PART I:
Synthesis and Characterization of Alkalides

and Electrides that Contain l5-Crown-5.

PART II:

Alkali Metal NMR Studies of Alkalides.

By

Mary L. Tinkha-

A DISSERTATION

Subaitted to Michigan State University
in partial fulfill-ent of the require-ents
for the degree of

DOCTOR OF PHILOSOPHY

Depart-eat of Cheaistry

1985

ABSTRACT
PART I: Synthesis and Characterization of Alkalides
and Blectrides that Contain 15—Crown—5.
PART II: Alkali Metal NMR Studies of Alkalides.
By

Mary L. Tinkham

Ten crystalline compounds that contain alkali metal
ions and l5-crown-5 (1505) were synthesized and analyzed.
These compounds were precipitated from solutions of alkali
metals and lS-crown-S in mixtures of dimethyl ether (MeZO)
and either diethyl ether or trimethylamine. Analyses were
in good agreement with the expected stoichiometries of
K+(l505)2-Na-, n+(1505)2-Rb', Rb+(15c5)2-e',
nb+(15c5)2-Na‘, ab+(15c5)2.ab' and Cs+(15c5)2-Na’ but
deviated somewhat from K+(1505)2-e—, K+(1505)2-K-,
Cs+(l5C5)2-K- and Cs+(15C5)2-Rb—.

Pressed powder D.C. conductivity measurements indicated
that these compounds are extrinsic semiconductors with
apparent band gaps of 0.5 to 1.7 eV. EPR spectra of
K+(1505)2-e- and Rb+(l5C5)2-e— showed a narrow line at g =
2.003 for each electride. Magnetic susceptibilities showed
that a high fraction of electrons are unpaired in each com-
pound. A ”jump" in the electronic xm of Rb+(l5C5)2-e_

between 120 and 140 K may indicate a phase transition.

Tinkham, Mary L.
Solid state alkali metal NMR was used to identify alka-
lides that were synthesized both by the author and by

others. Magic angle spinning (MAS) spectra were obtained

for 23Na, 39K, 87Rb and 133Cs NMR. Eleven compounds were

identified as sodides and each had the 23Na— NMR peak of in

the range of -56 to -63 ppm. The NMR signal of the

complexed cation was observed for the homonuclear sodides

Na+c222-Na' and Na+(lzc4)2-Na’. 393 MAS-NMR studies of

potassides provided the first NMR observation of K- chemi-
cal shifts at -105 to -115 ppm. Quadrupolar broadening

prevented the observation of a complexed cation signal

both for K+ and for Rb+. The Rb- anion had a frequency

dependent NMR peak at -187 to ~199 ppm. 1330s MAS-NMR was

used to identify three alkalides and each had a 133Cs NMR

peak at +24 to +29 ppm which corresponded to Cs+(1505)2.

Some samples had additional peaks which corresponded to

Cs+(1505)2-e- at +505 ppm and Cs— at -263 ppm. 39K and

87Rb solution NMR spectra of alkalides in MeZO are also

reported.

ACKNOWLEDGMENTS

The author wishes to express her sincere gratitude to
Professor James L. Dye for his guidance, encouragement and
support throughout this study. I would like to thank the
members of Dr. Dye’s research group, both past and present,
for their cooperation, suggestions and aid. In particular,
I wish to acknowledge Zexia Barnes, Steven Dawes, Margaret
Faber, Odette Fussé, Rui Huang, Stephan Jaenicke and Iraj
Behbahani for their help and moral support. Thanks are
also extended to the newer members of the group, Lauren
Hill, Jineun Kim, Mark Kuchenmeister and Joseph Skowyra. I
also wish to thank Dr. Dye and his group for the fishing
trips, the great Asilomar adventure and other group outings
which helped to make working together much more enjoyable.

Thanks go to the glass blowers Scott Bancroft, Manfred
Langer and Keki Mistry for their speedy and excellent ser-
vice. The secretaries, Naomi Hack, Carol link and
especially Margy Lynch, are also acknowledged for their
help. The cheerfulness and the helpful attitude of Dorothy
Boetteger is also greatly appreciated. I also wish to

thank the members of the electronics and machine shops.

-ii-

Special thanks go to Kermit Johnson for his help with
the solid state NMR work at Michigan State University.
Drs. Klaas Hallenga and Long Dinh Le are also gratefully
acknowledged. A special "thank you" goes to Dr. Pat Smith
at Dow Chemical Corporation at Midland, Michigan for his
initial aid and continuing help in magic angle spinning NMR
studies. Thanks go to Professor Eric Oldfield and his
group at the University of Illinois for their aid in the
87Rb high field NMR work.

Financial support is gratefully acknowledged from the
Department of Chemistry, Michigan State University and from
the National Science Foundation (grants DMR 79-21979 and
DMR 84-14154).

I would like to thank my family, in particular, my
sister Linda, for their faith and encouragement throughout
this study. Finally, I would like to thank my friends,

Ruby Freed, Bob Kean and especially N.R. Nirmala, for their

patience, understanding and support; for without their
friendships, this road would have been impossible to
travel.

-iii-

PREFACE

The work covered by this dissertation can be divided
into two areas -- each heavily dependent on the other. The
first is the synthesis and characterization of new in-
organic salts that have either an alkali metal anion (alka-
lide) or a trapped electron (electride) as the negative
species. In both cases the positive ion is an alkali metal
cation complexed by macrocyclic organic ligands. Species
identification is essential in the characterization of
alkalides and electrides to correctly distinguish between
pure compounds, mixtures and "junk". Historically, alka-
lides and electrides have been identified via stoichiometry
and optical spectroscopy. Although these techniques have
proved satisfactory for most cases, some ambiguity arose
for others.

The second portion of this work describes the applica-
tion of alkali metal NMR as an identification tool for
alkalides and electrides. This nuclear specific technique
was applied to all alkali metals except lithium. For the
most part, it was restricted to the solid state where the
"magic angle" sample spinning technique was employed,

although some solution experiments were also performed.

-iv—

Because of the major differences in the background
required to understand these two facets of this research,
the material is presented in two separate sections, Parts I
and II. Each section has historical, experimental and
results chapters. A final chapter is used to relate the
conclusions that can be drawn from these two sections and

to present suggestions for further work.

-v-

List of

List of
Part I:

Chapter
A.
B.
0.

Chapter

O’HWU

Chapter

HNUOW>

TABLE OF CONTENTS

Figures.

Tables.

Synthesis and Characterization of Alkalides

and Electrides that contain 15- -crown— 5.

One. Introduction. . . . . . . .

Metal Ammonia Solutions. . . .
Alkali Metals in Amines and Ethers. . .
Properties of Alkalides and Electrides.

Two: Experimental.
Reagent Preparation.
1. Metals.
2. lS-crown- 5.
3. Solvents. .
a. Dimethylether and trimethyl-
amine. . . . . . . .

b. Diethylether.
Synthesis Techniques.

1. Glassware preparation.

2. Improved synthesis technique.
Analysis. . . . . . . .
1. Hydrogen evolution.

2. pH titration.

3. Flame emission.

4. Proton NMR.

Optical Spectra. .
Pressed Powder Conductivity.
Electron Paramagnetic Resonance.
Magnetic Susceptibility.

Three: Results and Discussion.
Synthesis.

Analysis. . . . . .
Optical Spectra. . . . .

Pressed Powder Conductivity.
Electron Paramagnetic Resonance.
Magnetic Susceptibility.

-vi-

PAGE

.viii

xi

. 30
.30
30
30
32

.32
32
. 32
.33
33
38
40
42
43
. 43
.44
.47
.48
.48

. 50
.51
. 55
.60
.73
.82
.87

PART II:

Chapter

UGU>

Chapter

DO
Summary

List of

Results from other Sources.

1. Differential Scanning Calorimetry.
2. X- ray Diffraction. . . . . . .

3. EXAFS and XANES.

Alkali Metal NMR Studies of Alkalides.

One: Introduction.

Nuclear Properties of Alkali Metals.

NMR of Alkali Metal Anions in Solution.
Solid State NMR.

Identification of a Ceside and an Elec-
tride by 13303 MAS NMR.

Two: Experimental. . . . .
Sample Preparation. . . . . . . . . . .
1. Solution Samples.

2. Model Salts. .

3. Solid State Samples.
Instrumentation.

1. At Michigan State University.

2. At Dow Chemical Corporation.

3. At the NSF Regional NMR Center.

Three: Results and Discussion.
23Na NMR Spectra.

39K NMR Spectra. .

1. Solution studies.

2. MAS studies.

a'7Rb NMR Spectra.

1. Solution studies.

2. MAS studies.

13303 NMR Spectra.

and Suggestions for Future Work.

References.

—vii-

PAGE

. 94
.94
.95
.97

102

.102
103
.105
108

.115

.120
.120
.120
121
121
121

.122

123

.124

.125
.125
132

.133

135

.140

.141
.143
152
157

.164

FIGURE
1

2

10

ll

12

13

LIST OF FIGURES

Models for Species of Stoichiometry M‘.

Optical Spectra of Alkali Metals in
Ethylenediamine. . . . . . . .

Cryptand-2,2,2 and 18-crown-6.

Log K versus Cation Radius in Methanol
at 25'C.

Packing of Na’C222 and Na” (solid
circles) in the crystalline Solid
Na’0222-Na'. . . .

Born-Haber Cycle for Formation of
M‘L-N‘.

Vacuum Distillation Apparatus for
15-Crown-5 Purification.

Synthesis Apparatus: H Cell.

Analysis Scheme for Alkalides (A) and
Electrides (B). . . . . . . . .

Hydrogen Evolution Apparatus.

Optical Absorption Apparatus.

Optical Absorption Spectra of Thin Dry
Films Made From Solutions of

K’(15C5)2-K‘ (A) and K‘(15C5)2-e’

(B and C) in Methylamine. . .

Optical Absorption Spectrum of 8 Thin Dry

Film Made From a Solution of
K(15C5)1 3 in Methylamine.

«viii?

PAGE

.11

.14

.16

.19

20

.31

34

39
41

45

62

.63

FIGURE PAGE

14 Optical Absorption Spectra of Thin Dry
Films Made From Solutions of
Rb’(l505)2-Rb (A) and Rb’(l5C5)2-e
(B)in Methylamine. . . . . . . . . . .65

15 Optical Absorption Spectra of Thin Dry
Sodide Films Made From Solutions of
Cs'(1505)2-Na' (A), Rb’(15C5)2-Na‘ (B),
and K’(15C5)2-Na‘ (C) in Methylamine. . . . . . 66

16 Optical Absorption Spectrum of a
Thin Dry Film Made From a Solution
of Crystals with Stoichiometry

KRb(1505)2 in Methylamine. . . . . . . . . . . .69
17 Optical Absorption Spectra of Thin Dry

Films Made From a Solution of

Cs’(15C5)2-K‘ in Methylamine. . . . . . . . . . 72
18 -Ln (current) versus 1/T for Rb’(1565)2-e'. . . 79
19 EPR Spectrum of Rb*(1505)2-e' at -103°C. . . . .84
20 EPR Spectrum of K’(15C5)2-e' at -97'C. . . . . .85
21 l/xm versus Temperature for Rb‘(15C5)2-e'. . . .90
22 l/xm versus Temperature for K‘(1505)2-e‘. . . . 92
23 DSC Trace for Rb*(15C5)2-Na'. . . . . . . . . . 96
24 K-edge X-ray Absorption Spectrum of

Rb’(1505)2-Na'. . . . . . . . . . . . . . . . . 98
25 Diagram Illustrating the Motion of a

Typical Internuclear Vector, ri., when a

Solid is Rotated with an Angulag Velocity,
0r , about an Axis Inclined at Angle

Br to Ho . . . . . . . . . . . . . . . . . . . . 111
26 133Cs MAS-NMR Spectra of the Two Crystal-

line Compounds, Cs’(18C6)2-Cs’ and

Cs*(18C6)2-e' at 44.24 MHz. . . . . . . . . . .118
27 23Na NMR Spectra of Crystalline

Na’0222-Na‘ at 52.94 MHz. . . . . . . . . . . .127
28 23Na MAS- NMR Spectrum of Na*(1204)2-Na at

47. 61 MHz. . . . . . . . . . . . . . . . 129

FIGURE PAGE

29 39K NMR Spectrum of K’(15C5)2-K in

Dimethylether (0.06 M) at 220 K and

8. 40 MHz. . . . . . . . . . . . . . . . . . .134
30 39K NMR Spectra of Polycrystalline

K*(15C5)2-K' at 180 K. . . . . . . . . . . . . 136
31 87Rb NMR Spectrum of Rb’(1505)2-Rb' in

Dimethylether (0.068 M) at 235 K. . . . . . . .142
32 37Rb NMR Spectra of Polycrystalline

Rb’(15C5)2-Rb‘ at 65.4 MHz. . . . . . . . . . .150
33 133Cs MAS- -NMR Spectrum of an Impure

Sample of Cs’(15C5)2-K‘. . . . . . . . . . . . 153
34 133Cs MAS- -NMR Spectrum of Cs*(1505)2-K‘

- at 23. 62 MHz. . . . . . . . . . . . . . . . .155

TABLE

II

III

IV

VI

VII

VIII

IX

XI

XII

XIII

LIST OF TABLES

Alkali Metal Solubilities in Various
Solvents.

Crystallization Conditions and Relative
Stabilities of 15-Crown-5 Alkalides and
Electrides. . . . . . . . . . . .

Analysis Results of Alkalides and
Electrides. . . . . . . . . . . .

Results of Optical Absorption Studies of
Thin Dry Films. . . . . . . . .

Results of Pressed Powder DC Conductivity
Measurements. . . . . . . . . . .

Relative Areas of Rubidium K—Edge
Absorption Threshold Resonances.

Nuclear Properties of Alkali Metals.
Calculated Chemical Shifts for Gaseous
Alkali Metal Atoms and Anions and Observed
Chemical Shift of M‘ in Solution.

Results of 23Na MAS-NMR Studies.

Results of 39K MAS-NMR Studies.

Results of 87Rb MAS-NMR Studies at Three
Frequencies. . . . . . . .

Results of 87Rb NMR With and Without Proton
Decoupling.

Results of 133Cs MAS-NMR Studies.

-xi-

PAGE

10

54

57

68

77

.99

104

.109
128

.138

144

.149

.156

PART I:

Synthesis and Characterization of Alkalides
and Electrides that Contain l5-Crown-5.

PART I

CHAPTER 1 -- INTRODUCTION

The synthesis of crystalline alkalides and electrides
is a direct outgrowth of attempts by Dye and coworkers to
extend the study of alkali metal-ammonia solutions to other
solvents. This chapter will cover a brief history of the
development of alkalide and electride chemistry starting
with metals in liquid ammonia. The use of macrocyclic
complexing agents to increase the solvent range will be
discussed as well as the effect of these complexants on
general properties. Finally, the synthesis and charac—
terization of alkalides and electrides, prepared as powders
or thin films, as well as precipitated crystals, will be

reviewed.

A. Metal Ammonia Solutions

The unpublished notebooks of Sir Humphrey Davy show
that alkali metal-ammonia solutions were prepared as early
as 1808 [1]. These solutions were first described in the
open literature by Weyl in 1864 [2]. Since that time, many

papers have been written on the properties of metal-ammonia

-1-

-2-

solutions. The proceedings of six international
conferences [3-8] and a number of review articles [9-20]
have been published. Although metal-ammonia solutions have
been studied rather intensely, several questions remain
unanswered. The area of metal-ammonia chemistry is still a
very active field both on the fundamental and applied
level. For example, the high reducing power of such solu-
tions has been utilized to create new organic and organo-
metallic intermediates [21].

Discussions of metal-ammonia solutions are usually
restricted to one of three concentration ranges: dilute
(infinite dilution to ~l.0 mole percent metal (MPM)),
intermediate (1+8 MPM) and concentrated (>8 MPM). These
distinctions are drawn on an arbitrary basis since the
properties of the solutions change gradually with concen-
tration. In general, dilute solutions have electrolyte
behavior and concentrated solutions are metallic. There is
a gradual transition from electrolyte to non-metal to metal
as the metal concentration is increased. These transitions
are usually seen in the intermediate range since the non-
metal to metal transition is usually complete at a concen-
tration of 8 mole percent metal.

Liquid ammonia dissolves the following metals: Li, Na,
K, Rb, Cs, Ca, Sr, Ba, Yb and Eu. It is universally agreed
that the major species in dilute solutions are solvated
electrons and metal cations. Indeed, one may avoid most of

the cationic effects by introducing electrons by means of

-3-

ionizing radiation from an accelerator. The majority of
metal ammonia solution studies involve alkali metals. This
review will be restricted to such studies.

Metal-ammonia solutions may be studied over a wide
range of temperatures and concentrations. Even at the
normal boiling point of NH3, -33'C, very concentrated solu-
tions of the alkali metals can be made. Solubilities range
from 15 MPM for potassium to 65 MPM for cesium for
saturated solutions at -33'C. The saturation concentra-
tions change very little with temperature, with the
exception of that of (cesium, which is found to be
completely miscible with NH3 near 25'C and above. There
are miscibility gaps in the phase diagram for the other
alkali metals. It has been speculated that all alkali
metals would be completely miscible with liquid ammonia at
the melting point of the metal. However, most metal-
ammonia work is carried out at temperatures at or below the
normal boiling point of ammonia. Even with this as a high

temperature limit, the temperature range available is quite

large. The eutectic temperature of lithium in NH3 is
-184.6'C -- a freezing point depression of over 100
degrees! The concentration is surprisingly high at the

eutectic temperature (20 MPM).

The electrical conductivity of a saturated solution of
lithium in liquid ammonia exceeds that of liquid mercury at
room temperature. The other alkali metals in NH3 have con-

ductivities almost as high. In concentrated solutions, the

-4-

electrical conductivity increases approximately as the cube
of the metal content. The conductivity also increases in a
nearly linear fashion with temperature, except for the most
concentrated solutions of lithium and cesium. In these
solutions a slight decrease is observed as the temperature
increases.

A second similarity of concentrated metal ammonia solu-
tions and metals occurs in the plasma absorption edge
observed by optical reflectance studies. This plasma edge
occurs at approximately 10,000 cm.1 at the metal-nonmetal
transition but shifts to the blue as the concentration is
increased. The color (by reflectance) of concentrated
metal solutions is a polished bronze.

Magnetic data for M-NH3 solutions give rather confusing
results. The static magnetic susceptibility shows a weak
temperature dependence even in the most concentrated solu-
tions. The derivatives of the EPR signal show an asymmetry
of the low field maximum compared to the high field mini-
mum. The change in the A/B ratio is reminiscent of
metallic systems. A and B are the respective amplitudes of
the low and high field lobes of the first derivative EPR
spectra. These changes are observed at relatively low con-
centrations. The magnetic behavior of metal-ammonia solu-
tions, at any concentration, has not been explained in a
satisfactory manner, although Thompson has put forth

several models [19].

-5-

Dilute solutions (concentrations (1 MPM) have a charac-
teristic blue color at concentrations of 10.6 molar or
more. The optical maximum for the solvated electron lies
in the near infrared region at 6900 cm—1, but has an asym-
metric high energy tail into the visible. This line is
completely unaffected by the type of solute at low concen-
trations. The absorption band is also observed when elec-
trons are produced by ionizing radiation or when ND3 is
substituted for NH3. The spectrum is affected by
concentration and temperature, and shifts to lower energies
as either is increased. Because of the concentration de-
pendence, pulse radiolysis is the preferred method for
optical studies of isolated solvated electrons in liquid
ammonia.

The specific conductance of M-NH3 solutions increases
with the metal concentration. In very dilute solutions
((10.2 M), the equivalent conductance decreases as expected
for a solution that can form ion pairs. One can infer that
the cation and electron are truly separated only at concenr
trations below about 10.4 M. Other properties such as sus-
ceptibility, activity coefficient and density are nearly
the same from one metal to another when compared at the
same concentrations. The EPR spectrum of a very dilute
solution features a very narrow line with a g—value near
the free electron value of 2.0023. As the concentration is

increased some interaction between the cation and solvated

electron can be observed by the paramagnetic shift in the

-6-

alkali metal NMR. Although several models have been
postulated for the species in dilute M-NR3 solutions, only
one is in reasonable agreement with the various properties
mentioned above. This model proposes the formation of
ionic aggregates which retain the basic characteristics of

M+ and e; 1 . Species with stoichiometries such as M, M-

o v

and M2 could form without changing the optical spectrum or
molar volume of the solution. This model would then allow
for the spin pairing observed in the static susceptibility
and EPR measurements as the concentration is increased.
Figure 1 gives three models that have been proposed for
species of stoichiometry M-. The first is an example of
the ion aggregate model and is EPR active.

Perhaps the most interesting properties of metal-
ammonia solutions are found in the intermediate concentra-
tion range. In this range, a transition from non-metallic
to metallic behavior is observed as the metal concentration
is increased. The phase diagram for sodium ammonia solu-
tions shows a liquid-liquid phase separation at 4 MPM
sodium at temperatures below 230 K. The more concentrated,
less dense metallic phase floats on top of the dark blue,
less concentrated electrolytic phase. As the temperature
is increased, these phases become miscible. This phase
separation has been compared to the liquid vapor phase
separation of a metal near its critical point. A review
article by Edwards and Sienko [22] discusses the results of

supercritical alkali metal studies and the onset of metal-

H
’ \ I I
I, \ h; \‘
gum-I ° . "" v ‘
x w" .4/ \z. ,I

‘ mnarnwniziuri
000--~\ H ’----
I H \\ H ' H \,' ‘/
@ H u H . \N, ’ ‘
I \N/ \ é \“
I .
I H‘ H /o ‘
' H ’ /-H\ H-\N'- M’ III-(H .2 ,-——
-N M I '
|H PK I / H
I H/ H, H . ‘ ’I

\ 7 \ ,’
‘~ A 69' (in /‘~--" \

I
EXPANDED ORBITAL ION PAIR WITH OIELECTRON M".

Figure 1. Models for Species of Stoichiometry
M-O

-3-

lization. For a very general but thorough discussion on
the metal to non-metal transition, the reader is referred
to Thompson [19].

As a final note on metal-ammonia solutions,. the com—
pounds isolated from concentrated solutions will be
discussed. In general, these compounds retain the golden
color of the concentrated solution. Li(NH3)4 is the only
truly crystalline alkali metal-ammonia compound that has
been isolated. The others are of the general formula,
M(NH3)6, and contain Ca, Sr, Ba, Eu or Yb. Structures of
the hexaamines show that the metal atom is octahedrally
situated among the six nitrogens. Powder neutron diffrac-
tion spectra show that not all N-D distances are equal in
M(ND3)6 and the pseudo trigonal axis of rotation of the ND3

is 13' from being colinear with the M-N bond. Li(NH has

3’4
at least two crystalline forms with a solid-state phase
transition at 82 K. Both structures are body-centered
cubic. Magnetic susceptibility measurements also confirm a
structural change at 82 K and suggest magnetic ordering at

low temperatures. The 82 K phase transition was not

present 1n L1(ND3)4.
B. Alkali Metals in Amines and Ethers
The range of solvents available for nonaqueous metal

solutions is limited by low solubilities and by solvent

reduction [5-8,19,20]. Only with lithium in methylamine

-9-

can metal concentrations be obtained that are on the order
of those obtained in liquid ammonia. For other solvents,
concentrations range from millimolar to undetectable. As
an example, the solubility of sodium in ethylenediamine is

only 2.4 x 10—3

M at 25'C. Table I gives the solubilities
of alkali metals in various solvents [23].

The characteristic blue color of dilute metal-ammonia
solutions is also observed in metal-amine solutions with
concentrations of 10-5 M or greater. The optical absorp-
tion spectra of these solutions feature either one or two
maxima. One of these bands may be attributed to the
solvated electron, as in liquid ammonia. The second is
shifted to higher energy, and is strongly metal dependent
but only moderately dependent on the solvent. The species
responsible for the second absorption band has been identi-
fied as an alkali metal anion. This species is not simply
one of those shown in Figure 1 but rather a true anion with
no solvent molecules between the nucleus and the valence
electrons. The optical spectra of alkali metals in
ethylenediamine solutions are given in Figure 2 [24].
Except for that of lithium, these spectra all show the
metal dependent peak attributed to M-. In the case of
lithium, only a solvated electron peak at 7810 cm"1 is
observed. The spectra of K, Rb and Cs show the maximum due

to M- and a shoulder due to e- Sodium solutions show

solv'

only the peak of Na—.

-10-

TABLE I. Alkali Metal Solubilities in Various Solvents.

 

 

 

 

Metal Solvent Solubility Temperature
(M11. ('0)
Li ethlenediamine 0.29 25
methylamine xLi = 0.15 --
Na ethylenediamine 2.4 x 10-3 25
methylamine >2 x 10‘4 -50
K ethylenediamine 1.04 x 10-2 25
tetrahydrofuran 5 x 10.6 24
Rb ethylenediamine 1.31 x 10’2 25
Cs ethylenediamine 0.054 25

methylamine >2 x 10‘ -5o

-11-

.mcwfimwnmcmahnum :H macaw: Hamxa< mo muuommm HMOfiumo .N shaman

n.o-. A - .Eu-m

 

 

 

 

 

 

 

ON 0. 0. . 0
_ I q
.. and
.664
s’._ .
T - H.0-
00 0.0 . . N0 0 .0 0.. 0N-. 0. ._ W-

E:- x

-12-

Further evidence of the existence of M— is found in the
EPR spectra of solutions which contain only the metal
dependent optical band. In contrast to the strongly para-
magnetic ammonia solutions, these solutions were dia-
magnetic with very weak EPR signals. The filled ns2
orbital of M- would satisfy the conditions necessary to
give these spectra.

Other evidence is found in the oscillator strength,
electrical conductivity and solubility studies. The oscil-
lator strength has been measured for Na in ethylenediamine
(EDA) and was found to be twice that of e;°1v. The conduc-
tance of these solutions decrease with increasing concen-
tration as expected for electrolytes. Although this
behavior indicates ion pairing as with normal salts, the
extent is less than expected. This would occur for ions
with large closest approach values. The distance for
sodium in methylamine (MA) exceeds 6 A, an indication that
Na- is a large ion. Solubility studies show that sodium

metal will not dissolve in 1,2-methoxyethane alone.

However, the equimolar Na-K alloy does dissolve. This can

readily be explained by the formation of K+ and Na-. The
most conclusive evidence for M- is found in the alkali
metal NMR spectra of these solutions. The NMR spectra

feature a peak with a chemical shift that corresponds well
to the theoretical value of M-. Further background and the
results of alkali metal NMR studies are deferred until Part

II of this dissertation.

-13-

Solvation of alkali metals in amines and ethers may be

described by the following equilibria:

2M(8) <____ M + M (1)
M_ _-—-> M + e- (2)
<---- solv solv
____) + _
solv <-—-- M + esolv (3)
where Msolv is the "monomer" and probably represents a
contact ion pair between M+ and e;°1v [25]. This species

has a characteristic EPR signal with alkali metal hyperfine
coupling, and is observed to be present only in very low
concentrations. To increase the solubilities of alkali
metals, organic complexing agents were used. These

ligands, L, introduce a fourth equilibrium

M + nL M L (4)

which drives equations (1) and (3) to the right. Two main
classes of complexants have been used for this purpose,
crown ethers and cryptands, originally developed by
Pedersen [26] and Lehn and coworkers [27], respectively.
Figure 3 features a representative molecule from each
class. Most commonly, crown ethers are cyclic ethers with
alternating -CH2-CHZ-groups and oxygens. Their abbreviated

names reflect the ring size and number of oxygens. For

-14..

2,2,2 C rypto nd

(C222)

(WW
0 O
(0‘ .3
pov’
2
18-C row n- 6
(18-C-6)

Figure 3. Cryptand-2,2,2 and
l8-Crown-6.

-15-

example, 18-crown-6 (18C6) is an eighteen

containing 6 oxygen atoms.

alkyl or benzo groups attached to the ring.

are three dimensional complexants that consist

ether linkages between two tertiary nitrogens.

abbreviation for cryptands is C-m,n,o where m,

numerical representations of the number of

in each strand between the two nitrogens.

have been used not only to increase alkali metal

but also as solvents for alkali metals. Sodium

to dissolve in 12-crown-4 to form stable, dark

tions. Potassium, rubidium and cesium form

solutions in 12-crown-4 [28]. Edwards and

studied Na— and Rb_ in these two component metal
[29].
ethers

Crown

The hole

The crown ethers may also

diameter

member ring

have

The cryptands

of ethyl

The generic
n and o are
-CRZCH20-groups

Crown ethers

solubility
was found
blue solu-
metastable

coworkers

solutions

and cryptands have ion selectivity based

of 18-

1argely on
crown-6 is
18-crown-6
studies of
distortion
The

cavity and

The complexation constant as a function of cavity

graphically represented in Figure 4 [30].

larger

degree of cation specificity than the

larger

complexation

their cavity sizes.
large enough to accommodate either Na+ or K+ but

shows a preference for K+. X-ray diffraction

the complexes of alkali halides show a ring

and loss of symmetry as one goes from K+ to Na+.

cations, Rb+ and Cs+, are unable to enter the

can give metal:complexant ratios other than 1:1.

size is

Because of their

energies, cryptands show a higher

crown ethers. The

-16-

 

 

 

2107
.0
0 1006
.0
20 m 1505
.A
1.0'I I I I I I
Na" Ag*- K" Rb" Cs+

 

 

I

1.00 1'20 1.110 1.00 1.00
Cation Radius (A)

‘Figure 4. Log K versus Cation Radius in Methanol
at 25°C. -

-17-

formation constants (log K) for cryptands range from 1.8 to
almost 10 in methanol. The maximum for crown ether in
methanol is slightly more than 6 for K+1806. Since the
cryptands have 3-dimensiona1 cavities they tend to form
only 1:1 complexes. However, in some cases, more than one
ion can occupy a single cavity.

The optical absorption spectra of metal solutions with
crown ethers or cryptands present are temperature depen-
dent. The absorption maxima shift to higher energies as
the temperature is decreased as for dilute metal-ammonia
solutions. The cryptand or crown ether concentration also
influences the observed spectra. In solvents where the
metal is insoluble, addition of complexant gave M- and/or

e optical bands. When the concentration of complexant

solv

solv band

exceeded the amount of dissolved metal, the e
increased. The conditions needed for this observation are
described by equilibria 1-4.

The use of complexing agents in nonaqueous metal solu-
tions not only provided accessibility to new solvent
systems but also opened up the possibility of the formation
of new compounds. When a solution of a pure metal in
ammonia or other solvent is evaporated, the electrons
recombine with the cation to form metal films or powders.
The presence of a complexing agent inhibits this reaction
because of the energy required to break the cation-ligand

bonds. The synthesis and characterization of compounds of

this type are discussed in the next subsection.

-13-

C. Properties of Alkalides and Electrides

A golden precipitate forms from a solution of sodium
and cryptand-2,2,2 (0222) in ethylamine as it is cooled
below -l5'C [31,32]. The precipitate, which was first
observed in 1974, was the first crystalline compound which
contained an alkali metal anion and has the formula
Na+C222-Na_. The parent compound has been fully charac-
terized; even the crystal structure has been determined.
The packing of Na+C222 and Na- is represented in Figure 5
[32]. The sodium cation is located in the center of the
cryptand, which has its three strands twisted to give an
antiprismatic arrangement and to allow a closer fit to the
sodium cation. These "expanded" cations are closest-packed
in ABCABC repeating layers. The sodium anions are located
in the pseudo-octahedral holes .of the lattice. The
van der Waals contact distances in the crystal suggest that
the radius of Na- is about 2.6 A. Since Na+C222-Na- forms
isothermally at O'C from sodium and 0222 in solution, the
Gibbs free energy must be negative for the following

reaction:

+ 0222 --—-> Na+C222-Na

2N8 (3) (8) (5)

(6)
Estimates of the enthalpy, free energy and entropy were
made based upon a modified Born-Haber cycle [33]. The

cycle is diagrammed in Figure 6 and initial calculations

-”7\
I

 

 

~19-

 

-20-

 

 

 

 

 

 

AH'TOTAI. -
LI.) “ Mm * NIsI " MiL‘ N m
(AH 80b).
(AI-I «ML I (AHNNN AHLATTICI
Mm
A"'Iomz.
- I Aflu. -
e (9) NIq) ’ N10)
I " “
AHIcmMphm
LIo) * M'Ig) 0" M“-

 

Figure 6. Born-Haber Cycle for Formation of M+L-N-.

-21-

suggested that Na+0222oNa— is unstable to disproportiona-
tion to sodium metal and cryptand by 22 kJ mol_1, a measure
of the error in these estimates, since the compound is
stable. Schindewolf and coworkers developed a sodium elec-
trode [34-36] which permitted the cell potential of
reaction (5) to be measured directly. It was found that
the estimated entropy value was 100 J .01-1 K“1 too low,
probably due to errors in the estimation of the solvation
or crystal entropy [37]. The observed value of A6298 was
-7.1 kJ mol-l.

Estimates of thermodynamic stabilities were made for
several alkalides which utilize cryptands [23] or crown
ethers [38]. Cation radii were estimated by Lehn’s method
using the ligand thickness so that the Kapustinskii equa-
tion could be used to calculate the lattice energies.
Because of errors in the method, these stability estimates
serve only as guidelines for synthesis rather than as hard
and fast predictions. For example, Li+(1204)2-Cs- is cal-
culated to be twice as stable as Cs+(12C4)2-Li-. There-
fore, if a lithide is desired, a different complexing agent
and cation should be chosen.

Subsequent alkalides were prepared by one of three
different methods [25,39]. The first is the gradual
cooling with or without a solvent composition change to
precipitate crystals [39-44]. The stoichiometry of the

resultant crystals was not necessarily the same as that of

the solution. The second method is slow solvent evapora-

-22-

tion to yield powders and films with the same stoichiometry
as the solution [45]. The third method is direct vapor
deposition of the metal and complexant onto a substrate
[49,50]. The latter method requires either a solid state
reaction or a gas phase reaction near the surface of the
substrate. By increasing the complexant to metal ratio,
electrides may be made by any of these methods.

Films for optical absorption spectra of alkalides and
electrides were first made by flash evaporation of solu-
tions of metals and complexant in stoichiometric amounts.
Later, crystalline compounds were introduced directly into
the optical apparatus and redissolved [51-53]. Films were
made by flash evaporation as before. There exists the pos-
sibility that the films do not contain the same species as
the crystals that are dissolved to produce the solution.
The maxima of the resultant spectra were compared to the
spectra of pure alkali metals in solution. Films from
Na+C222-Na- solutions yield spectra featuring an absorption

1

band at 15,000 cm- with a pronounced shoulder at 18,900

cm-1. A small but distinct peak at 24,500 cm.1 was also
observed. Decomposition of the films resulted in the loss
of all spectral features. Films of potassium, rubidium or

cesium with C222 were prepared and are considerably less
stable. The spectrum of each of these films consisted of a
single asymmetric peak and showed neither a high energy
shoulder nor an additional high energy peak. The absorp-

tion maxima were at 11,900, 11,600, and 10,500 em‘1 for x‘,

-23-

Rb-, and Cs-, respectively. These locations correspond

well to the M_ bands in ethylenediamine (EDA) at 12,000,
1. Na- in EDA solutions has an absorp-

tion band at 15,400 cm-1 but no other band or shoulder.

11,200 and 9,800 cm-

The similarity between the homonuclear alkalide films and
the solution spectra lead to the utilization of optical
spectra as an identification tool for heteronuclear
alkalides.

Films from solutions of K-Na-1806 in methylamine gave a
broad absorption band at 13,300 cm.1 with a shoulder at

9,000 e-‘1

[52]. The main band lies intermediate to the
expected locations of Na- and K-. The crystals were found
to have up to 18X solvent content. This system was resyn-
thesized by the author to produce solvent-free crystals
with 1:1:1 stoichiometries [51]. Films made from solutions
of the pure crystals had a narrower optical maximum at
14,000 cm-1 with no low energy shoulder. Although this
compound was presumed to be a sodide, a clear assignment
could not be made. The position of the maximum required a
1000 cm-1 red shift of Na- from the homonuclear 0222 film
or an 1100 cm“1 blue shift for K-. The maximum position is
even further shifted from the 16,700 cm_1 peak observed for
films made from solutions of sodium 18-crown-6 in a 2:1
mole ratio. The compound was presumed to be a sodide
because of the selectivity of lB-crown-6 for potassium

rather than sodium and the greater stability of Na-

compared with K—.

-24-

Other heteronuclear alkalides also have changes in the
position of the absorption maxima from those in solution.
Cs+(1806)2-Na- has a sharp peak at 14,600 es'l while
Rb+1806-Na— has a maximum at 13,800 cm-l. These
compounds were also thought to be sodides. In the case of

rubidium and cesium with 18-crown-6, two peaks are

observed. One covers a 1000 CI_1 range with its maximum at

11,500 cm—l. This band is asymmetric with the center at
12,000 cm_1. The second peak occurs at 9000 cm-1 and was
assigned to trapped electrons. Neither Rb+ nor 0s+ can

enter the crown ether cavity completely and the anion
assignment cannot be made without ambiguity. Shifts in the
absorption maxima may be due to cation-anion interaction.
Crown ether complexes can have axial symmetry, which would
allow exposure of the cation above and below the ring.
Cations complexed with 2 crown molecules should be more
shielded and may disturb the anion environment less. That
this is not necessarily the case is seen with 12-crown-4
sandwich sodides. Dry films of K+(1204)2-Na_ showed two

maxima at 13,100 and 17,000 cs’1

[38]. Films prepared from
Na+(1204)2-Na- also had two absorption peaks but these were
slightly shifted from the heteronuclear peaks at 13,500 and

17,500 cm,"1

[54]. The origin of the higher energy peak is
unknown, but it is reminiscent of the shoulder seen with
Na+0222oNa_.

The cation-anion interaction should be less for cryp-

tated cations than for those salts with the cations

-25-

complexed by crown ethers. Since in the former case the
cation is complexed by a three dimensional complexant, one
would expect less interaction of the cation with its
surroundings and smaller shifts in the absorption maximum
location of the anion for these salts. However, this is
not the case as seen in the M+0222-Na_ compounds. The peak
location for K+0222-Na— is red-shifted 300 cm"1 from that
of Na+0222-Na- while in Rb+0222oNa_ the absorption is
shifted to 14,000 em’l.

The first optical absorption spectrum of an electride
was that of a film made from a solution of potassium and
cryptand-222. The maximum at 7400 cm"1 was in the near
infrared, as expected for a trapped electron. As the
potassium/cryptand mole ratio increased from 1 to 2, a
second band corresponding to K_ appeared. At a mole ratio
of 2 only the K- peak was observed [55]. Since that time a
number of electride films have been prepared. The observed
spectra fall into one of two categories -- localized or
"metallic" behavior. The localized electrides have absorp-
tion maxima which correspond to traps of at least 0.4 to
0.6 eV depth. These absorption bands tend to be broad,
indicative either that electron-electron interactions may
be important or that the excited state is delocalized.
These electrons may also undergo fast exchange with
adjacent trapped electrons but not on the optical time

scale. The "metallic" electrides give absorption spectra

similar to those of concentrated metal-ammonia solutions.

-25-

The absorbence remains high in the infrared, out to at
least 3000 cm-1. This suggests electron delocalization or
at least shallow traps. This behavior has been observed
from certain K+0222-e— films and films from solutions with
stoichiometries of Li20211 and 032 1806. It has not been
possible yet to verify the "metallic" behavior of crystal-
line electrides by D.C. conductivity measurements because
of their instability. However, powders of these stoichio-
metries have large microwave power absorptions that would
be expected for conducting samples.

The golden appearance of Na+0222-Na- suggested that it
may be metallic. However, a plot of log resistance versus
1/T had a slope that corresponded to a semiconductor with a
band gap of 2.4 to 2.5 eV. Extrapolation to infinite tem-

perature gave a specific conductivity greater than 106

ohm"1 cm-l, an indication that Na+0222-Na_ is an intrinsic
semiconductor. D.C. conductivity measurements of pressed
powders indicate that all alkalides behave as semi-
conductors. Reported band gaps range from 0.89 to 2.6 eV.
However, most alkalides have much lower specific conduc-

2 to 10 ohm.1 c111-1

tivities of 10— at infinite temperature
than expected for intrinsic semiconductors. The observed
electrical conductivity is probably due to defect electrons
which are trapped at energies much closer to the conduction
band rather than from the valence electrons in M_, thereby

giving extrinsic semiconductor behavior. Electrides,

prepared as powders by solvent evaporation, also behave as

-27-

semiconductors with band gaps of 0.5 to 0.8 eV. The con—
ductivity of electride samples may also arise from "defect
electrons" rather than from the bulk of trapped electrons
in the sample. The first crystalline electride,
Cs+(1806)2-e-, had a specific conductance intercept at
infinite temperature of ~102 ohm.1 cm”1 with a band gap of
0.9 i 0.1 eV. Therefore, Cs+(1806)2-e- has characteristics
that lie between extrinsic and intrinsic semiconductivity
[42,53].

The ns2 electronic configuration of alkalides predicts
that they are EPR inactive. Samples of Na+0222-Na_ are
indeed diamagnetic and have no appreciable EPR signal.
Some samples of K+1806-Na- [51] and Rb+1806-Na- [53] showed
comparable strong EPR signals with the presence of fine
structure. This structure is due to the hyperfine
splitting of the signal by the interaction of trapped elec-
trons with the K+ or Rb+ nucleus. The presence of hyper-
fine splitting is preparation-dependent and is not evident
in all samples. In most cases the EPR spectrum of an alka-
lide sample consists of a single, narrow line near the free
electron g-value. Sometimes a second broader band is
superimposed on the first, indicating the presence of two
or more trapping sites. In general, the EPR spectra of
alkalides appear to arise from trapped electrons rather
than from intrinsic structural features of the crystals.

In contrast to alkalides, most electrides have very

strong EPR signals. Powders of Li+0211-e— had microwave

-28-

power absorption similar to those of powdered metals [56].
The 0s+(1806)2-e_ EPR spectra featured a single Dysonian
line at g = 2.0023. The line became more symmetrical at
lower temperatures. The width of the line was only 0.48 G,
which indicated well-defined local trapping of the
electrons [53]. Samples of crystalline K+0222-e- and
Rb+0222-e- gave such strong EPR signals that only a very
few crystals would saturate the detector of the spectro-
meter [50].

Magnetic susceptibility measurements of alkalide
crystals also show that alkalides are diamagnetic. Some
paramagnetism of defect electrons is usually evident at low
temperatures. The electrides shows a wide range of
magnetic properties. Cs+(1505)2-e- has low temperature
deviations from the Curie—Weiss law that allow it to be
classified as an antiferromagnetic material [57].
Crystalline K+0222-e- is best described as a Pauli para-
magnetic compound. At low temperature there is some
adherence to the Curie law but the susceptibility quickly
goes to a metal-like temperature-independent value as the
temperature is increased. Rb+0222-e- behaves very much
like the potassium electride but the Curie-Weiss behavior
extends over a larger temperature range. The electrical
conductivities of these two salts were too high to measure
by the available methods, a further indication that
K+0222-e- and Rb+0222-e- are either metallic or near the

metal-nonmetal borderline [50].

-29-

The object of Part I is to describe the synthesis and
characterization of new alkalide and electride salts.
Characterization techniques include elemental analysis and
optical absorption spectroscopy as identification tools and
D.C. conductivity measurements for all compounds. Magnetic
susceptibility measurements and EPR spectra were recorded
for those compounds thought to be electrides. The results
of these studies for 10 new alkali metal—crown ether

systems are reported.

PART I

CHAPTER TNO -- EXPERIMENTAL

A. Reagent Preparation

1. .Mbtals - Sodium, potassium and rubidium under argon
were purchased from Alpha Ventron Products in 5 gram break-
seal ampules. Total purity was 99.95% for each metal.
Cesium metal was obtained as a gift from Dow Chemical Corp.
and had been previously transferred into ~5 gram sealed
glass ampules. Each metal was further distributed to
smaller tubing following the procedure outlined by less
[52]. Desired quantities of metal were obtained by
isolating calculated lengths of metal inside of tubing that
had premeasured inner diameters.

2. 15-Crown-5 - (1505 or IUPAC 1,4,7,10,13-pentaoxacyclo-
pentadecane) was distilled at 85-90'0 under dynamic vacuum.
The apparatus for the distillation is given in Figure 7.
The crown ether was pipetted into the bottom of the
apparatus without further purification. The apparatus was
evacuated to ~2 x 10-5 torr before submerging the lower
half into a preheated oil bath at the desired temperature.

The cold finger was cooled by flowing dry nitrogen gas

-30..

-31_
To Vacuum Manifold

/ Cold N2 Gas

‘I

 

 

 

 

 

.\

Trap :0
protect
Vacuum
Manifold

 

 

   

 

 

t:

Figure 7.

 

 

 

 

cu
I p

0" Crow n

 

Vacuum Distillation Apparatus for lS-Crown-S
Purification.

-32-

at 0’0. The temperature of the nitrogen should be warm
enough to avoid condensation of any unwanted impurities
(such as H20) onto the cold finger. When the crown ether
had completely distilled into the bucket, the apparatus was
closed, removed from the oil bath and allowed to achieve
thermal equilibrium. The entire apparatus was brought into
an inert atmosphere box and the crown was transferred with
a disposable pipet from the bucket to a storage bottle.
3. Solvents - a. Dimethyl ether (MeZO; Matheson) and tri-
methylamine (TMA; Matheson) were first stirred over calcium
hydride at -20'C for several hours. The solvent was frozen
in liquid nitrogen, pumped to <4 x 10-5 torr, thawed and
stored at -78'C overnight. The solvent was then
transferred to flasks containing an excess of NaK3 alloy
and about 0.5 gram diphenyl ketone, frozen, pumped and
stored overnight, before transferring to another NaK3
bottle. After the solution maintained the characteristic
blue color of the benzophenone dianion-ketyl mixture for 2
consecutive cycles, the solvent was transferred to heavy
walled glass storage bottles or stainless steel storage
tanks. The solvent was then subjected to several freeze-
pump-thaw cycles until the initial pressure at pumping was
less than 1 x 10_4 torr.

b. Diethyl ether (ethyl ether, anhydrous, Mallinck-
rodt, Inc.) was treated in the same manner as dimethyl

ether and trimethylamine. Final storage was over NaK3

alloy without diphenylketone.

-33-

B. Synthesis Techniques

1. Glassware - All glassware was first rinsed with a hydro-
fluoric acid cleanser solution (53 HF (28M), 33% HNO3 (16M)
and 628 deionized water, by volume) then immediately rinsed
six times with deionized water. The glassware was filled
with freshly prepared aqua regia (3 H01:l HN03, by volume)
and left overnight. The aqua regia was discarded and
glassware was rinsed six times with deionized water
followed by six rinses with conductance water (house
deionized water which had been further deionized with a
Crystalats Deeminizer and distilled through a high reflux
ratio column to less than 1 ppm impurity) and dried in a
450°F oven overnight.

2. Improved synthesis technique - The general synthesis
scheme for the preparation of alkalides and electrides has
been fully described [51]. This section is used to point
out the improvements made over this general scheme by the
use of a simpler apparatus called an H-cell (Ace Glass) and
a new solvent system. The H-cell, shown in Figure 8, was
modified from the factory issue to better accommodate the
low temperature synthesis of alkalides and electrides.
First, the angle between chambers A and g was increased
from 90' to about 120' with respect to the horizontal con-
nection. Second, the distance between the stopcock and

horizontal was increased by 7-8 cm to allow complete

 

 

 

 

 

 

 

Figure 8.

Synthesis Apparatus:

H Cell.

-35-

cooling of the frit without cooling the stopcocks. Lastly,
a 9 mm OD side arm was added to one of the chambers for
metal introduction. The size of the H-cell is its main
advantage. H-cells are small enough to allow entry into
the inert atmosphere box so that the complexants are never
exposed to atmosphere but can be loaded into the synthesis
apparatus under the dry helium atmosphere. Syntheses
involving lithium are also greatly simplified by direct
loading of the lithium in an inert atmosphere box [50].

The 15-crown-5 was loaded through the stopcock into
chamber A with a disposable pipet. The amount of crown
ether was determined by mass difference of the storage
vial. A premeasured amount of metal, sealed in a 3 or 4 mm
OD scribed glass tubing was broken and placed in the metal
sidearm. The H-cell was closed and sidearm capped with a
3/9” Ultra-torr Union (Cajon) and a closed end glass cap.
The H-cell could then be removed from the dry box and
evacuated to 2 x 10-5 torr. The Ultra-torr union and glass
cap were removed by making a vacuum seal on the sidearm. A
metal mirror was made in chamber B by gently heating the
remaining sidearm until all metal had distilled into B. A
final seal was made at the constriction and the H-cell
allowed to thermally equilibrate.

The use of dimethyl ether instead of methylamine
improved the stability of the alkalide and electride solu—
tions. Working temperatures were increased from -40'0 with

MA to -10' to 0'0 with MeZO. This allowed for faster dis-

-35-
solution of the alkali metal and generally shorter synthe-
sis times. Chamber A of the H-cell was cooled to <-30'C
and ~10 ml dimethylether was distilled onto the now frozen
15-crown-5. The apparatus was closed and warmed to -10'C
in a dry ice-isopropanol bath. When the crown was
completely thawed and dissolved, the solution was poured
onto the metal mirror in A. Me20 was distilled from A to A
and then poured back into A several times to insure
complete solution of the crown. The metal mirror was
dissolved by gentle agitation of the H-cell. Complete dis-
solution took 0.5 to 3 hours at -10'C. When no trace of
the metal mirror could be seen, the entire solution was
poured into chamber A. The total volume of the solution
was reduced to a small viscous puddle (~l-3 ml) by distil-
ling the excess Mezo to a waste bottle. The cell was
cooled to less than -30'C. If there was any precipitation
of material, more MeZO was added to yield a homogeneous
solution.

Diethyl ether (DEE) or trimethylamine (TMA) were added
as co-solvents for crystallization. It is important to
keep the DME solution temperature well below the DME
boiling point of —27'C to avoid reverse distillation. The
co-solvent was distilled, a small portion at a time, with
the H-cell being shaken between additions to insure proper
mixing. Total volume of co-solvent did not exceed 7 ml.

The H~ce11 was then packed in dry ice for 3 to 12 hours.

-37-

The H-cell was then checked for the formation of
crystals. If no crystals were present, the solution was
further concentrated by removal of the solvent to a waste
bottle. The H-cell would again be stored at -78'C for
several hours. After the formation of crystals had
stopped, the mother liquor was poured into A and then
distilled into a waste bottle.

The wash solvent for the crystals was the same as the
co-solvent. Fresh solvent was distilled onto the crystals,
the cell shaken to insure mixing and the solvent decanted
to chamber A. The solvent was then distilled from A to A
and decanted to A to remove any soluble impurities from the
crystals. This distill-decant procedure was repeated
several times until the solvent became colorless. Finally
the solvent was distilled from A to a waste bottle until
the H-cell was solvent free. The crystals were further
dried by dynamic pumping to 2 x 10_5 torr. If the crystals
had sufficient thermal stability, the H-cell was allowed to
warm to ambient temperature for several minutes while
pumping. During pumping, the cell was repetitively shaken
and tapped to free crystals from adherence to the walls.

Crystals were harvested by transferring into either
evacuable storage tubes or glass vials. The most stable
compounds could be taken into the inert atmosphere box and
stored at -40'C in the freezer that is located in the box.
Compounds with greater temperature sensitivity were

transferred to evacuable storage tubes in a nitrogen glove

-38-

bag. The storage tubes were evacuated and individual
sample fingers were sealed off. These compounds were
stored at -80'C.

Some syntheses were made in the more complex synthesis

" which features an additional

apparatus called the "cow
sample storage tube chamber. This apparatus has the advan-
tage of keeping the crystals under vacuum until characteri-
zation studies were performed. The synthesis procedure is
identical to the H-cell preparation with the exception that
the crystals are moved to the storage chamber prior to
washing. After the crystals are washed, the wash solvent
is frozen in one of the other chambers. The apparatus is
evacuated to (4 x 10.5 torr, as in the case of the H-cell,

and the storage chamber is sealed off. A detailed discus-

sion of the final stage has been given elsewhere [51].
0. Analysis

The general analysis scheme for determining the
stoichiometry of alkalides is given in Figure 9A. The
crystals were first allowed to react with water to yield a
basic residue and hydrogen gas. The volume of the hydrogen
is directly proportional to the reducing power of the
crystals. A calculated excess of H01 (at least three times
the number of moles of 0H_) was added to the decomposed
material. The resultant solution was divided into 3 par-

tions; one for flame emission spectroscopy, one for pH

-39-

A. Mt (I505),- N"
+

2 14,0

i /

M" “505). 1‘ N+ 4' 20H" * Ham
.+

'icnIsunun

i
I I . 1 dried

flame pH titration Quantitative
emission with base 'H NMR

/

M" 0505);? + I-|20-">M"(|5<35)u+ OH- I V" Ham

Figure 9. Analysis Scheme for Alkalides (A) and
Electrides (B).

-4o_

titration with base and the third for quantitative proton
NMR measurements. Figure 98 gives the equation for the
reaction of an electride with water. Note that the amount
of hydrogen gas evolved is exactly half the amount of elec-
tride used. The same analysis scheme was used for elec-
trides. This procedure gives three checks for the metal
content and one for the crown content of the crystals.

1. HYdrogen evolution - Care was taken that the crystals
decomposed only according to the equations given in Figure
9. Sample tubes were scribed with a glass knife and
weighed quickly to avoid thermal decomposition. The hydro-
gen evolution apparatus is shown in Figure 10. Crystals
were loaded under a dry nitrogen atmosphere by breaking the
scribed sample tube open and placing both pieces through
the 1/2" tubing at the top of the apparatus. The bottom
surface was precooled in liquid nitrogen to avoid thermal
shock of the crystals. The 1/2" opening was then capped
with a sealed glass tube and an Ultra-torr union. The
apparatus was evacuated to ~10—5 torr on the vacuum system
designed for hydrogen collection. This system consists of
a gas buret and a modified Toeppler pump and is described
by Van Eck [23]. The entire system had been previously
evacuated to less than 2 x 10—5 torr. Conductance water
which had been degassed several times was used for the de—
composition reaction. The water was allowed to distill
through the hydrogen collection system and freeze on the

walls above the crystals. The vessel was warmed spot-by-

-41....

Ca|on
Ultra-Torr
Unlon

 

 

 

 

 

Figure 10. Hydrogen Evolution Apparatus.

-42-

spot to melt the ice and allow only a small amount of
liquid water to contact the crystals at any time. After
the reaction began, a dry ice—ispropanol bath was used to
cool the vessel in order to slow or halt any further reac-
tion. By repeating this thaw-quench procedure, the
crystals were slowly decomposed without thermal decomposi-
tion to yield a white residue. At this time, additional
water was rapidly distilled onto the decomposition product
to form a solution and insure complete reaction. The
evolved hydrogen was pumped through a double liquid nitro-
gen trap into a chamber of known volume by means of a
manual Toeppler pump. After several cycles, the pressure
of the hydrogen was measured by noting the height of the
mercury in a U tube opened on one side to atmosphere and
the other side to the known volume of hydrogen. The end
point was determined to be the pressure at which five addi-
tional pump cycles did not change the mercury level. By
noting the temperature of the hydrogen chamber and the
atmospheric pressure, the ideal gas law could be employed
to calculate the males of hydrogen evolved. The pump
cycles also served to evaporate the basic solution to yield
a dry residue.

2. pH" titration - The residue from the hydrogen evolution
was dissolved in a known amount of H01 and conductance
water under a nitrogen atmosphere, to give a solution of pH
~3. The number of moles of hydrogen evolved was used to

estimate the number of equivalents of base present. The

-43-

solution was divided into 3 portions, as described above.
The pH titration portion was further divided into 2 or 3
parts. Each part was titrated with an NaOH solution that
had been freshly standardized with H01. Phenolphthalein
was used to indicate the titration endpoint. To prevent
CO2 absorption by the base, the titration buret had a glass
sheath allowing dry nitrogen gas to flow over the solution.
The results of the different titrations were averaged to
give the number of equivalents.

3. Flame emission - A Jarrel-Ash Atomic Absorption/Flame
Emission spectrometer was used to determine the alkali
metal content of the compound. Only the flame emission
feature was used. The flame emission portion of the resi-
due solution was further diluted to give solutions of ~35
ppm for each metal. Standard solutions of each metal in
the 10-100 ppm range were used to obtain the calibration
curve. The reading obtained with conductance water was
noted between each standard in order to give a baseline
correction. Calibration curves of ppm vs digital output of
the amplifier circuit were used to find the concentration
of the unknown.

4. Proton MMR‘~ The third portion of the hydrogen evolution
residue solution was neutralized by adding the appropriate
amount of the NaOH solution. The solution was covered with
tissue and allowed to air dry at room temperature. When
dry, the resultant residue was further dried in a vacuum

desiccator for 30-60 minutes. More rigorous pumping

-44-

resulted in evaporation of crown and led to errors in the
early determinations. A known amount of dry sodium acetate
and ~0.5 ml 020 were added to the residue and the solution
was transferred to a 5 mm OD NMR tube. The sodium acetate
served as an internal standard. A 1:1 ratio of acetate
protons to crown protons was approximated by using the
amount of hydrogen evolved as an estimate of the number of
moles of 15-crown-5. Spectra were taken with a Bruker 250
MHz Fourier Transform NMR spectrometer equipped with an
Aspect 2000 computer. Individual peaks were isolated and
fitted to either a Lorentzian or Gaussian curve by a line
fitting program provided by Bruker [58]. This program
gives the amplitude, full width at half height and the
standard deviation for the frequency of the fitted curve.
A ratio of the area under the curves gave the ratio of

acetate to crown protons.

D. Optical Spectra

Optical transmission spectra were taken of films made
from solutions of crystals dissolved in methylamine.
Figure 11 shows the apparatus used for optical studies.
Crystals were loaded through the stopcock into the pre-
cooled optical cell under a dry nitrogen atmosphere. The
apparatus was closed and evacuated to 2 x 10-5 torr.

Methylamine was distilled onto the crystals to fill the op-

tical cell. Films were made by two methods. The first

-45-

 

 

 

 

Graded
Seal C.
H
Quartz 4
Optical *
Cell

 

 

Figure 11. Optical Absorption Apparatus.

-46..

method consists of pouring the solution from the cell to
the side chamber which is held at liquid nitrogen tempera-
ture. The solution must be concentrated enough to leave a

good film with one pour. The resultant films were fairly

uniform to the naked eye. The second method is flash
evaporation. The solution was poured into the side chamber
leaving 1 or 2 ml in the optical cell. The side chamber

was frozen in liquid nitrogen while the apparatus was
vigorously shaken to spread the solution over the walls of
the cell. These films were not as homogeneous as those
made by the first method but less concentrated solutions
could be used. Spectra were recorded on a Beckman DK-2
double beam recording spectrophotometer that had been modi-
fied to permit temperature control of the sample compart-
ment between 0' and -l30'C. Rough temperature control was
achieved by adjusting the flow rate of nitrogen gas that
had passed through a cooling coil immersed in liquid nitro-
gen. A copper-constantan thermocouple was placed in the
cell compartment to monitor the temperature. Spectra were
recorded in the 2800 to 400 nm range with the reference
beam passing though air. The spectra were normalized by
subtracting the baseline correction of an empty cell and
then rescaling the lowest absorbence to zero and the maxi-
mum to 1.0. Temperature dependence of the optical trans~

mission spectra was also studied.

-47-

E. Pressed Powder Conductivity

Powder conductivity measurements were made under pres-
sure in an apparatus designed by J.L. Dye and M.R. Yemen
[20,59]. Under a dry nitrogen atmosphere, a sample was
loaded into a precooled 2 an inner diameter heavy-walled
fused silica tube which rested on a stainless steel elec-
trode. A second stainless steel electrode was inserted
into the top of the tube and was used to compress the sam-
ple by means of a spring with a measured force constant.
The sample cell was loaded into a cryostat which was cooled
by controlled liquid nitrogen boil-off. Ohm’s law
obedience was checked by measuring the current as a
function of voltage. At a constant voltage, the crystals
were slowly cooled to less than -50'0 and warmed again to
0'0. The current was read at. approximately two degree
intervals during the temperature cycle. The apparatus was
then disassembled to check for decomposition. If the
crystals had maintained a dark blue color, the system was
reassembled and taken through another temperature cycle.
For some samples a check was made for polarization by
reversing the current flow. After completion of the second
temperature cycle the height of the pressed powder was
measured so that the specific resistivity could be

estimated.

-43-

F. Electron Paramagnetic Resonance

The samples for EPR spectra were loaded into evacuable
4 mm OD quartz tubes in a dry nitrogen atmosphere. Each
tube was evacuated to 4 x 10-5 torr and sealed. Spectra
were recorded on an X—band spectrometer (Bruker model 200).
Temperature regulation was unavailable except by reduction
of the flow rate of cold nitrogen gas. To avoid condensa-
tion of water onto the sample tube, the gas flow was main-
tained at a high rate. Spectra were taken at -105 to
-90‘0. A copper-constantan thermocouple with a digital
readout (Doric model DS-350) was placed under the sample

tube to observe the temperature.

C. Magnetic Susceptibility

The magnetic susceptibilities of the electrides were
measured by an S.H.E. Corp. computer controlled variable
temperature susceptometer equipped with a superconducting
quantum interference device (SQUID). Samples were loaded
under a dry nitrogen atmosphere into thin wall Delrin
buckets with snap-on lids. The buckets had an overall
height of 10 mm and 6 mm outer diameter. The buckets were
stored in liquid nitrogen until loading into the airlock of
the susceptometer. The airlock was evacuated to ~500
millitorr and flushed with helium gas. This evacuate-flush

process was repeated 2 or 3 times and the sample bucket was

-49-

visually checked to be sure no frost remained before the
sample was lowered into the SQUID at 5 K. The magnetic
moment of each sample was measured at various temperatures
between 1.7 and 240 K. The apparent ”susceptibility", x’,
was given as the magnetic moment in emu divided by the
field strength in gauss. Ten readings were taken at each
temperature. The software automatically discarded any data
points that had more than a preselected percentage
difference between two successive moments. This value was
usually chosen to be 10%. After a complete temperature
cycle, the sample was ejected and allowed to thermally
decompose in the helium atmosphere of the air lock. The
apparent susceptibility of the decomposed sample was
measured at the same temperatures.

Mass determinations of the decomposed sample and empty
bucket were made so that the molar susceptibility, x",
could be calculated. If the sample mass was known prior to
the measurement of the moments, the S.H.E. computer could
calculate xM directly. However, the weighing procedure
introduces the chance of premature decomposition of the
sample. The electronic contribution to the molar suscepti-

bility was then calculated by the following equation:

I _ I
z xsample-tbgcket xdecomposed sample+bucket (6)
xM moles of sample

PART I

CHAPTER THREE -- RESULTS AND DISCUSSION

The utility of any new compound is dependent on the
ease of preparation and total cost. Electrides and alka-
lides are excellent reducing agents with one or two eleC+
trons readily available to donate. Although the environ-
ment must be rigidly controlled to avoid decomposition of
these salts prior to reaction, the use of alkalides and
electrides as electron donors for difficult reduction reac-
tions seems feasible. Some of the problems encountered in
the economical utilization of alkalides and electrides are
the length of the synthesis procedure, decomposition of the
solutions during synthesis, the cost of complexants and the
thermal stability of the crystals.

The major goal of Part I of this dissertation is to
describe the synthesis and characterization of new alka-
lides and electrides. A new solvent system was implemented
for the synthesis, and the synthesis procedure was modified
to reduce total time. The use of dimethylether as a
dissolution solvent greatly increased the stability of the
solutions so that fewer reactions decomposed. A new

complexant, 15-crown-5, is relatively inexpensive compared

-50-

-51..

to most crown ethers and to all cryptands was tried. The
success achieved with 15-crown-5 as a complexant is
reported in this chapter. In all, 10 crystalline systems

were isolated and characterized.
A. Synthesis

Due to the ease of preparation and thermal stability of
Na+0222-Na-, the first system to be attempted with 15-
crown-5 was Na+(1505)2-Na_. The sodium cation, with a
radius of 1.02 A, is slightly too large to enter the cavity
of 15-crown-5 which is estimated to be 1.7 to 2.2 A in
diameter. Several syntheses attempts were made with crown
ether to sodium mole rations ranging from 0.88 to 1.14 in
the hope that we could mimic the l8-crown-6 systems.
Cesium forms "sandwich" complexes with 1806 in a 1:2, Cs+
to 1806 mole ratio [53] whereas rubidium forms compounds
with metal to crown ether mole ratios ranging from 0.5 to
2.0. Both cations are too large to enter the l8—crown-6
cavity. In the case of sodium with 15-crown-5, no crystal-
line precipitates could be isolated from any of the
synthesis attempts. One problem was difficulty in
dissolving sodium metal in dimethylether (MeZO). Even in
the presence of lS-crown-S, dissolution took from several
hours to two days. Methylamine (MA) was used either alone

or with Mezo to decrease the dissolution time. Although,

the solubility of sodium was increased in the presence of

-52-

MA, the solution stability decreased. Lithium was used to
give solution stability in those synthesis in which methyl-
amine was a dissolution solvent. The second problem was
the precipitation of metal at the time of the addition of
the crystallization solvent. Although the presence of
lithium reduced the chance of decomposition, its presence
compounded the metal precipitation problem. In one
synthesis which utilized MA and lithium, long, flat, golden
crystals were observed. These crystals appeared to be
highly reflective and were reminiscent of the K+1806-Na—
crystals which had ~14X solvent content [51]. The sodium
15-crown-5 crystals decomposed before they could be
isolated and analyzed.

The next system attempted was K+(1505)2-Na_. The
potassium cation with r = 1.38 A is too large to fit into
the l5-crown-5 cavity even if one uses the largest size
estimate. The metal mirror of K-Na dissolved readily in
dimethylether 15-crown-5 solution at -10'C to form a dark
blue solution. The solution left bronzy-red films on the
surface of the reaction vessel. Crystallites began to form
immediately when diethylether was added to the concentrated
Me20 solution. Precipitation of crystals continued for
several hours at 78'C. The crystals are brick red in color
and are stable at room temperature for several hours.

Five other heteronuclear systems were tried and the
attempts resulted in similar success. In addition, preci-

pitates formed from potassium or rubidium solutions that

-53-

had a metal to l5-crown-5 mole ratio of 0.5 or 1.0.
Dimethylether was used as the dissolution solvent for all
of the syntheses and complete salvation of the metal
required 30 minutes to 4 hours. Diethylether (DEE) was
initially tried as the co-sovent for crystallization. In
these syntheses in which no precipitate formed in the
presence of diethylether (DEE), trimethylamine (TMA) was
used as the co—solvent. The initial stoichiometry,
crystallization conditions and general comments on color
and stability of 10 metal 15-crown-5 systems are given in
Table II. In general, the precipitates were polycrystal-
line or clusters of crystallites and the morphology cannot
be described.

Several generalizations may be made for these systems.
First, the heteronuclear compounds which contain sodium are
easier to prepare than those without sodium. Precipitation
began with the addition of the co-solvent even at tempera-
tures as high as -30°C. The formation of crystals was
complete within 30 minutes for the systems with initial
stoichiometries of RbNa(1505)2 and CsNa(1505)2. These
precipitates were microcrystalline rather than clumps of
crystallites. Second, the thermal stability of the com-
pounds that contained sodium was greater than that of the
other compounds which did not. The precipitate of presumed
stoichiometry RbNa(1505)2 was stable for over 24 hours at
25'C. Third, the precipitates from solutions with metal

stoichiometry of K1505 and Rb1505 were more stable than

_
4
5

_

.swmoo oa masumhho ho sawusaflaMoaha ham moawsvou unawawpsoon

 

 

 

 

.maasswl

oH sow comm an anodes .onsoLA105HA mas: oembt .nason NHtoH .<29 «Amomnvnmmo
.mousswl

om mom Down as cannon .ousosolvoa ass: comet .nsso: mated .<29 «Amumuvamu
.mssoa

N ooh comm as cannon .owsosntvos and: vocal .ounvaIIM .mm: «Amomdvnzmo

.nasoa an new osmN an vacuum .owsosn mom cecal .ouswvollw .mm: «Amomuvsznm

.aaso: N mom comm as ounces .owsoao man: comb: .nsso: m .mu: Anomavom
.noasswl

om mom cecal as cannon .xosnntosam comet .maso: mum .<29 «Anomavnm

.mausswl on now comm an cannon .moslosgm Dewbt .nasos m .<za «Anomavnmx
.muso:

~nso>oa mom comm us cannon .vos magma comet .ounwvallw .mmn «Amomuvsza
.:«8 ca mom

assasaonloa loan as magnum .owsoho mas: osmbl ..msa NHIoH .mma Amomuvx

.cwa ca sow cecal an cannon .xosdnlosum comet .nsso: mum .<za «Amomfivs

Newawonum qsoHoo oamflmsommou
.olwu .uoo>uomoo
musolfloo adamawvsoo haaoEOMAOMOHm
seawuswwaananmso unwawsu

. .sovfihuoo~m can
mokusaad mlm30h01m~ mo nowawumasam obmusmom can anomawmmoo sowuswmufisuahho .HH mamdh

-55-

those of the compounds from solutions with a higher 15-
crown-5 content. The colors of K(1505)2 and Rb(1505)2 were
deep blue-black whereas those of K1505 and Rb1505 were dark
bronze. Apparently two distinctive compounds are prepared
from solutions with metal to l5-crown-5 mole ratios of 0.5
and 1.0. Fourth, the ease of preparation and relative
stability of RbNa(1505)2 indicate that this compound is the
best candidate for use as a reducing agent for other reac-

tions.

B. Analysis

Precipitates which form during the synthesis of alka-
lides and electrides may or may not have the same stoichio-
metry as the solutions from which they form. An elemental
analysis scheme has been used which has three checks for
metal content and one test for crown ether content. This

scheme has been reviewed in Chapter Two of this disserta-

tion. Two main problems were encountered with the use of
this scheme. First, premature decomposition caused low
amounts of hydrogen to be evolved. Decomposition would

occur for the more thermally sensitive compounds during the
determination of the mass of the scribed sample tube.
Great care was taken to prevent decomposition before reac-
tion with water but in most cases, the amount of hydrogen
evolved was low in comparison to that predicted on the

basis of OH—titrations and M flame emission studies.

-55-

Second, l5-crown-5 is volatile and is removed by rigorous
drying of the analysis residue for 1H NMR studies. The
crown ether ring can also be destroyed during thermal
decomposition and low crown ether content is often
encountered.

To avoid these problems some samples were analyzed by a
shortened scheme. The weighed, scribed sample tube was
opened in a nitrogen atmosphere and the crystals were
sprinkled into a dilute solution of H01. Portions of the
solution were titrated and the rest of the analysis scheme
was followed as described in Chapter Two. The large volume
of water ensured that the decomposition occurred via the
reaction with water rather than by thermal pathways.

The results of the analysis of the precipitates are
give in Tables IIIA and B. Thermodynamic calculation have
been made for compounds of the formula M+(15C5)2-N- and can
be used to estimate the relative stabilities of the poten-
tial salts [38]. These estimations have been used to pre—
dict the species within the compounds from the stoichio-
metries. The analytical results have been divided into two
categories. The first category is comprised of those
results which have a high degree of consistency among the
number of moles obtained by the various techniques. The
second category contains those compounds whose stoichio-
metries can be attributed to a unique compound. Sample
sizes were determined by the OH- titration results based on

the presumed stoichiometry. The percent deviation from the

-57-

.N

.lsson> amisshv hops: pawns smalsm
.-=o mo sowassafla an mosmasoaom sewn sadism

n «\s + -so + s-mom-.2 All ohm + -s.s-moe-.z

o
A

so a + -ION + .z + samomav.z All Oman + -z.samomav.z assume mo campuses any so panama

6.51
o.ml
m.ct
m.¢+
w.m+
m.m+

 

:22 me NA
mIGSOMOImH

m.oHI
o.H+
m.oHI

m.m+
m.m+

mas-s14m

m.m+ tn-
m.et It-
m.e- «.m1
-.-t -.-N1
s.mt It:
mahsuxm: .m
«.m- It-
e.-+ it-
m.mt m.--
m.m- m.mm-
m.m+ It-
m.m+ m.m+

am::OAIou sham

clean .2 IINMI

mono-owmowoam volsnosm loam sawasw>oa a

sea assassa~< «0 usesauns .HHH aaa<a

. [Q‘MHHOOdH

N.m+
c.b+

.<

mmm.~
Nem.m
mm.o
mmN.H
mum.N
om.~

 

«OH x oHoml
anew oaalnm

IDmsNAWQWHvOQU
IMsNAmUWHvoflo
-sz.s-mom-v.eo
IN.- NAQUDHV 4M
IQsNAWOWHv+¥

-nz.~Amom~v.nu
-nm.~Amom~v.Am
-sz.smmomsv.sm
-o.«AmomHv.nm
IDmoNAmUOHvog
Iflz-NAWUQAV+¥

 

manomoflao«09m
melanoma

-53-

titration results is given for each of the other tech-
niques.

In general, the amount of hydrogen is lower than
expected based on the OH- titration and flame emission
results. The deviation increases with decreasing thermal
stability which is indicative of premature decomposition of
the sample. The mass determinations indicate that there is
more compound present than is indicated by the other tech-
niques. Loss of sample during transfer or loss of glass
chips from the sample tubes would give higher than actual
results from the mass determinations.

The presence of an electride is presumed from the stoi-
chiometry Rb(l5C5)2. The assignment of Rb+(1505)2-e- also
is supported by the blue-black color of the crystals and
the isolation of crystals with stoichiometry corresponding
to Rb+(l505)-Rb_. All compounds listed in Table IIIA have
stoichiometries which suggest that the cation is complexed
by two l5-crown-5 molecules as expected from the size of
the l5-crown-5 cavity and cation radii.

Four of the compounds appear to be mixtures or have
slight deviations from a strict M+(1505)2-X_ stoichio-
metry. The expected stoichiometries were K+(l5C5)2-e-,
K+(15C5)2-K-, and Cs+(1505)2-Rb-, and the analysis results
are given in Table IIIB. In addition, Table IIIB contains
the analysis results for a sample of Cs+(l5C5)2-Na- which
was dried under dynamic vacuum before the 1H NMR determina-

tion of the crown ether content. The homonuclear potassium

-59-

compounds appear to be mixtures of K+(l505)2-e- and
K+(l505)2-K- but any one synthesis yields predominantly
either the electride or alkalide with only a slight
contamination of the other compound. The crown ether
content greatly exceeded the expected value in
Cs+(1505)2-Rb- and trapped electrons may be present in this
sample so that the concentrations of Cs+(l505)2-e— and/or
Rb+(1505)2-e- may be signficant. Several preparations of
this compound were made and the crown ether to metal ratio
varied with the preparation. The tendency to form mixtures
may also be present in Cs+(1505)2-K— where the cesium and
potassium content deviate ~93 with respect to one another
and the 15—crown-5 content is 11% low compared to that
expected from the OH-titration results. Nuclear specific
characterization techniques are needed to correctly
identify the compounds.

In addition to the tendency to form mixtures, the
vacuum drying of the analysis residue before the determina-
tion of l5-crown-5 content can give misleading results.
Compare the results obtained for Cs+(1505)2-Na- with and
without vacuum drying, Tables IIIB and IIIA, respectively.
The sample dried by dynamic vacuum has a crown ether
content which is 48.3% lower than expected. The sample
analyzed by the abbreviated scheme has a crown ether
content which is within the acceptable range of the number
of moles calculated from the OH- titration and flame

emission results.

-60-

The result of the analyses indicate that eight new
alkalides and two electrides were synthesized. Exact iden-
tification of the species within these compounds has been
attempted by optical absorption spectroscopy and is
reported in the following section. The elimination of the
hydrogen evolution step in the analysis scheme and the slow
drying of the analysis residue has simplified the analysis
and gives acceptable results. There is a tendency to form
mixtures of electride and homonuclear alkalides when the
synthesis of either is attempted. The tendency is greater
for the potassium systems than for those salts which con;
tain rubidium. The levels of contamination are low, how-
ever, which is suggestive that the compounds may be treated

as pure compounds in the characterization studies.
0. Optical Spectra

The "pour" method, rather than the flash evaporation
method, was preferred for making films for optical studies.
Uniform films could be made much more easily by pouring the
solution from the optical cell to the side chamber. Solu-
tion stability was sufficient to allow the pouring to be
done with the apparatus immersed in a dry ice-isopropanol
bath. Films left on the walls above the cell and on the
side arm of the apparatus showed no signs of decomposition
during the time period of the experiment with the exception

of K+(l505)2-e— and Rb+(1505)2-e-. Decomposition of these

-51-
films occurred within 30 minutes so that films had to be
made quickly and kept cold.

The optical spectrum of the film with stoichiometry
K-l505 is given in Figure 12A. A single maximum at 12000
cm—1 with a low energy shoulder was observed. The optical
absorption maximum for K- in ethylenediamine occurs at

12000 cm‘1

and the assignment of this film is K+(1505)2-K—
rather than K+(l5C5)-e-. The low energy shoulder indicates
the presence of some trapped electrons but the major
species is seen to be K-. The assignment of a potasside
for this compound is further supported by the spectrum of
K(l505)2 films shown in Figure 12B-0. The maximum at 7800
cm— corresponds to trapped electrons and therefore the
assignment of this compound is K+(1505)2-e-. The two spec-
tra of K+(1505)2-e- were obtained from different films made
from the same solution. The extreme narrowness of the
spectrum in Figure 120 would require very narrow bands for
the initial and final states. It is likely that this peak
is an artifact. Spectrum B is more typical of electride
spectra with a width that corresponds to wider energy
bands.

Mixtures of K+(1505)2-K- and K+(1505)2-e- were easily
obtained in the preparations of the electrides. A typical
spectrum is given in Figure 13. The peaks, located at
11000 and 6900 cm_1, correspond to K- and e_, respectively.

In general, the excited state energy bands of an alkalide

metal anion are narrower than those of electrons. Analysis

-52-

 

 

B. J "/ \~_____,I‘ ‘
I "- _________
(3. .1’ 1 1 1 l ij
5 IO IS 20 25

WAVE NUMBER x 10“. cm"

Figure 12. Optical Absorption Spectra of
Thin Dry Films Made From §olutions _
of K+(15c5)2-K‘ (A) and K (15C5)2'e
(B and C) in Methylamine.

-53-

 

 

 

 

1 1 l l 4
5 ‘ IO I5 20 25

WAVE NUMBER x lo". cm"

Figure 13. Optical Absorption Spectrum of a Thin
Dry Film Made From a Solution of
K(15C5)1.3 in Methylamine.

-54-

of the residue left from the optical absorption studies
showed that the stoichiometry was K1(1505)1.3.

The compounds of rubidium with lS-crown-S had spectra
similar to those of the potassium compounds. Crystals of
stoichiometry Rb1505 yielded spectra that featured a single
sharp absorption maximum at 9600 cm.1 as shown in Figure
14A. These crystals were identified as Rb+(l505)2-Rb—.
However, crystals of stoichiometry Rb(1505)2 did not yield
the simple trapped electron spectrum seen in the analogous
potassium system. The spectra from several Rb(l5C5)2 syn-
theses were recorded and all featured peaks at both 9600
and 6500 cm-1 which correspond to Rb- and e—, respectively.
The relative amplitudes of the two peaks were sample- and
film-dependent. A typical spectrum of Rb(1505)2 is shown
in Figure 148. Decomposition of films on the walls above
the optical cell led to decomposition of the solution when
attempts were made to produce additional films. The
assignment of these films is tentatively Rb+(1505)2-e- with
some Rb- present.

The heteronuclear salts were more difficult to identify
with optical absorption spectroscopy. Those salts which
contain sodium could be confidently identified as sodides
due to the close proximity of their maxima. K+(1505)2-Na-,
Rb+(1505)2-Na- and Cs+(1505)2-Na- had absorption maxima at
14300, 13900, and 13900 cm_1, respectively. The spectra of
these films are given in Figure 15A,B and C. In addition

to the main peaks, the spectrum of 0s+(l505)2-Na— featured

-65-

 

 

 

J
l I l L l
5 IO I5 20 25
WAVE NUMBER :1 to", cm"
Figure 14. Optical Absorption Spectra of TAin Dry
Films Made From Solutions of Rh (15C5) 'Rb
' (B) in Methylamine.

(A) and Rb+(15C5)2-e

-55-

 

 

 

 

1. ij. L L. J
5) It) l5» 2!) 2!)
WAVE NUMBER :1 to", cm"
Figure 15. Optical Absorption Spectra of Thin Dry

Sodide Films Made From+Solutions 9f
Cs+(1§c5)2-Na' (A), Rb (15C5)2'Na (B)
and K (15C5)2°Na' (C) in Methylamine.

-57-

a shoulder at 15400 cm-l. This shoulder was barely
discernible at -40' but became more distinct as the tem-
perature decreased. The lowest temperature achieved was

-106'C and the shoulder did not completely resolve into a
separate peak. The presence of the shoulder suggests there
may be two or more sites for Na- with perhaps an exchange
mechanism at high temperatures. The shoulder became less
distinct as the temperature was increased from -106'C.
Only one temperature cycle was done and it is not known if
the appearance of the shoulder is reversible.

The results of the optical studies are given in Table
IV. Assignments were made based on the peak location,
stoichiometry and thermodynamic estimations. The main
ambiguity lies in the identification of Rb-K-l5052 whose
optical absorption spectrum is shown in Figure 16. The
thermodynamic estimations predict K+(1505)2in- to be
slightly more stable than Rb+(l505)2-K-. These calcula-
tions assume that the enthalpy change for the complexation
of the second crown ether with potassium is equal to that
with rubidium. The optical spectrum of films made from
solutions of these crystals featured a main peak at 10200
es’l with a 12300 cm‘1 shoulder.) As in the Cs+(l505)2-Na-
spectrum, the shoulder became more distinct as the tempera-
ture was decreased. The main peak lies between the loca-
tions observed for the maxima of the two homonuclear salts.
The peak could either be a composite of K- and Rb- peaks or

represent a fast exchange of electrons between the two

-68-

.nsaooaa “sewage can» Rosana nasalsmss Owlnsmmolaosu so some: woulouana<

 

 

 

Ac
....... wmmewn---------------------------------- ------------- --
N so N
avian. Amomuv+no moons chem Anomavommo
handsome
N usomsoaomsw a cabNH N
(a. AmomHv+no anon can! omNb Amomuvxno
N sovasosn comma N
102. AmomHV+mo anon can! oomw~ Amumgvnzuo
-ez.-moe-+em asses eeme- N-mom-vszss
tom.NAmum~V+nm manna comm Amomfivnm
.oualnn a» cannon
N flosw hsn> exams mo N
Is. Anomav+nm novsumaasn o>wusaou comm can comm Amomfivnm
mad) as owl n a as
assuxwa mac~o>ov sovfisosn oomNH
N no vsoso N
slam. Amomdv+a hso> .xsoa can: coNo~ Anomavomx
)sz.-mom-+= asses eeme- «Amos-Vezs
(s.N-mom-+s soc-sees Assess :e- eeem- Amos-es
us.-mum-s+u sense some . «Amos-Va
unassumnn< sawuawhommm lo ddbwusooq zoom asuomow£o«0uw
HI vssoaloo

.enses as: swan «e soaesam seAsssess< ”sesame we assess: .>H man<a

-69..

 

 

I I I I _I
5 IO I5 20 25

WAVE NUMBER 11 I0", cm"

Figure 16. Optical Absorption Spectrum of a
Thin Dry Film Made From a Solution
of Crystals with Stoichiometry
KRb(15CS)2 in Methylamine.

-7o-

nuclei. If the peak were due to only one of these anions,
the cation would have a much larger effect on the Rb_ or K-
environment than is the case for Na_. The Na- optical
absorption maximum shifts to lower energies by only 400
cu“1 as the cation is changed from n+(1505)2 to Rb+(l505)2
and there is no change in its location as the cation is
changed from rubidium to cesium. The observed maximum for
RbK(l505)2 would require an 1800 cm"1 blue shift for K- or
an 800 cm-1 red shift for Rb—. This compound is most
likely a mixture of K+(1505)2-x‘ and Rb+(1505)2-Rh' and/or
the heteronuclear salts. Only a nuclear-specific technique
can remove this ambiguity in the identification of this
system.

One of the problems associated with the use of optical
absortion spectroscopy as an identification tool is the
effect of the solvent. The anion in the film may or may
not be the same as the anion in the crystals used to make
the solutions. This is because the time scale of crystal-
lization is much longer than that of film preparation so
that a kinetically more favored species may be present in
the films. For example, a metal anion may dissociate in
solution to form the solvated monomer or cation and elec—

trons according to:

-71-

N + e (5)

Films made from these solutions would have absorption maxi-
ma corresponding to trapped electrons if reaction (6) lay
far to the right and if the kinetics of metal anion forma-
tion in the solid state is slow. Anions with low electron
binding energies would be more likely to undergo this type
of dissociation. For example, films made from solutions of
Cs+(1505)2 in the presence of either rubidium or potassium
had absorption maxima that corresponded to trapped elec-
trons rather than alkali metal anions. Films of
Cs+(l505)2-K- showed the presence of both K- and e_.
Figure 17 represents two spectra of dry films made from the
same solution. The first film, shown as Figure 17A, has a
single peak at 12700 cm.1 with a low energy shoulder. This
film was prepared shortly after the solution was made. The
second film was prepared after a one hour time interval.
The main absorption occurs at 7250 cm-1. Presumably,
decomposition of the solution or precipitation of metal
upon standing caused pronounced changes in the nature of
the films produced by solvent evaporation. The relative
amplitudes of the K- and e- maxima were independent of time
and temperature which suggests that the processes in solu-
tion are more important than the solid state reactions.

Equilibria such as those given by Equation (6) probably are

-72-

 

 

 

I ' I I
5 IO I5 20 25

WAVE NUMBER :1 I0", cm"

Figure 17. Optical Absorption Spectra of Thin
Dr Films Made From a Solution of
Cs (15C5)2'K‘ in Methylamine. (A)
spectrum of film prepared shortly
(B) spectrum

after solution was made;
of film prepared one hour later.

-73-

the cause of the broad, ambiguous peak seen for the RbK-
(1505)2 system.

Three conclusions may be drawn from the optical absorp-
tion studies. First, electrides may be distinguished from
the homonuclear alkalides but not without some uncertainty.
Second, sodides can be easily identified by the location of
the optical absorption maximum. Third, dissolving the
crystals may introduce dissociation (or association) reac-
tions and thereby yield films that have different anionic
species than those present in the crystals. To provide
clearer identification of compounds, a technique is needed
which does not alter the crystalline state of the alkalides

and electrides.

D. Pressed Powder Conductivity

Intrinsic semiconductors have temperature dependent
electrical conductivities. The temperature dependence may

be described by

a = A exp[-Eg/2kBT] (7)

where a is the conductivity, A is the conductivity at infi-
nite temperature and kB is the Boltzman constant. The band
gap, 88' can be calculated from the temperature dependence
of the conductivity. the resistivity, p (p = l/a) ranges

from 10.6 ohm cm for copper to 1018 ohm cm for fused silica

-74-

at room temperature. The resistivity for intrinsic semi-
conductors is of the order of 10-3 to 109 ohm-cm.

Extrinsic semiconductors also show temperature depen-
dent conductivities but usually over a much shorter tem-
perature range. The conductivity arises from an impurity
band which can either donate or accept electrons and lies
between the valence and conduction bands of the bulk mater-
ial. Electrons may be thermally excited either from the
donor band to the conduction band or from the valence band
to the acceptor band. At sufficiently high temperatures,
saturation is reached and the conductivity returns to in-
trinsic behavior. Extrapolation to infinite temperature
generally yields resistivities which are much lower than
those expected for an intrinsic semiconductor.

The current flow at constant voltage was measured at
several temperatures for each of the new compounds.
Adherence to Ohm’s law was checked by measuring the current
at different voltages at constant temperature. This
insures that the voltage used for subsequent measurements
was in the linear region of the current versus voltage
plot. Substantial polarization has been observed in single
crystal conductivity studies of Na+0222oNa-. The polariza-
tion is due to conductivity from ionic mobility followed by
electrochemical reactions at the electrodes rather than
from electronic conductivity. The contribution was
observed to "freeze out” at temperatures below -24°C where

the conductivity for the single crystal behaved as in the

-75-

pressed powder studies for Na+0222-Na_ [60]. All of the
new 15-crown-5 compounds obeyed Ohm’s law with the excep-
tion of Cs+(l5C5)2-K—. The plot of current versus voltage
in this case had a nonzero intercept of l x 10”8 amp at
zero volts. This deviation from Ohm’s law could be due to
many reasons such as l) Polarization of the powder by ionic
conductivity followed by electrochemically produced
current; 2) Offset of the zero by l x 10-8 amp; or 3) Tem-
perature variation during the measurements. The plot was
linear and the most likely cause of the deviation is a zero
error. In the case of Cs+(1505)2-K—, a voltage comparable
to that used for other alkalides was chosen for the current
measurements.

Previous conductivity measurements required two full
temperature cycles for a complete experiment. The data
from the second cycle were used to calculate the band gap
and resistance at infinite temperature in order to avoid
packing effects that could be present in the first tempera-
ture cycle [51]. In this study, the entire loaded cell
assembly was cooled to <-150'C by immersion in liquid
nitrogen and then allowed to warm up to -10'C before
measurements were taken. This procedure had the effect of
packing the crystals as in the first temperature cycle of
the previous experiments. In addition to this process, for
most compounds the current was measured through one com-
plete temperature cycle and then measured again as the

temperature decreased to insure that the packing density

-76-

did not change. For five of the compounds, the polarity of
the electrodes was then reversed and the current as a func-
tion of temperature was measured. The data were fit by the
equation

-Eg/2kBT

1=R—‘-’-e (8)

where I is the current at temperature T, V is the voltage,
Eg and kB are the band gap and Boltzman’s constant as
before, and R“ is the resistance at infinite temperature.
In this procedure the nonlinear, least squares program

KINFIT [61] was used. The least-squares estimates of E

8

and R are given in Table V. Each direction of the change
Q

of temperature was treated as a separate data set. For

example, K+(1505)2-Na- has four values listed for Eg’
These values, presented in order of measurement, correspond
to: 1) current as temperature decreases, 2) current as tem-
perature increases, 3) current as temperature decreases and
4) current as temperature decreases with reverse polarity
of the electrodes. An estimate of the room temperature
resistivity was made for each data set and is also pre-
sented in Tablee V. A representative plot of logn current
versus l/T for Rb+(1505)2-e_ is given in Figure 18.

The band gaps calculated for the alkalides and elec—
trides correspond to those typical of semiconductors. The

resistivity at 298 K is however, much higher than that

expected for intrinsic semiconductors with such small band

-77-

.nosswuooo > manna

 

mod N mm.m
can x am.m
cod x hm.H
voH x Nm.m
«ca x m¢.m
can x mb.m
moa x hm.u
meg N am.“
nod x hm.N
now x mc.m
mod N vm.N
mod x bc.v
ncu x NN.m
me“ x mm.N
meg x N¢.~
mag x m~.H
moa x «H.A
voH x mN.N
cod x Nm.v
mod x mm.~

ooH
«IOH

v-oH
v-cH
s-oH
nroH

area
area
A-oH

~-°H
m-oH
~-c~
~-o~

m-oH
m-cH
o-oH
o-oH

v-o~
«IOH
«IoH

X
NN

X

XX XX
O
+I+I+I+I

X
HOOI-II-I

X

x
HCIN
o

OOOO

+I+LH-H

QVDCO

XXXX

XXXX

X

C

+l +l +l

Ncooa c~a3muo firqndm
O

+I +I

+| +l +I

+I +I +I 'H

s
H

mmco NNNO Q‘M
IO

arenas

I-ICO

mI-IN QQ‘Q'O')
NIO

 

-sz.-momsv+sm

-o.-eoe-s+se

tem.N-moesv+m

Ix.~AmomH-+a

Isz.mflmom~v+a

to.N-momss+m

 

 

mo.o A mm.o
mo.o A ¢A.H
mo.c A o-.-
mo.e a as.-
mo.o A em.~
mo.o A on.“
mo.o w em.o
meo.o A so.~
Ne.o A sm.o
moc.c a ea.-
Ho.o A A-.-
moc.o A we.s
~o.e A mo.~
mo.e N am.-
mo.c « em.-
ao.o A om.~
ma.o H mm.-
Nc.c a me.-
Ao.o A mo.s
we.o A we.“
so «.m

vasomloo

.aanoIOLQIsOI huw>maosvnoo on Revlon

washout ho newsman

.> Nflfldh

-73-

moa x ¢H.N
mcu x mc.N
mca x mc.~

vo~ x mN.N
noH
sag

xx
1000
I-Isl'
I-lco

day
may
vcH
voH

XXXX
N
l‘
O
b

 

 

A-cH x m.o a H.H No.o w mm.o

«cm x >.o H N.H Ho.c M mm.o N

neg x N H N mo.o w em.o (om. Anomav+mo

A-o~ x m.o H ¢.v moc.o H mm.o

NIoH x N H m moo.° H mm.o N

oo~ x H.o w N.H voo.c u m>.o IN. AmomHv+no

o-oH x m H m mo.c H om.~

m-oH x w w m No.c a mN.H N

~-o~ x H w m No.9 H bo.H Isz. Amomfiv+mo

m-oH x m H m ec.o H m~.~

~-oH x N H N ec.o w mm.o

~-c~ x >.o H v.m no.9 w om.o N

~-o- x A A s so.o A mm.e use. Anom-+sm
) «8 JWII
ago a >0 N .Illfldflddflddll

 

.eumssssoo > manna

-79-

CD
1'6

 

.Im.NAmUmHV+Qm Mom B\H msmum> Auswuuso. :4: .ma munmwm
.-x.no.x+
b r r h b b D b P — P h b b P n h P P
d 4 q d] d d I q d — d d 1 d — d q q d
4

 

06—

OK.—

O.—N

DAN

OfiN

OKN

OdN

U1—

(I)

-80-

gaps. This indicates that the conductivity of these sam-
ples is an extrinsic property. Electrons trapped in
shallow traps would give rise to the small band gaps and
high resistivities. The resistivities at 298 K for these
electrides are an order of magnitude lower than those for
the alkalides which indicates that there are more carriers
in electrides. The band gaps of the electrides are not
significantly different from those of the alkalides. The
electrides may, however, still be extrinsic semiconductors
with the most important carriers trapped in shallower wells
compared with the wells for the majority of the trapped
electrons. In support of this picture, the optical spec-
trum of Rb+(15C5)2-e- for example, also indicated the
presence of electrons at many energy levels because of the
breadth of the optical absorption peak.

There is a general trend for many of the compounds to
have an increase in the apparent energy gap as the experi-
ment proceeds. This increase of 83 may be due to a
"bleaching" of the shallow traps by the current flow or by
an annealing or decomposition. The results for
Cs+(15C5)2-Na_ indicate that electrons from the shallow
traps are removed by the first set of measurements and that
the conductivity observed for subsequent temperature runs
is due to electrons in deeper traps. There is very little
change in the band gap calculated for the second and third
data sets. This explanation does not hold for other com-

pounds in which the Eg values remain fairly constant,

-31-
(e.g., K+(15C5)2-e-), or which have erratic values from one
temperature run to the next, e.g. Rb+(1565)2-e-.

One of the problems encountered in the estimation of
the band gap by conductivity measurements is the available
temperature range. The conductivity of many materials can
be measured over a wide range of temperatures. However, in
alkalides and electrides thermal instability of the
crystals prevents current measurements above -lO'C. Low
current readings limit the low temperature value so that a
100' span is the maximum obtainable range. Slight varia-
tions in the energy gap estimation can result in wild
deviations for the estimates of resistance at infinite te-
mperature. The values of R“ for Cs+(1505)2-Na— differ by a
full order of magnitude between two temperature runs even
though both have the same calculated energy gap. Data from

the KINFIT output show a strong correlation between E and

8
R as expected from Equation (8). Small errors in either
Q
estimation result in errors for the other. These errors

may be plainly observed by the randomness of the R. values.
The computer-calculated curves fit the data very well as
reflected by the small error in the individual estimates of
Eg for each data set. However, the correlation of 88 with
R” is too strong and the temperature range too small to use
the calculated uncertainty for E8 as the true error of the

calculated values. A much more realistic value for the

uncertainty of the band gaps should be 10.1 eV or more.

-32-

In conclusion, all the compounds measured behaved as
extrinsic semiconductors with relatively small band gaps.
The conductivities for K+(15C5)2-e- and Rb+(1505)2-e— swere
higher than those of the alkalide salts which indicates the
presence of more carriers in the electrides. The tempera-
ture range covered by the conductivity experiments is too
small to give accurate estimations of the band gap for such
extrinsic semicoductors. It would be be interesting to
measure single crystal conductivities for these crystals to
see if any ionic conductivity occurs and perhaps obtain

more reliable values for the band gaps.
E. ‘Electron Paramagnetic Resonance
Electron paramagnetic resonance (EPR) spectra give in-

formation about the environment of unpaired electrons.

Pure alkalides have an ns2 valence electron configuration

and would be EPR inactive. Indeed, Na+0222-Na- has no
appreciable EPR signal. Paramagnetic impurities, such as F
centers or metal atoms, would give EPR signals. Previous

studies of alkalide systems yielded EPR spectra of trapped
~electrons that featured hyperfine splitting [51,53]. The
splitting of the signal was due to the interaction of the
trapped electron with the nucleus of the cation. The
amount of interaction and extent of the splitting were pre-

paration-dependent for each compound.

-33-

Preliminary EPR studies were made for K+(l505)2e- and
Rb+(15C5)2-e— to determine the line shape and g-value for
each compound. Both electrides had very strong.microwave
power absorption. A very small amount of sample was used
to avoid saturation of the microwave receiver. Even so,
power studies showed distortion of the EPR line at power
levels of 800 nanowatt or higher. The g-values were calcu-

lated from the equation

hv = g'p'H (9)

where h is Planck’s constant, v is the spectrometer fre-
quency, p is the Bohr magneton and H is the field of the
resonance in gauss. Rb+(15C5)2-e- and K+(1505)2-e- had
very similar spectra and are shown in Figures 19 and 20
respectively. Both compounds have g-values of 2.003 t
0.001 which is very close to the free electron value of
2.0023. The linewidths for the spectra were quite narrow
and equaled 0.45 G for n+(15c5)2-e’ and 0.58 a for
Rb+(15C5)2-e_ and showed no evidence of hyperfine
splitting. The g-value and linewidth indicate that the
electron is fairly isolated in its environment and has
little interaction with neighboring nuclei that would lead
to a g-shift.

When the conduction electrons in metals give rise to an
EPR signal, the lineshape of the spectrum is highly asym-

metric and is referred to as Dysonian. For thick metal

plates in which the skin depth, 6, is small compared to the

-34-

.UOMOHI HM

 

m
m.. “momav+nm mo asuuommm mam .mH magmas

 

Tic—Iv.

 

 

~85-

.oosm- um m.NAmUmHV+s mo asuuomom mam .om museum

 

 

_...mvu.l+

 

 

 

-86-

sample thickness, the A/B ratio exceeds 2.7 [62]. The A/B
ratio will have a value between 1 and 2.7 for spherical
metal particles in which the skin depth is on the order of
the particle size. However, the packing density of pow-
dered samples may also affect the A/B value [53].

The line shapes for both K+(15c5)2-e' and ab+(15c5)2-e'
were asymmetric with A/B ratios of 2.31 and 1.67 respec-
tively. The asymmetry of the line may be due to the
intrinsic conductivity of the sample even for nonmetallic
samples. For small particles with high conductivity, a
Dysonian line shape may be obtained where the skin depth is
on the order of the particle size. The A/B ratio usually
ranges between 1 and 2.7 for these types of samples.

The temperature dependence of the A/B ratio has been
used to estimate the apparent band gap of Cs+(15C5)2-e_
[53]. The band gap was estimated to be 0.1 eV from the EPR
experiments and 0.9 eV from the pressed powder D.C. con-
ductivity experiment. The cause of the difference of E3
between the two methods is not known. Perhaps this
difference arises from intergrain resistance or the
intrinsic differences in the methods of conduction. The
microwave "conductivity" is really a measure of the AC
impedance rather than the resistance only and the contribu-
tions of both resistance and capacitance components may
give an usually small apparent band gap.

A number of additional EPR experiments should be done

to permit us to understand the cause of the asymmetry of

-37-

the derivative of the signal. First, temperature studies
need to be made to determine if there is a temperature de-
pendence of the line shape. Temperature dependence would
probably confirm the 0.0. conductivity conclusions that

K+(1505)2-e- and Rb+(1505)2-e- are semiconductors rather
than metals. Second, the EPR spectra could be obtained for
samples which have been diluted with an EPR inactive

material. Diluted samples should have Lorentzian line

shapes with A/B 1 if the distortion is due to the bulk
conductivity of the sample with conduction across the
intergrain boundaries. Third, single crystal EPR could be

used to study the effects of orientation on the line posi-

tion and shape.

F. Magnetic Susceptibility

Pure paramagnetic materials have susceptibilities that
are inversely proportional to the temperature. The propor-
tionality constant, C, is called the Curie constant, named
after the man who proposed this relationship. For a para-
magnetic material with lOOX unpaired spins in the state S =
1/2 and in the absence of orbital angular momentum of the
electrons, C = 0.3760. This ”law" was modified by P.R.
Weiss to include the mean local field effects generated by
ferromagnetic and antiferromagnetic materials above their
transition temperature. The resultant relationship is

given by

-33-

- _Q
X - T-G (10)

and is called the Curie-Weiss law where C is the Curie
constant and e is the Weiss constant. The high temperature
susceptibility may be used to predict whether a material is
ferromagnetic (9)0), paramagnetic (9:0) or antiferromag-
netic (e<0). For antiferromagnetic materials, a tempera-

ture, T called the Néel temperature, may be approximated

N’
by TN = -9. Below the Néel temperature the spins of the
material will attempt to align perpendicular to the applied
field. If the spins were completely free to move, a tem-
perature independent susceptibility would result. However,
this is never the case due to crystal field effects or
other mechanisms which constrain the motion. In single
crystal studies the crystals may be so oriented that the
ideal behavior is observed. However, in powdered samples,
deviations will be seen from Equation (10) near and below
the Néel temperature [63]. Deviations from pure Curie law
behavior may also be seen for systems where local interac-
tions affect the environment of the electron but in which
no distinct transitions occur. Many materials do not have
the rigid spin lattice required for them to be classified
as ferromagnetic or antiferromagnetic. For these
materials, 9 serves as a "catch all" correction parameter.

Unfortunately, there is no straightforward way to analyze

data that deviate from the simple Curie-Weiss law.

-39-

The temperature dependence of the susceptibility was
measured for Rb+(1505)2-e— and K+(15C5)2-e- up to 225 K.
The data were fit, with the use of KINFIT [61], to a
modified version of Equation (10) to which a temperature
independent correction term, B, was added. Diamagnetic im-
purities, such as M—, would give negative values for B and
would decrease the observed value of the susceptibility. A
correction for paramagnetic impurities, such as precipi—
tated metal, can also be made by the use of the B correc-
tion term. The plot of l/xM versus T for Rb+(l505)2-e_ is
given in Figure 21 for measurements over a temperature
range of 1.3 to 224 K. The susceptibility has been
corrected to remove contributions from the bucket, crown
ether molecules and nonvalence electrons of the rubidium
so that only the electronic contribution to the magnetic
susceptibility is reported. Two features of the l/xM
versus T plot are remarkable. First, at temperatures of
20 K and below, the data deviate from the straight line
expected for strict adherence to the Curie-Weiss law.
This deviation is in a direction that suggests a decrease
in the antiferromagnetic interaction at low temperature.
The calculated value of 9, based upon all of the data, was
-4.42 x,‘ which indicates that Rb+(15C5)2-e_ exhibits anti—
ferromagnetic interactions.

The second notable feature of the l/xM versus T plot
for Rb+(l5C5)2«e- is the discontinuity in the slope

observed between 120 and 140 K. Above 140 K, the plot is

-90-

SN

.Iw.NAmUmHv+nm How wnsumummEmB msmum> EX\H .HN munmwm

ONN

c_\,.ov..h

DON oo— 8— Ov— ON— 8—

O.

on 8 a? ON

ll con

 

LI 8N—

-91-

linear, but is displaced from the line observed in the 20
to 120 K temperature range. This change was observed as
the temperature was decreased as well as increased, which
is indicative of a reversible transition. The calculated
Curie constant for the entire temperature range was 0.2068.

Since the diamagnetic background has been subtracted, we

can say Rb+(15C5)2-e- behaves, to first order, as a
compound with 55% free spins. The remaining spins may be
associated with rubidium nuclei to form Rb-, lie in

diamagnetic ground states or be removed by decomposition of
the sample. However the deviations suggest that the Curie-
Weiss equation is not the correct one to use, so that one
should not take the 55X figure too seriously. The correc-
tion term, 8, was calculated to be -65 x 10—6 moln1 for
this sample. These results might be used to imply the
presence of Rb- but, more likely, are a result of an
inapproriate equation.

Instrument trouble prevented the measurement of the
susceptibilty of K+(1505)2-e- below 5 K. As in the case of
Rb+(15C5)2-e-, the high temperature data points had good
adherence to the Curie-Weiss law but no discontinuity in
the slope was observed. The apparent Curie constant is
0.3548 for the electronic susceptibility of K+(15C5)2-e—
and corresponds to ~95% "unpaired spins". The experimental
points deviate from the calculated line at T g 15 K as seen

in the l/xM versus T plot give in Figure 22. The Weiss

constant is -19.5 which, if real, corresponds to signifi-

-92-

.Im.NAmUmHV+M How wwsumummfima mnmuo> EX\H .NN masmfim

c_>_ox..w

OO— OO 8 O¢ ON

0
O
N
—b8
3;.
-_3
.-
O
2

D
.-
'IIP
uh
un-
cli-
——
uln-
an.
d-
-:
III-
III-

—
d

u-
.-
C.
‘-
di-

--
4.
cu-
--

1 OO.

1094.

[Sn

18v

1. com

I COO

IOOh

 

E

-93-

cantly higher antiferromagnetic interactions for

K+(l505)2-e- than for Rb+(15C5)2-e-. The temperature
independent correction term is also much larger and is
calculated to be -500 x 10-6 mol~l for K+(15C5)2-e_. The
magnitude of this correction term is too large to be
attributed to K- or other impurities. Although variations
of the B term from one sample to another have been observed
in the past, the values of the B term obtained for
different samples of the same compound vary by no more than

6 mol-1

a factor of 2 and are on the order of ~40 XIO-
[44,50]. The magnitude of the correction term for
K+(15C5)2-e- was such that it was first thought to be
caused by a change of the zero in the susceptometer between
the pristine and oxidized sample measurements. However
neither data set, when considered individually, showed
evidence of zero drift or discontinuity during the
susceptibility measurements. A more plausible cause of the
large 8 term is the low temperature susceptibility devia-
tions from the Curie-Weiss law. The line fitting program
probably overestimates the temperature independent correc-
tion terms in order to achieve a good fit with the wrong
equation.

Careful studies of Rb+(15C5)2-e_ and K+(15C5)2-e-
should be made at low temperatures with varying field
strengths in order to determine the extent of the anti-

ferromagnetic behavior of K+(l5C5)2-e- and Rb+(15C5)2-e-.

In addition, the susceptibility of Rb+(15C5)2-e- should be

-94-
carefully examined in the temperature range of 100 to 170 K
in an attempt to verify the existence of the discontinuity.
This transition may arise from the ”quenching" procedure
which occurs when the sample is loaded into the suscepto-
meter at 5 K. Also, more samples of each electride need to
be measured to get more reliable estimates of B. Care must
be taken that only freshly prepared samples are used for
susceptibility measurements due to the apparent decomposi-

tion with time during storage at -80'C.
G. Data From Other Sources

The unusually high thermal stability of the compounds
reported in this chapter makes them ideal candidates for a
variety of studies. The results of these types of studies,
made by investigators other than the author, are included
in this section in order to complete the presentation of
information gathered for these new compounds. The reader
is referred to the publications of the original investiga-
tor(s) for full experimental details and data manipulation
techniques.

1. giffbrentja] Scanning Cglarigetry —- The thermal stabi-
lity of the new compounds that contain sodium is greater
than that of the other new alkalides. In general, most
sodides are stable up to their melting point (if heated
rapidly) and decompose irreversibly at higher temperatures.

Two of the new compounds, K+(l5C5)2-Na- and Rb+(1505)2-Na-,

-95-

were studied by differential scanning calorimetry (DSC) by
J.L. Dye at AT&T Bell Laboratories, Murray Hill, New Jersey
[43]. The DSC trace for Rb+(l5C5)2-Na- is given in Figure
23. There is a small endothermic transition at 35’C which
may be due to the melting of excess crown ether, evapora-
tion of remaining traces of solvent or the presence of
electride. The compound melts at 75' and then decomposes
irreversibly. Na+0222-Na- melts at 73'C and all other
alkalides and electrides tested were observed to melt at
temperatures below 70'C. Therefore, Rb+(15C5)2-Na- is the
most stable alkalide to date. Rapid decomposition begins
at ~100°C and is marked by the large exothermic peak at
121'C in the DSC trace. K+(15C5)2-Na- was observed to melt
at 45°C with no prior endothermic peak. Decomposition was
only 10* at 97'C and reached a maximum rate at 108°C. The
thermal stability and ease of synthesis of Rb+(1505)2-Na-
and K+(15C5)2-Na_ make these compounds the most feasible
for possible utilization as two-electron donors for
difficult reduction reactions.

2. gEray Diffraction Studies -- The crystal structure of
Rb+(15C5)2-Na- was determined by X-ray diffraction studies
of a single crystal. Measurements were made by 0. Fussé
and D. Ward with a Nicolet P3F Diffractometer. The
rubidium cation is sandwiched between two 15-crown-5 mole-
cules as suggested by the stoichiometry. The crown ether
rings are staggered such that the Rb+ serves as an inver-

sion center. The ten oxygen atoms are more-or-less at

-95-

.uaz.~ .
AmUmH-+nm you mouse own
.mm onsmwm
cc- 3. ON 3.. wcah<¢wa5mh
. - O: 00. OO 00
on 8
-

a1]. [a
d d
— q
- u
- . - .
A d

J -. a

10m

 

..
.-
-.
..
- -
. n
.- -
. ._

.- - no?
u a
. '

Il‘lIl-u‘inillllul. \\\\ III‘
1 cm

\\
loo

 

~
N
.
~
18.

 

(Mm) MO‘IJ .LVBH

-97-

equal distances from the Rb+ and the ether rings lie planar
and parallel to one another. The "expanded" cations are
stacked in ABAB repeating layers with the crown ether rings
tilted a few degrees from the column of stacking. The
sodium anions are situated in the pseudo-octrahedral hole
with a radius of 2.4 to 2.6 A. The reader is referred to a
forthcoming paper for the full crystallographic description
of Rb+(15C5)2-Na_.

3. ‘EZAFS and 144;; -- Extended l—ray Absorption Fine
Structure (EXAFS) and fi-ray Absorption gear Edge Structure
(XANES) give information on the physical and electronic
environment of a nucleus, respectively. The K-edge X-ray
absorption spectrum of Rb+(1505)2-Na- is given in Figure
24. The spectrum has three features that were used to
identify the environment of the rubidium nucleus. First,
the location of the X-ray absorption edge can be used to
distinguish between two oxidation states of a nucleus.
Second, the relative area of the threshold resonances up to
40 eV from the edge may also be used to identify the elec-
tronic environment of the nucleus (XANES). Third, the
modulations at 40 eV and beyond, after Fourier transforma-
tion may be compared to those of compounds with known
structures. By the use of EXAFS, structural information
may be obtained even when the crystals are too small or of
too poor quality for single crystal X-ray diffraction
studies. This information is usually limited to neighbors

within a 4 A distance from the nucleus under study.

-93-

PX

 

\J

LLLLLLLLIlllLLllllL-LLIIIIIAJ

I4750 ISOOO ~I5250 I5500 I5750 IBOOO
Photon energy in W

 

Figure 24. K-edge X-ray gbsorption Spectrum of
Rh+(15c5)2-Na . -

-99-

O. Fussé and coworkers [64,65] have studied the
rubidium environment of Rh+(15c5)2-e", Rh+(15c5)2-Na“.
Rb+(1505)2-Rb+ and be(1505)2. The energy of the K—edge
absorption threshold of Rb- is only 2 eV lower than that of
complexed Rb+ salts. Therefore, the relative edge posi-
tions are not good indicators of the formal oxidation state
of rubidium. XANES is used to assign the oxidation states.
The area, A, of each absorption threshold resonance ("white
line") was normalized to unit edge jump and compared to
that of complexed rubidium model salts and known rubidides.
The samples were allowed to react with air and the X-ray
absorption of the decomposed sample was measured. The

results of these studies are given in Table VI below.

Table VI. Relative Normalized Areas8 of Rubidium K-Edge
Absorption Threshold Resonances.

 

 

 

Compound A (pristine)b A (oxidized)?
Rb+(l505)2-e_ 36 36
ah+(15c5)2-Na‘ 29 34
Rb+(l505)2-Rb- 18 29
Cs+(15c5)2-Rh’ 1+6 36
K,Rb(1505)2 16 34

8Relative uncertainty is :4 x 10'2 cm.
b
cm x 100

-100-

The area of the white line for the complexed model
salts and the oxidized alkalides and electrides ranged from
25 to 36. ah+(15c5)2-e‘ and Rh+(15c5)2-Na' have areas
equal to 37 and 29 respectively which indicates that the
majority of the rubidium is in the +1 oxidation state in
these compounds. This assignment is further confirmed by
the small change in the area upon oxidation. 0n the other
hand, the white line of Cs+(15C5)2-Rb_ has a very small
area of only 1 to 6 which is on the order of that observed
for the pure rubidide, Cs+18C62-Rb— (A = 1). Upon oxida-
tion the area changes to 36 for Cs+1505)2-Rb- and to 30 for
Cs+(18C6)2-Rb-

The compound with stoichiometry Rb(1505) was thought to
have rubidium in both the +1 and -1 oxidation states. The
white line area of 18 lies in between the values observed
for pure rubidides and compounds with only complexed
rubidium cations. The value of A increased to 29 upon
decomposition thereby confirming the assignment of
Rb+(15C5)2-Rb- for this compound. It was hoped that the
ambiguity over the anion assignment of RbK(15C5)2 could be
resolved with rubidium XANES. Again, the area of the white
line had an intermediate value of 16 which increased to 34
upon oxidation of the sample, indicative that there is a
mixture of oxidation states as in Rb+(1505)2-Rb-. From
this result and the stoichiometry, it is evident that there
is a mixture of crystals having both potassium and rubidium

present as sandwich cations and anions.

-101-

The crystal structure of Rb+(1505)2-Na— allowed the
EXAFS of this compound to serve as the model result. The
Fourier transform of the spectrum of Rb+(15C5)2-e- is
virtually identical to that of Rb+(1505)2-Na- where the
rubidium is sandwiched by two crown ether rings. The EXAFS
and stoichiometry' of this compound identify it as a true
electride. Likewise, the EXAFS spectra of Rh+(15c5)2-Rh'
and the mixture, RbK(15C5)2, are consistent with the
presence of both Rb+(l5C5)2 and Rb-. The EXAFS spectrum of
Cs+(1505)2-Rb- shows only very weak modulations as observed
for krypton gas. Due to the large radius expected for Rb-,
the rubidium anion is expected to behave as an isolated ion
in the extended X-ray absorption region. Cs+(1806)2-Rb-
also has a very weak fine structure such that no structural
information may be obtained for the Rb_ environment [64].

At this time EXAFS and XANES have been successfully
applied to rubidium alkalides and electrides. Although
sodium and potassium EXAFS studies are feasible, the K-edge
of these elements lie at energies too low for the available
experimental set up. Cesium K-edge X-ray absorption does

not yield useful information due to fast core-relaxation

effects. Some studies were made for the L-edge of cesium
with fluorescence detection. The signal to noise ratio
was, however, too low to provide useful information about

alkalides and electrides. Although EXAFS and XANES can
clearly identify rubidium compounds, another nuclear speci-

fic identification tool is needed for the other alkalides

and electrides.

PART II:

Alkali Metal NMR Studies of Alkalides.

PART II

CHAPTER ONE -- INTRODUCTION

The identification of alkalides and electrides has
traditionally been based on the stoichiometry and optical
absorption spectrum of each compound. However, the species
in the films which are used for optical studies, may or may
not, be the same as those present in the crystals.
Recently, X-ray absorption near edge structure (XANES)
studies have been used to identify crystalline alkalides
and electrides that contain rubidium [64,65]. Unfor-
tunately, due to instrument availability and other experi-
mental limitations, this technique cannot be used to probe
the environment of the other alkali metals.

Nuclear magnetic resonance provides a direct probe of
the electronic environment of a nucleus since the resonance
frequency of the nucleus is dependent on the shielding by
the electrons. This chapter will cover some nuclear
properties of the alkali metals, the application of alkali
metal NMR to study alkali metal anions in solution, solid
state NMR and finally, an example of magic angle spinning

(MAS) NMR as an identification tool.

-102-

-103-

A. Nuclear Properties of Alkali Metals

Each alkali metal has at least one magnetically active
isotope. Table VII gives some nuclear properties of the

alkali metals. Most commonly, NMR studies involve the iso-

topes 7Li, 23Na, 39K and 133Cs due to their high natural
abundances. In the case of rubidium, however, the less
abundant isotope is the nucleus of choice. The relative

sensitivity is dependent on the cube of the magnetic moment
as well as the natural abundance and nuclear spin, so the

larger p" for 87Rb determines the larger sensitivity for

that isotope than that of 85Rb.

The Sternheimer antishielding factor, 1+7“, is a
measure of the amplification of the field gradient of the
nucleus produced as a result of polarization effects in the
electron cloud induced when the atom, as a whole, is
exposed to an electric field gradient. Values of 1+Y“ are
determined from the fine structure in electron spectra of
atoms by the use of quantum chemical calculations similar
to those leading to electrical quadrupole moments. The
values listed in Table VII refer to the free cations and
will change, if the the electronic wavefunction becomes
significantly altered by strong ion-ion or ion-solvent
interactions.

For several of the alkali nuclei the quadrupole moments

are large enough to make quadrupole effects dominate the

NMR spectra. Linewidths are a reflection of the reaction

 

.bm mososcmom

.mlsuo> enema

case use saw: was macaw ofiaosusfl onus on» an em: mo gunman :22 caucus one 09 vessmmomfin
.Om mososomom

A0

 

 

 

---------..-----------iii}-----------....-----------:1}---..---------.i-------u- ........ hm
~HH ~-c~ x «p.v moo.on vmm.N cofl «\e moons

N.m¢ 5H.O NH.O ~¢p.N m.vN «\n amps

N.m¢ ~-O~ x mO.H ON.O mvm.~ N.Nh «\m ammo

«.mH n-o~ x ¢.m smo.o seem.o mm.m «\s use

% m.m~ e-oH x N.w so.o1 mmN.~1 Nfic.o e was
w N.NH v-OH x NO.m OO0.0 _ON.O ~.NO «\o goo
-

H.m ~-OH x ON.O OH.O OHN.N OOH «\m Mme
¢>.o mN.o Neo.ou mmN.m m.Nm «\n “An
vb.O n-O~ x m.m v-O~ x m.¢ NNN.O ¢.b H “no

Nsx+av .souomH mafi>wuwmsom mason usoaoz . llzm a H .swmm msoaosz
wswvaownuwaso .oo>waoaom sodomssvozo .usvloz .oossvssn< mummfiosz
omelwoassmum A auguoswnz masssaaz

E"l'I'l'I'll||'|l-'ll‘l'llsll'lll'l'1'"-.II'"|--|"""'-"'lll-'||l'"l""I'l'|"|l"|lll"l:'
:l|"lll|'I'Il'-""llIl.I-I"Illl|sllIll'llllllll'llIII-"I'IIII"IIIII|I‘IIIl|l|'|'|||"||l"l"l'l'é

.muflaol wuflxud «0 mvwahomchm huvuosz .HHD MAO<B

-105-

rates and the efficiency of the quadrupole relaxation and
depend upon the square of Q(l+y~). The quadrupole moment
and Sternheimer antishielding factor are so small for 7Li
that the relaxation is relatively inefficient and the
natural linewidths are often less than 1 Hz. In contrast
to 7Li, the linewidths of 85Rb and 87Rb cations in aqueous
solutions are on the order of 150 Hz. The extremely small
quadrupole moment of 133C3 makes the study of cesium as
favorable as that for sodium.

Numerous NMR studies of alkali metals in solution have
been made to investigate ion-ion and ion-solvent interace
tions [67-69]. These investigations also include the study
of alkali metal complexation by organic macrocycles such as
crown ethers and cryptands. Due to the low sensitivity of

39K [70] and the large linewidth of 87Rh [71], the majority

of these studies have been made with 7Li, 23Na and 13303.

B. NMR of Alkali Metal Anions in Solutions

The first direct proof of the centrosymmetric anion as
the species with stoichiometry M_ in amines and ethers was
obtained by 23Na NMR studies of nonaqueous solutions that
contained cryptand-2,2,2 and excess sodium metal [72,73].
Two signals were observed at ~-10 and -62 ppm from Na+(aq)
at infinite dilution and were assigned to Na+C222 and Na-,

respectively. The Na- peak was shifted paramagnetically by

only 1-2 ppm from the calculated values for the gaseous

-106-

atom [74]. Calculations based on Lamb’s complete expres-
sion for atomic diamagnetic shielding [75] and analytic
Rartree-Fock wavefunctions yielded a chemical shift for
Na_(g) of -63.1 ppm [73]. Pyper and Edwards have combined
the Hartree-Fock predictions with electron-electron
correlation corrections to obtain chemical shift values for
all alkali metals [76]. They predict a value of -63.3 ppm
for the chemical shift of Na-(g).

The report of the first observation of Na- by alkali
metal NMR was followed by other studies of Na- in non-
aqueous solutions with and without cation complexants
present [29,77-80]. Edwards and coworkers have observed
Na- in two component solutions of sodium metal in neat 12—
crown-4 [29], in hexamethylphosphoric triamide (HMPA) [77]
and in N,N-diethylactamide (DEA) [78]. No cation signal
was observed in any of these solutions due to the extreme
broadening of the line by cation-solvent interactions
and/or by interactions with solvated electrons. Both Na+C
and Na- were observed in solutions of cryptand-2,2,2 and
sodium metal in methylamine (MA), ethylamine (EA) and
tetrahydrofuran (THF) [73]. For solutions of sodium metal
and 18-crown-6 in MA, the two NMR signals collapsed into a
single broad peak at a chemical shift intermediate to those
of Na+(18C6) and Na_. This temperature dependence of the
spectrum is due to higher rates of exchange at higher tem-
peratures. The kinetics of the sodium cation-sodide ion

exchange has been investigated by Phillips, et a1. [80].

 

-1o7-

The anions of rubidium and cesium have been observed by
alkali metal NMR for solutions of cryptand 2,2,2 and excess

37Rh‘ has been

metal [73]. In addition, the signal of
obtained for rubidium in neat lZ-crown-4 [29]. The
complexed cation was not observed for either metal in these
solutions, except for Cs+C222 in THF [81]. This is not
surprising in the case of Rb-C since the linewidth is
extremely large for Rb- and the cation is expected to be
less shielded and therefore have broader peaks. The line-
width of Rh. ranged from 15 Hz in THF to ~1000 Hz in 12C4.
Khazaeli, et 81. reported the absence of a Rb+ signal when
the complexant-rubidium ratio was sufficiently high to

87

ensure complete complexation of Rb+ [71]. The Rb+ signal

broadened as the complexant-rubidium ratio increased until

it was too broad to measure.

133Cs+C signal in most

solutions is not clear. Cs+C222 was observed by 1330s NMR

The reason for the absence of a

in an 0.08 M solution of Cs+0222-I_ in methanol. The
chemical shift of the Cs+C was ~161 ppm more paramagnetic

than the signal observed for a 0.08 M methanol solution of

Cs+I-. The linewidths were <3 Hz for Cs+ and ~30 Hz for

Cs+C222 and are indicative that the cation is truly com-

133

plexed at 25'C. P. Smith [81] observed a Cs+C222 signal

in THF solutions of cryptand—2,2,2 and cesium metal at low
temperatures (-64 to -86'C). The peak was probably
broadened by rapid exchange between inclusive and exclusive

133C

complexation of the cation. The 3+0222 peak became

 

-108-

narrower as the temperature decreased, indicative of an
exchange reaction.

Table VIII gives the calculated chemical shifts for
"(g) and M-(g) as well as the values observed for M— in
solution. All previous attempts to observed K- have been
unsuccessful [73,79,82]. The chemical shifts deviate from
the calculated values of M-(g) as the atomic number
increases. This deviation may be due to more polarization
by the anion—solvent or anion-cation interactions or by
greater mixing of p and/or d character into the ground

state 3 function of M_ as the atomic number increases.
C. Solid State NMR

A notable feature of NMR is that the linewidths in the
spectra of solids are much broader than those observed from
solutions. An excellent example of the extreme difference
that can occur between the solid state and solution line-
widths is given by the proton spectra of H20. Liquid water
has a linewidth of 0.1 Hz while that of ice is on the order
of 105 Hz. This substantial difference arises from the
static anisotropic interactions in solids which are
averaged‘ by rapid isotropic motions of the nuclei in fluid
samples. Sometimes there is sufficient motion in solids to

narrow the NMR signals, but these cases are the exceptions,

not the rule [83].

Solution.8

1!]

Calculated Chemical Shifts fgr Gaseous Alkali Metal Atoms and Anions and

Observed Chemical Shift of M

TABLE VIII.

 

Reference

Solvent

II a

n M E
II -H
"‘68
now

II >
Ilhv-I A
1100 >
not: H
llD’H O
IIO I m
II 0 v
H Al

N o:
u

u

n

H

MD +3
“'66:
IIO-HI
u a.:

II can
IIH
"5H
IIUU A
Made) u
Ila-H v
u 0 El

u a):
II .A

ll 0

n

n

u

u

IUD H
"'64:

H 0-H
Il+hfl I
H flea
IIH

u s-«

n U o
"HO A
n o-u 1»
u o I ~/
II a):
II .A

ll 0

Nucleus

77
29
73

HMPA

-61.7

-63.l

1

-60.5

23Na

73

1204
MA
EA
THF
DEA

—61.8
-61.9
-62.1
-62.8
-62.9

-109-

78

not observed

-103.4

.5

-101.1

39K

73
29
73

EA
12C4
THF

~185.4

-191
-197.2

-213.6

.1

-211.6

87Rb

73

THE

-292

-346.4

6

-344.3

13308

at infinite dilution.

(aq)

elative to M

R
)Reference 76.

)Reference 74.

C

"a A

-110-

Several experimental techniques have been implemented
to reduce the anisotropic broadening in solids. One of
these, which is called magic angle spinning (MAS), will be
discussed in this section. The types of anisotropic
nuclear interactions that are of interest are magnetic
dipolar, chemical shift anisotropy and electronic quadru-
polar interactions. Numerous review articles have been
written on the NMR of solids and the magic angle spinning
technique [84-88].

The truncated dipolar interaction Hamiltonian, for all
nuclear pairs i,j in the solid is given by:

3

2 -
: 1 . _
7( Z /2Yirjfi ri.(I. I. 31

2
D i<j J ~1 “J iszz)(3 cos eij—l) (11)

where Yi’ Yj are the nuclear magnetogyric ratios, rij is

the internuclear distance, and eij is the angle between rij
and the direction of the field Ho' The isotropic average
of coszei‘j is 1/a, so that the rapid isotropic motion in
fluids yields no net dipolar contribution.

When a solid specimen is rotated uniformly with angular
velocity "r about an axis inclined to H0 at an angle 8,
every internuclear vector, rij’ may be described by the
motion illustrated in Figure 25. The angle ei’ may be

J
expressed by other angles in the system as follows:

_ I . . I
cosei‘j — cosB cosBi‘j + Sine s1nBiJ cos(urt+¢oij). (12)

Figure 25.

 

-lll-

 

Diagram Illustrating the
Motion of.a Typical Internuclear

Vector, ri-, when a Solid is
Rotated With an Angular Velocity,
wr, about an Axis Inclined at
Angle B to Ho'

-112-

By substituting equation (12) into equation (11) we get,
after rearrangement:
3

.(I.-I.-3I I )

2 -
= 1 . .
){ 2 /2Yinh rij ~1 ~j 12 jz

D i<j

{1/s(3 cosZB-l)(3 coszegj—l)

3 . o I
+ /2 sinZB ainZBi‘j COS(urt+¢OiJ)

+ 3/2 sinZB sinZB;J cos2(urt+¢oij)} . (13)

Note that, the first term in the brackets is independent of
time while the second and third are periodic with mean
value of zero. If the angle of rotation,B, is chosen such
that 3 cosZB-l = 0, then the dipolar effects will be seen
only as rotational sidebands at multiples of or from the
central peak. The value of the appropriate angle is
54'44’8” and is referred to as the magic angle. The
effect of high speed rotation of the sample about an axis
at the magic angle on chemical shift anisotropy and quadru-
polar interactions is similar and is described below.

The electron shielding or shift interaction of nuclei

in nonmetals may be written as:

:fs = fi§(7i°izzno) (14)

-113-

If the principal values of the chemical shift tensor are

1,2,3) and the direction cosines of

represented by °p (P

its principal axis with respect to “0 are represented by

x , then
P
a 3 2X20 (15)
27- P P
P
for nucleus i. The isotropic average of each x: is 1/3 so
that the value of 622 in a normal fluid is simply a, the

scalar chemical shift encountered in the high resolution
NMR spectra of fluids.

When a solid is rotated with angular velocity at about
an axis inclined at angle 8 to Ho and at angles xp to the

principal axis of a, the direction cosines become

xp = cos 8 cos xp + sinB sinxp cos(urt+Qp) (16)

Substituting (15) and (16) into (14) gives an equation that
is time independent and periodic in Or. The time indepen-

dent chemical shift anisotropy becomes:

6 = 3/2o sinZB + 1/2(3 coszfi-l)Za coszx . (17)
22 p p p

When 8 is chosen to be the magic angle, 322 reduces to the
scalar isotropic value a and the shift anisotropy is

removed from the central line. The periodic terms of ff:

contribute to the spinning side bands just as for the

-114-

dipolar time-dependent terms.
All alkali metal nuclei, except 6Li, are quadrupolar.
The electric quadrupolar interaction Hamiltonian may be

written as

c 8°“
} e z 1 I.-V.-I. (18)
Q i 611(211-13 1 1 1

where eQi is the nuclear electric quadrupole moment of
nucleus i and Vi is the electric field gradient tensor at
the nucleus. V1 is a second rank tensor that is symmetric
and traceless. The effect of high speed sample rotation on
the quadrupolar interaction in solids has been treated by
Cunningham and Day [89]. The first-order time-averaged
diagonal terms of the quadrupolar Hamiltonian are given by
ezoq

5E0::161(21‘1) (3IZ-Iz)(l+n)(l—3 cos28)(l-3 cosze) (19)

 

where n is the asymmetry factor and e is the angle between
the principal axis of the electric field gradient tensor
and Ho' At an angle 8 = 54°44’, the first order quadrupole
interaction becomes zero in the limit of infinite rotation
frequency.

The second—order quadrupole effect is not averaged to

zero in the MAS experiment. In fact, in some solid state

-115-

experiments, angles other than the magic angle are more
effective in linewidth reduction [90]. The area of solid
state NMR spectra of quadrupolar nuclei has been a field of

great interest, especially for the study of inorganic salts

[90-97].
D. Identification of a Ceside and an Electride by 1330s
MAS-NMR

Historically 23Na was the first nucleus in crystalline
alkalides to be studied by solid state NMR [23,98]. The
results of these studies are given in Part II-Chapter Three
of this dissertation. Later, Ellaboudy applied 1330s MAS-
NMR to the identity problem surrounding the compound with
stoichiometry Cs(1806) [42,99]. This section describes the
problem and the NMR results as an example of the usefulness
of MAS-NMR as an identification tool.

The first Cs(18C6) crystals were isolated by Issa
[41,52]. The optical absorption maximum was located at
6700 cm—1, which corresponds to trapped electrons. The
magnetic susceptibility data indicated that this compound
was virtually diamagnetic with only IX unpaired spins. The
single crystal spectrum featured two lines, one temperature
independent centered at g = 2.0023, the other, which

broadened with increasing temperature, at g = 2.0016. The

crystals were stable for hours at room temperature and

-116-

melted to a dark blue liquid at ~65'C before decomposing
irreversibly. D.C. conductivity measurements yielded a
small apparent band gap of ~0.6 eV.

The results obtained from the characterization studies
of Cs(1806) suggested that if this compound was the elec-
tride, it had extensive spin pairing. Spin pairing had
been observed for other electrides [48,55], so the nearly
diamagnetic behavior was not surprising. This compound was
considered a possible electride because: (1) The single
peak at 6700 cm_1 observed in the optical spectrum corres-
ponds to e- trapped. (2) Crystals of Cs+(1806)-Na— had
been isolated and characterized [49]. Therefore, it was
surmised that other cesium 18-crown-6 salts would have a
Cs:18C6 ratio of 1:1. (3) The observed band gap was too
small for a ceside salt.

Ellaboudy synthesized a second compound from solutions
of cesium and 18-crown-6 which had the stoichiometry
Cs(1806)2. The optical absorption spectrum of films made
from solutions of Cs(18C6)2 in methylamine featured a
single maximum at 6700 cm—1. As in the spectrum of
Cs(1806), this peak corresponded to the trapped electron.
However, the magnetic susceptibility data were radically
different. Cs(1806)2 had a calculated Curie constant that
corresponded to ~75% unpaired electron spins and a very
small Weiss constant, indicative of weak electron-electron
interactions. The EPR spectrum had a single intense signal

with a g value of 2.0023 and a peak to peak width of only

-117—

0.48 gauss. independent of temperature. Based on these
results the two compounds, Cs(18C6) and Cs(1806)2, were
tentatively assigned to Cs+(1806)2-Cs- and Cs+(1806)2-e-,
respectively.

Definite proof of these assignments was obtained by

1330s MAS—NMR. Figure 26 shows the spectra of
Cs+(18C6)2-Cs— and Cs+(l8C6)2-e-. The former salt exhibits
two cesium peaks at -61 and -228 ppm. The more para-
magnetic chemical shift corresponds well with those

observed for model salts with a Cs:1806 ratio of 1:2. The
second peak at -228 ppm is reminiscent of the Cs- peak at
-292 ppm observed in THF solutions [73]. The peak at
-228 ppm was assigned to 05-. By contrast, the spectrum of
Cs(18C6)2 had only a single peak at +81 ppm. The large
paramagnetic shift is probably due to the high concentra-
tion of unpaired electrons in this compound. The identifi-
cation of this compound is clearly Cs+(1806)2-e_.

The 1330s

MAS-NMR results proved unequivocally that
Cs(1806) was the ceside, Cs+(1806)2-Cs-. The optical spec-
trum maximum must arise from the reaction during film

deposition, such as

+ — + —
Cs (18C6)2-Cs -—-) Cs (18C6)2-e + Cs(s) (20)
and suggests that the electride is thermodynamically
favored. Extrapolation at the D.C. conductivity results to

infinite temperature indicated that Cs+(1806)2-Cs- is an

-118-

.N32 VN.¢¢ um

m
it. -coma- no can u N
use + t o. -m0ma- no .n :
.Hacunsuo 039 use no nauocem+mzzimmzammsoo .
ommH cm enemas
Eon .
con - zen-.8 Ease-em 8-526
1 CON- 8.1 O 00.
oo~ OOn

 

 

 

- . - 14 ;-- . - .4. 1a fil
‘1- .---
. . . -

 

    

 

nan-coarse

 

 

 

 

 

.au .n-mum: .8

  

 

 

 

-1l9-
extrinsic semiconductor and that the apparent band gap is
due to the presence of trapped electrons as impurities.
This resolution of the "identity crisis" for Cs(1806)
demonstrates the usefulness of MAS-NMR as an identification

tool for alkalides and electrides.

PART II

CHAPTER TWO -- EXPERIMENTAL

A. Sample Preparation

The reagents, solvents and crystalline alkalide samples
were prepared as previously described in Part I, Chapter
Two of this thesis. The present chapter will describe the
exact handling of the NMR samples starting with crystalline
alkalides and the instrumentation used to record the
spectra.

1. Solution samples - Crystalline alkalide samples were
loaded through the stopcock of an evacuable 10 mm thin wall
NMR tube under a dry nitrogen atmosphere. The apparatus
was closed, removed to a vacuum line and evacuated to 2 x
10_5 torr. The scribed sample tube had been weighed before
breaking so that a mass determination could be made. For
samples in which an excess of l5-crown-5 was desired, the
evacuated apparatus was brought into a helium-filled inert
atmosphere dry box and a weighed amount of crown ether was
pipetted through the stopcock onto the crystals. Dimethyl
ether was distilled onto the crystals or crystal-crown

ether mixture until a solution height of about 4 cm was

reached. The walls of the apparatus were washed to

-120-

-121-

dissolve any crown ether or crystals adhered to them. The
apparatus was then immersed in a dry ice-isopropanol bath
at -78'C and opened to a rough vacuum. A flame seal-off
was made at a constriction above the NMR tube at the vacuum
of 9;. 4 x 10—3 torr. The sealed tubes were stored at
-80’C until use.

2. .Mbdel salts - Conventional alkali metal salts were
complexed by crown ethers or cryptands by procedures simi-
lar to those of Pedersen [26]. The salts, together with
stoichiometric amounts of complexant (mole ratio 1:1 or
1:2), were dissolved in either hot methanol or 2-propanol.
The solutions were allowed to cool until crystals formed
and were then filtered to yield the 18—crown-6 and cryp-
tand-2,2,2 salts. The 15-crown-5 model salts were allowed
to dry at room temperature. Model salts were further dried
in a vacuum desiccator. Conventional salts without
complexant were reagent grade and were used without further
purification.

3. Solid state samples - Crystalline alkalides or elec—
trides were loaded into either Andrew Beam type rotors made
from Delrin or cylindrical A1203 rotors with Eel—F turbine
and end caps. The cylindrical rotors required special Kel-
F caps with holes in the centers and a Kel—F bolt and nut
for low temperature experiments. Each type of rotor was
loosely filled with crystals while cold under a dry nitro-
gen atmosphere. The rotors were then capped and brought

slowly from static to spin rates of 1.2 to 4 kHz. This

-122-

allowed for centrifugal packing of the samples within the

rotors. Model salts could be tightly packed in the cylin-
drical rotors prior to spinning by use of a plunger. Spin
rates are significantly higher when the samples are

uniformly packed.

B. Instrumentation

Spectra were recorded at three field strengths with
instruments located at Michigan State University, Dow
Chemical Corp. (Midland, Michigan) and the National Science

Foundation Regional NMR Center (University of Illinois;

Urbana, Illinois). Single 90' pulse sequences were used
for all experiments. Downfield (paramagnetic) shifts are
positive. A brief description of the instrumentation at

each location follows.
I. At Michigan State university - A Bruker WH 180 super-
conducting multinuclear spectrometer equipped with a

Nicolet 1180 computer system was used. All experiments for

39 87

K and Rb in solution were made without a deuterium

lock. The 39K resonance was operated at a frequency of
8.403 MHz at a field strength of 4.227 Tesla. A Bruker
10 mm probe tunable in the range of 4.8 to 12.2 MHz was
used for the 39K solution studies. 89Rb has a resonance of
58.9 MHz in this field strength and a homebuilt, broad
band, 10 mm probe tunable in the range of 25 to 103 MHz was

used for the solution studies.

-123-

Solid state samples were studied by the magic angle
sample spinning (MAS) technique. A commercial MAS probe
from Doty Scientific was used. This probe required cylin-
drical rotors with dimensions of 18 mm x 7 mm OD and was
tunable to the 45 to 72 MHz range. Additional capacitors
were used with the tuning impedance to bring the probe

range to lower frequencies for measurements of 133Cs at

23.62 MHz and 39K at 8.403 MHz. An external standard was
used to determine the chemical shift from the value of the
reference salt in aqueous solution at infinite dilution.
Spectral drift did not exceed 3 Hz during the time period
of the experiment. The Doty MAS probe also featured proton
decoupling capabilities at 180 MHz. Decoupling experiments
could be studied only at low frequencies due to additional
noise present at higher frequencies.

Samples were rotated at speeds up to 3 kHz by precooled
nitrogen gas. Variable temperature experiments in the
range of 160 to 300 K could be made with a Bruker variable
temperature control unit. In the case of alkalide samples,
spinning speeds were limited by sample packing and the
effect of the center bolt of the low temperature turbine
caps.

2. At Dow Chemical Corporation - A Bruker CXP-ZOO wide bore
spectrometer equipped with an Aspect 2000 computer was
used. Andrew-Beam type rotors with a volume of ~0.4 cm3
were used in a standard commercial MAS probe. Spectra were

23 87

taken of Na and Rb at frequencies of 52.94 and 65.4 MHz

-124-

respectively, in a field strength of 4.7 Tesla. The
spinning gas (nitrogen) was cooled by passing it through a
copper cooling coil submerged in a dry ice-isopropanol
bath. No temperature control was available and sample tem-
peratures were in the range of ~20 to -10'C. Spinning
rates up to 4 kHz were achieved.

3. At the NSF Regional NMR Center - A "home built" Fourier
transform NMR spectrometer with an 11.74 Tesla field
strength was used. This system consisted of an Oxford 3.5
inch bore solenoid, a Nicolet 1280/293B computer system and
a Control Data Corporation disk memory. Andrew-Beam type
rotors were also used with this system. As at Dow Chemical
Corporation, the spinning gas was cooled by a dry ice slush
bath but temperature regulation was not available. Spin

rates up to 4.5 kHz were observed.

PART II

CHAPTER THREE -- RESULTS AND DISCUSSION

Alkali metal nuclear magnetic resonance spectroscopy is
a sensitive technique for the study of the ions present in
alkalides and electrides. MAS—NMR has been demonstrated by
Ellaboudy, et al. to be a powerful identification tool for
cesium compounds [42,53]. This chapter covers the results

of MAS-NMR studies that include the nuclei: 23Na, 39K, 87Rb

1330s. 87

and In addition, some advances in 39K and Rb NMR

studies of solutions have been made and are reported.

A. 23Na NMR Spectra

23

The 100 percent natural abundance of Na and its rela-

tive NMR sensitivity permits easy data collection. The
success of 23Na NMR studies in solution made sodium a
natural first choice for solid state NMR studies. The pre-

23Na NMR spectrum of Na+C222-Na-, obtained without

liminary
spinning, consisted of a broad peak, with linewidth of
~2600 Hz, centered at ~-60 ppm relative to Na+(aq) at infi—
nite dilution [23,98]. A barely discernible, broad

shoulder could also be seen on the low field side of the

-125-

 

~126-

peak at the peak location of Na+C222-I-(s) (~-13 ppm).
These peak locations can be easily compared to the spectrum
of Na+0222-Na- in various nonaqueous solvents, as described
in Chapter One-Part II. The ~-60 ppm peak and low field
shoulder resolve into two clearly discernible peaks upon
rotation of the sample at an angle of 54.74' from the
direction of the field. Figure 27 demonstrates the drastic
reduction of the linewidth of Na+C222-Na- upon spinning at
this "magic angle". The two peaks at ~22 and ~61 ppm are
assigned to Na+0222 and Na-, respectively. The location of
the Na- peak is within experimental error of the chemical
shift calculated for Na_(g) [73,76].

The 23Na MAS-NMR spectra were recorded for several
alkalides that contain sodium. Table IX gives the chemical
shifts and linewidths of some conventional salts, both with
and without complexing agents, and a number of crystalline
alkalides. All heteronuclear alkalides had a single 23Na
NMR peak in the range of ~56 to ~63 ppm which corresponded
to the peak at ~61 ppm for Na- in Na+C222-Na—. The results
prove that these compounds are all sodides. The concentra-
tion of Na+, either free or complexed, is too low in these
salts to give a detectable NMR signal. The 23Na MAS-NMR
spectrum of crystals with stoichiometry Na(1204) feature
two peaks at ~14 and ~62 ppm. Analogous to the case of
Na+C222-Na-, these peaks correspond to the complexed
cation, Na+(1204)2, and Na_, respectively. The 23Na MAS-

NMR spectrum for Na+(1204)2-Na- is given in Figure 28.

—127-

 

 

 

 

 

 

 

Figure 27.

8 [ppm FROM NO+ (oq.)]

23

MOI 0222 . ~01
Na
(NON-SPINNING)
NO+ Ic3222
(SPINNING)
1 l L l 1 l 1 l 1
+100 0 -|OO -zoo

Na NMR Spectra of Crystalline Na+C222-Na-

at 52.94 MHz. Top - static spectrum;

bottom - MAS spectrum.

 

~128-

TABLE IX. Results of 23Na MAS-NMR Studies.

Compound Q, ppm Apl/z, Hz Reference

NaCl +8 140

NaBr +6 200

NaI ~3 160
Na’(12C4)2-Na‘ -14: 760
Na’(12C4)2-Ng' ~62 170

K’(lZC4)2-Na‘ ~61 51 38
Na’lSC5-Br’ +4 ‘ 580

K’(1505)2-Na‘ ~61 90
Rb‘(1505)2-Na' ~61 90
Cs’(15C5)2-Na‘ ~61

Na’lBC6-SCN' ~22 500

E’lBCS-Na' ~56 120

Rb’lBCS-Na' d ~60 90
Cs’(1806)2-Na’ ~63 75 53
Na‘CZZZ-SCN' ~21 1120

Na’C222-I' ~21a 1300

Na’C222-Na' ~24b 1200

Na’CZZZ-flp' ~61 290

K’CZZZ-Na' ~61 400

Rb’0222-Na' ~61 360

8Chemical shifts are referenced to Na’ aqueous at infinite
dilution. Chemical shifts at separate samples were

breproducible to :1 ppm.

Attributed to Na‘L

Attributed to Na‘.

First reported as Cs‘18C6-Na'. Later analysis showed this

compound had Cs*(1806)2-Na' as the correct stoichiometry.

d

~129-

.ums Ho.se on tnz.-so~H-+nz mo asuuomsm mzzumss n2mm .mn magmas
can .33 ea: soc..- .E-xm 43-5-6
co»... oo-.. 0 oo-+
— d u _ _ _ d d

 

 

 

 

 

~130-

The chemical shift of Na‘ is remarkably insensitive to
the nature of the cation of the sodide salt. Of the eleven
sodides examined, seven have identical shifts of ~61 ppm,
which is very close to the value of ~63 ppm calculated for
Na-(g) [76]. Nine of the sodides are within experimental
error of the value of ~62.3 ppm observed for Na- in solu-
tion. The Na- NMR peaks for K+(18C6)-Na- and Rb+(18C6)-Na—
are shifted downfield to ~56 and ~60 ppm, respectively.
The small paramagnetic shifts observed for these sodides
might be caused by the presence of trapped electrons that
have been detected by EPR spectroscopy [43,51,53]. The
small deviations of the Na- peak location from that of
gaseous Na- reflect the high degree of shielding of the 2p
orbitals by the filled 3s orbital.

The environment of Na+ affects its chemical shift much
more than that of Na-. The chemical shift ranges from ~24
ppm for Na+C222 to +8 ppm for Na+Cl-. Most of the com~
plexed sodium cations have more diamagnetic chemical shifts
than uncomplexed sodium. The exception is Na+(1505)-Br_
with 6 = +4 ppm. This peak is most likely due to partially
complexed Na+ which would allow the cation to participate
in more interactions with the environment. An extremely

poor signal to noise ratio was obtained for this spectrum.

Oldfield has shown that only the central (m = ~1/2, m =
1/2) spin transitions of 23Na can be observed for solids
when the quadrupolar coupling constant, equ/n, is

appreciably larger than the sample spinning frequency

~131-

[90,100,101]. The residual linewidths and frequency shifts
are determined by second order quadrupole effects [101].
These shifts could amount to 15-20 ppm for Na+0222 and up
to 8 ppm for Na-. The narrow spread of the chemical shift
values for Na_ and the good agreement with Na_ in solution
and in the gas phase suggest that these effects are
minimal.

Ellaboudy and Dye [102] have made line shape studies of
23Na NMR spectra of solid Na+C222-Na-. Their findings pre-
dict eZQO/n to be 1.2 t 0.1 MHz with an asymmetry para~
meter, n. of about 0.1 for Na+C222. Calculations of the
quadrupolar line shapes indicate that the residual dipolar
coupling contribution to the linewidths is approximately
300 Hz for a coupled MAS experiment. The dipolar contribu-
tion to the static linewidth was calculated to be 3340 Hz.
Combination with the predicted quadrupolar effects yields a
linewidth of 3670 Hz which is in good agreement with the
observed linewidth of 3700 t 300 Hz for Na+C222.

The linewidths of the sodides that contain crown ethers
when compared to Na- in M+C222oNa— suggest greater asym-
metry of the field gradient for Na- in the cryptate salts.
The relaxation processes may be enhanced by interaction of
the hydrogen or nitrogen atoms in the cryptand with the
sodium anion to give broader signals. The broadening of
the line by greater relaxation rate enhancement of the sig-

nal is also indicated by results obtained by Ellaboudy

[53]. Proton decoupling of the signal gave linewidths on

~132-

the order of 30-60 Hz for the cryptate sodides and 40-80 Hz
for the crown ether sodides. The linewidths of the crown
ether sodides are not drastically narrowed by proton
decoupling as that of Na- in M+C222-Na-. This result
further indicates a more symmetric environment and greater
isolation from the neighboring atoms for Na- in sodides
which contain crown ethers.

In summary, 23Na NMR has been shown to be a very effec-
tive identification tool for alkalides and electrides. One
new homonuclear salt with stoichiometry Na(lZC4) [54] was
shown to be the sodide, Na+(1204)2-Na-, rather than the
electride, Na+(lZC4)-e-. In all, eleven compounds were
examined and proven to be sodides. The chemical shift of
Na_ at ~56 to ~63 ppm is in good agreement with the calcu-
lated value of Na-(g) at ~63 ppm and the observed value for
Na- in solution at ~62 ppm. The chemical shift values
reflect the shielding efficiency of the two 3s electrons
for the 2p orbitals.

B. 39H NMR Spectra

The relatively low NMR sensitivity of all potassium
isotopes has caused potassium to be the least studied of
the alkali metals by NMR. In addition, the potassium anion

had never been observed by solution NMR studies. These two

factors made the use of MAS-NMR for 39K more challenging

than for 23Na or 133Cs.

~133-

1. Solution studies. The previous failure to observe K-

39

in solution by K NMR may have resulted from the equili-

brium:

K = K+ + 2e‘(solv) (21)
since K_ is more dissociated than Na—, Rb- or Cs- [20,25].
Rapid equilibration to give paramagnetic solvated electrons
(or intermediate species such as K-) would substantially
broaden the K‘ NMR peak. Therefore, a solvent with low
donicity, dimethyl ether, was chosen to dissolve crystal~
line K+(1505)2-K- for solution NMR studies. Dimethyl ether
(MeZO) dissolves potassium metal only in the presence of a
complexing agent and reaction (21) is expected to lie far
to the left of these solutions.

The spectrum of a 0.06 M solution of K+(15C5)2-K- in

Me O is shown in Figure 29. The peaks at ~9.9 and ~99.3

2
ppm from K+(aq) at infinite dilution are assigned to
K+(1505)2 and K_, respectively. The measured chemical
shift for the gaseous potassium atom is ~lOl t 5 ppm [74]
and the anion is calculated to be slightly more diamagnetic
at ~107 ppm. The negligible solubility of potassium in
Me20 rules out the presence of uncomplexed potassium. The
narrower linewidth of the K_ peak (20 Hz) compared with
that of K+(15C5)2 (50 Hz) reflects a more spherical
environment for K_. Edwards and Ellaboudy have also

recently observed K- in solution by 39H NMR [103]. Their

~134-

.umz os.s can x e- an .2 ce.o- uosuo
lamenessa as -s.-momH-+s so saucepan «:2 son .o~ mucosa

Tao- ..v- E9: Eon H—w

.009: nle 0%.. Av .00 AX!

b p p p -

 

 

w -
SUD-ox

 

 

~135-

solutions consisted of a 08K alloy dissolved in mixtures of
12~crown~4 and tetrahydrofuran. The resultant spectrum
featured a single peak at ~98.2 ppm with a linewidth of 180
Hz, which corresponds to the K- peak observed in this work
for solutions of K+(15C5)2-K-. The larger linewidths for
the 1204 solution indicates more interaction of the K- with
the environment than in the MeZO solutions. The cesium
cation is much too large to enter the cavity of the 12-
crown-4. Even in a "sandwich" conformation, the cesium
cation would be partially exposed and be able to partici~
pate in cation-anion interactions. Interactions that per~
turb the valence electrons of K- would give broader lines
rather than affect the chemical shift. The l5~crown~5
sandwich of K+ should effectively shield the cation and
minimize cation-anion interactions. Also, THF has higher
donicity than MeZO so the extent of solvent—anion interac—
tions is greater in the THF solutions. Therefore, the
greater linewidths observed for K— in the THF solutions is
probably due to the increased solvent- and cation-anion
interactions. The observation of K— in solution indicated

that it might be possible to detect K— in the solid state.

2. MMS studies. The first compound studied by 39K MAS-NMR
was K+(1505)2-K-, which was expected to give two 39K NMR
signals. Unfortunately, only the K— signal was observed at
~105 ppm. The static 39K NMR spectrum without 1H de-

coupling and the decoupled static and MAS spectra of poly-

crystalline K+(15C5)2iK- are shown in Figure 30. The full

~136-

 

 

 

 

L l 1 L 1 L 1

+IOO O ~IOO ~200
8[ppm from K*(oq.)]

Figure 30. 39K NMR Spectra of Polycrystalline K+(15C5)2°K-

at 180 K. (TOp) Static spectrum; (middle)
static, 1H decpupled spectrum; (bottom) MAS
spectrum with H decoupling. All Spectra have
60 Hz exponential broadening.

~137-

widths at half height of the K— peak are 300, 175 and 40
Hz, respectively. The residual width of the central
transition (m = 1/2, m = ~1/2) upon spinning and 1H de-
coupling give a maximum quadrupole coupling constant,
eZQO/b, of about 0.1 MHz. This estimate is based on the
calculation of the singular positions following the method
of Kundla, et al. [92].

The 39K MAS-NMR data for several other crystalline
alkalides that contain potassium were collected and are
presented in Table X. The chemical shifts of two conven~
tional potassium salts are also presented. The chemical
shifts observed for the alkalides are all within the ex-
perimental error of the value expected for K-(g). These
compounds are therefore assigned to be potassides. The
linewidths increase with cation size in M+(15C5)2-K-, which
suggests increasing cation-anion interactions. The
increase in the interactions is probably due to the
decrease in the ability of the two 15~crown~5 molecules to
effectively shield the cation from the environment as the
cation radius increases. The larger linewidth of K- in
Cs+(1505)2-K- compared to that of Cs+(18C6)2-K- is also
indicative of greater cation-anion interactions as was seen
in the 15~crown~5 versus l2~crown~4 solution spectra.

Rubidium XANES indicated a mixture of Rb+ and Rb_ for
KRb(18C6) and it was expected from the stoichiometry that a
39

K_ NMR signal would be observed. However, no signal was

detected, possibly due to strong Rb+~K— or K+~K- interac-

~138-

TABLE I. Results of 39! NMR-MAS Studies.

Compound 6, ppm from K‘gag)a Apl/z, Hz

x01 +46.7(5) 28
KBr +54.5(5) 20
K’(15C5)2-K' -105(1) 70
Rb’(1505)2-K'b -105(2) 120
Cs’(1505)2-K' -105(5) 220
KRb(1806)b no signal

Cs‘(18C6)2-K‘ -115(10) 150
K’C222-Rb‘ no signal

aUncertainty of the last digit given in parentheses.
Probably contain both K‘ and Rb‘ as indicated by rubidium
XANES.

~139-

tions. The radius of the cavity of the 18~crown~6 molecule
is large enough to accommodate K+ and give axial symmetry
with access to the K+ from above and below the ring. A
rubidium cation is slightly too large to enter the cavity
of 18~crown~6 and therefore Rb+(18C6) may participate in
cation-anion interactions to a larger extent than K+(1806).
In contrast to the 18~crown~6 compound, a K- signal was
observed for crystals with stoichiometry KRb(15C5)2.
Rubidium XANES also indicated a mixture of Rb+ and Rb- in
KRb(1505)2 and the presence of K+(15C5)2 as well as K- may
be inferred from the stoichiometry. The observation of K-
for this compound indicates less cation-anion interaction
due to the more effective shielding of the cation by the
two 15~crown~5 molecules. The linewidth of KRb(1505)2 is
reflective of more perturbation of the K_ in this compound
than in the homonuclear potasside.

As in the case of Na+0222-Na-, the estimated quadrupole
coupling constant for the complexed cation is much larger

39K NMR studies. The absence

than that of the anion in the
of a peak signal for K+(l505)2 indicates that eZQO/n is at
least 0.5 MHz since a signal of linewidth $1000 Hz could be
detected. This value is not unexpected since Na+C222 has a
quadrupole coupling constant of 1.2 MHz. The complexed

39K NMR for any complexing

cation was not detected by
agent, even in the case of K+0222-Rb-. This compound was
observed to be a pure rubidide by rubidium XANES and the

environment of the cation should be more symmetrical than

~140-

that of the crown ether salts. In all cases the signal
must be substantially broadened by quadrupole interactions.

The 39K studies of K-, both in solution and in potas-
side crystals, show that K- is a "genuine" anion with two
electrons in the 4s orbital. The small chemical shift
range observed for K- in solids and the proximity of the
peaks to that value calculated for K-(g) indicates that the
filled 4s orbital effectively shields the 3p orbitals from
appreciable interaction with the surroundings. However,
the increase of linewidth with an increase of the accessi~
bility of the cation reflects that cation-anion interac—
tions do occur. A K+Cn signal could not be detected in
crystalline samples. Therefore, 39K MAS-NMR is useful for
the detection of K_ in crystals where the cation is effec-
tively shielded from the anion but this technique' is in-
effective for the complete identification of alkalides that
contain potassium.

C. 87Rb NMR Spectra

In contrast to previous potassium NMR studies, the

87Rb NMR

rubidium anion in solution was readily observed by
[29,73]. The study of the complexed cation, however was
hindered by extreme line broadening of the NMR signal [71].
Rubidium X-ray absorption spectroscopy has proved to be

very useful in the identification of alkalides and elec-

trides [64,65] but the data must be collected and analyzed

~141-

at sites other than Michigan State University. The MAS-NMR
studies of crystalline alkalides and electrides were

extended to include 87Rb in the hope of providing an

H

in
house" identification technique.

1. Solution studies. The first objective of the 87Rb NMR
studies was to observe a signal for Rb+Cn in solution. The
absence of a signal for the complexed cation in earlier
studies was thought to be due to either exchange broadening
with the use of 18~crown~6 as a complexant or asymmetry of
the Rb+ environment in Rb+0222. Since linewidths observed
for species in solution are generally much smaller than
those from the same species in the solid state, it was
presumed that if Rb+Cn could not be observed in solution,
then it would be impossible to observe a complexed cation
signal in crystalline samples. Analogous to the procedures
used for 39K NMR solution studies, crystalline
Rb+(1505)2-Rb- was dissolved in dimethyl ether. To ensure
the complete complexation of Rb+, excess 15~crown~5 was
added to the samples. The spectrum of a 0.068 M
Rb+(1505)2-Rb- in Me20 in the presence of excess l5~crown~5
(mole ration 15C5/Rb+ = 14) is given in Figure 31. The
peaks at +33 and ~192 ppm are assigned to Rb+(15C5)2 and
Rb‘, respectively. The Rb- peak position with the chemical
shift range of ~185 to ~197 ppm has been observed
previously [29,73]. The linewidth of Rb+(1505)2 is ~5000

Hz and demonstrates the much lower symmetry of the cation

environment relative to Rb- with a linewidth of 170 Hz.

~142-

l .mswcwpmosn HoHusocoaxw
n: com on: m mess; monotonous mnsusosoexo an as no; a .s mm~

 

on -2 mso.o- uosuosssuossa as use. -momH-+nm no souuoonm mzz omen .Hm masons
Eon .38 +9: Eat E5 -ou_Eo-o
nxumt. nxyN.. nXuT. Au nxV-
- u u 4 u u d u q 1
j\f.i(P.\I\c.S)asI.l.l1J.LIi.I.(.IIfl))l)1$\\(\.(f\5\?fkk(w‘
m

 

 

~143-

The negligible solubility of rubidium metal in dimethyl
ether rules out the possibility of uncomplexed rubidium in
the solvent. This study provides the first observation of
87

Rb+Cn by Rb NMR [104]. The extremely large linewidth

observed for Rb+(15C5)2 makes the prospect of the observa-

tion of Rb+Cn by 87Rb MAS-NMR doubtful (see later calcula-
tions).

2. MMS studies. Several crystalline alkalides were
examined by 87Rb MAS-NMR at one or more Larmor frequencies

(58.90, 65.44 and 163.6 MHz). Table XI gives the chemical
shifts and full widths at half height for the spectra of
three simple rubidium salts and six alkalides that contain
Rb-. All alkalides had frequency dependent signals in the
range of ~185 to ~199 ppm which correspond to Rb-. Only
for the simple salts could the signal of Rb+ be detected.
Extreme quadrupolar broadening or other mechanisms pre-
vented its detection in model salts such as Rb+(18C6)-SCN-
or in the sodide Rb+(1565)2-Na-. The latter compound is
known from the crystal structure to contain the sandwich
cation Rb+(15C5)2. In some cases, even the expected signal
of Rb- known to be present in the salt, was not detected.
For example, rubidium XANES studies indicate the presence
of Rb-. in compounds of stoichiometry Rb2(C222) and
KRb(0222); yet no 87Rb signal could be detected in the MAS

studies of these systems. As expected, the electride,

Rb+(1505)2-e-, showed no signal of either Rb+ or Rb—.

~144-

Aunuwos has: as new“: dash mo cusses! maso>osn mouso>o zoom

.sosss-sc reassess so usoosoo .se nose asses Hoosaoso

.uononusosmm :« sm>mu hasflmasoosa

Ass-cos A-vmm-u
seasons asvmm-u
assume on
Ao-voms As-sm-u
Aeneomm Adena-u
Aosvoom Asymm-n
Ao-vomm A-vem-u
AmvN.ONl
o Am.vm.ONl
Aosvoou An.em-+
mm|4s\ss4 oflddldfll
nuluns: Hm.mm-uuuu

nonvomm Aavvmal
Aomvomv ANvmmal
AOMVOO¢ AHvamHl
AOMVOOM A~v~m~1
AONVOO¢~ Am.vmb~+
AOHVOOH Am.vbN~+
a: .~\-M4. Alum .w
IIIII N22 m¢¢.mmlllll

Aomcomm AN-em-l
Aomvoom AN-mm-n
Aomvomm AH-mm-u
Aomvmmm A-cmm-u
Ace-cum Am.-me-+
ASH-ems Am.-mms+
mm|4s\ssd mumldws
s
...... as: om.mmuuuul

mom~nm.m
Oomunm

.sm.n-eomH-.no
lam-NAmUmHVOQD
-se.sflmomH-.sm

a-momH-ne.u

Zomba
Ham
Home

 

p:=omޢo

.nosososeous mouse as possess eszam<z sees me penance

.HN mandh

~145-

Salts that contain rubidium and 18~crown~6 with and
without potassium appear to be mixed systems that are pre-
paration dependent. Although XANES data indicate that
these alkalides contain both Rb+(18C6) and Rb- [64,65], not

87Rb MAS-NMR spectra. For example, one

all samples yielded
preparation of stoichiometry KRb(18C6) gave Egg Rb- signals
at 163.6 MHz indicating, perhaps, the presence of mixtures
of K+(lBCS)-Rb-, nb+(1806)-Rb‘ and probably the corre-
ponding potassides. However no Rb_ signal was observed
with subsequent preparations, nor was any 39H signal
observed. Likewise, only one preparation of nominal
stoichiometry Rb(18C6) showed the 87Rb peak of Rb_ and
crystals of stoichiometry Rb2(18C6) showed no 87Rb peaks.
However, every sample prepared from rubidium and 18~crown~6
showed evidence in the XANES studies of the presence of
Rh“. Thus, the absence of the NMR peak cannot be used as
evidence that the rubidide ion is not present. The varia~
bility from one preparation to the next in these systems
(18C6 with Rb and/or K) indicates that the factors that
determine which crystalline phases precipitate from solu~
tion were not under control in these cases. In contrast,
the compounds Rb+(1505)2-Rb‘, Cs+(1505)2-Rb‘ and
Cs+(1806)2-Rb- yield no such ambiguities.

The 87Rb MAS spectrum of the compound KRb(l5C5)2 indi-
cates the presence of Rb-. This result agrees with the
rubidium XANES data, optical spectrum and 39K MAS-NMR

studies, all of which indicate this system to be a mixture.

~146-

The crystal quality was too poor to visibly detect two or
more types of crystals.

In addition to species identification, 8'7Rb MAS-NMR
studies of crystalline rubides yield information about the
isotropic chemical shift and quadrupolar coupling constants
of Rb-. As for 23Na and 39K, only the central transition
(m = 1/2, m = ~1/2) can be observed. The second-order
quadrupolar shift of this transition affects both the line-
width and the chemical shift observed by MAS-NMR. For the
axially symmetric case, the frequency difference between
the singularities is given by

2

a» = 5%; ° [I(I+l) - 3/4] . (22)

L

C

 

C

in which we, the nuclear quadrupole frequency, is related
to the nuclear quadrupole moment 0 and the electric field

gradient eq by

2 2
- 3892 - e__92
”o ‘ 21(21—l)h ’ l/Zl ] ° (23)

 

Equation (22) may be used with the linewidth to estimate

the maximum value of Va. Of course, incomplete removal of

proton dipolar coupling and chemical shift anisotropy by
spinning at the magic angle will also contribute to the
linewidth. The absence of well-defined singularities
suggests a distribution of quadrupolar coupling constants.

87

The linewidths of the Rb MAS-NMR peaks of Rb- in the

~147-

various rubidides given in Table XI yield a maximum quadru-
polar coupling constant 2vQ, of 1.3 MHz, with a standard
deviation of $0.3 MHz.
The chemical shifts of Rb- decrease with increasing
Larmor frequency, as expected from the quadrupolar shift of
87

the central transition. For Rb with nuclear spin 3/2 the

shift in ppm is related to the Larmor frequency by [91]

- 105[§§]2 , <24)

6 = 80

in which 60 is the isotropic chemical shift. The data in

Table XI (excluding RbK(18C6)) yield 80 = ~186 t 2 ppm and
ZVQ = 1.2 $0.2 MHZ.
A quadrupole coupling constant (eZqO/b) of 1.2 MHz for

Rb is consistent with the estimates of $0.23 MHz for Na-
and >0.5 MHz for K- as discussed earlier in this chapter.
If the external field gradient is assumed to be the same
for Rb- as for Na_, then correction for the quadrupole
moment and Sternheimer antishielding factor (of the
cations) yields an estimate of 52.2 MHz for Rb-.

The absence of a signal for complexed Rb+ is not sur-
prising in view of the large quadrupolar broadening. Based
upon the quadrupole~broadened MAS-NMR signal of Na+ in
Na+(1806)-SCN_, the MAS linewidth of Rb+(1505)2 is esti-
mated to be ~60,000 Hz which is too broad to detect with
the available instrumentation. Another method to estimate

the linewidth for Rb+Cn is to examine the 87Rb NMR results

~148-

from solutions. The linewidth of 150 Hz for Rb” in solu-
tion, together with a nuclear quadrupole coupling constant
of 1.2 MHz estimated from the solid state spectra, yield an
approximate correlation time of ~1010 s for this species.
This estimation is extremely rough since the field gradient
in solution may differ appreciably from that in the solid.
The complexed cation, being larger, probably has a longer
correlation time than the anion. However, if the same
correlation time is used for the cation as for Rb_, the
linewidth yields a quadrupole coupling constant of ~7 MHz
for Rb+(15C5)2. This, in turn, would give a MAS linewidth
of ~24,000 Hz, which is also too broad to measure.

The static and MAS spectra of Rb—, with and without
proton decoupling, were measured at 65.4 MHz. The results
for four compounds are given in Table XII. The quadrupolar
contribution is expected to increase by a factor of 2.7 to
3.6, depending upon the asymmetry parameter, n, from MAS to
static spectra [96]. The static coupled and MAS coupled
spectrum of Rb+(1505)2-Rb- is given in Figure 32. Since
the linewidths observed by MAS-NMR were essentially
unchanged upon proton decoupling, the dipolar contributions
from coupling to protons must have been removed by spin~
ning. The increase in linewidth from spinning to static-
decoupled spectra falls within the range expected from the
quadrupolar contribution. This indicates that the contri~

bution from chemical shift anisotropy is small.

Results of 37E) I"! with and without Proton Decoupling at 65.4 “12.

TABLE XII.

 

1H Decoupled

1H Coupled

1H Decoupled

6, ppm from Rb’ infinitea lH Coupled

MAS

 

 

Static

 

Static

Dilpion

 

und

Co

~149-

AA
CO
NM
vv
O

108
00¢
AA
GO
0300
vv
OD
bCD
00¢
AA
OO
COM
VV
88
OQ‘
F4 F1
AA
00
MM
vv
GO
ON
Q‘m
-, Fl
AA
P! F!
VV
p4 Pi
0'30)
7"?

Rb‘(15C5)2-Rb'

KRb(15C5)2

490(30) 490(30)

1160(30)

1830(30)

~189(2)

Cs’(l505)2-Rb’

2210(30) 650(30) 680(30)

2450(30)

-194(l)

Inb-

Cs’(1806)

aUncertainty given in parentheses.

~150-

 

 

 

j J l

l
~IOO -200 -300
Chemical Shift from Rb+ (00). ppm

0)-

Figure 32. 87Rb NMR Spectra of Polycrystalline
Rb+(15C5) °Rb' at 65.4 MHz. Top, static
coupled spectrum; bottom, MAS coupled

spectrum.

~151-

Reduction of the static linewidth of Rb- upon proton
decoupling ranged from ~250 Hz for Cs+(18C6)2-Rb- to ~650
Hz for Cs+(15C5)2-Rb-. Calculation of the expected dipolar
contribution by the method of Van Vleck [105] would require
knowledge of the crystal structures, which are not avail~
able for these compounds. However, an estimate can be made
by increasing the known Na- to H distances for
Cs+(1806)2-Na- [57] to accommodate the larger size of Rb-.
These calculations yielded a dipolar contribution of
~900 Hz which is much larger than that observed for
0s+(1806)2-Rb‘. This suggests, as with the sodides [57],
rapid motion of the CH2 protons of the 18~crown~6 molecule
at the temperatures at which the NMR measurements were
made.

The use of 87Rh MAS-NMR provides an excellent confirma-
tion of the presence of Rb- in crystalline alkalides. How-
ever, the absence of a signal does not necessarily smean
that Rbfi is absent. The signal of complexed Rb+ in the
solid is too broad to detect but both Rb+(15C5)2 and Rb“
were detected in MeZO solutions. The quadrupole coupling
constant of Rb— in crystals is 1.2 t 0.2 MHz, while that of
complexed Rb+ is probably at least an order of magnitude
larger. Comparison of proton-coupled and proton-decoupled
static linewidths indicated the presence of ~CH2 motion in
the complexant, even in the solid state. Extrapolation of
the chemical shift to infinite frequency gives an isotropic

chemical shift of ~186 t 2 ppm for Rb- in crystalline

~152—
rubidides, which falls within the range of Rb- chemical
shifts in solution.

D. 1330s NMR Spectra

Extensive 133Cs MAS-NMR studies have been done by
Ellaboudy and coworkers [42,53,99] and Dawes [57] as
described in Part II-Chapter One of this disssertation. In
the present work, 13303 MAS-NMR was used to identify the
compounds whose synthesis and characterization were
described in Part I. The unusually small electric quadru~

pole moment of ~3 x 10.27 cm2 and 100 percent abundance of

133Cs provide ideal conditions for NMR studies. The chemi-
cal shifts of Cs-, Cs+(1806)2-e_ and Cs+(1806)2-X-, where
X- is a diamagnetic anion, are separated by several hundred
ppm [53]. It was hoped that the peaks of the corresponding
l5~crown~5 complexes would be similarly separated from Cs—
and each other so that exact identification could be made.
The 13303 MAS-NMR spectrum of Cst(15C5)2.5 represented
in Figure 33 clearly shows the separation of the various
cesium signals. The peaks at +505, +29 and ~263 ppm from
Cs+(aq) at infinite dilution are assigned to Cs+(l505)z-e-,
Cs+(1505)2-X- and Cs-, respectively. This sample had a
87Rb MAS-NMR peak at ~194 ppm which indicates that Rb_ was
also present. Obviously, this sample consisted of a mix-

ture of Cs+(1505)2-e-, 0s+(1505)2-Rb' and probably

Rb+(15C5)2-Rb_ with some Cs- contamination. Another sam-

~153-

MMH

. am NAmUmHV m0 m0 mamfimm masmEH as NO Esuuommm mEZImdz mo .mm musmflm
. +

Ema .33 too 30¢... Pain 440.8qu

2.- 2...-

J] 4 —

OONI O OON+ 00¢... 80... 80+
- . - . a - - _ - s - - -

 

 

~154-

ple, without excess 15~crown~5, gave a cleaner spectrum.

133Cs

The MAS-NMR spectrum of Cs+(l5C5)2-Rb_ is shown in

Figure 34. The major peak at ~29 ppm corresponds to

Cs+(1505)2-X-. A small but discernible peak at ~265 ppm
indicates a small amount of Cs- in the sample.
133

The results of the Cs MAS-NMR studies of the 15-
crown~5 alkalides that contain cesium are given in Table
XIII. Each compound has a chemical shift in the range of
24-29 ppm which corresponds to diamagnetic Cs+(15C5)2
salts. Often a small peak at ~260 ppm would be observed in
the spectrum of Cs+(15C5)2-K- or Cs+(1505)2-Rb- and indi-
cated a small amount of Cs- in these samples. The narrow-
ness of the Cs+(1505)2 peak (l30-270 Hz) compared to that
of Cs- in Cs+(18C6)2-Cs— (350 Hz) is unexpected for a
cesium cation in an environment with axial symmetry. Tem-
perature studies by S. Dawes show an increase of the line~
width for Cs+(1505)2 with a decrease in temperature [57].
The unusual narrowness of the complexed cation peak may be
due to molecular motion at higher temperatures as in the
case of Rb- previously observed.

The frequency dependence of 13305 MAS-NMR spectra
studied by Ellaboudy indicates that the quadrupole interac~
tions are minimal and that the chemical shift anisotropy
and magnetic dipole interactions are the main source of
line broadening in Cs+(1806)2 compounds [53]. In contrast
to the 87Rb MAS-NMR studies, the chemical shifts were inde—

pendent of the Larmor frequency. The quadrupole coupling

~155-

 

 

 

s
A —‘A__
v—w v—v 'v A~
A
v
" ——

 

l L l l L l l l J l l l l L 1
+000 +600 +400 +200 0 -200 ~400 ~600 _
CHEMICAL SHIFT FROM Gothic), ppm
. 133 + -
Figure 34. Cs MAS-NMR Spectrum of Cs (15C5)2-Rb at

23.62 MHZ.

~156-

TABLE XIII. Results of 13303 MAS NMR Studies.

6, ppm from Cs‘ at Infinite

 

Compound Dilption AM4/2L_E§
CsI +284 ~~~
Cs’(15C5)2-Na‘ +24 130
Cs*(1505)2-K‘ +24 270
Cs’(15C5)2-Rb‘ +29 250

constants for Cs+(18C6)2 and Cs+(1505)2 were calculated
from the location of the satellite peaks with respect to
the central transition for each spectrum.

133Cs MAS-NMR is very useful for the

In conclusion,
identification of alkalides and electrides since the com-
plexed cation as well as the anion are easily observed.
The chemical shift of Cs+(1505)2 is independent of the
anion as long as the anion is diamagnetic. Impurities such
as trapped electrons cause large paramagnetic shifts which
are indicative of cation-anion interactions. Small amounts
of Ca. could be detected in some samples. The chemical
shift of Cs_ was ~~260 ppm in contrast to the calculated
value of ~346 ppm for Cs-(g) [76]. The large paramagnetic
shift of Cs- from the gaseous anion indicates strong inter~
action of Cs- with the environment and the less effective

1330s nucleus by the ns2 electrons than

that observed for 23Na and 39K.

shielding of the

SUMMARY AND SUGGESTIONS FOR FUTURE WORK

Ten new crystalline systems were isolated from solu-
tions of alkali metals and l5~crown~5. Elemental analysis
and optical absorption spectroscopy identified two of these
compounds as the electrides, K+(15C5)2-e_ and
Rb+(15C5)2-e_. The remaining compounds are alkalide salts.
In all cases the alkali metal cation is thought to be
"sandwiched" between two 15~crown~5 molecules. This con-
figuration has been confirmed for Rb+ in Rb+(1565)2-Na- by
single crystal X-ray diffraction studies. The use of opti-
cal absorption spectroscopy as an identification tool was
found to yield ambiguity in the case of systems with the
stoichiometry KRb(l5C5)2 and Cst(1505)2. The assignments
of these salts as a mixed alkalide and Cs+(15C5)2-Rb_,
respectively were based on rubidium XANES data. In
general, only the sodide salts could be identified by opti-
cal absorption spectroscopy with a fair degree of cer-
tainty; the other alkalides had very broad absorption peaks
and it was difficult to make exact assignments of the
anions.

Pressed powder conductivity measurements indicate that
all of the new compounds behave as semiconductors with

relatively small apparent band gaps. The calculated resis—

~157-

~158-

tivities are much larger than those expected for intrinsic
semiconductors with the observed band gaps. The conducti~
vity arises from impurities in the crystal. The electride
salts, K+(1505)2-e‘ and ab+(1505)2-e’, both had higher con-
ductivities than the alkalides salts, indicative of the
presence of more carriers in the electrides.

Preliminary EPR and magnetic susceptibility studies
were used to probe the environment of the electron in
H+(15C5)2-e_ and Rb+(l5C5)2-e—. The EPR spectra of each
electride had very narrow, single lines at the free elec~
tron g-value. The narrowness of the line and the absence
of a g—shift shift indicate that the electrons have very
few interactions with neighboring atoms and undergo rapid
exchange. The lines were asymmetric with A/B ratios less
than 2.7. This lineshape is consistent with those observed
for small particles with high conductivity where the skin
depth is on the order of the particle size. Spectra were
recorded at only one temperature.

The molar susceptibility of each electride deviated
substantially from Curie-Weiss behavior at low tempera~
tures. The apparent Curie constants at high temperature
correspond to 94 percent unpaired spins for K+(15C5)2-e-
and 55 .percent unpaired spins for Rb+(15C5)2-e-. Each
electride had a temperature independent contribution to the
observed susceptibility. In the case of Rb+(l5C5)2-e-, a
small amount of Rb_ may be present. The temperature inde-

pendent contribution for K+(1505)2-e— is too large to be

~159-

attributed to K A probable cause of the magnitude of
this term is the use of an inappropriate equation to
describe the magnetic behavior.

In general, the new alkalides and electrides are rela-
tively stable. Most of the new compounds may be handled at
room temperatures for time periods up to 10 minutes. The
stability makes characterization studies relatively easy to
complete. Differential scanning calorimetry studies deter-
mined the melting points of Rb+(1505)2-Na_ and
K+(15C5)2-Na- to be 75 and 45’0, respectively. The higher
melting point of Rb+(15C5)2-Na- makes it the most stable
alkalide characterized to date. The ease of preparation
and stability indicates that Rb+(1505)2-Na- is the most
likely candidate to be used as a two electron reductant.
For the alkalides, the stablity decreases with an increase
in anion size. The electrides are sufficiently stable to
resist decomposition at room temperature for short periods
of time. However, decomposition of the crystals occurs at
a very slow rate when stored at ~80'C.

Identification of alkalides and electrides without
ambiguity has been a major problem. Alkali-metal magic
angle spinning NMR is a technique that can be applied to
each alkali metal for the identification of alkalides and
electrides. The MAS-NMR spectra have been recorded for
several alkalides. The generalizations, which can be made

from these spectra, are as follows:

~160-

~23Na NMR can be used to detect the presence of

both the complexed sodium cation and the sodium
anions in the solid. The chemical shift range
observed for Na— is ~56 to ~63 ppm, which indi~
cates very effective shielding of the 2p orbitals
by the filled 3s orbital.
~Potassium 39 was used for the NMR studies. The
first NMR observation of K—, both in solution and
in the solid state was made. The chemical shifts
observed were ~99.3 ppm for K_ in dimethylether
and ~105 to ~115 ppm for K— in crystalline potas~
sides. These shifts are in good agreement with
the calculated value of K22) at ~107 ppm and are
reflective of highly efficient shielding of inner
orbitals by the two 4s electrons. No signal was
observed for the complexed cation, presumably due
to the large quadrupolar coupling constant.
~87Rb MAS-NMR can detect Rb- in crystalline alka-
lides but the absence of a signal does not neces~
sarily mean that Rb- is absent. The signal of
complexed Rb+ in the solid was too broad to detect
even though both Rb+(15C5)2 and Rb- were detected
in MeZO solutions. The chemical shift of Rb- in
the solid state was dependent on frequency and
extrapolation to infinite frequency yielded an

isotropic chemical shift of ~186 r 2 ppm. The

~161-

calculated chemical shift value of Rb(g) is
~214 ppm. Therefore, Rb- is slightly more de~
shielded than Na” or K”

133Cs MAS-NMR for

~The chemical shifts observed by
Cs+(1505)2-e-, Cs+(15C5)2-X_ and Cs- are suffi-
ciently separated so that these species can be
readily identified. Therefore, 133Cs MAS-NMR can
be used to check the purity of a given alkalide
system. Three alkalides were identified by 133Cs
NMR. The chemical shift observed for Cs- was ~260
ppm compared to the calculated value of ~348 ppm
for CsEg). This value indicates even further
deshielding of Cs- relative to Rb-.
~The deviation of the chemical shift of “(s) from
the calculated value for MEg), calculated
increases with the atomic number for the alkalide
anions. This could be caused by less effective
shielding of p and/or d electrons by the valence s

electrons 95 by an admixture of p and/or d charac-

ter with the ground-state 3 wave functions.

Finally, seven of the new crystalline alkalides, which

were described in Part I, have been identified without
ambiguity by alkali metal NMR. They are K+(15C5)2-Na-,
x+(1505)2-x', nb+(15°5)2'"a', Rb+(15C5)2-Rb_,

0s+(1505)2-Na', 0s+(15c5)2-x’ and Cs+(1505)2-Rb_. The

eighth system, which had the stoichiometry KRb(l5C5)2, was

~162-
shown to be a mixture with K‘, nb‘, nb+(1505)2 and
K+(1505)2 present.

The work presented in this dissertation can serve as a
springboard for several other experiments., First, the
electrides, K+(l5C5)2-e- and Rb+(1505)2-e—, appear to have
fascinating magnetic properties. Careful, low field, low
temperature susceptibility measurements are needed to
determine the exact types of localized magnetic interac-
tions. The EPR studies of the electrides should also be
continued to include the temperature and dilution depen-
dencies of the line shape.

Second, a need for crystal structures is present and
these compounds, for the most part, have sufficient stabi-
lity so that X-ray diffraction studies are feasible. These
studies require high quality single crystals and therefore,
the crystallization techniques need to be improved.
Besides X-ray diffraction studies, high quality single
crystals could be used in investigations of the orientation
dependence of the EPR and MAS-NMR spectra and AC and DC
conductivity experiments.

Third, the presence of a Ca. signal in the 13303 MAS-
NMR spectra indicates that the final alkalide in the
series, Cs+(15C5)2-Cs-, may be synthesized.

Fourth, the field dependence of the K— signal should be
checked by 39K MAS NMR at higher field strengths. The

availability of appropriate instruments is limited by the

low Larmor frequency of the 39K nucleus and these studies

~163-

would have to be done at other NMR facilities at this time.
Finally, the synthesis of new alkalides and electrides
should be pursued with other complexants such as
12~crown~4, 21-crown-7 or "lariat" ethers. The most easily
available may be those compounds that contain 12~crown~4
due to its relatively low cost and commercial availability.
New electrides are especially desirable for the study of
magnetic properties and the understanding of the nature of

the trapped electron.

10.

11.

LIST OF REFERENCES

P.P. Edwards, Adv. Inorg. Chem. Radiocbem. 25, 135
(1982).

W. Weyl, Annalen der Pbysik und Chemie £92, 601
(1864).

G. Lepoutre and M.J. Seinko, Eds., "Metal-Ammonia
Solutions, Colloque Weyl I", W.A. Benjamin, New York,
1964.

J.J. Lagowski and M.J. Seinko, Eds., "Metal-Ammonia
Solutions, Colloque Weyl II", IUPAC Buttersworth,
London, 1970.

J. Jortner and N.R. Kestner, Eds., "Electrons in
Fluids, Colloque Weyl III", Springer-Verlag, Berlin,
1973.

J.L. Dye, Conference Organizer, "Colloque Weyl IV,
Electrons in Fluids ~~ The Nature of Metal-Ammonia
Solutions", J. Phys. Chem. 79(26), (1975).

B. Webster, Conference Organizer, "Colloque Weyl V,
The Fifth International Conference on Excess Electrons

and Metal-Ammonia Solutions," J. Phys. Chem. gg (10),
(1980).

J. C. Thompson, Conference Organizer, "Colloque Weyl
VI, Le Dernier Colloque Weyl, The Sixth International
Conference on Excess Electrons and Metal-Ammonia
Solutions," J. Phys. Chem. 88 (17), (1984).

C. A Kraus, J. Chen. Educ. §0, 83 (1953).

W. L. Jolly, Frog. Inorg. Chem. A, 235 (1959).

M. C. R. Symons, 0. Rev. 13, 99 (1959).

~164-

12.

13.
14.

15.

16.

17.

18.

19.

20.

21.

22.

23.

24.

25.

26.

27.

28.

29.

30.

~165-

, 165 (1968); Angew

U. Schindewolf, Angew Chem. 8
l 68).

0
Chem. Int. Ehg. Ed. Z, 190 ( 9
T. P. Das, Adv. Chem. Phys. A, 303 (1962).

J. L. Dye, Sci. Amer. 8A8 (77), (Feb. 1967).

M. H. Cohen and J. C. Thompson, Adv. Phys. AZ, 847
(1968).

R. Catterall and N.F. Mott, Adv. Phys. A8, 665 (1969).

L. Kevan and B. Webster, Eds., "Electron-Solvent and
Anion-Solvent Interactions", New York, 1976.

M.C. R. Symons, Chem. Soc. Rev. 8, 337 (1976).

J. C. Thompson, "Electrons in Liquid Ammonia”, Oxford
University Press, Oxford, 1976.

J. L. Dye, Prog. Inorg. Chem. 88, 237 (1984).

J. T. Lin, G. P. Hagen and J. E. Ellis, J. Am. Chem.
Soc. A88, 2296 (1983).

P. P. Edwards and M. J. Sienko, Acc. Chem. Res. A8, 87
(1982).

B. Van Eck, Ph.D. Dissertation, Michigan State
University, 1983.

R. R. Dewald and J. L. Dye, J. Phy. Chem. 88, 128
(1964).

J. L. Dye, Angew Chem. Int. Ed., Eng. A8, 587 (1979).

C. J. Pedersen, J. Am. Chem. Soc. 88, 7017 (1967); 88,
386 (1970).

B. Dietrick, J.~M. Lehn and P. J. Sauvage, Tetrahedron
Lett. 83, 2885 (1969).

R. R. Dewald, S. R. Jones and B. S. Schwartz, J. Chem.
Soc., Chem. Commun. A888, 272 (1980).

D. M. Holten, P. P. Edwards, D. C. Johnson, C. J.
Page, W. McFarlane and B. Wood, J. Chem. Soc., Chem.
Commun. A883, 740 (1984).

J. D. Lamb, R. M. Izatt, C. S. Swain and J. J.
Christensen, J. Am. Chem. Soc. A88, 475 (1980).

31.

32.

33.

34.

35.

36.

37.

38.

39.

40.

41.

42.

43.

44.

45.

46.

47.

48.

~166-

J. L. Dye, J. M. Ceraso, M. T. Lok, B. L. Barnett and
F. J. Tehan, J1 Am. Chem. Soc. 88, 608 (1974).

F. J. Tehan, B. L. Barnett and J. L. Dye, J. Am. Chem.
Soc. 88, 7203 (1974).

J. L. Dye, J. Chem. Ed. gag, 332 (1977).

U. Schindewolf and M. Werner, J. Phys. Chem. 88, 1123
(1980).

M. Werner and U. Schindewolf, Ber. Bunsenges. Phys.
Chem. 88, 547 (1980).

W. Cross and U. Schindewolf, Ber. Bunsenges. Phys.
Chem. 88, 112 (1981).

U. Schindewolf, L. D. Le and J. L. Dye, J. Phys. Chem.
88, 2284 (1982).

. N. Tientega, M.S. Dissertation, Michigan State

F
‘University, 1984.

J. L. Dye, C. W. Andrews and S. E. Mathews, J. Phys.
Chem. Z8, 3065 (1975).

B. Van Eck, L. D. Le, D. Issa, and J. L. Dye, Inorg.
Chem. 8A, 1966 (1982).

D. Issa and J. L. Dye, J. Am. Chem. Soc. A83, 3781
(1982).

A. Ellaboudy, J. L. Dye and P. B. Smith, J) Am. Chem.
Soc. A88, 6490 (1983).

J. L. Dye, J. Phys. Chem. 88, 3842 (1984).

D. Issa, A. Ellaboudy, R. Janakiraman and J. L. Dye,
J. Phys. Chem. 88, 3847 (1984).

J. L. Dye, M. R. Yemen, M. G. DaGue and J.~M. Lehn, J.
Chem. Phys. 88, 1665 (1978).

M. G. DaGue, J. S. Landers, H. L. Lewis and J. L. Dye,
Chem. Phys. Lett. 88, 169 (1979).

J. L. Dye, M. G. DaGue, M. R. Yemen, J. S. Landers,
and H. L. Lewis, J. Phys. Chem. 88, 1096 (1980).

J. S. Landers, J. L. Dye, A. Stacy and M. J. Sienko,
J. Phys. Chem. 88, 1096 (1981).

49.

50.

51.

52.

53.

54.

55.

56.

57.

58.

59.

60.

61.

62.

63.

64.

65.

66.

~167-

L. D. Le, D. Issa, B. Van Eck and J. L. Dye, J. Phys.
Chem. 88, 7 (1982).

M. K. Faber, Ph.D. Dissertation, Michigan State
University, 1985.

M. L. Tinkham, M.S. Dissertation, Michigan State
University, 1982.

D. Issa, Ph.D. Dissertation, Michigan State
University, 1982.

A. S. Ellaboudy, Ph.D. Dissertation, Michigan State
University, 1984.

F.N. Tientega, unpublished results.

M. G. DaGue, Ph.D. Dissertation, Michigan State
University, 1979.

J. S. Landers, Ph.D. Dissertation, Michigan State
University, 1981.

S. B. Dawes, unpublished results.
DISNMR, Disk Interactive Spectroscopy on the Aspect
2000 minicomputer, version 81091, Bruker instrument,

Inc., 1981.

M. R. Yemen, Ph.D. Dissertation, Michigan State
University, 1982.

J. Papaioannou, unpublished results, this laboratory.

J. L. Dye and V. A. Nicely, J. Chem. Ed. 38, 43
(1971).

R. S. Alger, "Electron Paramagnetic Resonance:
Techniques and Applications", Interscience Publishers,
New Ybrk, 1968.

D. C. Mhttis, "The Theory of.thnetism", Harper & Row,
New York, 1965.

0. Fussfi, S. Kauzlarich, J. L. Dye and B. K. Teo, J.
Am. Chem. Soc. A8Z, 3727 (1985).

O. Fussé, B. K. Teo and J. L. Dye, to be submitted to
J. Am. Chem. Soc.

R. K. Harris, "Nuclear Magnetic Resonance Spectro~
scopy", Pitman Books, Ltd., London, 1983.

67.

68.

69.

70.

71.

72.

73.

74.

75.

76.

77.

78.

79.

80.

81.

82.

83.

84.

~168-

B. Lindman and S. Forsén, "NMR and the Periodic
Table", (R. K. Harris and B. Mann, eds.), p. 129,
Academic Press, New York, 1978.

F. W. Wehrli, "Annual Reports on NMR Spectroscopy",
(G.A. Webb, ed.), p. 126, Academic Press, New York,
1979.

S. Forsén and B. Lindman, Meth. Biochem. Anal. 8Z, 270
(1981).

K. J. Neurohr, T. Drakenberg, S. Forsén and H. Lilja,
J. thh. Reson. 8A, 460 (1983).

S. Khazaeli, J. L. Dye and A. I. Popov, Spectrochimica
Acta 88A, 19 (1983).

J. M. Ceraso and J. L. Dye, J. Chem. Phys. 8A, 1585
(1974).

J. L. Dye, C. W. Andrews and J. M. Ceraso, J. Phys.
Chem. Z8, 3076 (1975).

A. Beckmann, K. D. B6klen and D. Elke, Z. Phys.
gzg,173 (1975). ,

W. Lamb, Phys. Rev. 88, 817 (1941).
N. C. Pyper and P.P. Edwards

P. P. Edwards, S. C. Guy, D. M. Holton and W.
McFarlane, J. Chem. Soc., Chem. Commun., 1185 (1981).

P. P. Edwards, S. C. Guy, D. M. Holton, D. C. Johnson,
J. Sienko, W. McFarlane and B. Wood, J. Phys. Chem.
Z, 4362 (1983).

m0:

'0

P. Edwards, J. Phys. Chem. 88, 3772 (1984).

R. C. Phillips, 8. Khazaeli and J. L. Dye, J. Phys.
Chem. 88, 606 (1985).

P. B. Smith, Ph.D. Dissertation, Michigan State
University, 1978.

J. L. Dye, in "Progress in Macrocyclic Chemistry", (R.
M. Izatt and J. J. Christensen, eds.), Wiley-
Interscience, New York, Vol. 1, p. 63, (1979).

E. R. Andrew, W. S. Hinshaw and A. Jasinski, Chem.
Phys. Lett. 88, 399 (1974).

E. R. Andrew, Prog. NMR Spectrosc. 8, l (1971).

85.
86.
87.

88.

89.

90.

91.

92.

93.

94.

95.

96.

97.

98.

99.

100.

101.

102.

103.

~169-

P. Mansfield, Prog. NMR Spectrosc. 8, 41 (1971).
U. Haeberlen, Adv. Magn. Reson., Suppl. (1) (1976).
M. Mehring, NMR Basic Principles & Prog. AA (1976).

"NMR Spectrscopy in Solids”, (Proceedings of a
Symposium), The Royal Society (London) (1981).

A. C. Cunningham and S. M. Day, Phys. Rev. A88 (287),
(1966).

S. Ganapathy, S. Schramm and E. Oldfield, J. Chem.
Phys. ZZ, 4360 (1982).

K. Narita, J.~I. Umeda and H. Kusumoto, J. Chem. Phys.
85, 2719 (1966).

E. Kundla, A. Samoson and E. Lippmaa, Chem. Phys.
Lett. 88, 229 (1981).

v.0. Mfiller, Annalen der Physik Z, 451 (1982).

D. Freude, J. Haase, J. Klinowski, T. A. Carpenter and
G. Ronikier, Chem. Phys. Lett. AA8, 365 (1985).

A. Samoson, Chem. Phys. Lett. AA8, 29 (1985).

H.~J. Behrens and B. Schnabel, Physics AAA8, 185
(1982).

E. Oldfield and R. J. Kirkpatrick, Science 88Z, 1537
(1985).

A. Ellaboudy, M. L. Tinkham, B. Van Eck, P. B. Smith
and J. L. Dye, J. Phys. Chem. 88, 3852 (1984).

J. L. Dye and A. Ellaboudy, Chem. Br. 88, 210 (1984).

E. Oldfield, S. Schramm, M. D. Meadows, K. A. Smith,
R. A. Kinsey and J. Ackerman, J2 Am. Chem. Soc. A88,
919 (1982).

M. D. Meadows, K. A. Smith, R. A. Kinsey, T. M.
Rothgeb, R. P. Sarjune and E. Oldfield, Proc. Nat.
Acad. Sci. USA Z8, 1351 (1982).

A. Ellaboudy and J. L. Dye, J. Magn. Reson., in press.

P. P. Edwards and A. Ellaboudy, Nature, 8~Z, 242
(1985).

-170-

104. M. L. Tinkham, A. Ellaboudy, P. B. Smith and J. L.
Dye, J. Phys. Chem., in press.

105. J. H. VanVleck, Phys. Rev. Z8, 1168 (1948).

TY

 

MICHIGAN STATE UNIVERSI LIBRARIIES
IIIIllfllilH‘IlfllllIlmlljllllililllllIllifllllHliHllHllI
I
3193 3177 594 I