
lysts and catalyst supports having the largest specific surface areas
were most effective in reducing the relaxation times.
The temperature dependence of the relaxation times of hydrogen
adsorbed on palladium has been studied by Burger.24 The spin-lattice
relaxation time (Tl) plotted as a function of temperature, shows a
minimum for all coverages of hydrogen while T2 exhibits a maximum at
low amounts of hydrogen. This maximum in T2 disappears as the amount
of hydrogen is increased. Similar studies on Pd-Fe alloys show that
T, has both a maximum and a minimum while T2 behaves in the same manner
as in the case of palladium. The minimum in the T1 curves is interpreted
as indicating that the protons are in rapid motion while the maximum in
the spin-spin relaxation time indicates that there is not a homogeneous
distribution of protons in the palladium.
The adsorption of water on alumina has been studied by Winkler.25,26
lHe found that the value of the spin-lattice relaxation time increases
linearly with increasing water coverage. Both the spin-spin and the
spin-lattice relaxation times show sharp breaks in plots of the relaxa-
tion times against the amount of adsorbed water. At higher coverages
the increase of relaxation times is greater for an incremental increase
in the amount of adsorbed water. The break points are associated with
the coverage necessary to completely fill the micropores and start
filling the macropores. The value of the correlation time found for
the first layer of adsorbed water is nearer to that of ice than it is
to that of liquid water.
In a study of water adsorbed on alumina by continuous excitation
n.m.r. Neikam27 observed a narrow line superimposed on a wide line at
very low water coverages. The narrow component does not change its
width or amplitude appreciably with addition of water. No reasonable
explanation of this component could be given. The broader component
of the line decreases in width with increasing water coverage. A plot
of line width against amount of water adsorbed shows two linear sections
separated by a region in which the line width changes only slightly
with addition of water. The second linear region, occurring at higher
coverages, is associated with rather loosly bound water molecules
adsorbed on the monolayer. This region is centered around the point
where the second layer should be completed. The first linear region
is attributed to those molecules which are adsorbed just after comple-
tion of the monolayer and are still interacting strongly with the
substrate. Whenever the water coverage was less than that necessary
for one monolayer the resonance signal could not be detected. The
transition region was postulated to be due to water molecules which
are adsorbed with their protons toward the monolayer and which thus
would not have as much motion as if adsorbed in the opposite direction.
The temperature dependence of the line widths shows two regions
with a sharp change in the slope at about -30 C. for low coverages
and at somewhat lower temperature for higher water coverages. This
abrupt change in line width is attributed to the "freezing out" of
motions which contribute to line narrowing.
The spin-spin relaxation process was treated as a first order
rate process and an apparent activation energy was obtained from plots
of the log of the line width against the reciprocal temperature.
Values around 1.7 kcal./mole were obtained for the region below the
transition temperature and values of less than one kcal./mole were
obtained for the region above the transition temperature. The values
for the upper region decrease with increasing water coverages. The
energies found indicate that the motions which were "frozen out" are
Dielectric studies were made on the same systems and the results
compared with the n.m.r. results. Plots of the tangent of the loss
angle (see Appendix C) against the amount of water coverage exhibit
four linear regions. These regions appear to have little relation to
the regions in the line width curves, however. In all cases the loss
tangent appears to reach a maximum at about the same coverage where
the second linear region appeared in the line width curves.
The adsorption of methylamine and CH3ND2 on gamma-alumina has
been studied by Hirota and coworkers.28 It is concluded from chemical
shift measurements and from infrared spectroscopic data that the amine
is adsorbed in such a manner that the molecules are bonded by the amino
group to the solid surface. The deuterated material gave similar results
except that the line width is less for the same amount of material
The same group of workers studied the adsorption of formic acid
on alumina and on silica gel.29 Only one signal is observed for low
coverages of the acid on silica and it is assigned to the hydrogen on
the carbon atom. This assignment was made by substituting deuterium
for the hydrogen on the carbon atom. At a coverage of about 2.7 layers
the signal from the acidic hydrogen is also observed. For formic acid
adsorbed on alumina the signal could not be detected at any coverage.
Infrared measurements indicate that the interactions between the acid
and the alumina substrate are very strong.
Several workers have studied water adsorbed on proteins by magnetic
resonance techniques. Brey and Fuller30 have studied water sorbed on
bovine serum albumin. This system shows a two component absorption
line. One component is very broad and apparently is independent of
temperature and unaffected by the presence or absence of sorbed water.
This component is evidently characteristic of the protein itself. As
water is added to the protein the line component of the adsorbed water
can be observed. The width is much less than that of the component
belonging to the protein protons. The water component decreases in
width with increasing temperature from about -95 C. to about 300 C.
Above this temperature range the width remains virtually constant for
high water coverages. For low water contents the width does not
reach a constant value.
Calculation of apparent activation energies for the reorientation
processes from the slopes of plots of the log of the line width against
the reciprocal temperature gives values ranging from one to six kcal./-
mole. For very low water coverages the energies are somewhat uncertain
but appear to be greater than six kcal./mole. As the coverage is
increased the energies decrease to about one kcal./mole but then at
even higher coverages larger values are again found. As the coverage
is increased even more the energy values decrease again. The maximum
in the energies, at the higher coverages, corresponds to an abrupt
change in the plot of the line width against the amount of adsorbed
water. These observations are explained by assuming that the first
water added is hydrogen bonded to the free polar groups in the protein
in such a way that one water molecule is attached to two groups. This
produces a rather large activation energy. As more water is added it
becomes necessary to have a one-to-one ratio of water molecules and
polar groups which results in a lower energy barrier to reorientation.
As the water coverage is increased still more it is assumed that some
type of hydration shell is formed causing another increase in the
activation energy. The constancy of the line width at high temperature
is unexplained but may result from magnetic field inhomogeneities
produced by the solid material.
Berendsen31 has studied the hydration of collagen in bovine
Achilles tendon. Measurements on partially dried samples show a water
absorption line separate from the collagen absorption line. The water
line was studied as a function of the angle between the magnetic field
direction and the fiber direction.
Three components of the water line are observed at some values of
the angle for all the samples studied except one which was exposed to
ten per cent relative humidity until equilibrium was reached. This
particular sample produces only a single component which is independent
of the angle. The separation and the angular dependence of the other
peaks allow certain hydration structures to be ruled out, at least in
this system. It is concluded that the hydration structures which are
not present are: (a) molecules bound to polar groups and rotating about
the bond axis, (b) water molecules bound to specific sites, by either
hydrogen or polar bonds, in such a way as to remain motionless and
having the direction of the proton interaction along the fiber direction,
and (c) free water molecules that translate in channels between the
The results are explained by assuming that the water molecules
form chains in the direction of the fibers. These chains reorient
about their lengths and have average lifetimes which are short com-
pared to the lifetime of a given spin state.
It is concluded that if the water has a three dimensional struc-
ture at high coverages it must have a large number of lattice defects
and is in a state somewhere between a solid and a liquid. The small
width of the observed lines makes this necessary.
B. Adsorption Studies by Dielectric Measurements
Extensive work has been done on the relation of the dielectric
properties of adsorbed systems to the nature of the adsorbed materials
and to the nature of the bonding between adsorbate and adsorbent.
Because of the large number of papers treating this subject only a
selected few will be discussed.
McIntosh and coworkers32 studied the dielectric properties of
ethylene oxide, butane, and ethyl chloride adsorbed on porous silica
gel. Curves obtained by plotting the change in capacitance against the
amount of adsorbed material consist of two or three linear segments.
Both butane and ethylene oxide produce curves with two linear
regions while three linear segments are found for ethyl chloride.
Comparison of the break points in these curves with the amount of
material necessary to produce one monolayer gave no correlation between
the two. However, several facts become apparent from the observed
results. These are: (a) all adsorbates appear to have two or more
portions having different dielectric constants. (b) The dielectric
constants of the adsorbed butane and ethyl chloride show a very small
temperature coefficient. (c) Adsorbed ethyl chloride has a dielectric
constant significantly lower than that of the bulk liquid. (d) Adsorbed
ethyl chloride behaves as a non-polar compound due either to restricted
mobility or to molecular association. (e) Butane appears to be adsorbed
in a liquid-like state.
A similar study33 with butane and ethyl chloride adsorbed on titanium
dioxide gave similar results. The temperature dependence of the capaci-
tance for adsorbed ethyl chloride is found to be essentially zero below
a monolayer coverage. This is interpreted to indicate the presence of
very rigid bonding in the monolayer. In a second investigation34 of
butane, ethyl chloride, and water adsorbed on silica gel no Debye type
dispersions were found down to temperature around -30* C. Mixtures of
water and ethyl chloride and water and butane were also studied and the
results indicate that both adsorbates share two types of sites on the
substrate. It is concluded from this study that water molecules undergo
oscillatory motion and ethyl chloride molecules, at least the first
quantities adsorbed, rotate freely.
In a more recent study35 McIntosh's group has investigated the
dielectric properties of methyl and ethyl chlorides and butane adsorbed
on porous vycor glass. Loss maxima are observed for low coverages of
methyl and ethyl chloride and are attributed to complexes being formed
between the hydroxyl groups in the glass and chlorides. Butane shows
no evidence of complex formation.
Two linear sections are observed in plots of capacitance change
against the amount of adsorbate with the second section having a lower
slope. It is concluded that the chlorides not included in the complexes
behave as an oscillatory system.
Heukelom and van Reijen36 have studied the adsorption of several
organic liquids and water on silica gel. For all materials, methanol,
water, methylethyl ketone, chlorobenzene, benzene, and diethyl ether,
high values of the real part of the dielectric constant are observed.
The values do not differ substantially among the several adsorbates or
for different amounts of the adsorbates. It is concluded that the
adsorbent is primarily responsible for the values of the dielectric
constant. Plots of the relaxation times as a function of the amount of
adsorbates indicate no special behavior at a coverage of one monolayer.
The value of the relaxation time decreases with increasing coverage for
Kurosaki22 has studied the dielectric behavior of water sorbed on
silica gel in the frequency range from two kilocycles to one megacycle.
Three different states of the sorbed water are found. In the first
stage of adsorption the very low specific polarization of the sorbed
water indicates that this water is firmly bound to the surface. In
the second stage, which ranges from about fifteen to ninety-five milli-
grams of water per gram of adsorbent, the water behaves as water
dissolved in an organic solvent. At high water coverages the water is
considered to be capillary-condensed and the value of the maximum loss
frequency indicates that hydrogen bonding is greater in this state than
in liquid water; in fact, it approaches that found in ice. The frequency
of the maximum loss increases as the amount of coverage increases.
Several groups of workers have studied the adsorption of various
liquids on alumina. Thorp37,38 has studied water, methanol, ethanol,
and benzene on alumina and on silica gel.
For water, ethanol, and methanol adsorbed on alumina, silica gel,
and mixed silica-ferric oxide gel37 a hysteresis phenomenon is observed
in both the adsorption isotherms and in the incremental capacitance
measurements. This phenomenon is attributed to the different electrical
properties of the molecules in the multilayers as compared to those in
the entirely capillary-condensed state which is present during desorp-
tion. In a study of water and benzene adsorbed on alumina,38 Thorp
found no hysteresis in the capacitance data for the water-alumina system.
This absence of hysteresis is attributed to strong adsorbate-adsorbate
interactions. The results obtained indicate that a layer of oriented
water molecules is physically adsorbed onto the substrate surface which
has been modified by some chemisorbed water molecules.
Benzene adsorbed entirely in the capillary-condensed state on both
alumina and silica gel has a dielectric constant which compares with
that of liquid benzene. The dielectric constant of a monolayer of
benzene on silica gel is lower than the value for the liquid but the
value for a monolayer on alumina is equal to that of benzene in the
solid state. These values are explained by assuming interactions
between the electrons of benzene and the hydroxyl groups in the
silica and by assuming close-packing of the benzene molecules on the
More recently Baldwin and Morrow39 and Dransfeld and coworkers40
have studied the water-alumina system. Baldwin and Morrow39 measured
the dielectric constant and loss tangent as a function of temperature
in the frequency range from fifty cycles per second to one hundred
kilocycles per second. They found that water in the monolayer is very
strongly bound while water in the multilayers has sufficient freedom
to orient in the applied field. The polarization of the molecules in
the monolayer is very small and it is concluded that it does not arise
from dipole orientations. The polarization of molecules in the upper
layers is much larger and it is concluded that the orientation primarily
takes place by means of activated jumps of the molecules from one
equilibrium position to another on the substrate. A very narrow distri-
bution of relaxation times is found at the low frequencies and a much
broader distribution at the higher frequencies. Both the dielectric
constant and the loss factor are essentially temperature independent at
low coverages but quite temperature dependent at high coverages.
In order to ascertain the nature of the freezing process of sorbed
water Dransfeld and coworkers40 have studied the dielectric relaxation
of water on gamma -alumina as a function of temperature. In all cases
studied the amount of adsorbed water was sufficient for capillary
condensation to take place and the major portion of the adsorbed water
could be considered to be capillary-condensed. No abrupt change in the
relaxation rate is observed as the temperature is varied through the
normal freezing point of water. It is concluded that the solidifica-
tion of capillary-condensed water can be considered as a transition
between a supercooled liquid and a glass-like state.
Ebert41,42 has studied the adsorption of water on alumina rather
extensively. Three linear regions are observed when the tangent of the
loss angle is plotted against the amount of water adsorbed. The tempera-
ture dependence of the first region is very small but the other regions
show significant changes with temperature. This behavior is attributed
to the rigid bonding present in the monolayer compared to that in the
multilayers. The first two linear regions are attributed to the water
molecules in the monolayer.
In a later study42 Ebert found linear portions in the plots of the
apparent dielectric constant versus the amount of coverage. As the
frequency at which the measurements were made was increased the change
in the dielectric constant and in the loss factor with addition of
water became less. Both the apparent dielectric constant and the
dielectric loss decrease with decreasing temperature.
Neikam27 investigated the alumina-water system and found that plots
of the tangent of the loss angle versus the amount of adsorbate exhibit
four linear regions. In the first region the value of the tangent
remains essentially constant. This region ends at a water coverage
necessary for the completion of a monolayer. The other regions are
attributed to the water molecules being bound to the substrate by one,
two or three hydrogen bonds. The possibility of active sites on the
substrate is proposed as a explanation of the various types of adsorp-
tion. Heats of adsorption values indicate that all bonding after the
completion of the monolayer is rather weak. Plots of the change in
capacitance against the amount of adsorbed water exhibit no linear
regions. The dielectric results were compared with results from nuclear
magnetic resonance measurements on the same system. This comparison
was discussed in the previous section.
"If we could first know where we are, and whither we
are tending, we could then better judge what to do,
and how to do it."
1. Preparation of thorium oxide samples
The thorium oxide samples used in this investigation were prepared
either by precipitation of the hydroxide from a thorium nitrate solu-
tion and then dehydrating thermally to the oxide or by direct thermal
decomposition of thorium oxalate.
The samples prepared by the precipitation method are designated
with a Roman numeral and a letter. The Roman numeral describes the
batch and the letter describes a subbatch which has been activated
under certain conditions. These samples were activated in a vycor tube
under vacuum after being powdered to 100 mesh. The temperature and
length of time of the activation process are given in Table 1 along
with other details of the preparation.
Several of the samples used in this study were prepared and used
in an earlier study of the catalytic activity of thorium oxide toward
the decomposition of ethyl alcohol.43
a. Batch VI. Two hundred and ten grams of thorium oxide was
dissolved in twelve hundred and fifty milliliters of water and to
this solution two hundred and fifty milliliters of fifty-eight per cent
ammonium hydroxide was added at a very rapid rate. The precipitate
was filtered in one portion, under suction, and washed with twenty-
eight hundred and fifty milliliters of water. The washings gave a
negative test to red litmus paper. The hydroxide was dried for twenty-
four hours at 1200 C. and powdered to 100 mesh before activation. The
details of the activation process are given in Table 1.
b. Batches I, IX, X and XI. Forty grams of thorium nitrate was
dissolved in five hundred milliliters of water and fifty milliliters of
fifty-eight per cent ammonium hydroxide was added with stirring. The
precipitates were filtered under suction and dried at 120" C. for
twenty-four hours. The details of the rate of precipitation, washing,
and activation are shown in Table 1. When the ammonium hydroxide was
added in one batch the time of addition is listed as zero.
c. Samples Oi-1 and 011-2. These samples were prepared in a
manner identical to that described above with the exception of a
different rate of addition of the hydroxide and the amount of washing.
Although these samples were used as the hydroxides, the surface areas
of activated portions were determined.
d. Sample N-2. One hundred grams of thorium nitrate was dissolved
in twelve hundred and fifty milliliters of water and one hundred and
twenty-five milliliters of fifty-eight per cent ammonium hydroxide was
added with stirring. The precipitate was filtered with suction and
washed with two hundred milliliters of 30-70 ammonium hydroxide-water
mixture. The precipitate was then dried for about twenty-four hours
at 1000 C.
e. Samples OX-3 and OX-4. These samples were prepared by the
thermal decomposition of thorium oxalate obtained from Fisher Scientific
Company. Attempts were made to activate the oxalate in vacuum and in
a nitrogen stream but in both cases the oxide obtained was contaminated
with a large amount of carbon. The method which gave best results was
as follows: The oxalate was placed in a vycor tube open at both ends
and the tube placed in and open-end electric furnace. A smaller vycor
tube containing a thermocouple was placed in the large tube so that the
thermocouple junction was adjacent to the oxalate sample. A small
blower was used to circulate air through the tube and to carry away
some of the water vapor for sample OX-3. This required a longer time
for the sample to reach 6000 C. In this manner the samples were
activated for six hours at 6000 C.
2. Preparation of alumina samples
a. Sample F-lA. The alumina used was Alcoa activated alumina
obtained in granules of eight to fourteen mesh. The material was used
as obtained without additional treatment with the exception of powdering
to 100 mesh. The activation was done under vacuum for a period of six
hours at a temperature of 6000 C.
3. Determination of surface areas
The surface areas of the samples were determined by nitrogen
adsorption using the method of Brunauer, Emmett, and Teller.44 Previous
to the adsorption of nitrogen the surface of the sample was cleaned by
heating under vacuum for one to two hours at 2000 C. This was done to
remove any sorbed material picked up in the process of handling and
transferring the sample.
The surface area of the sample was determined by using the linear
form of the adsorption isotherm. This equation is:
P/V(Po p) = 1/VmC + p(C l)/VmCpo [I]
p = pressure of the gas
po = saturation vapor pressure of the gas
V = volume of gas adsorbed per gram of sample
Vm = volume of adsorbed gas necessary to form a
C = a constant
A plot of p/po versus p/V(po p) yields a straight line from
which Vm and C can be determined from the values of the slope and the
intercept. In most cases the intercept value is very near zero and
only the slope need be used. Once the value of Vm is obtained the
surface area can be calculated. A typical plot is shown in Figure 1.
B. Vacuum System and Techniques Used in the
Nuclear Resonance Experiments
The vacuum system used for the preparation of the samples used
in the n.m.r. experiments is shown in Figure 2. The system is evacuated
by means of a Cenco Hyvac-2 vacuum pump and the pressure in the system
determined by means of the manometer shown at D. The system was cali-
brated, in terms of milligrams of adsorbate corresponding to a given
pressure in the system, in the following manner. A sample of activated
material, either thorium oxide or alumina, was allowed to adsorb an
increment of the adsorbate in question and the pressure change in the
system was observed. The sample was then weighed and the amount of
adsorbed material determined. A graph of pressure change in millimeters
versus milligrams of adsorbate was then constructed. From this graph
0 0.1 0.2 P/Po 0.3 0.4 0.5
Figure 1. A Typical B. E. T. Isotherm Used for Determination
of the Surface Area.
B C D E F G
Figure 2. Vacuum System Used in the N. M. R. Experiments.
the amount of material adsorbed could be determined by noting the
change in the pressure in the system. However, this graph was used
only as a guide and the actual amount of material adsorbed was deter-
mined by weighing the sample before and after each adsorption.
The sample tube containing the adsorbent is connected to the
system at A. The tube is fitted with a vacuum stopcock and is made
from a commercially available 5 mm. n.m.r. sample tube.
The various adsorbates were placed into the containers designated
B, E, and F. A drying tube filled with Drierite (G) was used to admit
air or nitrogen into the system. Each of the adsorbates were deaerated
by a process of freezing and pumping until no gas bubbles were visible
in the liquids. The container marked C is a vycor tube used for
activating the sample.
After being activated, the samples were placed in the sample tubes
and again heated to 400* C. to drive off any materials that might have
been adsorbed in the transfer process. To keep the samples in the tubes
during the degassing, a small piece of glass wool was placed in the top
of the sample tubes just beneath the stopcock. The amount of sample
in the tubes was determined by weighing the empty tube and then
reweighing after the sample had been placed into it and had been heated
to 400* C. In most cases about two grams of the material were sufficient
for a sample.
Whenever different adsorbates were used on the same sample care
was taken to insure that all the previous adsorbate had been removed.
This was done by heating the sample to a temperature just below its
activation temperature for six hours while under vacuum.
The temperature was kept somewhat lower than the original activation
temperature to prevent any change in surface area.
C. Description and Procedures of the Magnetic Resonance Studies
The magnetic resonance spectra in this study were obtained from
a Model 4300-2 Varian Associates spectrometer operated at a frequency
of fifty-six and four tenths megacycles. The spectra were recorded as
the derivative of the absorption signal on a Varian Model G-10 graphic
recorder. Several recordings were made for each measurement and the
values averaged. The rate of change of the magnetic field was deter-
mined by modulating the 56.4 mcs. oscillator with a low frequency sine
wave from a Hewlett-Packard oscillator which was monitored continuously
with an electronic counter. Either a sample of alumina saturated with
water or, in the case of the fluorine-containing materials, a sample
of sodium fluoride was used for calibration purposes. The modulation
produced side bands at known frequency differences from the main peak.
During all line width studies care was taken that excessive power
from the transmitter was not used while recording as this produces
broadening of the signal as a result of saturation. The power level
was adjusted until the effects of saturation could be detected and
then reduced several decibels below this point before recording the
signal. In addition, the amplitude of the sweep field used to modulate
the static magnetic field as it is being varied through resonance was
controlled in order to minimize artificial sweep broadening. While a
small magnitude of the sweep field must be used to prevent artificial
broadening a large magnitude is desired from the point of view of the
signal-to-noise ratio. In this study a sweep width which just produced
sufficient signal to record was used. This was usually one of the two
lowest available values. The lowest possible value of the line width
which would not be broadened excessively is estimated to be about one-
hundred and fifty cycles. This value is lower than most of the observed
line widths in this study. The frequency of the modulating field was
maintained at eighty cycles per second throughout the measurements.
Checks made at lower modulating frequencies indicated no change in the
The temperature dependence of the line widths was studied by using
a Varian Model V-4340 variable temperature probe unit. The low tempera-
tures were obtained by passing a stream of dry nitrogen gas through a
copper coil immersed in a liquid nitrogen bath or a carbon dioxide-
acetone bath. Previous to passing through the cooling coil the stream
of nitrogen was passed through a tube containing calcium sulfate to
remove any water vapors. This prevents the formation of ice in the
cooling coil. The temperature was varied by regulating the nitrogen
flow rate by means of a needle valve.
The temperature of the samples was determined by means of a copper-
constantan thermocouple placed in the probe adjacent to the sample tube.
The thermocouple was used in conjunction with a Leeds and Northrup
potentiometer using a standard Weston Cell and a General Electric gal-
vanometer as a null point indicator. An equilibrium mixture of ice
and water was used as the reference point and the potentiometer
readings were converted to temperature units by use of standard cali-
The temperatures above the normal probe temperature (30 C.) were
obtained by using a Varian V-4343 heater control unit to heat the
nitrogen stream. The thermocouple arrangement described above was
used to determine the temperature. The accuracy of the temperature
is probably within one degree. For both high and low temperatures
several minutes were allowed for the sample to come to thermal equilib-
rium after a change in temperature.
At each temperature the spectra were recorded a number of times
and the measurements were averaged. With the exception of the studies
of the line widths as a function of temperature all spectra were obtained
at the temperature of the probe which is 300 C.
The chemical shifts of the adsorbed hexafluoropropene were deter-
mined in the following manner. The signal from the reference material,
in this case trifluoroacetic acid, was displayed on the oscilloscope
and the scope trace calibrated in terms of cycles per inch by applying
sidebands of known frequencies to the reference signal. The reference
was then removed from the spectrometer and replaced by the sample being
studied without changing any of the controls. The separation between
the reference and sample signals was then determined from the scaled
face of the scope. By repeating this a number of times a shift value
which is accurate to plus or minus thirty cycles can be obtained.
D. Determination of T1 and T2
Under certain conditions the spin-spin relaxation time (T2)*
is inversely proportional to the width of the resonance line. The
exact relationship is
Line width = 3-1/2/TT2 [II]
*For a discussion of relaxation times see Appendix B.
when the line is Lorentzian and
Line width = (2)-1/2/T2 [III]
when the line is Gaussian.* In both cases the width is measured
between the points of inflection.45
The spin-lattice relaxation time (Tl) can be obtained from a
knowledge of T2 and a knowledge of the magnitude of the alternating
field H1.46 By determining the value of H1 which produces the maximum
signal TI can be obtained. The maximum signal is obtained when the
(yHi)2T1T2 = 1 [IV]
is fulfilled. In this expression y is the magnetogyric ratio which is
a constant for a given nucleus.
E. Measurement of Hi
In order to use the above expression to obtain T1 the values of
the alternating field H1l must be known for the various degrees of
attenuation which can be chosen. Values of Hi can be obtained by
modulating the magnetic field with a small audio frequency.46 Resonance
is obtained when the modulating angular frequency wm satisfies the
m = 2 ( o 11)2 + y2f112 [V]
This condition is satisfied by two values of H and two signals will
be observed. The value of Hi for a given amount of attenuation can
be obtained by measuring the distance between the two signals and
plotting the square of this distance against the square of the modu-
lating frequency. This plot yields a straight line with an intercept
*It is more appropriate to call this a line-width parameter since a
true T2 cannot be defined in this manner for a Gaussian line.
of y2H12. Since precision attenuators are used in the commercial
spectrometers, a few values of Hi can be found in this manner and the
values for other attenuations can be calculated. A typical measure-
ment by this method is illustrated in Figure 3. The distance is
plotted in arbitrary units and the different plots are for different
F. Systems and Techniques Used in the Dielectric Experiments
1. Vacuum system
The vacuum system used in the dielectric experiments is shown
in Figure 4. The pressure in the system is measured by means of the
Zimmerli gauge B or the McLeod gauge F. The system is evacuated by
means of a mercury diffusion pump (G) in series with a Cenco Hyvac-2
vacuum pump (H1). The dielectric cell is connected to the system by
means of a ground glass ball joint at A. The adsorbate was stored in
the container C. At E there is a carbon dioxide-acetone cold trap to
prevent vapors from reaching the pumps. D is a gas burette used in
determining the volume of the system. The entire volume of the system
was obtained and this value used to determine the amount of material
adsorbed on the samples.
2. Dielectric Apparatus
Measurements of capacitance and conductance were made with a
General Radio Twin-T Type 821-A Impedance Bridge. The Bridge was
used in conjunction with a General Radio Oscillator Type 1330-A and
a Hallicrafter SK-62A receiver as a detector. A Heathkit model OP-1
Oscilloscope and a Heathkit Model V-4 vacuum-tube voltmeter were also
used in balancing the bridge. The scope was preceded by a sixty-cycle
d2 4 lOdbs. AA 15dbs.
oo 20 dbs. oa 30dbs.
o / / /
0 2 4 6 8 10
f xl04 cps2
Figure 3. Determination of the Magnitude of H1 for Various
Vacuum System Used in the Dielectric Experiments.
The Twin-T bridge circuit is a null instrument with a frequency
range from 460 kilocycles to 30 megacycles although frequencies some-
what higher and/or lower can be obtained. The frequency range of
the oscillator is 5 kilocycles to 50 megacycles and that of the
receiver is 540 kilocycles to 109 megacycles.
The apparatus is used to measure the impedence of the sample in
terms of its susceptance and conductance by a parallel substitution
method. The susceptance B is related to the capacitance, which is
read directly from the bridge. The relation is
B = w(C1 C2) [VI]
where w is the angular frequency, and C1 C2 is the change in capaci-
tance when the sample is placed in the bridge circuit. The conductance
G'is also obtained directly from the bridge by using the relation
G = Co(f/f)2 [VII]
where Co is the dial reading, f is the operating frequency, and fo is
the frequency at which the frequency range switch is set.
From the conductance and susceptance the dissipation factor or
loss tangent can be obtained. This factor (see Appendix C) can be
expressed as the ratio of the loss current to the charging current
and for this bridge circuit is given by
tan 6 = G/B [VIII]
which is equivalent to the expression given in Appendix C.
3. Dielectric Cell
In previous studies20 the available dielectric cells have had
several features which made it difficult to obtain results which were
completely reproducible. These difficulties were due in part to the
length of the leads on the cell and to the increase in volume of the
solid material whenever water was adsorbed. The expansion of the sample
with uptake of water caused the sample to push from between the plates
of the cell. In order to remove as many of these difficulties as possi-
ble a new cell was constructed for this study. To overcome the prob-
lem of the volume change of the sample with water uptake special
electrodes were constructed. One electrode is made up in the shape of
a dish and the other electrode is simply a perforated metal plate.
The outer diameter of the dish electrode is two inches and that of the
plate electrode is one and seven-eights inches. The plate electrode
is perforated to allow rapid equilibration of the water vapor and the
sample. Details of the two electrodes are shown in Figure 5.
The dish electrode is constructed of stainless steel and the
plate electrode of monel metal. Stainless steel was used for the dish
electrode because the edges had to be turned to give the desired shape.
This electrode was constructed by the Reliable Metal Spinning Co.,
Newark, New Jersey.
The cell is constructed in such a manner that it is possible to
use as many as three sets of the electrodes at once. In this study
only one set was used and these had a separation of 0.025 inches.
The electrodes are separated by means of Teflon spacers. If more than
one set of electrodes are used the sets are connected in parallel by
means of a series of 0.025 inch Teflon and steel washers. Once the
tier is completed the top electrode is held in place by a steel and a
The tier of electrodes (or in this case the single electrode) is
enclosed in a cylindrical glass container with hemispherical ends.
A. Dish Electrode
B. Plate Electrode
C. Switching Device
Figure 5. Electrodes and Switching Device for the
The two halves of the glass container are connected by means of a
vacuum-tight flange joint. The top portion of the container is
fitted with a vacuum stopcock and a ball joint by which the cell is
connected to the vacuum system. The bottom glass section has two
nickel leads sealed into the glass. The electrodes themselves are
connected to the leads by screws and can be removed completely from
the cell. The assembled cell, with the exception of the external
bridge connections, is shown in Figure 6.
The cell is constructed in such a manner that it is not necessary
to have long leads from the cell to the bridge. Banana plugs are
connected to the electrode leads directly beneath the cell and this
allows the cell to be placed directly above the bridge.
In order not to remove the cell from the bridge during the entire
period measurements were being made, a special switching device was
constructed. This device is necessary since the cell must be removed
from the circuit in some manner during the initial balancing of the
bridge. A hole was drilled into the top of a banana plug and fitted
with a Teflon sleeve. One of the leads from the electrodes extends
into this sleeve. A hole was then drilled into the side of the
banana plug and through the Teflon sleeve. This hole was tapped and
fitted with a brass screw. This screw extends through the Teflon and
makes contact with the lead when tightened. Whenever it is necessary
to remove the cell from the bridge circuit the screw is simply with-
drawn until contact is broken. The reproducibility of the cell
properties was checked and found to be entirely satisfactory if the
screw is tightened snugly each time. The capacitance of the cell can
Figure 6. Assembled Dielectric Cell.
be reproduced to within 0.2 picofarads. A sketch of the switching
arrangement is also shown in Figure 5.
4. Loading of the dielectric cell
Before measurements of the dielectric properties were undertaken
the following procedure was carried out.
The entire cell assembly is weighed and then the activated sample
is placed into the dish electrode until all space between the two
electrodes is full. The cell assembly containing the sample with
some water which it had adsorbed from the atmosphere in the handling
process is now weighed. By comparing the weights of the empty and
filled cell, the weight of the "wet" sample is obtained. To prevent
any leakage it is necessary to seal the flange joint with Sealit
obtained from Fisher Scientific Company. The cell is now placed onto
the bridge and connected to the vacuum system described above. After
pumping on the sample for a period of 48 to 72 hours at a pressure of
about 10-5 millimeters of mercury the cell is closed off and removed
from the bridge and weighed again. The amount of dry sample is then
obtained from the difference in weights. Corrections are made for
the weight of the air contained in the cell during the original
weighing. Once the weight of the sample is found the cell is fastened
to the bridge and not removed until all measurements are completed.
The cell contained from six to eight grams of sample in most cases.
PRESENTATION AND DISCUSSION OF RESULTS
"For my thoughts are not your thoughts, neither
your ways my ways."
A. Magnetic Resonance Results
1. General Comments
A macroscopic approach to the phenomenon of nuclear magnetic
resonance has been presented by Bloch47 by means of a set of phenom-
enological equations which describe the time behavior of the trans-
verse and longitudinal components of the magnetization. Even though
assumptions are made concerning the nature of the various relaxation
processes these equations appear to offer a simple and in many cases
a completely adequate manner of viewing these processes.
In the ideal cases the shapes of the absorption lines obtained
in nuclear resonance can be described by line shape functions which
are either Lorentzian or Gaussian. In actual practice the line shapes
may not be described exactly by either of these functions.
The solution of the Bloch equations leads to a shape function
which is Lorentzian. The Lorentzian line is characteristic of and
traditional in the theory of collisional- and radiation-broadened
spectral lines. The shape function which describes a Lorentzian line
(in nuclear resonance) is
g(u) = 2T2/[11 + (W wo)2T22 I1]
where T2 is the spin-spin, or in Bloch's nomenclature, the transverse
relaxation time. From absorption lines described by this shape func-
tion the two relaxation times can be obtained by the method described
in the previous chapter.
To justify the use of the described method of obtaining relaxa-
tion times the lines obtained in this investigation were tested by
the method of Pake and Purcell48 to determine to what extent they
vary from true Lorentzian or true Gaussian lines. This test consists
of determining the ratio of the large maximum to the small minimum
in the derivative presentation of the dispersion curve. For a
Lorentzian line the ratio of the two is 8:1. For a Gaussian line the
ratio is 3.5:1. The lines obtained from the adsorbed water in this
study have ratios which range from 4.9:1 to 7.5:1. This indicates,
as expected, that the lines are neither pure Lorentzian or pure
Gaussian but have a shape which is somewhere between the two. The
average of all the obtained ratios is 6.2:1 which possibly indicates
that the lines may best be described by a Lorentzian function.
While the relaxation times obtained here may not be the true
values characteristic of these systems it is believed that comparison
of the obtained values, for identical conditions, will afford valuable
information about the systems. It should be noted that the main points
of interest in this study are the changes and the differences in
relaxation times rather than absolute values.
One disadvantage of the method of determining relaxation used in
this study should be pointed out and kept in mind. Since the spin-
spin relaxation time is derived from the widths of the resonance line
it is affected by all factors which produce line broadening. In
addition to the "natural" broadening produced by the relaxation
processes "artificial" line broadening may result from (a) inherent
inhomogeneities in the magnetic field resulting from the properties
of the of the magnet itself, (b) inhomogeneities produced in the
magnetic field by the difference in the magnetic susceptibilities of
the solid and the surrounding medium, and (c) the effect of the
magnitude and frequency of the modulating field.
In this study the factor described in (a) is very small compared
to the line widths and does not present a problem. However, the
factors described in (b) and (c) are not always small compared to the
line widths and must be considered. As a result of the broadening
produced by factors (b) and (c) the observed spin-spin relaxation times
are not the true times needed to describe the relaxation processes
but are related to them in the following manner:
1/T2,obs = 1/T2 + l/'f2 [II]
The factor 1/T2 describes the total broadening produced by factors
(b) and (c).
An attempt was made to determine the magnitude of these two
factors. An excess of water was added to several thorium oxide samples
until a nearly homogeneous mixture was obtained. The line width of
the water was then measured with the results shown in Table 19. These
measurements were made using the smallest magnitude of the modulating
field possible on this particular spectrometer. The frequency of the
field was maintained at eighty cycles per second.
From the data in Table 19 it is noted that the minimum line width
obtained from any sample is 137 cycles per second. The major portion
of this value is considered to be the combined results of factors (b)
and (c) since the true line width of liquid water should be only a
few cycles at most. To determine what part of the observed widths
resulted from the inhomogeneity broadening, the same samples were
then run by the normal high-resolution technique to eliminate the
effect of the modulating field. In each case the line width was
reduced by about a factor of two. From these results it is concluded
that each of the factors (b) and (c) contributes about fifty or sixty
cycles to the observed line widths. The contribution from (b) remains
essentially constant from measurement to measurement but the contri-
bution from (c) depends upon the true line width and should approach
zero for lines with widths greater than about three or four hundred
cycles per second.
As mentioned previously the frequency of the modulating field
was maintained at eighty cycles per second. On previous measurements
this frequency was reduced to twenty cycles per second but there
was no reduction in the width of the line. It is believed that a
reduction in line width does not occur in this case because of the
larger effects of the magnitude of the modulating field and of the
No attempt will be made to correct the observed line widths and
relaxation times since definite correction factors cannot be obtained
but it should be kept in mind that the observed T2 values are smaller
than the true values and consequently the observed T1 values are
larger than the true values.
Prior to a discussion of the experimental results it may also
be informative to point out some characteristics of the surface of a
solid such as thorium oxide both before and during the adsorption
process. The physical condition of an adsorbing surface is very
difficult to determine even before an adsorbate has been added and
even more difficult after adsorption has taken place. However, one
fact is sure, the surface of a metal oxide is not homogeneous through-
out but must certainly consist of sections which are characteristic
of the oxide crystal structures, sections characteristic of the
hydroxide structure, and portions which have other characteristics.
After adsorption has taken place to some extent the already hetero-
geneous surface is made more so by the changes brought about by the
addition of the adsorbate. For example, the adsorption of water by
a metal oxide may result in ions being present on the surface as
well as water molecules in various states. As a result of these
changes in the surface with adsorption the upper layers of adsorbate
will not be adsorbed on the same kind of surface as were the first
layers. In the following discussion it should not be assumed that
the author has a definite geometrical arrangement of atoms or molecules
in mind when referring to the structure of the oxide surfaces.
One point should also be kept in mind when considering the sur-
face area of a material such as thorium oxide. Surface areas deter-
mined by using different adsorbates are not in general the same. For
example, the surface areas in this study were determined by nitrogen
adsorption and would probably not be the same if determined by the
adsorption of water or other materials.
2. Adsorption Studies
The adsorption of water has been studied on several different
thorium oxide samples. In Figure 7 the data for four different oxide
samples, all prepared from thorium hydroxide, is shown. In this figure
the line width is plotted as a function of the number of layers of
coverage. The surface areas of the samples are given in the figure.
The amount of water necessary for a monolayer is determined by taking
the surface area occupied by a water molecule as 14.0 A2. This value
was chosen from values found for water adsorbed on aluminum oxide27
and appears to be near the average of all values reported in the
literature. It should be pointed out that the area occupied by an
adsorbate molecule may depend to some extent on the nature of the
surface of the adsorbent.27
All the samples appear to show the same general behavior with
only small differences. In each case the line widths are large at
low coverages and decrease with increasing coverage until an essenti-
ally constant value is reached. This constant value is obtained only
after several layers of water have been adsorbed. The values of
the line widths at low coverages are of the same order of magnitude
as the line width of thorium hydroxide itself (Table 18). This
indicates that the first layers of water are bound quite tightly to
the surface. For most samples it is not possible to obtain an
absorption line whenever the coverage is less than one monolayer.
This is in part due to the small surface areas of the samples which
dictates the amount of water necessary for a monolayer and in part
to the degree of bonding of the first water layer. From the data
ooo 14.1 m2/g.
,- 9.4 m.2/g.
0 00 28.9 m2g.
***oo31.1 m.2 /g.
4 8 12 1(
Line Width of Water on Thorium Oxide Prepared from Thorium Hydroxide.
in Figure 7 it is apparent that the influence of the surface extends
out several molecular layers. Even at the highest coverages and after
application of the correction factor discussed in the previous section
the line widths do not approach that of pure liquid water. To indicate
the differences, liquid water has a line width of about one cycle per
second and ice has a line width of about sixty thousand cycles per
Since the water coverage is plotted in terms of the number of
monolayers it is reasonable to expect that the curves would very nearly
superimpose; however, large differences are found. These differences
can be attributed to at least two things. One is that some residual
water remains on the samples after activation. This can be as high
as one to two per cent for an activation temperature of 6000 C. and
even higher for lower activation temperatures.49 In view of this it
should be noted that the samples having surface areas of 28.9 m.2/g.
and 31.1 m.2/g. were activated at lower temperatures than the other
two. Another explanation for the lower line widths of the sample
activated at the lower temperature may be in the condition of the
surface itself rather than in the amount of residual water. Since it
is apparent that these samples will not adsorb as many layers of water
as the samples activated at higher temperature even after considering
the residual water it appears quite reasonable to attribute part of
the differences in line widths to the condition of the surface.
A peculiar behavior was found for the sample having a surface
area of 28.9 m.2/g. (Sample VI-E). The line width was found to be
dependent upon the amount of elapsed time between the adsorption
process and the time of measurement. The break in the curve near
four monolayers occurred when the sample was allowed to stand for a
period of about thirty days between measurements. Similar behavior
was observed for other samples activated at 5000 C. but usually to a
lesser extent. Since the line width increases rather than decreases
with time it must be attributed to a very slow hydration process or
possibly some type of exchange process, say between water molecules
on sites of different energies. Another possibility is the diffusion of
water into the pore structure. Other evidence of this "time factor" has
been found in adsorption studies of thorium oxide.6
The behavior of the observed relaxation times of water adsorbed
on a thorium oxide sample prepared from thorium hydroxide (Sample IX-A)
is shown in Figure 8. The spin-spin relaxation time (T2) increases
with increasing water coverage and approaches a constant value. The
values (10 to 50 milliseconds) suggest that the water is rather tightly
bound to the surface and has restricted mobility even at the highest
coverage. For this sample there appears to be some difference in the
values obtained on adsorption from those obtained on desorption. On
other samples no differences were noted. The behavior on this partic-
ular sample may be related to the pore structure of the surface of
The spin-lattice relaxation time (T1) for this sample also increases
with increasing coverage but only slightly. This increase can probably
be taken as an indication that the adsorbed water approaches a more
"liquid-like" state as the amount of water increases. In some samples
the values found for T, are larger than the T1 value of liquid water
The Observed Relaxation Times of Water on Thorium Oxide Prepared from
but in light of the previous discussion of the magnitudes of the
observed T1's this is expected.
Several differences can be found between the data for the samples
prepared from thorium hydroxide and that for the samples prepared from
thorium oxalate. In Figure 9 results for the adsorption of water on
oxide samples prepared from thorium oxalate are shown. The most
prominent difference between these and the previous results is the
difference in the width of the lines. The lines from the samples
prepared from the oxalate are in general not as wide, especially at
very low coverages, as those found for samples prepared from the
hydroxide. This indicates that the interactions between the surface
and the adsorbed molecules must be weaker in these oxides. This could
indicate that catalysts prepared from thorium oxalate should have more
activity for such chemical reactions as the dehydration of alcohols
since they should not be as easily poisoned and the products of the
reaction should not be bound as tightly to the surface. In general
this is found to be true.
Both sets of data (Figures 7 and 9) indicate that the samples
having the largest surface are less efficient in producing relaxation,
i.e., smaller lines, for a given water coverage (in terms of the T2
relaxation). This is in accord with the catalytic behavior since
larger surface areas, all other factors being equal, usually indicate
The differences between the two kinds of oxides are reflected
quite strongly in the relaxation times. In Figure 10 the observed
relaxation times of water adsorbed on an oxide sample prepared from
oo o12.6 m2/g.
A A 37.0 m.g.
Figure 9. Line Width of Water on Thorium Oxide Prepared from Thorium Oxalate.
2 4 6 8 10 1:
The Observed Relaxation Times of Water on Thorium Oxide Prepared
from Thorium Oxalate.
thorium oxalate are shown (Sample OX-3). A comparison with Figure 8
shows that the spin-lattice relaxation time (Tl) of water adsorbed on
an oxide prepared from the oxalate is much smaller than that for water
adsorbed on an oxide prepared from the hydroxide. A comparison of the
T2 values also shows a difference but not as great as that for the TI
values. The smaller values of T, could result from the effect of
paramagnetic centers, on the surface of the oxalate oxides, perhaps
introduced into the oxide surface during the preparation. It has been
found by infrared studies50 that thorium oxide samples prepared from
thorium oxalate usually have a surface layer of thorium carbonate
which results from adsorbed carbon dioxide. It is quite conceivable
that the differences in the T, values can, at least in part, be
ascribed in the effect of this layer of carbonate. In general tempera-
tures greater than those used in the preparation of the samples used
here are necessary to completely remove the adsorbed carbon dioxide.
Several interesting facts become apparent whenever other adsor-
bates besides water are used. In Figure 11 the resulting line widths
of adsorbed water, methanol, ehtanol, and butylamine on a thorium
oxide sample (Sample VI-D) are shown. It is apparent that both water
and methanol behave in a similar manner but different from ethanol and
butylamine, which themselves behave somewhat differently. The line
widths of the latter compounds, on this sample, are correspondingly
less than the width of the methanol and water lines. This could
indicate that this oxide sample has a pore structure such that the
pores can accommodate the smaller molecules but not the larger ones.
Comparison of the results in Figure 11 with the results obtained from
Line Widths of Various Adsorbates on Thorium Oxide.
a sample activated at 500 C. (see Table 3) shows that the line widths
of the same adsorbates on the sample activated at the lower temperature
are all approximately the same. It is likely that the lower activation
temperature results in a more porous structure which is also indicated
by the larger surface area. The fact that the oxide samples appear
to have different degrees of interaction between the surface and various
adsorbates can possibly be used to gain knowledge of their catalytic
activity. As an example, the sample from which the data in Figure 11
is taken should not be a good catalyst for the dehydration of ethanol
since water is bound to the surface more tightly than ethanol.
The behavior of the line width of adsorbed butylamine differs
from that of adsorbed ethanol in that the changes with increasing
coverages are quite small. This can be seen in Figure 11 although the
small surface area of the sample allowed only three determinations.
The fact that the line width does not change appreciably with increas-
ing coverage can be attributed in part to the fact that the hydrogens
on the nitrogen are chemically shifted from the other hydrogens in
the molecule. At the higher coverages where the line width should
decrease this chemical shift prevents it to some extent.
It appears that the line widths of all the adsorbates are approach-
ing a constant limit which probably represents the result of the
artificial broadening mechanisms discussed previously.
The differences between water and the other adsorbates become
more apparent upon comparison of the observed relaxation times. In
Figure 12 the observed relaxation times of methanol adsorbed on thorium
oxide (Sample VI-D) are shown. Although the values of T2 are approx-
imately the same as those for adsorbed water, the values of T1 are
The Observed Relaxation Times of Methanol on Thorium Oxide.
(s e c.)
much higher. Granting that the T, values are somewhat too large
because of the reasons discussed previously, it is apparent that the
true values are higher than those for the adsorbed water. The relax-
ation times of ethanol (see Tables 2 and 3) and butylamine (see Tables
2 and 4) are also higher than the values found for adsorbed water.
The spin-lattice relaxation times (Tl) of the alcohols differ in
one other respect from those of adsorbed water. The values for water
always increase as more water is added but the values for the alcohols
and butylamine increase in some cases and decrease in others although
the few points obtained in most cases make some of these trends
The fact that the trend may be different for different samples
should not cause alarm since it has been found21 that T1 may show a
minimum if the values are obtained over a sufficient range of coverage.
It is quite possible that the coverages here may be on opposite sides
of the minimum and thus would show this particular behavior.
The larger values of the T1's for the alcohols and butylamine
indicate that these adsorbed molecules are not as sensitive to the
thermal motions which produce relaxation or that the adsorbed materials
behave more like a normal liquid. A liquid-like behavior could result
from the manner in which the molecules are adsorbed. If the adsorption
takes place through the polar groups the non-polar part of the molecules
would be free to move about and thus behave more like a liquid state
than if all parts of the molecules were strongly bound to the surface.
Further evidence of this type of interaction will be seen in a later
discussion of temperature effects upon the line widths and relaxation
times of these molecules.
3. Temperature Studies
The temperature dependence of the line width and relaxation times
has been studied for all adsorbates. A typical temperature dependence
of the line width of adsorbed water is shown in Figure 13 and other
data given in Tables 8 12. In Figure 13 the changes in line width
with temperature are plotted for three different water coverages. For
the lowest coverage, which is just less than two monolayers, the line
width begins to increase quite sharply at about 100 or 15 C. For
larger amounts of water the width remains essentially constant down
to about -300 to -50 C. The fact that the line width increases
before the normal freezing point of water, for low coverage, indi-
cates that some type of motion is being frozen out rather than the
normal formation of ice. In fact the data for the high coverages
indicates that if ice is formed at all it is formed much below the
normal freezing point of water. Several experiments have shown51 that
adsorbates tend to freeze at temperatures much below their normal
freezing points. The data in Figure 13 offers further proof of this
since the line width does not approach that of ice (60-80 kcs.) even
at -800 or -900 C.
The sharp increase in the line width at low water coverages
indicates that under these conditions it is the adsorbate-adsorbent
interactions which are primarily responsible for the line width.
At higher coverages the adsorbate-adsorbate interactions may also
play a substantial part in determining the relaxation times and line
A typical temperature dependence of the observed relaxation
times of adsorbed water is shown in Figure 14. This data is for a
L5 2 4.44
-70 -50 -30
The Temperature Dependence
Water on Thorium Oxide.
of the Line Width of
Dependence of the Observed Relaxation Times of Water on
water coverage of 16.7 mg. H20/g. ThO2 adsorbed on sample IX-B. As
the temperature increases the observed value of T2 also increases
indicating an increased amount of mobility in the adsorbed molecules.
The observed values of T1 become less as the temperature increases
as is expected since the spin-lattice relaxation time is a measure
of the time for thermal equilibrium to be established between the
spin system and the thermal motions of the lattice. As the tempera-
ture increases more and more motions having components with frequencies
near the resonance frequency of the nuclei are present thus producing
more efficient T1 relaxation.
By considering the change in T2 with temperature, an apparent
activation energy for the processes responsible for the decay of
the transverse components of the magnetization can be obtained.
Theoretically, the activation energy is obtained by considering the
change in the correlation time (Tc) with temperature (see Appendix
B for a short discussion of the correlation time). The longer the
correlation time becomes the more rapid is the relaxation process
and the shorter T2 becomes. The decrease in T2 results in a wider
line width since the two are related inversely. Thus the line width
and the correlation time have the same directional behavior.
The correlation time is of course governed by thermally activated
motions and should follow an equation of the type
Tc To exp(Eact/kT) [III]
Plots of the log of the correlation time against the reciprocal of
the temperature should yield straight lines from which the activation
energy can be obtained from the slope. From the above discussion it
is apparent that a plot of the line width against l/T should also
yield straight lines from which the activation energy can be obtained.
Plots of this type for the two different kinds of thorium oxide samples
are shown in Figures 15 and 16. In each case a number of linear
segments are found for each water coverage. These segments range in
number from two to four. In Table 17 the energy values corresponding
to the various regions are listed.
Several features of Figures 15 and 16 and Table 17 are of interest.
One fact which is of considerable interest is the existence of differ-
ences between the two different kinds of oxides. The data from the
oxide samples prepared from thorium hydroxide shows linear portions
which have increasing slopes as the temperature decreases. The data
for the oxides prepared from the oxalate shows linear portions which
have alternating values of the slopes with decreasing temperature.
It can be seen from the figures that for low water coverage the first
linear portion, which has a small slope, is not present. The higher
the coverage becomes after this section is present the larger becomes
the temperature range covered by this segment. It can also be seen
from the data that the slope of the third linear segment, for the oxides
prepared from the hydroxide, becomes greater with addition of water.
At a very high water coverage this portion starts at a higher temper-
ature than it does for lower water coverages.
Upon first considering the data in the two figures and in Table
17 there appear to be inconsistencies. If the numbers are accepted
at face value it appears that those motions having the lowest activa-
tion energies are the first ones to be frozen out. This is certainly
oo8.20mg./g. AA 12.8 mg./g.
0024.9mg./g. oo33.1 mg./g.
3.7 4.2 1 0O 4.7 5.2 5.7
Log of the Line Width versus the Reciprocal Temperature
for Water on Thorium Oxide Prepared from Thorium Hydroxide.
an 11.4 mg./g.
oo 29.3 mg./g.
4.2 _3 4.7
Log of the Line Width versus the Reciprocal Temperature
for Water on Thorium Oxide Prepared from Thorium Oxalate.
not reasonable if the numbers represent activation energies. However,
the data can be explained in a reasonable way and in a way which leads
to consistent results.
In the first linear region, which occurs when the line widths are
quite small, the widths are controlled not by the molecular motions but
by the mechanisms discussed previously. As evidence of this it should
be noted that when the water coverage is low this portion is not
present. This is because a low water coverage results in a wide line.
In other words the actual width of the line in the absence of the
artificial broadening is such that the effects of the modulating field
and the inhomogeneities are too small to broaden the line further. As
a consequence of the artificial broadening the line width is virtually
temperature independent and it is not until the line becomes suffi-
ciently broadened by the decrease in molecular motions does the observed
line width change. This is the beginning of the second portion of the
plots. In the second and the third, for the oxides prepared from the
oxalate, regions the obtained energies are more nearly true activation
energies of the various kinds of motions. For water on the oxides
prepared from the hydroxide there appears to be only one general kind
of motions since there is only one straight line portion in this
region. On the other hand the data for water on the oxides from the
oxalate shows two portions in this region. The first has the greatest
slope and represents the motions having the largest activation energies.
These are expected to freeze out at higher temperatures than are the
motions having lower activation energies.
The last linear portions of the plots, which appear to represent
motions having activation energies higher than the ones frozen out at
the higher temperatures, are in fact due to the beginning of an ice-
like structure. At this temperature the molecular motions become
sufficiently slow so that the water begins to take the ice structure
and line broadening occurs by dipole-dipole interactions. In general
the more water there is on the sample the more nearly like liquid it
becomes and the greater the amount of water which takes part in the
freezing process. This results in the increase in slope of this
portion of the plots with increasing water coverage.
The most important feature of this data is the fact that there
appear to be different kinds of motions present in the oxides prepared
from the oxalate while in the other samples there is indication of
only one general kind of motion. It is possible that there are
"active sites" of some type present in the oxides prepared from the
oxalate which result in exchange type motions in addition to the
normal rotations and vibrations.
Studies of the temperature dependence of the line widths of
adsorbed alcohols and butylamine have also been done. In Figure 17
the line width of adsorbed methanol is shown for the temperature
range from 300 to about -1150 C. For both coverages an apparent
maximum occurs in the line width near 0 C. This maximum is more
pronounced at the lower coverage. At temperatures below the point
where the maximum occurs the line width decreases to a value lower
than that found at room temperature and remains virtually constant
until the temperature reaches about -1150 C. This behavior appears
to be reasonable if it is assumed that the methanol molecules are
adsorbed onto the surface through interactions between the hydroxyl
groups and the surface. Since in methanol the chemical shifts of
ooo 23.9 mg./g.
AAA 1 8.9 mg./g.
i 0.9 A
-110 -90 -70 -50 -30 -10 10 30
CFigure 17. The Temperature Dependence of the Line Width of Methanol on Thorium Oxide.
Figure 17. The Temperature Dependence of the Line Width of Methanol on Thorium Oxide.
the two kinds of hydrogens are not sufficiently far apart to observe
two distinct resonances, the line that is observed is made up of two
components. These two components come from the two different kinds of
hydrogens. At room temperature and at all temperatures above the
apparent maximum the observed line results from both of these
components. As the temperature is decreased the hydroxyl component
broadens out considerably since it is bound strongly to the surface
and at the temperature at which the maximum occurs begins to completely
disappear. Below this temperature only the component from the methyl
group is observed and this remains quite sharp since movement of the
methyl group is hindered only slightly by the surface. This "rotation"
of the methyl group persists until very low temperatures but at about
-115 C. the motions become sufficiently slow that dipole-dipole
interactions become a factor and cause the line to broaden. The data
in Figure 17 is for an oxide sample prepared from thorium hydroxide
but the same behavior is found for samples prepared from thorium
oxalate (see Tables 13 and 14).
In Figure 18 the variation with temperature of the line width of
adsorbed ethanol is shown. The general behavior is the same as that
of methanol. The difference between the two is attributed to the
presence of the methylene group in ethanol which interacts with the
surface in a manner different from either the methyl or the hydroxyl
group. The slight increase in line width between -50 and -80 C. is
attributed to the affect of the surface on the methylene group. This
interaction becomes greater as the temperature decreases.
The temperature dependence of the line width of adsorbed butyl-
amine is shown in Figure 19. For the higher coverage the behavior is
ooo 6.7 mg./g.
Figure 18. The Temperature Dependence of the Line Width of Ethanol Adsorbed on Thorium Oxide.
(kcs.) o 0 12.9 mg./g o
1.3 o o
AA 16.9 mg./g. o
A 0 0 0
wA A o
i A A A
1.1 A A
w A 0
0.9 A o o A
-110 -70 -30 10 50
Figure 19. The Temperature Dependence of the Line Width of Butylamine Adsorbed on Thorium Oxide.
similar to that of the alcohols except that two apparent maxima are
present. The first of these is attributed to the broadening of the
component of the line belonging to the NIl2 group. It is believed that
this group is interacting quite strongly with the surface in a manner
similar to the hydroxyl groups in the alcohols. The second apparent
maximum can possibly be attributed to the effects of the surface upon
the methylene group adjacent to the NH2 group. That these interactions
are different from the interactions of the methylene group of ethanol
is to be expected since the different geometries of the OH and Nil2
groups would cause the two groups to be in different positions relative
to the surface. The line width decreases quite sharply below the
second maximum before dipolar broadening begins. This is again expected
since the methyl group is now quite far removed from the surface.
The behavior of the line at the lower coverage is somewhat
different. While the values at the higher coverage are completely
reproducible over the entire temperature range there is a region
between 20 and -500 C. in which the line width of the lower water
coverage cannot be reproduced. Above and below this range the measure-
ments can be repeated well within the experimental error. The scatter
of the values in this region is attributed to the fact that the NH2
group interacts strongly with the surface producing a broad line which
is also chemically shifted from the line of the other hydrogens. Evi-
dence of this is shown in Figure 20. In this figure the absorption
line, here presented as the derivative, is shown at three different
temperatures. The line is not symmetrical at any temperature but
becomes less symmetrical in the temperature range mentioned above.
Figure 20. Presentations of the Absorption of Butylamine Adsorbed on Thorium Oxide.
The increased width of the lower-field part of the line is attributed
to the component of the line belonging to the NiH2 group which is shifted
to lower field. The shift appears to be maximum in the region where the
line width is not reproducible and it is almost possible to pick out
the two components here. The measured line width in this region is
not a true measure of either of the components but more or less an
average of the two. Below this temperature region the component
belonging to the NH2 group becomes broader and the sharp component
belonging to the alkyl group, or part of the alkyl group, governs the
width of the line. The sharpness of the line at low temperatures
indicates that the alkyl part of the molecule interacts only weakly
with the oxide surface and has considerable freedom.
Because the line widths are made up of different components,
the behavior of the relaxation times of adsorbed alcohols and butyl-
amine with temperature cannot be readily interpreted. For the same
reason it is not possible to obtain an apparent activation energy as
in the case of adsorbed water. The observed relaxation times are
given in Tables 13 17, but aside from the fact that the T1 values
are much larger than the values found for water little information can
be gained from them. It should be noted however, that all the relaxa-
tion times as well as the line widths are averages. For water they
are the average of the relaxation times of the different kinds of
water molecules; those bound tightly to the surface as well as those
which have considerable movement. For the alcohols and butylamine
this is also true with the additional averaging of the relaxation
times of the various parts of the molecules. This averaging process
may in part account for the larger values of the relaxation times,
especially T1, of the alcohols and butylamine. Certainly part of the
larger T1 values is due to the mobility of the alkyl parts of the
4. Studies of C3F6 adsorbed on aluminum oxide
To study more fully the apparent differences in the interactions
between the surface of a solid and various parts of an adsorbed
molecule it was decided to study a material containing fluorine adsorbed
on an oxide. A fluorine-containing material was chosen because of the
magnitude of the chemical shifts of fluorine. These are sufficiently
large to allow the observation of the lines belonging to the different
groups in the molecule. Several attempts to adsorb an observable
amount of a fluorine-containing material on thorium oxide proved
unsuccessful apparently because of the small surface area of thorium
oxide. Success was attained, however, with a high surface area
aluminum oxide sample and hexafluoropropene as the adsorbate.
The F19 spectrum of the adsorbed hexafluoropropene is shown in
Figure 21. This molecule has four different kinds of fluorine atoms
and the lines belonging to the different kinds can be clearly distin-
guished. In no case is there overlap of the lines. The assignment
of the various bands can be made by considering the high-resolution
spectrum obtained from the liquid material.52 The assignments are as
follows: CF3 line, Fa line, Fb line, and Fx line, with increasing
The widths of the various bands were obtained at several
different coverages and the temperature dependence studied at two
different coverages. The changes in the line width as a function of
the amount of adsorbed material are shown in Table 20. The widths
Fluorine Spectrum of Ilexafluoropropene Adsorbed on Aluminum Oxide.
of the CF3 and Fa lines decrease with increasing coverage but the
widths of the Fb and Fx lines remain essentially constant over the
entire range of coverage. The lack of change in the latter lines is
attributed to the presence of spin-spin interaction between the two
atoms. This interaction results in a doubling of the lines with a
separation of about 120 cycles per second.
Although it is not entirely conclusive it appears that the lines
tend to increase in width at very high coverages. This behavior, if
it is real, may indicate that adsorbate-adsorbate interactions
become important at these coverages. This could arise, for example,
if one CF3 group is surrounded by molecules which are tightly adsorbed
and another CF3 group is surrounded by molecules which are bonded very
weakly to the surface. The two groups would thus be in different
The observed relaxation times are listed in Table 21. These times
reflect the changes described above. The small values of the T1's
along with the small line widths indicate that the interactions
between the adsorbed material and the surface are weak. This is also
indicated in the ease in which the hexafluoropropene is desorbed.
The chemical shifts of the adsorbed material as well as the
neat liquid are shown in Table 22. The shift of the CF3 group is
essentially the same as that for the liquid and remains nearly constant
over the entire range of coverage. All the other lines are shifted
toward higher applied field by about ten parts per million on the
average. These values are probably no more accurate than t 1 part
per million. The observed shifts indicate that the three fluorine
atoms are influenced by the surface to a greater extent than the CF3
group is. To explain this it must be assumed that the double bond
itself is affected by the surface in such a manner that the electron
distributions about the fluorine atoms are changed. The distribution
in the CF3 group would be changed to a lesser extent since it is one
more bond removed. It is difficult to separate an effect of this
type, however, from the effect of the surface upon the fluorine atoms
just from being in proximity with them.
The temperature dependence of the relative chemical shifts (CF3
line taken as zero) is shown in Table 23. No major changes in the
shift values occur over the entire temperature range although it
appears that the lines may tend to approach each other as the temper-
ature is decreased. The lack of change of the shifts indicates that
regardless of the temperature the adsorbed molecules must remain in
very nearly the same positions relative to the surface. The fact that
these are relative shifts makes definite statements about the shifting
of the entire spectrum impossible.
The temperature dependence of the line widths is shown in Figures
22 and 23 for two different coverages. For the higher coverage
(Figure 22) all lines show a similar behavior except the Fb line. For
the lower coverage (Figure 23) both the CF3 and the Fb lines show a
behavior different from the other two lines. For low coverages it
appears that the widths are essentially temperature independent until
the temperature becomes low enough for dipolar broadening to occur.
For the higher coverage the changes in the line widths are somewhat
greater and no reasonable explanation of the behavior can be given.
-G0 "C -30
The Temperature Dependence of the Line Widths of 56.0 mg.
. F F
b 0 0 0 0 0
0 0 0
Figure 23. The Temperature Dependence
of the Line Widths of 93.6 mg.
It is quite possible that the changes may be due to changes in the
various contributions to the line width. The dominant contribution
may very well change with temperature. The most important point to
be gained from this study is that different parts of an adsorbed
molecule interact in specific ways with the adsorbing surface and
that it is possible to observe this by magnetic resonance techniques.
B. Dielectric Results
1. General Comments
The investigation of the dielectric properties of any substance
involves a study of the polarization produced in the substance by
the influence of the applied electric field. If the applied field
is static, the dielectric properties can be described in terms of the
dielectric constant but whenever alternating fields are used the
dielectric loss must also be considered. The dielectric loss of a
substance is obtained by measuring the component of current in phase
with the applied potential. This loss current is the result of the
absorption of energy from the field and consequently any mechanism
which results in an energy absorption contributes to the dielectric
2. Adsorption of water
The investigation of the dielectric properties of the systems
reported here consists of the measurements of the changes in capaci-
tance (which is proportional to the dielectric constant) and the
conductance or loss current. In these measured quantities are
contributions from the thorium oxide sample, the adsorbate, and
vapors surrounding the solid particles. In addition the properties
of the cell and its connections are involved but these remain essen-
tially constant from sample to sample and will not effect any inter-
pretation made in regards to differences between samples and/or
The capacitance changes with frequency are shown in Figure 24
for the empty cell, the "dry" sample, and a sample with adsorbed
water. The "dry" sample is the oxide sample which has been pumped
on for a period of time without the application of heat. It is very
probable that the oxide prepared from thorium hydroxide has a layer
of chemisorbed water on it even after pumping. It has been found in
preparing the n.m.r. samples that temperatures of 200* to 3000 C. are
required to completely remove all the water.
The capacitance of the empty cell is approximately twenty-four
picofarads and remains essentially constant over the entire frequency
range. A slight decrease is observed in the region around twenty
megacycles but this is attributed to the bridge circuit since it is
also present in the capacitance curve of a R-C network (see Table 25).
The cell shows no loss over the entire frequency range.
The capacitance of the "dry" sample also remains essentially
constant over the frequency range with the exception of the small
decrease near twenty megacycles. The data shown in Figure 24 is
for the oxide prepared from thorium oxalate (Sample OX-4). The
capacitance of the "dry" oxide sample prepared from thorium hydroxide
is higher by about twenty picofarads. This increase is attributed to
the water which remains on the sample after pumping.
The capacitance of the oxide sample with adsorbed water (Sample
OX-4) decreases with frequency over the range covered if the peculiar
o o 13.64 mg. H20/g.ThO2
o o dry sample
14) (pI Oqjo (p Op- aJ 00 ri 0 0--
Capacitance Changes with Frequency for the Cell, Dry Sample, and
Sample with Adsorbed Water.
behavior of the bridge at the higher frequencies is neglected. The
capacitance decrease becomes more obvious in terms of the dielectric
constant (C/Co). However, no additional information can be gained
from this transformation. This decrease is expected since the dielectric
constant decreases with frequency until at the high frequency limit
the optical dielectric constant is reached. The same trend is observed
at all water coverages on both samples (see Tables 26 and 27).
The behavior of the capacitance along with the frequency dependence
of the conductance indicates the presence of a dispersion region at a
frequency that is somewhere above thirty megacycles. The conductance
of the two samples is shown as a function of frequency in Figures 25
and 26. The conductance increases with frequency over the entire
frequency range and the increase becomes greater as the amount of
adsorbed water increases. The appreciable loss in this frequency
range indicates that the surface is producing some type of ordering
in the water layers. If the water remained as liquid water the region
of loss would appear at much higher frequencies.
A dispersion region has been found at low frequencies for the
alumina-water system39 but occurs in the kilocycle frequency range.
If the dispersion region is above thirty megacycles as suggested by
these results this would indicate that water is bound much tighter
to the alumina surface than to the thoria surface. This is accordant
with the values of the n.m.r. line widths of water on alumina and on
The conductance data of the two different kinds of oxides show
some differences. The conductance of water on the oxide sample
prepared from thorium hydroxide (X-B) (Figure 26) increases more
0o 9.43 mg./g.
ee 5.35 mg./g.
00 0.00 mg./g.
10 f 20 30
Conductance of Water on Thorium Oxide Prepared from Thorium Oxalate.
oa 1 5.76 mg./g.
D 0 5.28 mg./g.
o 0 0.00 mg./g.
0 10 f 20 30
Figure 26. Conductance of Water on Thorium Oxide Prepared from Thorium Hydroxide.
rapidly at the higher frequencies than it does for water on the oxide
prepared from thorium oxalate (OX-4) (Figure 25). This may indicate
that the region of maximum loss is at a lower frequency for the
former oxides. This would again be consistent with the n.m.r. data
which also indicates that water is in general bound tighter to these
Behavior of the tangent of the loss angle as a function of
frequency is shown in Table 31. The data for the oxide prepared from
thorium hydroxide roughly suggests the presence of a minimum in the
curve if the intrinsic perturbations of the bridge are ignored. The
data for the other sample is similar at the lower frequencies but
does not increase again at the higher frequencies. The decrease in
the loss tangent with frequency is the normal result of a mechanism
which produces a constant loss. This frequency dependence is shown
in equation IV, Appendix C.
The two sets of data can be reasonably interpreted in the
following manner. The dielectric loss results from at least two
mechanisms. In both oxides there is a component of conductance due
to an ohmic current which is attributed to ionic conductance. Evi-
dence of this component can be seen if the curves in Figures 25 and
26 are extrapolated to zero frequency. When this is done some
conductance is still present. This ionic component appears to be
greater in the oxide prepared from thorium oxalate and should be if
carbonate ions are present on this surface as postulated previously.
In addition to this ionic conductance it appears, from the large
increase in conductance with frequency, that dielectric loss due to
some other mechanism is present and that a region of dispersion is
being approached as the frequency is increased. This loss mechanism
is most probably due to dipole orientation. The dispersion region
appears to be nearer thirty megacycles for the oxide prepared from
the hydroxide than for the oxide prepared from thorium oxalate. This
is expected since water is bound tighter to the former oxides.
The changes in the capacitance with water coverage for several
frequencies shown in Figure 27 for the oxide sample prepared from
thorium hydroxide (X-B) and in Figure 28 for the sample prepared from
thorium oxalate (OX-4). In each case the amount of water which was
possible to adsorb prevents accurate deductions about its physical
state. When the amounts adsorbed on the dielectric samples are
compared to the amounts adsorbed on the n.m.r. samples a large differ-
ence is noted. This difference is attributed to the physical setup
of the dielectric cell in relation to the vacuum system. In order to
connect the cell directly to the bridge, thus eliminating long leads,
it was necessary that long glass plumbing be used from the vacuum
system to the cell. This resulted in a decrease in the rate of adsorp-
tion and for very high water coverages the time required to add an
increment became a matter of weeks.
The values of the capacitance of the "dry" samples indicate that
the oxide sample prepared from the hydroxide has some water remaining
on the surface even after it has been pumped on for a long period of
time. Few changes are observed in the slopes of the curves anywhere
in the range of available coverages. The only behavior which can be
considered a break occurs in the data in Figure 28 at very low cover-
ages. However, some indications of breaks are observed in the plots
4 8 12 16
Capacitance of Water on Thorium Oxide Prepared from
2 0.0 mc.
Capacitance of Water on Thorium Oxide Prepared from