STUDIES OF HYPERPOLARIZED 3He

RELAXATION AT GLASS SURFACES

by

Richard Emil Jacob

A dissertation submitted to the faculty ofThe University of Utah

in partial fulfillment of the requirements for the degree of

Doctor of Philosophy

Department of Physics

The University of Utah

May 2003

Copyright c© Richard Emil Jacob 2003

All Rights Reserved

THE UNIVERSITY OF UTAH GRADUATE SCHOOL

SUPERVISORY COMMITTEE APPROVAL

of a dissertation submitted by

Richard Emil Jacob

This dissertation has been read by each member of the following supervisory committeeand by majority vote has been found to be satisfactory.

Chair: David C. Ailion

Carleton DeTar

Dennis Parker

Brian T. Saam

Clayton Williams

THE UNIVERSITY OF UTAH GRADUATE SCHOOL

FINAL READING APPROVAL

To the Graduate Council of The University of Utah:

I have read the dissertation of Richard Emil Jacob in itsfinal form and have found that (1) its format, citations, and bibliographicstyle are consistent and acceptable; (2) its illustrative materials includingfigures, tables, and charts are in place; and (3) the final manuscript issatisfactory to the Supervisory Committee and is ready for submission toThe Graduate School.

Date David C. AilionChair, Supervisory Committee

Approved for the Major Department

Z. Valy VardenyChair

Approved for the Graduate Council

David S. ChapmanDean of The Graduate School

ABSTRACT

Enormous nonequilibrium nuclear polarizations, of order 10%, can be achieved

in certain noble-gas nuclei via spin-exchange optical pumping (SEOP). Applications

of such hyperpolarized (HP) gases depend critically on the ability to maximize and

maintain the polarization. Both the polarization level and longitudinal relaxation

time are limited by relaxive interactions between the gas and glass vessels, or cells,

that contain the gas. An understanding of the interactions is critical to consistent

production of nonrelaxive cells. Magnetic resonance image quality and sensitivity,

using HP gas as the signal source, benefit greatly from maximally polarized gas.

This thesis addresses nuclear longitudinal relaxation mechanisms of 3He on glass

surfaces.

Much information about the glass–3He interactions can be obtained by measuring

3He polarization loss by periodically sampling a 3He free induction decay signal using

a pulse nuclear magnetic resonance spectrometer. Over time, the initial amplitude

of the signal decays with a characteristic time constant T1. The HP 3He T1’s are

very sensitive to surface relaxation mechanisms because of the high mobility of the

gas.

Wall relaxation is a very complicated, multivariable problem. We show that

relaxation in bare Pyrex glass is mainly due to 3He dissolving into the glass and

interacting with paramagnetic Fe3+ ions in a predictable way. We describe the first

experimentally confirmed predictive model of relaxation rates of 3He in bare-glass

cells. However, once Rb metal is added to a cell for SEOP, the relaxation interactions

change significantly. Our data suggest that, when Rb is present, interactions with

Fe3+ ions no longer contribute significantly to relaxation so that other mechanisms

take over.

One surprising and significant mechanism in Rb-coated cells is interactions of the

3He with magnetic particles. We found that the 3He T1’s can be reduced significantly

due solely to exposure of a cell to a high (several thousand Gauss) magnetic field,

an effect termed T1 hysteresis. The magnetized cells can be degaussed to restore the

original T1. We present a model that predicts approximately 104 magnetic sites on

the surface of a typical spherical 50 cm3 cell, or a few sites per square millimeter.

This astonishingly low site density underscores the sensitivity to surfaces of 3He

relaxation measurements. The model also predicts a linear dependence of T1 on

pressure at a given temperature, which we confirm experimentally.

v

to Dad

who showed me the roads to take

and

to Connie

who kept me on them

It is a profound and necessary truth that the deep things in science

are not found because they are useful;

they are found because it was possible to find them.

– Robert Oppenheimer

CONTENTS

ABSTRACT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iv

LIST OF FIGURES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi

ACKNOWLEDGEMENTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiv

CHAPTERS

1. INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.1 Hyperpolarized Noble Gases . . . . . . . . . . . . . . . . . . . . . . 11.1.1 Overview of Polarized Gas . . . . . . . . . . . . . . . . . . . 11.1.2 Uses of Polarized Gas . . . . . . . . . . . . . . . . . . . . . . 1

1.2 Background and History . . . . . . . . . . . . . . . . . . . . . . . . 21.2.1 Spin-exchange and Optical Pumping . . . . . . . . . . . . . . 21.2.2 Current Issues . . . . . . . . . . . . . . . . . . . . . . . . . . 21.2.3 Basic Physics of Spin-exchange Optical Pumping . . . . . . . 31.2.4 3He Relaxation . . . . . . . . . . . . . . . . . . . . . . . . . . 61.2.5 Spin-exchange Vessels . . . . . . . . . . . . . . . . . . . . . . 8

1.3 Thesis Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

2. NMR SPECTROMETER AND TECHNIQUES . . . . . . . . . . . . . . 13

2.1 100 kHz Pulse NMR Spectrometer . . . . . . . . . . . . . . . . . . . 132.1.1 The Applied Field . . . . . . . . . . . . . . . . . . . . . . . . 132.1.2 The Spectrometer . . . . . . . . . . . . . . . . . . . . . . . . 142.1.3 The NMR Coil . . . . . . . . . . . . . . . . . . . . . . . . . . 16

2.2 T1 Measurement Techniques . . . . . . . . . . . . . . . . . . . . . . 16

3. WALL RELAXATION OF 3HE IN SPIN-EXCHANGE CELLS . 18

3.1 Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183.2 Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203.3 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203.4 Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213.5 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . 233.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303.7 Addendum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

4. 3HE SPIN-EXCHANGE CELLSFOR MRI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

4.1 Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 324.2 Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 334.3 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 344.4 Wall Relaxation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 354.5 Cell Fabrication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 374.6 Cell Preparation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 404.7 Cell Filling System . . . . . . . . . . . . . . . . . . . . . . . . . . . 444.8 The Polarizer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 474.9 Experimental Results . . . . . . . . . . . . . . . . . . . . . . . . . . 50

4.9.1 T1 Measurements . . . . . . . . . . . . . . . . . . . . . . . . 504.9.2 Polarimetry . . . . . . . . . . . . . . . . . . . . . . . . . . . 524.9.3 Overall Performance . . . . . . . . . . . . . . . . . . . . . . . 55

4.10 Transit Time of 3He in the Capillary . . . . . . . . . . . . . . . . . . 564.11 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 594.12 Addendum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60

5. MAGNETIC FIELD DEPENDENCE OF 3HE RELAXATION . . 63

5.1 Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 635.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 635.3 Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

5.3.1 High-field Hysteresis . . . . . . . . . . . . . . . . . . . . . . . 645.3.2 Low-field Hysteresis . . . . . . . . . . . . . . . . . . . . . . . 65

5.4 Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 665.5 Results/Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

5.5.1 High-field Hysteresis . . . . . . . . . . . . . . . . . . . . . . . 695.5.2 Low-field Hysteresis . . . . . . . . . . . . . . . . . . . . . . . 75

5.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78

6. FUNDAMENTAL MECHANISMS OF 3HE RELAXATION ONGLASS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79

6.1 Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 796.2 Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 806.3 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 806.4 Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81

6.4.1 T > Room Temperature . . . . . . . . . . . . . . . . . . . . 816.4.2 T < Room Temperature . . . . . . . . . . . . . . . . . . . . 87

6.5 Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 896.6 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . 90

6.6.1 T > Room Temperature . . . . . . . . . . . . . . . . . . . . 906.6.2 T < Room Temperature . . . . . . . . . . . . . . . . . . . . 92

6.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93ix

7. 3HE RELAXATION IN BARE AND RB-COATED GLASS . . . . 95

7.1 Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 957.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 957.3 Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96

7.3.1 Aluminosilicate Glass . . . . . . . . . . . . . . . . . . . . . . 967.3.2 Quartz Glass . . . . . . . . . . . . . . . . . . . . . . . . . . . 997.3.3 Rb-coated Pyrex Glass . . . . . . . . . . . . . . . . . . . . . 101

7.4 Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1017.5 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . 103

7.5.1 Aluminosilicate Glass . . . . . . . . . . . . . . . . . . . . . . 1037.5.1.1 T > Room Remperature . . . . . . . . . . . . . . . 1037.5.1.2 T < Room Temperature . . . . . . . . . . . . . . . 104

7.5.2 Quartz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1057.5.3 Pyrex Glass . . . . . . . . . . . . . . . . . . . . . . . . . . . 107

7.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113

8. CELL RINSING . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114

8.1 Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1148.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1148.3 Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1158.4 Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1178.5 Results/Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118

8.5.1 Reducing-agent Rinse . . . . . . . . . . . . . . . . . . . . . . 1188.5.2 Rb Rinses . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1198.5.3 Acid Rinses . . . . . . . . . . . . . . . . . . . . . . . . . . . 122

8.5.3.1 HF Rinse . . . . . . . . . . . . . . . . . . . . . . . . 1228.5.3.2 Aqua Regia Rinse . . . . . . . . . . . . . . . . . . . 1248.5.3.3 HF and HCl Rinse . . . . . . . . . . . . . . . . . . 1258.5.3.4 Intervening HCl Rinse . . . . . . . . . . . . . . . . 127

8.5.4 Potassium Rinse . . . . . . . . . . . . . . . . . . . . . . . . . 1298.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129

9. MRI OF FLOWING POLARIZED 3HE . . . . . . . . . . . . . . . . . . . . . 131

9.1 Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1319.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1319.3 Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1339.4 Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1359.5 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . 1379.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140

APPENDIX . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141

REFERENCES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147

x

LIST OF FIGURES

1.1 Depopulation optical pumping . . . . . . . . . . . . . . . . . . . . . . . . 4

2.1 100 kHz NMR spectrometer . . . . . . . . . . . . . . . . . . . . . . . . . 15

3.1 T1 hysteresis of cell 5A . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

3.2 T1 pressure dependence . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

3.3 Rb dependence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

4.1 A Pyrex valved spin-exchange cell for generating hyperpolarized 3He . . 38

4.2 Diagram of a cell manifold . . . . . . . . . . . . . . . . . . . . . . . . . . 40

4.3 The oil-free high-vacuum system used for cell fabrication . . . . . . . . . 41

4.4 Gas-handling system used to fill cells with 3He . . . . . . . . . . . . . . . 45

4.5 Polarization and decay transients . . . . . . . . . . . . . . . . . . . . . . 48

4.6 Relaxation rates for several cells . . . . . . . . . . . . . . . . . . . . . . . 51

4.7 Polarimetry free-induction decays . . . . . . . . . . . . . . . . . . . . . . 54

5.1 A sketch of a typical hysteresis loop showing the relationship betweenmagnetic moment M and applied field H . . . . . . . . . . . . . . . . 66

5.2 Gas transfer manifold . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

5.3 T1 hysteresis loop of an aluminosilicate cell . . . . . . . . . . . . . . . . . 70

5.4 A T1 hysteresis loop of a bare (no Rb) Pyrex cell . . . . . . . . . . . . . 71

5.5 A T1 hysteresis loop of Pyrex cell 9A . . . . . . . . . . . . . . . . . . . . 72

5.6 A T1 hysteresis loop of Pyrex cell 10A . . . . . . . . . . . . . . . . . . . 73

5.7 A T1 hysteresis loop of Pyrex cell 18A . . . . . . . . . . . . . . . . . . . 73

5.8 Low-field T1 hysteresis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76

6.1 Temperature dependent relaxation rates for three bare (no Rb) Pyrexcells above room temperature . . . . . . . . . . . . . . . . . . . . . . 91

6.2 Relaxation rate vs 1000/T for two bare Pyrex cells at ≈4 amagats . . . . 92

7.1 Relaxation rate vs. 1000/T for two bare aluminosilicate (GE-180) cellsfor T ≤ 300 K . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103

7.2 Relaxation rate vs. 1000/T for two bare aluminosilicate (GE-180) cellsfor T ≥ 300 K . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104

7.3 Relaxation rate vs. 1000/T for two bare quartz (GE fused silica) cells at≈ 4 atm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106

7.4 Wall T−11 measurements of several spin-exchange cells for T > 295 K . . 107

8.1 Cells rinsed with a chemical reducing agent . . . . . . . . . . . . . . . . 119

8.2 Relaxation rates for three different cells before and after Rb is rinsed out 120

8.3 Relaxation rates for cell 11A when bare (new, no Rb), with Rb, and withthe Rb rinsed out . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121

8.4 T1 hysteresis of HF-rinsed cells . . . . . . . . . . . . . . . . . . . . . . . 122

8.5 An AFM image of an untreated Pyrex sample . . . . . . . . . . . . . . . 123

8.6 An AFM image of a Pyrex sample rinsed with a 10% HF solution forseveral minutes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124

8.7 T1 hysteresis of cells rinsed with aqua regia . . . . . . . . . . . . . . . . . 125

8.8 T1 hysteresis of cells rinsed with HF and HCl . . . . . . . . . . . . . . . 126

8.9 Atomic force microscopy of a bare Pyrex sample rinsed with HCl . . . . 127

8.10 T1 hysteresis of Rb and HCl rinsed cells . . . . . . . . . . . . . . . . . . 128

9.1 The velocity and diffusion sensitive gradient sequence used to makeimages in Fig. 9.3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134

9.2 Experimental set-up for MRI flow imaging . . . . . . . . . . . . . . . . . 136

9.3 Velocity map (left) and ADC map (right) for flowing HP 3He through atube . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137

xii

9.4 Visualization of real-time MRI of 3He flowing through a tube . . . . . . . 139

A.1 Box 1 of the intermediate-field spectrometer . . . . . . . . . . . . . . . . 142

A.2 Box 2 of the intermediate-field spectrometer . . . . . . . . . . . . . . . . 143

A.3 Box 3 of the intermediate-field spectrometer . . . . . . . . . . . . . . . . 145

A.4 Intermediate field spectrometer pulse generator and low-pass filter sec-tion details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146

xiii

ACKNOWLEDGEMENTS

The individual most responsible for my success and progress as a graduate student

is my adviser, Brian Saam. He had an undying patience for my incessant questions,

hair-brained ideas, and myriad mistakes. And the fact that we were both new to

the game – I was his first graduate student – meant we were in it together. I learned

good experimental practices, some plumbing, some electronics, a little squash, and

how to make my hands and brain do things I never thought possible. It was only

one year after we entered a completely empty room we called a laboratory that we

found a ground-breaking and critically important new phenomenon which led to my

first publication: a Letter to the Physical Review. I will always have fond memories

of the time we spent together doing, talking, and thinking physics.

David Ailion was very helpful in getting me through the Common Exam orals

and in setting up my committee. He and Gernot Laicher introduced me to NMR

and took me on as a student prior to Brian’s arrival. Several others made important

contributions to the research, namely: Steve “Just Drill The Hole” Morgan, who

helped build the vacuum systems and acquire data; Ben Anger, who put together

the intermediate-field spectrometer and took the initial goofy field-dependence data;

and Ryan Stapley, who helped take data one summer. The guys in the machine shop,

Bob Fernelius, Jack Pitts, and Ed Munford, sure helped make my job easier. Mark

Conradi, Thad Walker, and Will Happer made significant intangible contributions,

mostly through conversations with Brian. Jason Leawoods was instrumental in

making the PRL complete, and continues to have good ideas for finding out more

about T1 hysteresis. The CPL was motivated by the brilliant thinking and cooper-

ation of Bas Driehuys, and the flow images would not have been possible without

Kevin Minard and the facilities at PNNL. Finally, Janice Kyle’s expert glassblowing

contributed significantly to all of our incredible cell-making success.

Thanks to my folks for always showing support and interest in whatever I did.

Just knowing they expected a lot was great motivation to achieve a lot. And knowing

that I would finally accomplish something that Kris, John (though it’s just a matter

of time), and Becky have not accomplished was additional motivation. They really

help me live up to my potential. And Scott and Dawnette Palmer provided a great

stress relief valve through evenings of games and watching ASU football (go Devils!).

My dear wife, Connie, to whom I have been married for over seven years, has

been incredibly patient and understanding over the five and a half years of graduate

school. She supported the family both financially and emotionally. She gave me the

encouragement, and shoulder, I often needed. The fact that she has stayed with me

astonishes me, and the prospect of being together forever elates me. I am, indeed,

a lucky man.

xv

CHAPTER 1

INTRODUCTION

1.1 Hyperpolarized Noble Gases

1.1.1 Overview of Polarized Gas

Enormous nonequilibrium nuclear polarizations (of order 10%) can be achieved

in certain noble-gas nuclei, specifically 129Xe and 3He. Gases with such polarizations

are referred to as hyperpolarized (HP), because the net magnetization of a sample

is several orders of magnitude higher than at thermal equilibrium in a several-Tesla

magnetic field. These polarizations can be achieved through optical pumping meth-

ods, either spin-exchange optical pumping (SEOP) [1] or metastability-exchange

optical pumping (MEOP) [2]. The former requires an alkali metal intermediary, and

the latter polarizes the 3He nuclei directly. Both MEOP and SEOP allow for the

polarization of large quantities of gas (∼ 1 atm·L) to sufficiently high polarizations

(≥ 50%). This dissertation deals exclusively with 3He polarized via SEOP.

1.1.2 Uses of Polarized Gas

HP gas has several applications in physics, chemistry, and biomedicine. These

include the determination of the neutron spin-structure function by scattering po-

larized electrons from targets of highly polarized 3He [3], studies of fundamental

symmetries [4, 5], neutron polarizers and spin filters [6], and studies of surface inter-

actions [7]. An exciting, recent application is magnetic resonance imaging (MRI) of

2

lung air spaces [8, 9]. Pulmonary MRI in humans requires a large volume (≈ 0.5 L)

of polarized gas, and recent developments in high power, low cost, diode-array lasers

has allowed for production of liter quantities of HP 3He making such experiments

feasible.

1.2 Background and History

1.2.1 Spin-exchange and Optical Pumping

The term “optical pumping” refers to the use of light to produce a nonequilibrium

energy level population of a system, such as the electron spin distribution of alkali

metal atoms. A Nobel prize was awarded in 1966 for the pioneering optical pumping

work of A. Kastler [10]. “Spin-exchange” is a process of angular momentum transfer

from optically pumped alkali metal atoms to nuclei of noble gas atoms. The first

successful demonstration of large, nonequilibrium polarizations in noble gas nuclei

attained via SEOP was published in 1961 by M. Bouchiat, et al. [11]. Optical

pumping is a photon-limited process; thus the exclusive availability of discharge

lamps and low-power lasers initially limited the quantities and applications of HP

gas. Within the last 10 years, developments in inexpensive, high-power, efficient

diode-array lasers have helped open the flood gates of polarized gas research, since

they made polarizing liter quantities of 3He possible. See [1, 12] for thorough reviews

of optical pumping.

1.2.2 Current Issues

As research in HP gas has progressed, many interesting physics problems have

arisen. This is a good sign, indicating a realm of physics gaining the interest of a

growing number of researchers. Several of the problems have affected, and arisen

3

from, my research on HP 3He. One issue is the production of the glass cells used

for SEOP. 3He relaxes to thermal equilibrium mainly through interactions with

the cell walls. Consistently making cells with long lifetimes has proven frustrating

and problematic, and some groups have developed elaborate methods of preparing

cells. Through the production of dozens of cells, we have developed successful

protocols for consistently making quality cells. Another problem is that a basic

understanding of the relaxation interactions between 3He and the glass is lacking, as

is an understanding of exactly what the 3He is interacting with. Studies of cell-wall

relaxation have been the crux of my research. Closely related to wall relaxation is

the problem that I (very unexpectedly) discovered involving ferromagnetic inclusions

in the glass. We would like to know exactly where these inclusions come from, how

much magnetic material is present, and, most especially, how to eliminate them. An

exciting consequence of this discovery is the possibility of a new application of HP

3He: an ultra-sensitive probe for surface magnetism.

1.2.3 Basic Physics of Spin-exchange Optical Pumping

SEOP is a two-step process that involves an alkali metal and circularly-polarized

laser light [1]. The first step, optical pumping, polarizes the valence electrons of

the alkali-metal vapor, typically Rb. A sufficient Rb vapor pressure is achieved by

heating the cell to 160–180C [13]. The circularly polarized laser light is resonant

with the transition from the 5S1/2 ground state to the 5P1/2 excited state. Allowed

transitions of the valence electron are from the m = ±1/2 to the m = ∓1/2 sublevels,

obeying the ∆m = ±1 selection rule for σ± light. Pressure broadening of the

absorption line allows for use of a laser with a broad linewidth (1 - 2 nm fwhm), such

as a diode-array. For example, as depicted in Fig. 1.1, left-circularly polarized (σ+)

light will excite the m = −1/2 (spin-down) sublevel to the m = +1/2 (spin-up)

4

50% 50%

5P1/2

5S1/2

collisional mixing

m=-1/2 m=1/2

σ+

Figure 1.1. Depopulation optical pumping. The interaction between left-circularlypolarized (σ+) light and alkali-metal atoms causes the m = −1/2 sublevel to bedepleted into the m = +1/2 sublevel. The ∆m = +1 selection rule causes the spinsin the m = +1/2 state to be invisible to the laser light.

sublevel with the absorption of +h of angular momentum per atom. Collisional

mixing in the excited state causes relaxation from the 5P1/2 state to both m sublevels

of the 5S1/2 state to occur with equal probability; thus an average of two photons are

required to polarize an atom. The m = +1/2 sublevel, transparent to left-circularly

polarized light, will increase in population until global saturation is reached, which

occurs on a time scale of 10’s of microseconds. Various relaxation mechanisms,

such as Rb–3He spin-rotation interactions or dipolar Rb–Rb interactions, cause Rb

polarization loss on the time scale of milliseconds, so the laser must remain on

continuously to maintain saturation. This process is termed “depopulation optical

pumping.”

The second step, spin-exchange, is the process of spin angular momentum trans-

fer from the alkali-metal electron to the nucleus of the noble gas. The angular

momentum is collisionally transferred via a Fermi contact interaction, leaving the

now-unpolarized alkali metal free to scatter another photon or two and continue the

5

SEOP process. The binary collisions between the 3He and Rb last about 10−12 s

and the spin-exchange cross section is quite small, ≈ 1 barn; thus spin-exchange is

an inherently slow process; the characteristic spin-exchange time is typically 4–6 h.

Equation (1.1) describes the time dependence of the 3He polarization under SEOP

conditions:

PHe(t) = 〈PA〉 γse

γse + Γ

[1− e−(γse+Γ)t

], (1.1)

where 〈PA〉 is the time- and volume-averaged alkali-metal electron polarization,

γse is the spin-exchange rate, and Γ is the 3He relaxation rate. Based on direct

Rb polarization measurements by other researchers, we estimate that the 〈PA〉is maintained at ≈ 100% in our cells [14]; thus the long time limit (t → ∞) of

3He polarization depends on the fraction γse/(γse + Γ). Since binary collisions are

responsible for Rb–3He spin exchange, the spin-exchange rate may be written:

γse = [Rb] 〈σv〉 , (1.2)

where [Rb] is the Rb number density, which is strongly temperature dependent,

and 〈σv〉 is the velocity-averaged spin-exchange cross section, which is weakly tem-

perature dependent. Since a macroscopic amount of Rb is in our cells, we can

approximate [Rb] by using the saturated vapor-pressure curve [13]:

[Rb] =10(10.55−4132/T )

kBT, (1.3)

6

where T is the temperature and kB is Boltzman’s constant. By increasing the

temperature, [Rb] can be increased dramatically in our temperature regime; thus

PHe should be easily maximized. However, SEOP is a photon-limited process: if [Rb]

is increased beyond the capacity of the laser to maintain uniform Rb polarization

in a given volume, then 〈PA〉 may drop well below 100%. Thus, 〈PA〉 γse can

be optimized, but the optimal value is essentially fixed. Hence, the final 3He

polarization is largely determined by the 3He relaxation rate Γ.

1.2.4 3He Relaxation

The 3He polarization decays to thermal equilibrium with a characteristic time

constant T1. The longitudinal relaxation rate T−11 of 3He (Γ ≡ T−1

1 ) is characterized

by:

1

T1

=1

T1 dd

+1

T1 G

+1

T1 wall

. (1.4)

T1 dd is the dipole-dipole relaxation rate due to interactions between colliding 3He

atoms and is given by [15]:

1

T1 dd

=[3He]

744hours−1, (1.5)

where [3He] is the 3He density in amagats (an amagat is defined as the measured

density per density at 0C and 1 atm). For a given 3He density, this relaxation rate

is fixed. At 8 atm of 3He pressure, T1 dd = 100 hours.

T1 G is relaxation due to diffusion in external magnetic field gradients [16]:

7

1

T1 G

= D(∇B

B

)2

, (1.6)

where D is the 3He diffusion coefficient, B is the mean longitudinal external field,

and ∇B is the external field gradient transverse to B. Equation (1.6) is valid in

the limiting case when the time required for a spin to diffuse across the cell is much

longer than the Larmor period. This condition was easily met in all of our cells. In a

homogeneous magnetic field, such as one created by a carefully adjusted Helmholtz

pair, T1 G is negligible (often as long as several thousands of hours at high 3He

pressure).

T1 wall is relaxation due to interactions with the cell wall, and depends on several

factors, including surface-to-volume ratio and concentration of relaxation sites in the

glass. Wall relaxation varies widely from cell to cell and ranges from several minutes

to several hundreds of hours. Thus it alone often limits maximum polarization and

determines the overall T−11 . The most basic model of surface relaxation assumes

ballistic collisions of the 3He with the surface. This essentially means that the 3He

interacts with a relaxive surface site for a time much shorter than the relaxation

time at that site. A cell containing N total atoms should have a measured relaxation

time given by:

1

T1 wall

=n

N× 1

Ts

, (1.7)

where Ts is the relaxation time for atoms under the influence of relaxation sites and

n is the number of atoms interacting with relaxation sites. That is, n is the number

of atoms within a characteristic distance δ of the surface such that δ = vts, where

v is a mean thermal velocity and ts is the surface–atom interaction time (≈ 10−13 s

at room temperature [17]). Therefore,

8

n =A

VNδ, (1.8)

where A/V is the surface to volume ratio of the cell. By substituting Eq. (1.8) into

Eq. (1.7), the relaxation time becomes:

1

T1 wall

=A

Vvη, (1.9)

where η = ts/Ts is the relaxivity, which is defined as the probability to relax per

surface encounter. In the ballistic limit ts ¿ Ts, and η is typically on the order of

10−6 for glass (Ts was estimated to be ≈ 10−7 s in glass [18]). The wall relaxation

time given in Eq. (1.9) is independent of the 3He number density and external

magnetic field strength, and is weakly dependent on temperature since v ∝ √T . In

reality, however, wall relaxation is not so simple. The bulk of this thesis addresses

the complexities of wall relaxation, including the effects of external magnetic fields,

temperature dependence, the effects of the presence of Rb, and history of cell

exposure to magnetic fields. By understanding wall relaxation mechanisms and

learning to minimize them, we can achieve the goal of consistent production of

reliable, long-lifetime spin-exchange cells.

1.2.5 Spin-exchange Vessels

The special vessels, or cells, used to contain HP gas must meet the following crite-

ria: they must be alkali-metal resistant, heat resistant (up to ∼ 200C), transparent

to laser light, able to withstand high pressures (∼ 15 atm), able to hold a quantity

of gas comparable to an average adult’s tidal volume (≈ 0.5 atm·L), and refillable

9

without affecting the relaxation properties. All of these conditions are best met by

using glass. Pyrex (Corning glass number 7740) is our glass of choice because it is

robust, easily worked by any glass blower, inexpensive, and readily available. As

discussed in Sec. 1.2.4, the characteristics of the cell walls are critical in determining

the maximum level of polarization and the rate at which the gas will relax. Most

of the work presented in this thesis addresses understanding and minimizing 3He

relaxation mechanisms in Pyrex spin-exchange cells.

Much work, mostly trial-and-error, has been done to consistently fabricate cells

that have long lifetimes or to apply various nonrelaxive coatings to cells. Important

contributions were made by Fitzsimmons et al. [19] in their study of relaxation

mechanisms in bare cells of various glass types. They determined that adsorption

(sticking) and absorption (permeation) of 3He contributes significantly to relaxation,

and they proposed some phenomenological theories that supported their findings.

To date, it is considered by many to be the definitive work on 3He relaxation. Timisit

et al. [17] investigated relaxation in several types of glass vessels that contained only

3He, as well as vessels that had been treated by irradiation or that contained various

materials, such as pieces of brass or silicon. Many of the results were inconsistent

but demonstrated the unique sensitivity of 3He relaxation to different surfaces. Heil

et al. [20] showed that coating storage cells (cells used to hold HP gas that are not

necessarily used for spin exchange) with various metals, especially Cs, resulted in

very long relaxation times compared to those of bare glass. Hsu et al. [21] used a

sol-gel coating to produce SEOP cells with lifetimes approaching the bulk dipole-

dipole limit [see Eq. (1.5)]. The technique was very tedious and time consuming.

For the most part, production of long-lifetime cells has proven inconsistent, with

much lore and tradition dictating techniques used to fabricate and prepare cells.

10

1.3 Thesis Summary

Three chapters of this thesis have been either published as articles (Chapters 3

and 4) or have been submitted for publication (Chapter 6). The other chapters, in

the interest of uniformity, were written in the same format. Thus, each chapter

should read like a self-contained document. The advantage of this is that the

chapters need not be read in any particular order. The drawback is that there

is some repetition of background information, especially in the introduction of each

chapter, although the reader may find this periodic review of concepts helpful.

Chapter 2 is a brief overview of the equipment and methods we used to make

T1 measurements. The high, field-independent nuclear polarization and relaxation

times of 10’s of hours require nonconventional approaches to NMR detection and T1

measurements.

Chapter 3 is an article published in the 1 October 2001 issue of Physical Review

Letters as an introduction to “T1 hysteresis.” T1 hysteresis is an effect characterized

by the dependence of measured 3He relaxation rates on the history of a cell’s

exposure to large magnetic fields (of order several thousand Gauss). We found

that the T1 of a cell can decrease dramatically solely due to exposure to a magnetic

field, and that the original T1 can be restored by degaussing the cell, a process of

rotating the cell in a slowly decreasing magnetic field. We attribute the effect to

magnetic inclusions in the glass, and we present a model for 3He relaxation due to

interactions with such inclusions.

Chapter 4 is an article published in the 1 August 2002 issue of the Journal of

Applied Physics. There has always been a lack of definite, proven techniques for

the consistent and reliable production of long-lifetime (∼ 40 h) spin-exchange cells

that can be routinely used to produce 3He polarized to 40% or more. We have been

very successful in reaching this “40/40” benchmark, and this chapter outlines our

11

protocols with specifications of our apparatus and procedures. This was the first

paper in the literature to provide a detailed discussion of successful cell fabrication

techniques. The contribution to the measured relaxation time due to diffusion in

the cell capillary is also discussed.

Chapter 5 addresses external magnetic field dependence of 3He relaxation. The

model derived in Chapter 3 predicts that T−11 is proportional to the square of a

site’s magnetic moment. This dependence is qualitatively seen in an aluminosilicate

cell that contains Rb and in a bare Pyrex cell. But in Pyrex cells coated with

Rb, a different dependence altogether is observed: a strong field dependence that

is independent of the size of the moments. This chapter describes this unusual and

unexpected behavior.

Chapter 6 is a manuscript submitted for publication to Chemical Physics Letters.

In this chapter we continue to use 3He as a surface probe to determine relaxation

mechanisms in bare (containing no Rb) glass cells. For the first time, we develop a

theory which accurately predicts the surface relaxation rate of 3Hein glass. In bare

Pyrex, above about 200 K, 3He relaxes mainly due to interactions with Fe3+ ions

while dissolved in the glass. At lower temperatures, adsorption to the cell wall is

the major cause of relaxation.

Chapter 7, a follow up to Chapter 6, extends the investigation of relaxation on

bare Pyrex. We show that the theory outlined in Chapter 6 is applicable to bare

aluminosilicate and bare quartz, but not to Pyrex cells containing Rb for SEOP.

Theory dictates that we should be able to achieve at least 80% 3He polarization for

a cell with a 40 h T1, but 50% is as high as has been reported. We show that this

polarization deficit is probably not due to increases in the wall relaxation rate at

the high spin-exchange temperatures (≈ 180C), as has been hypothesized.

Chapter 8 discusses effects of rinsing cells with acids or alkali metals. In an

attempt to eliminate the magnetic inclusions, we rinsed the cells with various acids

12

to dissolve the sites. We found that cells rinsed with acid are typically not very good,

and they tend exhibit T1 hysteresis more strongly than other cells. Atomic force

microscopy images of acid-rinsed Pyrex samples reveal increases in surface-to-volume

ratio due to etching. We also rinsed the Rb out of a few cells to learn about surface

chemistry that takes place between the Rb and the glass. We found that these rinsed

cells still show T1 hysteresis, whereas cells never exposed to Rb do not. As in other

chapters, we are essentially using the polarized 3He as a surface probe to determine

cell relaxation characteristics.

Chapter 9 discusses preliminary results of novel 3He flow MRI experiments done

in conjunction with the Virtual Lung project at Pacific Northwest National Lab-

oratory. This work was possible largely because of the success we have realized

in making long-lifetime cells. The project uses a computational fluid dynamics

model to predict particulate deposition and long-term disease progression in the

human respiratory system. As part of model validation, measurements of gas flow in

physiologically relevant phantoms must be made. The preliminary results described

in this chapter demonstrate the feasibility of HP 3He flow measurements, but there

are some limitations due to the high diffusivity of the gas.

CHAPTER 2

NMR SPECTROMETER AND

TECHNIQUES

2.1 100 kHz Pulse NMR Spectrometer

2.1.1 The Applied Field

All of the 3He T1 relaxation data, except where noted in Chapter 5, were taken

using home-built 100 kHz pulse NMR spectrometers and a ≈ 30 G magnetic field

provided by a Helmholtz pair. We constructed multiple spectrometers, each using

a different Helmholtz pair. The Helmholtz pair used during spin-exchange optical

pumping (SEOP) is ≈ 44 cm in diameter and ≈ 22 cm separation, has 200 turns of

14 AWG wire, and consumes about 80 W, which is dissipated to air. A second set

is attached to a wheeled cart for portability. It is conveniently made from ≈ 42 cm

diameter bicycle rims with ≈ 21 cm separation. Its 110 turns of 14 AWG wire

consume about 140 W, which is dissipated to air. The third set, also made from

bicycle rims, is ≈ 56 cm diameter and ≈ 28 cm separation. Its 115 turns of 12 AWG

wire consume about 180 W, also dissipated to air. We note that the third pair

typically runs very warm, and if a higher field were desired from it, water cooling

would be necessary. We found that using bicycle rims for coil forms is very cost and

time effective, but the channel depth is insufficient to hold many more than about

100 turns of 14 AWG wire.

14

2.1.2 The Spectrometer

The 100 kHz spectrometer is described in detail with circuit diagrams in Ref. [22].

It consists of three sections: the pulse generator, the transmitter, and the receiver.

The receiver consists of a duplexer, several amplifier stages, and a phase-sensitive

detector. The receiver amplifiers and duplexer are contained within one chassis box

and the other components within another box. Both boxes are grounded to the line

ground.

Figure 2.1 shows a block diagram of the spectrometer. The pulse generator

provides a 100 kHz pulse that can be continuously varied from 10 µs to 10 ms.

A pushbutton switch initiates the pulse sequence by enabling a digital gate. The

transmitter uses a 1 MHz crystal oscillator and ÷10 circuitry to obtain the 100 kHz

NMR frequency. The gate, whose duration is determined by the pulse generator

circuitry, controls the transmission of the pulse by opening an analog IC switch.

A switch at the RF amplifier allows the pulse amplitude to be switched between

≈ 3.5 V and ≈ 15.0 V. A typical pulse is 10 µs and 3.5 V. We estimate such a pulse

to result in a < 5 flip angle. The pulse length or amplitude can be increased if

necessary to provide stronger signal as the polarization diminishes. This is never

done during a single T1 measurement, only between different measurements.

A single NMR coil is used to transmit the RF pulse and to receive the NMR

signal. The duplexer consists of a cross diode gate, which conducts during the RF

pulse, and a tunable parallel LC circuit. Here Q ≈ 70, so the impedance is about

25 kΩ at the NMR frequency. The large Q is a result of the large inductance of the

NMR coil. An additional diode gate to ground protects the receiver circuit during

the RF pulse. During the free induction decay (FID), the diode gates channel the

signal to the receiver amplifiers. Four switchable operational amplifiers in series

provide a minimum of 44 dB and a maximum of 104 dB of gain. Phase-sensitive

15

pulse generator

100 kHz frequency generator

analog switch

RF amplifier

diode gates

receiver amplifier stages

phase

detector

audio-frequency amplifier

signal out to scope

gate 100 kHz

L

C

tuning circuit

TRANSMITTER

RECEIVER100 kHz in

from frequency generator

NMR

probe

Figure 2.1. 100 kHz NMR spectrometer. See [22] for a detailed description andcircuit diagrams of the spectrometer.

16

detection takes place by switching an analog switch at the resonant frequency, which

selects between the signal and its inverse, resulting in an audio-frequency signal. One

additional gain stage that can provide 0 dB or 20 dB of gain precedes the output.

The resulting FID can be viewed on an oscilloscope; typical signal amplitudes with

a total of 44 dB of gain and a sample of ≈ 40% polarized 3He are ∼ 5 V. Typical

noise levels are ∼ 20 mV. The enormous signal-to-noise ratio allows us to omit some

commonly used filtering techniques, such as quarter-wave cables.

2.1.3 The NMR Coil

Our NMR coils were constructed such that they would fit easily around the cell

stems, which are typically 6 mm o.d. The coils were wound on a form made from

plastic that can withstand the ∼ 200C operating temperature of the polarizing

oven. About 200 turns of copper Litz wire (25 strands of individually insulated

44 AWG wire) over a length of about 5 mm were used for the coils. Typical coil

inductances are approximately 600 µH. It is important that the RF pulse not have

a measurable affect on the total polarization of the cell, in order to be sure that the

measured T1 is accurate. We estimate that the coil was sensitive to about 0.25 cm3

of the ≈ 50 cm3 total cell volume, or about 0.5% of the cell volume. Thus, even

very large flip angles, up to 90, had only a small effect on the total polarization of

the cell, although such flip angles were always avoided when possible.

2.2 T1 Measurement Techniques

All data were taken using a single transmit/receive NMR coil placed around a

cell stem in order to minimize the volume of gas excited, thereby minimizing the

polarization loss with each pulse. The coil forms also provided convenient holders

17

for the cells during measurements. Very small flip angles, typically < 5, were used,

resulting in negligible polarization loss. Cells were carefully positioned at the center

of the Helmholtz field to minimize relaxation due to diffusion through field gradients.

Because of the highly nonequilibrium state of the gas, relaxation rate measure-

ments were made by sampling the initial FID height at different times and fitting

the data to S(t) = S(0) exp(−t/T1) to extract T−11 . Several T1 measurements could

typically be made on a single polarization of gas.

CHAPTER 3

WALL RELAXATION OF 3He IN

SPIN-EXCHANGE CELLS

3.1 Preface

This chapter is an article that was published in the 1 October 2001 print edition

of Physical Review Letters, and it describes T1 hysteresis, the effect that launched

my thesis research. I first noticed the effect on 11 October 2000 when I was making

polarization measurements on cell 5B by placing it in a ≈ 1 T magnetic field (see Sec.

4.9.2). The 3He relaxation time in the cell had been measured to be about 10 h, so

the trip to and from the magnet should have resulted in a negligible polarization loss.

To my surprise, upon returning from the electromagnet I was unable to detect any

NMR signal. An innocent yet foretelling comment was recorded in the lab notebook

that night at 10:00 pm: “Has T1 changed?” Over the following week, three other

cells were affected in a similar way after high-field exposure: the 3He relaxation

times appeared to immediately and inexplicably decrease. Finally, I began careful,

systematic tests using cell 5A and was able to determine with little doubt that

exposure to a high magnetic field was solely responsible for the sudden decrease in

the measured relaxation times in the cells.

At this point we were concerned that the changes would be permanent and that

our cells were doomed to ruin. Further experiment proved, however, that we could

degauss the cells and restore the original relaxation times. After discussing the

19

results with collaborators, namely M. Conradi and W. Happer, we were convinced

that this discovery was very significant. I must admit that there was some luck

involved. Both cells 5A and 5B, the first cells whose polarizations I attempted

to measure, were rinsed with hydrofluoric acid (HF) prior to being attached to a

manifold for Rb distillation. As we later learned, cells rinsed with acid, especially

HF, tend to show T1 hysteresis more strongly than unrinsed cells (see Chapter 8).

This fortunate coincidence made it easy for me to see the factor of 20–100 changes

in T1 that were occurring in cells 5A and 5B after the polarization measurements.

There are some important points that are not brought forward in the paper. One

is the fact that 3He does not relax at the surface of glass in the simple way described

in Sec. 1.2.4, even though a short correlation time is assumed. This is most pointedly

demonstrated by the pressure dependence discussed in Sec. 3.5 where a magnetized

cell shows a linear dependence of T−11 on pressure (at a constant temperature)

while the same cell has no such dependence when degaussed; the simple model

predicts no pressure dependence whatsoever. Another point is that these results

may affect all SEOP researchers, not just the few who periodically expose their cells

to high magnetic fields. The exquisite sensitivity of the relaxation rate to the wall

characteristics and the dependence of T1 on the square of a magnetic site’s moment

[see Eq. (3.1)] imply that the effect may be present in cells exposed to low fields

(≈ 30 G), such as those used during polarization. In fact, at the time this paper was

published, we had seen differences in T1’s as large as a factor of 2 in unmagnetized

cells depending only on orientation in a 30 G field (see Sec. 5.5.2).

My coauthors on this paper were S. Morgan, an undergraduate who assisted

in some of the data acquisition, J. Leawoods, who, at our request, independently

confirmed T1 hysteresis in cells made by his group at Washington University, and

B. Saam, my advisor.

20

3.2 Abstract

The 3He longitudinal spin-relaxation rate T−11 is crucial for production of highly

polarized 3He by spin-exchange optical pumping. We show that T−11 is increased by

a factor of 2–20 solely by exposure of spin-exchange cells to a few-kG magnetic field.

The original T−11 can be restored by degaussing the cell. The effect is attributed

to magnetic surface sites and has been observed in both Pyrex and aluminosilicate-

glass cells. Our results both advance the understanding of wall relaxation and

demonstrate the use of 3He as an extremely sensitive probe of surface magnetism.

3.3 Introduction

Large nonequilibrium nuclear polarizations can be obtained in certain noble-gas

isotopes by spin exchange with an optically pumped and polarized Rb vapor [1].

Polarizations O(0.1) are routinely achieved for liter-quantities (STP) of 3He and

129Xe at 350–450 K in applied magnetic fields B0 = O(10) G. These hyperpolarized

(HP) gases have been studied and applied in diverse scientific realms [23, 24, 25],

perhaps most dramatically as the signal source in magnetic resonance imaging (MRI)

of the lung air space [9]. Indeed, we are concerned here with HP 3He as prepared

for most MRI experiments, where one requires large (≥ 40 cm3) valved glass vessels

(cells) which can be repeatedly polarized, emptied, and refilled with 3He to pressures

approaching 10 atm.

The production and subsequent storage of highly polarized gas depends crucially

on the nuclear spin-lattice relaxation rate T−11 , which shorts out the delivery of an-

gular momentum by Rb–3He spin exchange. Since the characteristic spin-exchange

time for Rb–3He is at least several hours [26], a stable T1 of many tens of hours

is required to generate and preserve substantial magnetization. T−11 is usually

dominated by interactions with the cell surface (wall relaxation). Bulk relaxation

21

from 3He–3He collisions [15] also contributes at sufficiently high pressure (greater

than several atm). Despite decades of research, relatively little is known about

the nature of 3He wall relaxation at most surfaces. This has generally led to

irreproducibility in cell fabrication. Several types of glass have been tried with

varying degrees of success [19, 27, 28], but documented fabrication protocols yielding

consistent results are generally lacking, especially for large-volume valved cells.

In this Letter we present evidence that magnetic sites, showing remanence and

hysteresis, significantly affect, if not dominate, wall relaxation in spin-exchange cells.

We demonstrate that large reversible changes in T−11 , and hence in the corresponding

surface relaxation sites, are induced in such cells solely by exposing them to a large

(∼ 10 kG) magnetic field, and that this effect (termed “T1 hysteresis”) is correlated

with the presence of Rb in these cells. The presence of Rb is also strongly correlated

with reduced wall relaxation rates (by as much as an order of magnitude) compared

to bare-wall glass cells. Our results represent the first explicit evidence of the nature

of a surface-relaxation mechanism for 3He in spin-exchange cells. Further study of

this mechanism will likely yield vital information for the efficient and reproducible

production of highly polarized 3He by spin exchange.

3.4 Experimental

Our valved Pyrex cells have 10 cm of 0.5–1.0 mm i.d. capillary separating the

valve (glass stem with o-rings) from the ≈ 50 cm3 main chamber. Each cell was

attached to a glass manifold and baked (except for the valve) under high vacuum

(base pressure 2 × 10−8 Torr) for 2–4 days at ∼ 400C. Rb metal (100–300 mg;

> 99.93% pure) was then distilled in prior to flame-sealing each cell under vacuum

from the manifold. A separate gas-handling system was used to fill and refill the cells

with 3He to 8 atm at room temperature [29]. A sidearm protruding from the valve

22

body, normally used for gas filling and dispensing, defines two physical orientations

of a cell with respect to an applied magnetic field; these are termed “north” and

“south” according to whether the sidearm points to the north or south pole of the

magnet.

All relaxation measurements were made at room temperature using 100 kHz

NMR detection at H0 ≈ 30 G (see Chapter 2) . Very low flip angles were used to

generate large-amplitude free induction decays (FIDs) with neglible loss of longitu-

dinal magnetization. The initial height of the FID was recorded as a function of

time and fit to an exponential decay to extract T−11 .

The basic experimental sequence consisted of three pairs of T−11 measurements

made with the cell oriented north and then south (or vice versa). Each measurement

pair was made with no intermediate removal of the cell from the 30 G field, no

heating, and no exposure to laser light; the cell was simply rotated 180 about

its capillary axis and a new T−11 measurement was initiated. The first pair was

performed after the cell was fabricated and filled for the first time (before any

exposure to high field); the second pair was done after the cell was magnetized north

or south, i.e., exposed for ≈ 30 s to the 10 kG field of an iron-core electromagnet

in the specified orientation (n.b., the word “magnetize” here refers to the cell walls

and not to the 3He spins); the third pair was made after the cell was degaussed. A

magnetized cell is degaussed by rotating it at ≈ 1 Hz about the capillary axis in

the field of the electromagnet as the field is gradually lowered from 10 kG to the

electromagnet’s remanent field (≈ 30 G). The rotation is maintained as the cell is

slowly withdrawn from the magnet. The second and third pairs of measurements

were repeated after magnetizing the cell in the opposite cell orientation.

23

3.5 Results and Discussion

We have performed this sequence of measurements on 20 cells (40 cells as of

October 2002). All cells we have examined show significant and consistent increases

(factors of 2 to 20) in wall relaxation rate due solely to exposure to the 10 kG field.

All cells previously exposed to the 10 kG field show a nearly complete restoration

of the original relaxation rate after being degaussed. In addition, all magnetized

cells show a consistent dependence of T−11 on physical orientation (north or south)

in the 30 G field; this change is typically ≈ 20%, but factors of 2–3 have been

observed. Cells that have been magnetized north (south) at 10 kG have a larger

T−11 oriented north (south) with respect to the 30 G measurement field. These

results are reproducible over several exposures to the 10 kG field, several degaussing

procedures, several refills with 3He, and several repolarizations. Figure 3.1 is a plot

of relaxation rate vs. chronological history of magnetic-field exposure for a single

representative cell demonstrating all of the described effects. The initial lifetimes

vary among the cells from 10s of minutes to 10s of hours, but the qualitative behavior

shown in Figure 3.1 is the same for all.

We performed a number of checks to confirm that high-field exposure is the sole

and direct cause of the change in T−11 observed before and after magnetizing or

degaussing a cell. For most cells, several measurements are possible without need of

repolarizing the gas. In many cases, all that transpires between radically different

T−11 measurements at 30 G is that a cell is transported in a portable solenoid back

and forth from the 30 G Helmholtz coils to the electromagnet in order to be exposed

to the 10 kG field. We have verified that a partial or sloppy degaussing procedure

(e.g., slowly removing the rotating cell from the magnet without turning down the

field) only partially restores the original T−11 . One of us (J.C.L.) has observed T1

hysteresis in two valved Pyrex cells fabricated and filled using a different glass blower,

24

0.0

0.2

0.4

0.6

0.8

1.0

1.2

un

mag

n.

deg

aussed

mag

n. n

orth

mag

n. so

uth

mag

n. n

orth

deg

aussed

mag

n. so

uth

UnmagnetizedDegaussedMeasured NorthMeasured South

T1-1

(h

ou

rs-1

)

chronological order

Figure 3.1. T1 hysteresis of cell 5A. Relaxation rates at 30 G are plotted vs. thechronological history of intervening exposure to a 10 kG field for a single cell. Thecell was both magnetized and measured in each of two physical orientations, labeled“north” and “south.” Mere exposure to the large field increased the rate by about20 times. Rates were slightly greater when the magnetizing and measuring fieldswere in the same direction with respect to the cell’s orientation.

vacuum system, and filling system. The effect has also been observed unambiguously

in two aluminosilicate-glass cells, one of GE-180 (General Electric) and one of 1720

(Corning). The 1720 cell is a sealed 8 cm3 spherical cell containing 3.5 atm 3He (at

295 K) and has no valves or capillaries; it was prepared by a third research group.

Our results point to the existence of magnetic sites at or near the glass surface

of our spin-exchange cells. These sites are a major source of wall relaxation for cells

exposed to fields greater than several hundred Gauss. This conclusion is supported

by the data in Fig. 3.1, which shows that wall-relaxation rates T−11 in our cells

have all of the basic characteristics of magnetic hysteresis, including remanence,

orientation dependence, and the ability to be degaussed.

25

Previously, wall relaxation has almost always been ascribed to isolated para-

magnetic impurities at or near the surface [17, 19], but such a mechanism has

never been explicitly experimentally demonstrated. Surface paramagnetism may

well dominate relaxation in bare-wall cells, but it would not show the hysteresis,

reversibility upon degaussing, and orientation dependence of T−11 that we observe

in our Rb-coated cells. The large fractional change in T−11 for cells with a broad

range of initial lifetimes (tens of minutes to tens of hours) suggests that the size

and/or concentration of magnetic sites may be responsible for the wide variation in

relaxation rates that is often observed with spin exchange cells. Indeed, the data

in Fig. 3.1 actually understate the effect of T1 hysteresis on many of the cells with

longer initial lifetimes, since these also have a significant bulk contribution to the

wall rate (about 0.01 h−1 at 8 atm [15]), regardless of whether the cell is magnetized.

When a cell is first fabricated and Rb distilled in, the domains in each magnetic

site are randomly oriented, or perhaps slightly aligned. Exposure to the 10 kG field

aligns the domains and produces a large enhancement of the magnetic moment of

each site. A remanent magnetization exists in the cell after it has been removed to

30 G, where an increased T−11 is then measured. When the domains are randomized

by degaussing, the magnetic moment of each site is reduced, and T−11 returns to its

original value. We propose that the 3He spins relax by interacting with these sites

while diffusing near the cell surface. We assume N sites having magnetic moment µ

and radius R. In the weak-collision limit [30], where the interaction time τ is much

shorter than the 3He Larmor period at 30 G, the longitudinal relaxation rate for one

site is M2τ , where the second moment M2 ≈ (γµ)2/R6. Using τ = R2/6D, where

D is the diffusion coefficient, we obtain for the whole cell:

1

T1

=Nπγ2µ2

9RDV, (3.1)

26

where γ is the 3He gyromagnetic ratio, V is the cell volume, and we have factored

in the fraction of spins interacting with sites (≈ 2πR3N/3V ). This analysis assumes

that the mean free path λ for 3He atoms is much smaller than R (λ ≈ 24 nm at 8

atm [31]).

Equation (3.1) suggests a linear pressure dependence of T−11 through 1/D. We

have investigated this dependence by measuring T1 after each of several releases of

a known quantity of polarized gas from the cell. Prior to each measurement, the

capillary entrance to the cell was blocked by maneuvering a small bead of Rb metal

over the opening. Our results from one cell supporting the weak-collision theory are

shown in Fig. 3.2. By contrast, the limit τ À (γB0)−1 would produce an inverse

0.000

0.005

0.010

0.015

0.020

0.025

0.030

3 4 5 6 7 8

T1 magnetized

T1 degaussed

Wal

l T1-1

(h

ou

rs-1

)

Pressure (atm)

Figure 3.2. T1 pressure dependence. The appropriate He–He relaxation rate hasbeen subtracted from all data shown to yield the wall relaxation rate as a functionof pressure. The wall rate increases linearly with pressure when the cell is magnetized(supporting the weak-collision theory), but there is no pressure dependence after thecell has been degaussed.

27

linear dependence on pressure [16]. An upper limit on R can thus be calculated

from R2 = 6D/γB0 and yields R = 10 µm for 3He at 8 atm, where we have used

D = 0.23 cm2/s [32]. Since R must be at least several times λ (or else there is no

pressure dependence whatsoever), we place a lower limit on R of ≈ 0.1 µm.

For example, if the sites were metallic iron (see below), and we use V = 50 cm3,

R = 0.25 µm, and a magnetized T1 of 5 h, we obtain N = 4×104 sites. Here, we have

used the density of iron to obtain an estimate of 5.6× 109 atoms per site and have

assumed that each atom contributes one Bohr magneton at full magnetization. This

number of atoms is reasonable for producing the multidomain structure necessary

to generate T1 hysteresis.

Our hypothesis for the cause of the orientation dependence is that the 30 G mea-

surement field causes a slight deviation from the zero-field remanent magnetization

of the cell, thus slightly increasing or decreasing the magnetic moments (and hence

relaxivity) of the sites. The orientation dependence of T−11 we observe at 30 G is

consistent with this picture in all cells we have tested.

We have also investigated the dependence of wall relaxation and T1 hysteresis on

the presence of Rb in the cell. Two additional cells, otherwise identical to the others,

were prepared using the same protocol except that Rb distillation was omitted.

HP 3He was transferred to these bare-wall cells from another room-temperature

spin-exchange cell, and the measurement sequence described above was perfomed.

The bare cells exhibited no T1 hysteresis. We then remounted the cells to the high

vacuum system and distilled in the usual amount of Rb so as to visibly coat most

of the cell surface. Again, HP 3He was transferred from another spin-exchange cell,

and the standard measurement sequence was performed. Not only did T−11 decrease

significantly after the introduction of Rb, but T1 hysteresis was also observed; see

Fig. 3.3.

28

0.01

0.1

1

un

mag

n.

mag

n.

deg

aussed

un

mag

n.

mag

n.

deg

aussed

T1-1

(h

ou

rs-1

)

chronological order

no Rb

Rb

Figure 3.3. Rb dependence. Relaxation rate is plotted vs. the chronological historyof preparation of a single cell. This cell was tested before and after introducing Rbmetal. For all measurements, hyperpolarized 3He was transferred in from anotherspin-exchange cell. Rb both greatly reduces the wall relaxation rate and gives riseto T1 hysteresis.

The data of Fig. 3.3 suggest that the presence of Rb both inhibits wall relaxation

and gives rise to T1 hysteresis. The former conclusion is in line with earlier work [19,

20], and we have confirmed the effect using the same cell, thus reducing uncertainties

associated with cell-to-cell variation. We can speculate at present only about how

the Rb (itself or in a compound) beneficially affects the cell walls: it may, for

example, chemically neutralize paramagnetic sites. It may also act as a physical

barrier to surface sites or to helium permeation of the glass. It is further apparent

that Rb plays a role in creating magnetic sites, perhaps by acting as a reducing agent

on ionic iron impurities in the glass, catalyzing the formation of ferromagnetic iron

oxides or metallic iron. Alternatively, the 1 g Rb ampules we use [33] have Fe, Ni,

29

and Co impurities at the ≈ 10 ppm level, although these levels may be reduced

by distillation. The characteristic applied field at which cells become magnetized is

about 500 G, with saturation occurring at 1–2 kG – reasonable numbers for iron or

iron oxide impurities. The bare-wall cells we measured had T1’s between 5 h and

12 h, comparable to or longer than T1’s measured for most of the Rb-coated cells

when magnetized. It is therefore not likely that the sites are resident initially in the

glass and that the Rb is simply removing a more dominant nonhysteretic mechanism.

We have initiated studies of Rb-coated Pyrex using ESR, SQUID, and the magneto-

optical Kerr effect [34], in order to look for an independent confirmation of magnetic

hysteresis as well as to better quantify the size, concentration, and chemical identity

of the sites. Results so far are negative. However, we note that ESR and SQUID

suffer from decreased filling factor compared with our measurements, which are

exquisitely sensitive to the surface alone.

Our understanding of both T1 hysteresis and the importance of the Rb coating has

allowed us to make substantial progress toward reproducible fabrication of Pyrex

spin-exchange vessels. Early research suggested that the helium permeability of

Pyrex glass leads to large wall relaxation rates [19]. More recently, Hsu, et al.

[21] showed that long T1’s were possible even for simple Rb-coated Pyrex. Pyrex

remains attractive for spin-exchange cells despite its difficulties because it is rugged,

inexpensive, ubiquitous, and easy to work compared with most other glasses. Most

of our cells have T1 > 30 h when degaussed. Several cells have T1 > 60 h, from

which one infers wall relaxation times > 150 h using the bulk relaxation rate at

8 atm [15]. Absent exposure to high field, we find these T1 values to change very

little as the cells are repeatedly heated to 160–180C, exposed to the 40 W laser,

and repeatedly refilled with gas. We routinely produce 3He polarizations > 40% in

these cells; they are robust and well suited to the MRI experiments for which they

30

were designed.

3.6 Conclusion

We conclude that 3He T1 hysteresis is a robust, reproducible, and consistent

effect which should be observable to some degree in almost all spin-exchange cells.

The effect is observed only in the presence of the Rb needed for optical pumping

and may be due to ferromagnetic impurities which are either in the Rb itself or

are catalyzed by Rb at the glass surface. Our results suggest an approach to

making reproducible spin-exchange cells that greatly narrows the search for effective

fabrication techniques to those that are likely to affect the size, concentration,

and magnetic moment of the sites responsible for this effect. Our results also

demonstrate the first use of hyperpolarized 3He as an extremely sensitive probe

of surface magnetism.

We acknowledge helpful discussions with M.S. Conradi, P.A. Fedders, R.V. Cham-

berlin, and W. Happer, and the glass-blowing of J. Kyle. This work was supported

by a grant from the Whitaker Foundation.

3.7 Addendum

Although, at the time this paper was submitted, we strongly felt that the mag-

netic sites originated in the glass, we had little evidence. Later experiments provided

strong circumstantial evidence, but nothing conclusive, in support of our hypothesis.

Two of these experiments are discussed in Sec. 8.5.1 and Sec. 8.5.2, in which cells

are rinsed with a chemical reducing agent and rinsed of their Rb, respectively. All

of these rinsed cells, which contained no Rb, showed T1 hysteresis.

Further evidence against the sites originating with the Rb was found in two

independent Rb distillation experiments. In one experiment, a manifold was pre-

31

pared with the section containing the Rb reservoirs and retort bent down from the

horizontal at about a 30 angle. This forced the first two distillation steps to be done

up-hill, which prevented globs of Rb from flowing up the tube, as is common during

distillation. The idea was to prevent any clusters of iron that might be present in the

Rb from riding such globs, entering the cells, and causing T1 hysteresis. However,

the cells still showed T1 hysteresis, providing evidence that the sites do not originate

in the Rb, and were of shorter than average lifetime. The distillation process in this

experiment required excessive heating of the manifold, which caused some visible

Rb-glass reactions (i.e., visible orange and black stains formed inside the glass) and

required much more time than usual. Because there were no apparent benefits of

this distillation method, it has not been repeated.

The second experiment used a locally strong magnetic field (≈ 4 kG over an area

of ≈ 1 cm2) placed around the cell manifold during Rb distillation. The field was

created by two permanent-magnet discs attached to a steel U channel facing each

other with a 1.5 cm separation. If iron clusters were present in the Rb, the magnetic

field could have two effects during distillation as the Rb passed through the field:

it could trap iron particles that are present in the Rb, or it could magnetize iron

particles as they pass through. If the former occurred, then the cell should not

exhibit T1 hysteresis. If the latter, then the cell would start out magnetized with an

initial relaxation time that could be improved with degaussing. The cells showed

neither behavior, suggesting again that the sites originate in the glass.

CHAPTER 4

3He SPIN-EXCHANGE CELLS

FOR MRI

4.1 Preface

This chapter is an article that was published in the 1 August 2002 print edition

of the Journal of Applied Physics. This paper was motivated by our unprecedented

success at making long-lifetime spin-exchange cells, and the fact that there were

no previous publications detailing cell-making procedures. This chapter contains

specific information about the equipment involved in producing and filling the cells.

Most of the equipment designs and procedural protocols were based on work done

by my advisor, B. Saam, while he was a postdoctoral researcher at Washington

University in St. Louis. A few of my most significant contributions include the

following: the transition to spherical cells from oblong cells, the introduction of

the oven for cell baking, and the refinement of cell preparation protocols. We have

found that our spherical cells tend to have longer lifetimes than our oblong cells.

This may be due to the higher surface-to-volume ratio of spherical cells, however

we have no direct proof. The oven has greatly reduced bake-out set up time and

potential damage to manifolds due to over handling. Cell preparation protocols have

always been based on some degree of lore or legend. Through trial and error and the

production of many quality cells, we have established a protocol that results in good

cells routinely and reliably. Unfortunately, we still have not been able to decisively

33

pinpoint exactly what step(s) determine the cells’ quality. However, through my

work in producing about three dozen cells, we have been able to gather evidence

that a cell’s relaxation properties are set during cell preparation or introduction of

Rb. My coauthors were S. Morgan, an undergraduate assistant who did much of

the equipment assembly, and B. Saam.

4.2 Abstract

We present a protocol for the consistent fabrication of glass cells to provide hyper-

polarized (HP) 3He for pulmonary magnetic resonance imaging (MRI). The method

for producing HP 3He is spin-exchange optical pumping (SEOP). The valved cells

must hold of order 1 atm·L of gas at up to 15 atm pressure. Because characteristic

spin-exchange times are several hours, the longitudinal nuclear relaxation time T1

for 3He must be several 10s of hours and robust with respect to repeated refilling

and repolarization. Collisions with the cell wall are a significant and often dominant

cause of relaxation. Consistent control of wall relaxation through cell fabrication

procedures has historically proven difficult. With the help of the discovery of an

important mechanism for wall relaxation that involves magnetic surface sites in the

glass, and with the further confirmation of the importance of Rb metal to long

wall-relaxation times, we have developed a successful protocol for fabrication of 3He

spin-exchange cells from inexpensive and easily worked borosilicate (Pyrex) glass.

The cells are prepared under vacuum using a high-vacuum oil-free turbomolecular

pumping station, and they are sealed off under vacuum after ≥ 100 mg of distilled

Rb metal is driven in. Filling of cells with the requisite 3He–N2 mixture is done on an

entirely separate gas-handling system. Our cells can be refilled and the gas repolar-

ized indefinitely with no significant change in their wall properties. Relaxation data

are presented for about 30 cells; the majority of these reach a “40/40” benchmark:

34

T1 > 40 h, and 3He polarizations reach or exceed 40%. Typical polarization times

range from 12–20 h; 20% polarization can be achieved in 3–5 h.

4.3 Introduction

The past decade has witnessed vigorous progress in the study of HP noble

gases and their application to a broad range of problems in physics, chemistry,

and biomedicine. Advances are coming in areas as varied as neutron polarizers

[28], measurements of fundamental symmetries [35], NMR at surfaces [36, 37], and

magnetic resonance imaging of the lung air space [9, 38]. In HP gases, enormous

nonequilibrium nuclear spin polarizations (of order 0.1) can be attained at room

temperature in ordinary magnetic fields via optical pumping techniques [1, 39],

greatly enhancing the NMR sensitivity of these nuclei. We are concerned here with

SEOP [1] of 3He gas for application to pulmonary MRI. The advent of relatively

inexpensive high-power diode-array lasers has paved the way in particular for MRI

and other applications requiring large quantities (of order 1 atm·L) of polarized gas,

since the quantity is essentially limited by the available laser power.

HP 3He is produced (and often stored) inside a glass spin-exchange cell containing

3He at several or more atmospheres, 50–100 mbar N2 (a fluorescence-quenching gas

necessary for efficient optical pumping [12]), and a macroscopic amount of alkali

metal (typically Rb). The cell is heated to 160–200C to obtain the optimal Rb

vapor density. The laser light, circularly polarized at a frequency corresponding to

the D1 atomic transition in Rb (795 nm) and colinear with a small magnetic field

(of order 10 G) is trained on the cell, thus polarizing the valence electron of the Rb

atoms. The polarization is thence collisionally transferred to the 3He nuclei.

In this paper we present a protocol for the consistent fabrication of spin-exchange

cells which will provide liter quantities of highly polarized 3He for pulmonary MRI.

35

These cells must (1) hold a quantity of gas comparable to an average adult’s tidal

volume (≈ 0.5 atm·L), (2) be transparent to 795 nm laser light, (3) withstand

pressures of up to 15 atm at 200C (making efficient use of the spectrally broad

diode-laser array by suitably broadening the Rb absorption line [40]), and (4) be

valved and refillable for repeated use without altering important cell characteristics

(mainly the longitudinal nuclear relaxation rate at the cell surface).

4.4 Wall Relaxation

Controlling longitudinal nuclear spin relaxation is critical to optimizing both the

polarization and the useful storage time of the gas for applications. In MRI, for

example, polarization is directly related to image quality for a given amount of 3He,

and the gas often needs to be transported some distance to the MRI scanner without

significant polarization loss.

The noble-gas polarization transient PN(t) during optical pumping is given by:

PN(t) = 〈PA〉 γse

γse + Γ[1− e−(γse+Γ)t], (4.1)

where 〈PA〉 is the time- and volume-averaged alkali-metal electron polarization, γse

is the spin-exchange rate, and Γ is the 3He longitudinal relaxation rate, with the

corresponding relaxation time T1 ≡ 1/Γ. Since typical spin-exchange times are 5–

10 h, T1 must be several tens of hours to obtain noble-gas polarizations approaching

〈PA〉, which can normally be kept close to unity [14].

Contributions to Γ come from bulk He-He binary collisions, gas diffusion through

ambient gradients, and from wall collisions. The bulk rate is linear with the 3He

density and is usually only significant for cells above several atm; for our room-

temperature 8 atm cells it is 0.010 h−1 [15]. The Helmholtz coils we use for

36

SEOP (see Sec. 4.8) provide adequate field homogeneity so that the rate due to

diffusion [16, 41] for cells of the size and pressure discussed here is negligible. The

wall relaxation rate thus practically dictates the quality of a 3He spin-exchange

cell, and efforts have been made for more than thirty years to understand and

control it. Various glass types, surface treatments, surface coatings, and bakeout

procedures have been tried. The results vary widely among research groups and are

usually inconsistent and irreproducible. Paradoxically, consistently long polarization

lifetimes seem especially difficult to achieve for larger-volume cells, where one would

expect a lower surface-to-volume ratio to yield generally slower rates.

Work to understand and control 3He wall relaxation has generally proceeded

from the assumption that the major source of such relaxation is paramagnetic

impurities on the glass surface. In early work [17, 19], bare borosilicate (Pyrex)

and aluminosilicate glass surfaces (such as Corning 1720) were studied. These

workers used metastability exchange optical pumping [39], which does not require

an alkali-metal intermediary to produce HP 3He. A local maximum in T1 (at about

130 K) as a function of temperature in Pyrex [19] suggested the importance of

helium permeability, which brings the 3He into close and prolonged contact with

the surface. (Quartz is even more permeable than Pyrex [42], so much so that most

quartz cells would leak substantial fractions of their helium to the atmosphere in

days or weeks.) Lower permeabilities and overall better results generally led the

community toward the use of aluminosilicates [15, 28, 43], although these glasses

are generally more difficult and expensive to procure and are more difficult for a

glass blower to work than borosilicates. Good results were, however, reported for

sealed cells using Corning 7056, a high-alkali borosilicate glass with much lower

helium permeability than Pyrex [27]. An excellent review of the results of many

groups using various glasses and coatings for pumping and for storage cells is given

37

in Ref. [44].

We have chosen to continue working with Pyrex, due to its robustness, worka-

bility, and easy availability. Moreover, the presence of Rb (which surely coats the

cell walls to some degree) chemically alters the surface and inhibits wall relaxation

relative to bare Pyrex [19, 21, 44]; see Sec. 4.9.1. Indeed, in Ref. [19] there was

only one cell tested which contained Rb, and that cell had the longest T1 of all in

that work by a large margin. T1’s in the hundreds of hours have been observed in

HP 3He storage cells with macroscopic coatings of Rb and Cs metal [20]. These

developments, coupled with the discovery of a previously unknown relaxation mech-

anism involving magnetic surface sites [45] and some trial-and-error testing, have

led to our consistent achievement of two benchmarks, T1 = 40 h and PN = 40%, for

large-volume Pyrex 3He spin-exchange cells.

4.5 Cell Fabrication

Our cells are made of standard borosilicate glass (Pyrex), but we have also

experimented with quartz and aluminosilicate glasses. The cell body is either

spherical (≈ 4.5 cm i.d.) or cylindrical with rounded ends (≈ 3 cm i.d. × 5.5 cm

long). The typical total volumes are 50 cm3 and 35 cm3, respectively; see Fig. 4.1.

Recently, we have gone exclusively to spherical cells, as it is easier to produce a

surface of uniform thickness, thus minimizing lensing of the incident laser light.

The spherical cells generally yield longer T1’s and higher polarizations, although the

reasons for this are not clear. We use 32 mm heavy-wall tubing which is “reblown”

to the specified inner diameter, creating a freshly exposed inner surface. The cells

are shaped by blowing the glass on a lathe; cylindrical cells require the additional

use of a graphite shaping paddle on the outer surface. A capillary tube, valve, and

stem are then attached.

38

Figure 4.1. A Pyrex valved spin-exchange cell for generating hyperpolarized 3He.The spherical cell body shown here has a volume of ≈ 50 cm3. The capillary allowsthe o-rings in the valve to sit outside the ovens involved both with the initial bakeout(see Sec. 4.6) and with optical pumping (see Sec. 4.8). The penny is shown for scale.

The valve is a right-angle, high-vacuum, all-glass valve [46]. Perpendicular to

the valve is attached a threaded glass side arm [47], through which polarized gas is

dispensed and by which the cell is attached to a separate gas-handling system (see

Sec. 4.7) for filling with 3He. The valve is attached to the cell via a 10 cm length

of glass capillary, which consists of a 6 cm length of 0.5 mm i.d. tubing in series

with a 4 cm length of 1 mm i.d. tubing. The wider end is attached to the cell body

and helps to prevent the Rb metal from clogging the capillary; the narrow portion

is attached to the valve. The gas must pass through the capillary during cell filling

and dispensing, so it cannot be impractically narrow. The capillary allows the valve

to be kept outside of the oven during optical pumping and suitably lengthens the

transit time of a 3He atom from the cell body to the ≈ 1 cm3 volume near the valve.

Because of the unknown relaxation characteristics of the valve materials and the fact

that the valve cannot be baked out well, it is assumed that all 3He atoms that enter

39

the valve volume relax completely. The capillary plays a measureable role in wall

relaxation of our long-lifetime cells. By measuring T1 before and after maneuvering

a bead of Rb in the cell to block the capillary, we have estimated its contribution

to be Γcap = 0.002–0.004 h−1 at 8 atm pressure; see Sec. 4.10.

A 4 cm length of 6 mm standard-wall tubing (the “stem”) is attached opposite

to the capillary and connects the cell to a glass manifold for initial evacuation and

baking; see Sec. 4.6. After the cell is flame-sealed away from the manifold, the stem

accommodates an NMR coil for monitoring the production and decay of HP 3He;

see Sec. 4.8.

The manifold is basically a long tube (primarily 12 mm o.d. Pyrex) connecting

the high vacuum system on one end to an open vertical retort on the other end. The

retort, which is 15 mm o.d., accommodates a prescored 1 g ampule of 99.93% pure

Rb metal [33]. The cells (usually two at a time) are attached orthogonally to the

manifold by their stems; see Fig. 4.2. The manifold includes two small reservoirs

used in the Rb distillation process between the retort and the cells. A u-tube liquid

nitrogen (LN2) trap is located between the high-vacuum port and the cells. The

LN2 trap provides additional cryopumping of the manifold, limits backstreaming

contamination, and prevents Rb from migrating to the high-vacuum system.

The cells are fabricated and attached to the manifold by our chemistry depart-

ment’s glass blower. The completed manifold is annealed at 560C with a soaking

time (time for which the glass is held at the maximum temperature) of ≈ 10 min.

The manifold is then allowed to cool slowly for about 45 min. Upon removal from

the annealing oven, the open ends are covered with a self-sealing wax film to help

prevent ambient moisture or other contaminants from entering and adsorbing to the

inner surfaces of the manifold.

40

retort

reservoirscells

capillary

valveside arm

section bakedin oven

B

LN2trap

A

to high-vacsystem

stem

Figure 4.2. Diagram of a cell manifold. The glass manifold with two cells attached,as it appears just before being connected to the high-vacuum system (Fig. 4.3).The top of the retort is flame-sealed after a Rb ampule is dropped in. The cellsare attached to the manifold at a 55 angle out of the page. Constrictions in themanifold and stems allow for easy pull-off. Manifolds are labeled by numbers andcells by letters, starting with the furthest upstream.

4.6 Cell Preparation

The purpose of careful cell preparation is to remove impurities adsorbed to the

surface of the glass and to prevent contaminants from entering. To accomplish this,

the cells are baked under vacuum using an oil-free high-vacuum system; see Fig. 4.3.

The construction is stainless steel with copper-gasketed or swaged connecting seals

and packless, bellows-sealed valves [48]. The vacuum pump is a turbomolecular

drag pump backed by a diaphragm pump [49]. With the cell manifold attached,

the system reaches a base pressure of ≤ 4× 10−8 mbar, monitored at the inlet

by a combination cold-cathode/Pirani full-range gauge [50]. Connected opposite

the gauge is a residual gas analyzer (RGA) [51], which also functions as a helium

leak detector. The gauge and RGA are downstream from the 38 mm stainless

steel right-angle main valve. Upstream of this valve are large- and small-bottle

41

compressionseal

to large-bottlemanifold

RGAfull-range

gauge

pressuresensor

to dryroughing

pumpto small-bottlemanifold

to turbopump

mainvalve

to glassmanifold

Figure 4.3. The oil-free high-vacuum system used for cell fabrication. The glassmanifold (Fig. 4.2) is attached via the compression-seal fitting at right. Theconstruction is stainless steel with packless bellows-sealed valves. Nitrogen purgegas is provided as needed through the connection to the large-bottle manifold.

gas-handling manifolds, an additional diaphragm roughing pump [52], and a port for

connecting the system to the glass manifold via a 12.7 mm o-ring compression-seal

fitting. A 0–1.3 bar capacitance manometer [53] and a solid-state pressure sensor

[54] are used for fine and coarse monitoring of upstream pressures. High-purity

nitrogen, used as a purge gas, is available through the large-bottle manifold; the

attached gas-handling manifolds are generally used for making permanently sealed

spin-exchange cells and will not otherwise be described here.

The completed manifold with cells is attached to the vacuum system at the

manifold port. This is done while purging the system with research-grade nitrogen

gas to help prevent water vapor and oxygen from entering the manifold and vacuum

system. The valve stems, with lightly greased ethylene-propylene (alkali-metal

resistant) o-rings [55], are then seated, and the cell valves are closed. An ampule

42

of Rb is opened in the flow of N2 from the retort and is dropped, open end down,

into the retort. The purge gas is turned off, and the retort is flame-sealed shut.

With the manifold now sealed from the external environment, it is evacuated with

the roughing pump, opened to the turbo pump, and tested for leaks with the RGA.

Minor leaks can often be repaired on the spot, while more serious problems may

have to be sent back to the glass blower. Finally, with the manifold sealed and

leak-tight, LN2 is added to the dewar surrounding the trap.

The manifold is then baked continuously at about 400C for 2–4 days. We now

use a home-built, insulated, steel-walled oven designed especially for these manifolds.

We previously wrapped the manifold in heating tape and aluminum foil. The oven

heats stably and uniformly (within a few degrees), avoiding hot spots that can

develop from the use of heating tape; it also greatly reduces set-up time and the

risk of damaging the delicate manifold due to overhandling. The cells are attached

to the manifold at a 55 angle from vertical (see Fig. 4.2) to allow the valves to

protrude laterally through a cut-out in the oven (to avoid damaging the o-rings).

The oven is blanketed by 25 mm thick ceramic fiber insulation [56], which is covered

with high-temperature silica cloth [57]. This blanket keeps the valves near room

temperature, even though they are only a few centimeters from the oven. Heat is

provided by a 1.8 kW, 120 VAC ceramic strip heater [58] located on the oven floor,

which is controlled by a 2 kW, 120 VAC solid-state dimmer switch. A 50 VAC

input is sufficient to maintain the baking temperature. The retort and distillation

reservoirs also protrude out of the oven; these are wrapped in heating tape and

aluminum foil, which can be removed as necessary as Rb distillation progresses.

Temperatures are monitored with type-E thermocouples placed both in the oven

and on the manifold under the heater tape and foil.

After about 24 h of baking, the Rb metal is liquified for the first time by brief

43

exposure to a flame, which allows any trapped gases to escape and be pumped

away. After allowing the retort to cool, Rb is distilled from the retort to the first

reservoir. First, the foil and heat tape are unwrapped from the retort to 3 cm

beyond the first distillation reservoir, and this section is allowed to cool. The Rb

is then “chased,” i.e., heated and evaporated with a cool methane-oxygen flame

(not so cool that carbon deposits, which inhibit visibility, are left on the glass),

driving it into but not beyond the first reservoir. The idea here is to volatilize all

of the Rb that is eventually to be chased into the reservoir, leaving less volatile

contaminants in the retort. The flame is not held in one spot on the glass long

enough to produce any orange sodium glare; this avoids softening of the glass or

its reaction with Rb. Effective chasing requires 30–45 min and benefits from some

practice. When completed, contaminants (e.g., Rb oxides and RbOH) and about

10–20% of the Rb metal are left behind in the retort, which is then flame-sealed

away from the manifold.

After 12–24 h of further baking, Rb is chased to the second reservoir by a similar

procedure. The cells are allowed to bake for yet another 12–24 h before the oven is

turned off and allowed to cool completely. Rb is then chased into the cells, starting

with the one closest to the retort (cell “A”). The capillary is kept hot during this step

so that it will not become clogged with Rb. After 100–300 mg of Rb is distilled in,

the cell is flame-sealed from the manifold. The process is repeated for the remaining

cell(s), working downstream.

We have found a difference in convenience only, and not in cell quality, with

the introduction of the baking oven. We also find no difference in the case where

the Rb is distilled into all of the cells in one step before they are each sealed

from the manifold. We note that the amount of Rb distilled in does matter;

amounts significantly less than about 100 mg generally result in cells with poor

44

wall characteristics. The relationship between Rb and wall relaxation is discussed

further in Sec. 4.9.1. The baking time and temperature have not been optimized

experimentally, but those we use seem reasonable based on the ideas that (1) we

wish to bake as hot as possible without approaching the annealing point of the glass

and (2) the base vacuum pressure and RGA spectrum change very little after a

day or so of baking. Our ability to consistently produce quality cells has compelled

us not to experiment much with our bakeout parameters or other aspects of the

fabrication protocol.

4.7 Cell Filling System

A separate, home-built vacuum and gas-handling system is used to fill and refill

cells; see Fig. 4.4. This system is similar to one built previously by Saam and

Conradi [9]. It is constructed of 6.4 mm o.d., 4.8 mm i.d. stainless steel tubing with

weld and swage fittings. The packed, nonrotating-stem valves [59] are labeled by

letters, as shown in Fig. 4.4. The system, built vertically on two large aluminum

plates affixed to a relay rack, is divided into an upper gas-handling and purification

manifold and a lower vacuum manifold; these are connected at two points by valves

(I) and (G). The vacuum manifold has a dial gauge and a thermocouple gauge to

monitor pressure. Below valve (M) is a u-tube LN2 trap followed by a 150 L/min

rotating-vane mechanical pump. The pump has a Micromaze [60] trap at the inlet

to further inhibit oil backstreaming.

The upper manifold is essentially a fill path from the gas bottle to the cell. When

new, the lecture bottle contains 25 atm·L of research-grade 3He (99.99% pure) mixed

with 2% nitrogen, at a total pressure of 55.4 bar. The ≈ 3 cm3 volume bounded

by the bottle valve and valve (J) is the charging volume used for a new bottle. As

the gas is used up and the bottle pressure decreases, the ≈ 12 cm3 charging volume

45

A B C D

EF G H I

J

K L M

to cell

N2 line

purifiers

N2 line

tocollectionbottle

to LN2 trap andmechanicalvacuum pump

high-pressuregauge

vacuumgauge

t.c.gauge

3Hebottle

upperpanel

lowerpanel

Figure 4.4. Gas-handling system used to fill cells with 3He. This system iscompletely separate from the high-vacuum system (Fig. 4.3) used for bakeout andRb distillation. Cells can be refilled indefinitely with no change in the 3He wallrelaxation time. The valves and bottle are all mounted to two vertical plates ona relay rack. The dashed line marks the boundary between the upper and lowerplates. Open squares represent auxiliary ports. The collection bottle is used to save3He that would otherwise be discarded after cell filling.

up to valve (H) is employed. The gas is conducted through a purifier [61] bounded

by valves (B) and (C), and then to the cell through a 1 m length of 1.6 mm i.d.

stainless steel (ss) capillary tubing. The flexible ss capillary provides stress relief for

the cell and a low dead volume for filling. The cell is placed in a secure wooden box

(mounted to the side of the relay rack) with its glass capillary and valve protruding.

It is attached, using the compression-seal fitting on the side arm, to the ss capillary

tube via a custom-built tee connector. The other outlet of the tee is exhausted to

46

room air through a flow meter and one-way pressure-relief valve. Because of the low

conductance of the ss capillary, a N2 purge line with purifier is provided and used

in addition to evacuation to keep the system clean. The two identical gas purifiers

discussed here are designed for use with nitrogen but are also adequate for use with

noble gases.

Initially, the cell (with its valve closed) is attached to the tee connector with the

compression seal loose. A N2 purge is passed through the capillary at ≈ 0.2 L/min

for ≈ 20 min. After the purge, the compression seal is tightened, valve (A) is closed,

and the fill line is evacuated through valve (G). The system remains under vacuum

for tens of minutes to hours. A series of at least three backfill/evacuate sequences

(closing valve (G), filling the ss capillary and cell side arm with purge gas, then

re-evacuating) is completed to help remove room air that may have entered when

the cell was attached.

When the evacuation procedure is complete, valve (G) is closed, and the cell

valve is opened in preparation for filling. The cells are filled in a series of charges.

At the start of each charge, all valves except the cell valve are closed. The bottle is

opened to the charging volume and closed again immediately. The valves (H), (C),

and (B) are successively opened until the pressure measured by the 0–200 psig dial

gauge equilibrates. Valves are opened gradually, so that the gas can be metered;

this is particularly important in opening valves (C) and (B), since the purifier is

effective only at flow rates below 0.2 L/min. In our system, staying below this rate

is assured by watching the pressure gauge and keeping the rate of change in pressure

below about 2 psi/s. At equilibrium, all valves except the cell valve are closed again.

This procedure is repeated until the desired cell pressure is reached; we typically

fill cells to 8 atm (absolute pressure). When the desired final pressure is reached,

the cell valve is closed, and all other valves in the filling path are closed to preserve

47

the gas contained therein for future use. The filling path from the bottle to valve

(B) is evacuated and pumped on only occasionally, such as when the 3He bottle is

replaced. The remaining 3He is collected into a large ballast volume (an otherwise

empty gas cylinder) for recycling.

4.8 The Polarizer

The polarizer consists mostly of an aluminum-frame cart (≈ 2 m long, 0.6 m deep,

and 1 m high) with a top surface for mounting optical components and shelving

below. The cart has a built-in 45 cm dia. Helmholtz pair (200 turns of 14 AWG

wire per coil), which produces a 30 G field when driven in parallel with 12 V at 8 A.

A welded aluminum box covered with fiberboard insulation serves as an oven. It

is located at the center of the Helmholtz pair and is heated by air forced through a

filament-heater pipe [62] attached to the cart. The temperature is maintained to a

few tenths of a degree with a resistive temperature detector (RTD) and controller

[63]. A cradle at the center of the oven holds the cell body; the capillary protrudes

out one side of the oven to avoid heating the valve. The top plate and the top half

of the side plate above the capillary are welded together and can be removed as a

unit for internal access. Round windows (5 cm dia.) are located on four sides–for

laser entry, laser exit, on top, and laterally opposite the capillary. The latter two

are for monitoring fluorescence from the cell during SEOP. Window glass (6.4 mm

thick) is used throughout, double-paned for extra insulation except at the laser-

entrance window. (Laser transmittance through this window could be improved by

a few percent by using an optical flat, anti-reflection coated for 795 nm.) The oven

temperature is typically set for 160C, although based on the characteristic time we

observe to polarize cells (Fig. 4.5), the actual cell temperature is 170–180C, where

the saturated vapor density of Rb is about 2.5–4.5×1014 cm−3 [13]; see Sec. 4.9.2.

48

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0 500 1000 1500 2000 2500 3000

TIME (min)

T1 = 60 ± 2 h

Ch 1: 200 mVolt 2.5 ms

(b)

0

1

2

3

4

5

6

0 200 400 600 800 1000 1200 1400 1600

SIG

NA

L (

Vo

lts)

TIME (min)

(γse

+ Γ)-1= 10.0 ± 0.2 h

(a)

Figure 4.5. Polarization and decay transients. (a) Typical 3He polarization transientfor cell 8A at 8 atm in a 30 G magnetic field. The curve is a best fit to Eq. (4.1),yielding the characteristic “spin-up” time of 10 h. Although the oven temperaturewas set to 160C, the spin-up time corresponds to 170–180C. The temperatureincrease is due to the laser heating of the cell. (b) Subsequent room-temperaturedecay transient at 30 G for the same cell, measuring the combined wall and bulkrelaxation rates. The line is a best fit to a single exponential decay. One deduces awall-relaxation time for this cell of ≈ 150 h, since the bulk time is ≈ 100 h. Inset:A typical 3He FID acquired at 30 G (100 kHz); the inital FID amplitude providesthe data for spin-up and spin-down measurements.

49

The temperature difference is caused by the large laser power: the N2 gas in the cell

heats up due to increased collisional de-excitation of Rb atoms [64]. The laser and

optics train are mounted on optical rail to one side of the coils and oven. The laser

is a 795 nm, 40 W diode array [65] with a fwhm linewidth of about 1.7 nm. The

laser is resonant with the 795 nm D1 transition in Rb (see Sec. 1.2.3). It is mounted

to an aluminum block which is water-cooled by a closed-loop refrigerator [66] with

built-in temperature control to a few tenths of a degree. Cylindrical optics are used

to collimate the fast and slow axes separately. A 76 mm dia. mica quarter-wave plate

[67] circularly polarizes the light just before it enters the oven. Rubidium-vapor

absorption is monitored by a PC plug-in miniature spectrometer [68] with 0.3 nm

resolution coupled by a fiber optic to the light emerging from the laser exit window.

When the laser is at the proper frequency, the computer displays the laser line

with a central dip corresponding to absorption by the vapor. Laser fluorescence is

visually monitored with surveillance-system cameras and a small tv monitor. The

cameras are sensitive in the near-infrared and are equipped with interference filters

[69] for the Rb D2 resonance (780 nm). The D1 and D2 states are mixed after a Rb

atom absorbs a photon [12]; monitoring D2 fluorescence thus effectively separates

fluorescence from laser scatter. The beam profile can then be matched to the cell

dimensions by adjusting lens positions.

The 3He polarization is monitored by a home-built 100 kHz pulse NMR spectro-

meter [22]; the multiturn coil (inductance L ≈ 800 µH, quality factor Q ≈ 50) is

placed around the stem of the cell inside the oven. The aluminum walls of the oven

are grounded and provide adequate rf shielding. Figure 4.5 shows samples of a free

induction decay (FID), polarization transient (“spin-up”), and room-temperature

decay transient (“spin-down”) for 3He in a typical 8 atm cell.

50

4.9 Experimental Results

4.9.1 T1 Measurements

All measurements of the longitudinal relaxation time T1 for our cells were made

at room temperature and at 30 G, unless otherwise indicated. These are relevant

conditions for HP-gas production, since high fields are not necessary to generate

the polarization, and a 30 G Helmholtz pair is inexpensive and portable. Very low

flip angles (< 5) were used to generate a FID at appropriate time intervals with

negligible loss of polarization. The initial height S of the FID was recorded as a

function of time and fit to S(t) = S(0) exp(−t/T1) to extract T1. (The thermal

equilibrium signal is negligible for our conditions.)

We have recently discovered that T1 at 30 G is dramatically reduced (factors of

2–20) solely by intervening exposure of a cell to a large magnetic field (of order

0.1 T or greater). The original T1 can be restored by demagnetizing or degaussing

the cell, i.e., rotating it at about 1 Hz in a field which is gradually reduced to

zero from about 1 T. (Magnetizing and degaussing were conveniently done in a 1 T

iron-core electromagnet.) The effect, termed “T1 hysteresis,” is due to multidomain

magnetic sites at or near the glass surface. These sites become magnetized in a

large field and have a significant remanent magnetization when the cell is returned

to 30 G, leading to a stronger interaction with colliding 3He spins and a shorter

T1. The effect has been observed in borosilicate, aluminosilicate, and quartz glasses

and is correlated with the presence of the Rb necessary for SEOP. Indeed, the effect

should be observable to some degree in almost all spin-exchange cells. The details

of T1 hysteresis are discussed in Chapter 3.

We have consistently produced cells with T1’s in excess of 40 h. The measured

T1’s are generally robust and reproducible (to 10% or so), although the cell is refilled

several times and/or repeatedly exposed to the 40 W laser at temperatures of 160–

51

200C. The results of T1 measurements on many of our cells are shown in Fig. 4.6.

The wall relaxation rate Γw is shown for each cell before exposure to a high field

(unmagnetized), magnetized, and degaussed. The bulk contribution (0.010 h−1 at

8 atm [15]) has been subtracted from the data. T1 hysteresis occurs to some degree

for all cells over a broad range of initial unmagnetized lifetimes (a few hundred

minutes to tens of hours), indicating that the variation in rates from cell to cell may

be due to differences in the size and/or concentration of magnetic sites. In any case,

10-3

10-2

10-1

100

101

CELL DESIGNATION

UnmagnetizedMagnetizedDegaussed

WA

LL

RE

LA

XA

TIO

N R

AT

E

Γ w (

h-1

)

5A 6A 7B 10B8B 9B 15B14B12B11B5B 7A 8A 11A9A 10A 15A14A12A

Figure 4.6. Relaxation rates for several cells. The wall relaxation rate Γw at30 G plotted vs. cell designation for 19 of approximately 30 3He spin-exchangecells fabricated in our laboratory. These cells all contain Rb, except for 12A and12B, which contain potassium. The manifolds are numbered chronologically; “A”and “B” refer to a pair of cells made on the same manifold (Fig. 2). The bulk3He–3He relaxation rate of 0.010 h−1 for 8 atm has been subtracted for each cell.Prior to being magnetized at 1 T or after degaussing, most cells have Γw ≈ 0.01 h−1

or smaller, meaning a measured T1 ≥ 50 h.

52

it is necessary to avoid exposure of most cells to large magnetic fields (the onset of

T1 hysteresis occurs at a few hundred gauss [45]) or to degauss them before they are

used for SEOP.

In addition to its crucial role in optical pumping, the presence of alkali metal in

SEOP cells inhibits wall relaxation while simultaneously giving rise to T1 hysteresis

[45]. Almost all of the cells represented in Fig. 4.6 contain Rb (cells 12A and

12B contain potassium). We also made several otherwise identical cells containing

no alkali metal. Polarized 3He was then introduced in order to measure T1 on

bare Pyrex. For these cells T1 was typically less than 10 h, and no hysteresis was

observed. The Rb may be chemically neutralizing paramagnetic sites or inhibiting

permeation of 3He into the glass; see Sec. 4.4 and Refs. [17, 19]. At the same time,

the Rb introduces T1 hysteresis, presumably by creating magnetic sites. Our current

working hypothesis is that the Rb acts as a reducing agent, converting iron ions and

oxides to multidomain metallic iron. Section 8.5.1 explores this possibility further.

4.9.2 Polarimetry

Using our benchmark T1 of 40 h, the measured Rb–3He spin-exchange rate at

180C [26, 70], the saturated vapor pressure curve for Rb [13], and Eq. (4.1), the

theoretical limit of attainable 3He polarization is about 80%, given 〈PRb〉 = 100%.

We estimate (and other research groups have shown directly [14]) that 〈PRb〉 can be

maintained at nearly 100% under these conditions with a diode-array laser such as

the one we use; yet there are no reports in the literature of 3He polarizations above

about 50%. This polarization deficit remains unexplained at present.

We measure the absolute polarization of HP 3He by comparing the NMR signal

from the 3He cell to a proton signal provided by a water sample in thermal equi-

librium. The water sample has a geometry similar to the 3He cell and contains a

53

sufficient amount of dissolved CuSO4 to reduce the proton T1 to less than 100 ms.

The comparison is done at a common NMR frequency, high enough for sufficient

proton signal. Since the proton polarization can be calculated, and the spin densities

of both samples are known, the 3He polarization can be determined by measuring

the ratio of NMR signals from the two samples. (The polarization is independent of

applied field for hyperpolarized gases.) In the low-flip-angle limit, the signal ratio

S3/S1 is given by [22]

S3

S1

=P3

P1

n3

n1

γ23

γ21

, (4.2)

where P is polarization, γ is gyromagnetic ratio, n is spin density, and the subscripts

“3” and “1” refer to 3He and protons, respectively. The spin density for the 3He cell

can be calculated from the pressure measured when it is filled. When solved for P3

using water, Eq.(4.2) can be expressed

P3 = (3.76× 10−4)f

p

S3

S1

, (4.3)

where f is the common NMR frequency in Megahertz, and p is the cell pressure

in atmospheres at room temperature. For our cells, S3/S1 is typically 40–50 dB.

Fig. 4.7 shows a pair of FID’s at 32.5 MHz on the same oscilloscope voltage scale.

The 3He FID is a single acquisition from an 8 atm cell with 50 dB attenuation in the

signal line; the proton FID is four averaged signals with no signal-line attenuation.

Using Eq. (4.3) and factoring in the slight difference in the two FID amplitudes, the

3He polarization is P3 = 50 ± 4%. The uncertainty comes from the measurement

of the proton FID height and from small losses in transporting the cell from the

polarizer to the electromagnet.

54

>

Ch 1: 100 mVolt 1 ms

dY: 487 mVolt

(a)

1 >

Ch 1: 100 mVolt 1 ms

dY: 503 mVolt

(b)

Figure 4.7. Polarimetry free-induction decays. Two FID’s acquired at 32.5 MHzusing the exact same NMR equipment and settings; only the field is different.The flip angle is < 10 in both cases. The peak-to-peak voltage of the first fulloscillation is marked by the solid horizontal cursor lines. (a) Four averaged waterproton signals acquired at 0.763 T. (b) A single acquisition at 1.00 T from an 8 atm3He cell with 50 dB of signal-line attenuation relative to (a). Using Eq. (4.3), the(field-independent) 3He polarization is 50 ± 4%. The transverse coherence timeis dominated by field gradients and is longer in (b) because of the better fieldhomogeneity in the electromagnet at higher fields.

55

4.9.3 Overall Performance

We now routinely fabricate SEOP cells that reach a “40/40” benchmark: 3He

polarization ≥ 40% and a 8 atm relaxation time T1 ≥ 40 h. The maximum

polarization is achieved in 12–20 h in about 0.5 atm·L of gas, although polarizations

as high as 20% are achieved in 3–5 h. The system can be left to run overnight

unattended in order to achieve maximum polarization. The apparatus described

here cost $35–40k to build; about $15k of that total was spent on the high-vacuum

system.

A few other research groups and at least one company, Amersham Health (AH),

have put some effort into high-volume, high-throughput devices to generate HP

3He. With the exception of Gentile et al. [44], there are few details of these systems

described in the open literature. Recent papers from the University of Virginia

group describe the AH system (which also uses SEOP but is not yet commercially

available) as capable of up to 35% polarization in ≈ 1 atm·L of gas after several

hours [38, 71]. Our system is roughly comparable, although our spin-exchange rates

are typically somewhat lower. We note that while we have yet to use our system

for human studies, the group at Washington University has obtained the necessary

FDA exemption for a system very similar to ours.

Groups at NIST in Gaithersburg, MD and at Mainz University in Germany have

employed the technique of metastability exchange optical pumping (MEOP) [39].

The Mainz group can produce ≈ 1 atm·L of 55% polarized 3He in about 2 h [6].

As with all MEOP systems, the gas must be compressed from a few Torr up to

atmospheric pressure with minimal polarization loss. A two-stage titanium piston

compressor is employed for this purpose. The disadvantages of this system are its

size, complexity, and nonportability. Our system is portable enough to have been

recently driven from Salt Lake City to Richland, Washington for collaborative MRI

56

experiments at Pacific Northwest National Laboratory. At NIST, a compact and

portable device for gas compression involving a modified diaphragm pump has been

developed. For MRI applications, it is expected to produce ≈ 1 atm·L of 20%

polarized gas in about 2 h [44]. We achieve substantially higher final polarization

than the NIST system at the expense of considerable pumping time.

4.10 Transit Time of 3He in the Capillary

As discussed in Sec. 4.5, the valve on our refillable 3He spin-exchange cells is

separated from the cell body by a glass capillary tube (see Fig. 4.1), so that the

potentially relaxive components of the valve are isolated from the bulk gas. Here we

calculate Γcap, the 3He relaxation rate due to the capillary and valve. The analysis

here is based in part on notes from discussions with both H.L. Middleton and M.S.

Conradi. We assume a capillary of length L, radius r, and cross section A = πr2.

We assume that the diffusion time, both across the cell and between the cell and the

valve, is short compared to the polarization lifetime T1, that T1 is dominated by wall

relaxation, and that the cell has a uniform relaxivity η, so that the wall-relaxation

rate Γw is given everywhere by

Γw = ηS

V, (4.4)

where S/V is the surface to volume ratio, equal to 2/r for the cylindrical capillary.

We consider the case in which the magnetization M0 in the cell body may be

considered constant. We note that it is quite possible that η in the capillary is larger

than in the body (due, for example, to the low conductance during the bakeout or

to the capillary coming from a different batch of Pyrex). We deal with this possible

57

difference below. In the capillary, the diffusion equation for 3He magnetization M(x)

in one dimension is:

∂M(x)

∂t= D

∂2M(x)

∂x2+ Q(x), (4.5)

where D is the diffusion coefficient of 3He atoms in the cell and Q(x) is a source

term. Under steady-state conditions with Q(x) = −ΓwM(x) and using Eq. (4.4) we

obtain:

d2

dx2M(x) =

ηS

DVM(x) =

2η

DrM(x), (4.6)

The general solution to Eq. (4.6) is

M(x) = C sinh(qx) + K cosh(qx), (4.7)

where

q2 = 2η/Dr. (4.8)

We assume that the valve at x = 0 instantly relaxes all spins with which it comes

in contact. The boundary conditions are thus M(0) = 0 and M(L) = M0. Hence,

we must have K = 0, and the particular solution is

M(x) =M0

sinh(qL)sinh qx. (4.9)

58

The flux of magnetic moment J(x) through a plane of constant x in the capillary

is thus

J(x) = −DdM(x)

dx= − DM0q

sinh(qL)cosh(qx), (4.10)

and the total magnetic moment per unit time flowing into the cell at the capillary

opening is

AJ(L) = −πr2DM0q coth(qL). (4.11)

Hence, the effective relaxation rate due to the capillary and valve is

Γcap = −AJ(L)

M0Vc

=πr2Dq coth(qL)

Vc

, (4.12)

where Vc is the cell volume. In our case, η is usually small, and in the limit qL ¿ 1,

Γcap =πr2D

VcL. (4.13)

Note that in this limit, Γcap is independent of η, and a modest increase in η in the

capillary compared to the cell body would be irrelevant.

Based on T1 = 40 h for a spherical cell 4.5 cm in diameter, we estimate η ≈5 × 10−6 cm/s. For an 8 atm cell, D = 0.23 cm2/s for 3He at 295 K [32]. Using

r = 0.025 cm and L = 6 cm for the narrow portion of the capillary and Eq. (4.8),

we obtain qL = 0.25. Using Eq. (4.13), we calculate Γcap ≈ 0.008 h−1. This number

59

is a factor of 2–4 greater than the measured range for Γcap given in Sec. 4.5 for our

cells. The discrepancy has at least two potential sources: the additional 4 cm of

1 mm i.d. capillary in our cells was not included in this calculation, and the valve

may not be perfectly relaxing (i.e., we may have M(0) > 0).

4.11 Conclusion

We have developed a successful protocol for fabrication of large-volume, valved

3He spin-exchange cells for MRI from inexpensive and easily-worked Pyrex glass. We

have identified an important mechanism for wall relaxation that has been directly

confirmed experimentally by studies of T1 hysteresis, and we have confirmed the im-

portance of Rb metal (in amounts of order 100 mg or more) for long wall-relaxation

times.

We have yet to reach the ultimate goal of understanding the physics of the cell

fabrication process at each step, but we have detailed here some progress away

from cell-making “voodoo.” The filling of cells has been separated from the rest of

the process and is done on a separate gas-handling system. Cell properties are

determined and, so far as we know, fixed by one or more of the earlier steps

(glass blowing, evacuation, baking, and Rb distillation). Once sealed from the

high-vacuum system, cells can be refilled indefinitely with no significant change

in their wall properties. Based on the comparative previous experience of one of us

(B.T.S.) with several cell-fabrication systems, we can make an educated guess that

the important elements of cell-preparation include the clean, oil-free turbomolecular

pump, the u-tube LN2 trap included on the glass manifold, and the multistage

distillation of the Rb metal into the cells. Our vacuum system does not qualify as

UHV and is not entirely metal-sealed, although the stainless-steel construction and

60

good vacuum practice (e.g., keeping air and water out of the system at all possible

times) presumably help to further minimize contaminants in the manifold.

The discovery of T1 hysteresis has opened the door to learning more about cell

fabrication by finally providing a concrete lead as to the dominant mechanism

involved in glass-surface relaxation of 3He. The detection of magnetism in Rb-coated

glass with a second method (ESR or vibrating-sample magnetometry, for example)

would confirm the effect and, in conjunction with further NMR measurements,

potentially allow better determination of the chemical identity, size, concentration,

and magnetic moment of the magnetic sites. It may eventually prove possible to

eliminate the sites altogether, perhaps improving T1 still further and making cells

even more robust in high-field environments, such as in or near an MRI magnet.

We gratefully acknowledge many useful discussions with M.S. Conradi and J.C.

Leawoods at Washington University, as well as the expert glass blowing of J. Kyle.

This work was supported by a grant from the Whitaker Foundation.

4.12 Addendum

As discussed in this paper, much attention is paid to cleanliness during cell

preparation: the baking under high vacuum and high temperatures, the three-step

Rb distillation process, and the separate 3He filling system with gas purifiers all

help prevent cell contamination. It is well established that gaseous oxygen can relax

3He very rapidly [72], but relatively little is known about relaxation properties of

rubidium oxides. It has always been assumed in cell preparation “lore” that clean,

pristine cells have the best chance at long lifetimes. Recent evidence, however,

suggests that oxidized Rb could be beneficial [73]. Heil et al. [6] show that the

relaxation rate of 3He interacting with a substrate is strongly dependent on the

substrate’s work function. Qualitatively this makes sense, Heil et al. argue, because

61

substances with low work functions have loosely bound electrons which repel the

He atoms more effectively than a substance with a higher work functions. This is

a possible explanation of the relatively long lifetimes observed in Rb and Cs coated

cells. Alkali metals have comparatively low work functions, and alkali oxides can

have even lower work functions. For example, Rb oxide at room temperature has a

work function of about 0.9–1.2 eV [74], which is substantially lower than the 2.26 eV

for metallic Rb [75]. An alkali oxide layer may, therefore, be quite beneficial in a

spin-exchange cell, as long as a sufficient Rb vapor can be achieved for SEOP.

We are conducting experiments in which we add oxygen to spin-exchange cells and

investigate the effects on 3He relaxation rates. These experiments are ongoing at the

time of this writing, but preliminary results indicate that relaxation rates actually

improved somewhat by adding oxygen. However, T1 hysteresis is not eliminated, as

we hoped that the iron sites would oxidize and become nonmagnetic. Any oxygen

added to a cell reacts rapidly with the Rb. Depending on how much oxygen is added,

we have observed Rb turn, in progression, slightly dark, then a dark-bronze colored

liquid, then a very dark brown solid, and finally a yellowish solid. Cells with Rb in

all conditions except the last can be optically pumped because enough metallic Rb

remains to provide sufficient vapor pressure.

One example is cell 11A′, which is discussed in Sec. 8.5.3. This cell had a

relaxation time of ≈ 40 h prior to the introduction of oxygen. The cell was opened at

the valve to release the 3He, then it was attached to an oxygen bottle. About 5 psi of

oxygen was added to the cell. All of the visible metallic Rb turned dark brown and

solid at room temperature; the oxidized Rb does not melt at SEOP temperatures.

We then attached the cell to the high vacuum system (see Fig. 4.3) with the cell

valve open and evacuated any excess oxygen over several days. After filling the cell

with 3He we measured a relaxation time of about 52 h. After subtracting dipolar

62

He–He relaxation, 0.01 h−1 at 8 atm and room temperature [see Eq. (1.5)], this cell

showed an increase in the wall relaxation time of about 60%. This significant increase

indicates a promising technique for improving wall relaxation times in existing cells.

Further investigations should include quantifying the amount of oxygen required

to optimize a cell’s relaxation rate, investigating long-term behavior of the cells,

and investigations of applied field and temperature dependence. By quantifying

and optimizing the amount of oxygen needed, a step could be easily added to the

cell preparation protocol. Long-term cell behavior could be influenced by the ability

of the Rb to getter oxygen or water vapor that might leech in from the glass. We

typically rely on the Rb to neutralize such otherwise harmful contaminants that

cannot be removed during baking or evacuating or that might be inadvertently

introduced during cell filling. If insufficient Rb remains to take care of this task,

then oxygenated cells may have a limited useful lifetime or may require periodic

evacuation, both undesirable. Finally, studies of temperature and field dependence

of T−11 in oxygenated cells may provide clues about the nature of the Rb-oxide–3He

interactions.

CHAPTER 5

MAGNETIC FIELD DEPENDENCE

OF 3He RELAXATION

5.1 Abstract

An observed external magnetic field dependence of measured 3He relaxation

rates is characterized by a dramatic increase in T−11 with decreasing external field

magnitude and is inconsistent with a previous relaxation model. The effect is

observed only in Pyrex spin-exchange cells (cells containing Rb for spin-exchange

optical pumping). This field dependence is not observed in aluminosilicate or quartz

spin-exchange cells or in bare (containing no Rb) Pyrex cells. The effect could

be caused by the 3He remaining near relaxation centers much longer than the

typical ballistic interaction time of ≈ 10−12 s, but possible reasons for the very

long interaction times are not given. In all cells measured at high fields, local T−11

minima at coercive fields and an asymptotic approach to a maximum value of T−11 at

high fields is expected and observed. The lack of the dramatic increase in measured

T−11 ’s in aluminosilicate and quartz is the only qualitative difference that we have

observed in 3He relaxation in different types of glass.

5.2 Introduction

There have been few attempts in the past to investigate field-dependent nuclear

relaxation of 3He in glass vessels [19, 76]. These attempts have generally been at

64

low, positive external fields (≈ 0 to 430 G) with bare, sealed cells at low pressures

(generally a few Torr). A slight field dependence in measured relaxation rates was

observed, but no attempts at an explanation were made. In [19], observed field

dependent relaxation in bare cells was attributed to paramagnetic centers, the major

cause of relaxation in bare cells (see Chapter 6 for a detailed discussion of 3He

relaxation in bare cells). In Chapter 3 we showed that ferromagnetic sites are largely

responsible for 3He relaxation in cells containing an alkali metal for spin-exchange

optical pumping (SEOP) and that such sites appeared to be absent in bare cells.

We studied the dependence of T−11 on high external fields, in the relatively broad

field range of −2000 G to +2000 G. In this chapter we show that Pyrex cells that

contain Rb demonstrate a strong field dependence independent of the size of the

magnetic sites’ moments. This effect is absent in aluminosilicate and quartz spin-

exchange cells and in bare Pyrex cells. We also show that bare Pyrex cells appear

to demonstrate field dependence characteristic of the presence of magnetic sites.

The possibility of observable T1 hysteresis at low fields (on the order of tens of

Gauss) is more pertinent to most HP gas researchers than effects due to high fields.

All spin-exchange cells are exposed to low fields for optical pumping or polarized

gas storage, whereas few researchers expose cells to higher fields. If field-dependent

effects appear in such fields, then T1 hysteresis will have broader importance. For

example, careful attention may have to be paid to the orientation of a cell during

polarization and HP gas storage.

5.3 Theory

5.3.1 High-field Hysteresis

As discussed in Chapter 3, T1 hysteresis is characterized by a sometimes dramatic

increase in T−11 measured at low field (≈ 30 G) due solely to intervening exposure

65

of the cell to a several-kG magnetic field. The original T−11 can be restored by

degaussing the cell by rotating it in a gradually decreasing magnetic field. In Chapter

3 we proposed that N spherical, magnetic sites of moment µ at or near the surface

of the cell are the main source of relaxation in magnetized cells. We presented a

model for magnetic-site relaxation (relaxation due only to interactions of 3He with

the sites) of 3He in spin-exchange cells; see Eq. (3.1). According to this model, T−11

should be directly related to the cell pressure (at a given temperature) through the

gas diffusion coefficient D and inversely related to the square of the moment of a

site, µ2. We experimentally demonstrated the former relationship in Chapter 3, so

in this chapter we attempt to demonstrate the latter.

The total magnetization M , the magnetic moment per unit volume, of a typical

ferromagnetic substance is a multivalued function of the applied external magnetic

field H, best described graphically by a hysteresis loop (see Fig. 5.1). If the model

for relaxation in Eq. (3.1) is correct, then the field-dependence of T−11 should look

qualitatively like the square of the hysteresis loop (that is, M2 vs. H). Notably, there

should be local T−11 minima at the coercive fields, where the net magnetization is zero

for a nonzero applied field, and T−11 should “saturate,” or asymptotically approach a

maximum value, as the magnetization saturates at large positive or negative fields.

5.3.2 Low-field Hysteresis

If low-field (≈ ± 30 G) T1 hysteresis is observable in unmagnetized or degaussed

cells, the field dependence of the measured T−11 ’s should behave similarly to that

described above, except saturation will be absent, and the coercive field may be too

small to measure without the relaxation being affected by field gradients [see Eq.

(1.6)]. However, the 60 G change in field may be too small to cause any measurable

T1 hysteresis if the loop between −30 and +30 G is very narrow or is a straight line.

66

M

H

coercive fields

Figure 5.1. A sketch of a typical hysteresis loop showing the relationship betweenmagnetic moment M and applied field H. Of note are the coercive fields: nonzeroapplied fields at which the net magnetization is zero. Also, the magnetizationapproaches saturation values as the magnitude of the external field becomes large.Different ferromagnetic materials have different characteristic hysteresis loops.

In this case, if a cell is perfectly degaussed and if a 30 G field is strong enough to

change the magnetization, then a plot of T−11 vs. applied field would be single-valued

parabolic function with a minimum at H = 0 (where M = 0).

5.4 Experimental

Most cells studied were Pyrex, ≈ 50 cm3, spherical, valved, and prepared ac-

cording to our cell-making protocols (see Chapter 4). The aluminosilicate cell

(GE-180) used was ≈ 40 cm3, cylindrical with rounded ends, valved, and prepared

by collaborators at Washington University. All cells contained ≈ 8 amagats of 3He,

except for the bare Pyrex cell which contained ≈ 4 amagats of 3He.

The high-field measurements were made in an electromagnet. The electromagnet

has automatic field reversal capabilities, but field spikes that occurred during field

reversal forced us to use a different procedure. To reverse the field we reversed

67

the cell orientation by lowering the field to zero, removing the cell, rotating the

cell 180 degrees in earth’s field, and replacing it. This avoided rotating the cell in

the magnet’s ≈ 60 G remanent field. The best method for making field-dependent

measurements is to measure relaxation rates at the fields of interest without field

cycling (or cycling the field to and from a measurement field). This insures that

measurements are made on only one hysteresis loop and at the right position on

the loop. This necessitated the design and construction of an NMR spectrometer

(see the Appendix) which has the capability of measuring 3He NMR in the range of

−3000 G to +3000 G (a cell side-arm pointing to the magnet’s north pole defines

a positive field). A low-Q probe consisting of a series LC resonator in series with

a 50 Ω resistor was used for NMR detection. The coil, which accomodates the cell

stem (see Fig. 4.1), was placed in a fixed position in the magnet such that the body

of the cell was in the center of the magnet. Very small flip angles were used to

minimize polarization destruction. Because intervening polarization of the gas was

necessary to complete a loop, each cell was degaussed prior to reintroduction to the

field. This helped assure that we would find our way back to the same hysteresis

loop each time.

The measurements made on the bare Pyrex cell were made possible by trans-

ferring polarized gas into it from a spin-exchange cell of similar size. This was

done on a special gas-transfer manifold (see Fig. 5.2). The manifold is made of

1 mm i.d. Pyrex capillary tubing, to minimize the volume, with two ports where

cells attach, a valved port to attach the vacuum system, and a valved port for a

pressure sensor. The cell ports are separated by about 5 cm to to allow the cells

to be adjacent to each other. The manifold is positioned in a Helmholtz pair such

that the cell bodies are on the longitudinal axis of the coils and the entire manifold

is in a nonzero field. This helps prevent 3He relaxation due to field gradients during

68

X

X

to pressure

sensorto vacuum

system/purge gas

valve

valve

to cell

to cell

Figure 5.2. Gas transfer manifold. This manifold, made of 1 mm capillary Pyrextubing, is used to transfer polarized gas from a spin-exchange cell to a bare cell.Either cell can be in either position.

the transfer procedure. The cells attach via o-ring compression fittings. After the

cells are attached, the manifold is evacuated. It is then backfilled with nitrogen

and evacuated in sequence several times to help remove moisture and oxygen. After

about 20 minutes of pulling vacuum, the vacuum-system port valve is closed, the

valve of the cell into which gas will be transferred is opened, and the high-pressure

cell is opened. After equilibrium is reached, usually after a few seconds, both cell

valves are closed and the cells can be removed.

As with the high-field measurements, the low-field measurements were preceded

by degaussing the cells. To avoid any intervening exposure to high magnetic fields

and to assure thorough degaussing, we used a tuned, 60 Hz series LC circuit,

controlled by a variable AC transformer, capable of producing fields up to ≈ 1400 G.

Four 35 µF 440 V capacitors in a series/parallel configuration provide a total of

35 µF capacitance. The inductor cavity is ≈ 10 cm in diameter and ≈ 7.6 cm long,

69

large enough to easily accommodate our cells, and has an inductance of 200 mH.

The highest degaussing field used was ≈ 100 G, with the idea of avoiding exposure

of the cell to any field significantly higher than the maximum 30 G field at which

measurements would be made while still using a field high enough to assure thorough

degaussing. The low-field hysteresis measurements were made in a Helmholtz coil

using the 100 kHz pulse NMR spectrometer described in Chapter 2 and a coil on the

cell stem. Very small flip angles were used to avoid polarization destruction. Field

reversal was accomplished by reversing the cell orientation in zero field. All low-field

data were taken at 100 kHz, so the field was cycled along a hysteresis loop each time

data were acquired. The entire cycle took about one minute, so an insignificant

amount of time was spent away from the field of interest.

5.5 Results/Discussion

5.5.1 High-field Hysteresis

Measurements of an aluminosilicate (GE-180) spin-exchange cell (Fig. 5.3) show a

behavior which is qualitatively consistent with the field-dependence of T−11 through

µ given in Eq. (3.1). This figure shows a symmetric, closed loop with local minima

at the coercive fields of ≈ ±200 G, as expected. The initial measurement is at a

low, positive field with the cell degaussed (M ≈ 0 and H ≈ 0). As the external

field increases to +2000 G, the T−11 increases and gradually approaches saturation.

Hysteretic changes in T−11 consistent with ferromagnetism can be seen as the field is

decreased from +2000 G through zero to −2000 G and back up again to +2000 G.

Unfortunately, a quantitative assessment of the relationship between T1 and H is

not possible due to the unknown nature of the hysteresis of the magnetic inclusions.

We note that the relaxation rates measured at the coercive fields are lower than the

initial rate when the cell was degaussed. This may indicate that storing polarized

70

0.03

0.04

0.05

0.06

0.07

0.08

0.09

0.1

-2000 -1500 -1000 -500 0 500 1000 1500 2000

T1-1

(h

ou

rs-1

)

Field (Gauss)

Figure 5.3. T1 hysteresis loop of an aluminosilicate cell. The relaxation rate isplotted vs external field. The solid line with arrows was added to help guide theeye. The data were taken at the fields of interest with no field cycling. Notice thelocal minima at the coercive fields (≈ ±200 G). These results are consistent withthe model in Chapter 3.

gas in a magnetized cell at the coercive field may be better than storing it in a

degaussed cell at low field.

Measurements were also made on a bare Pyrex cell, as shown in Fig. 5.4. Polarized

gas was transferred into the cell from a high-pressure (≈ 8 atm) spin-exchange cell, as

discussed in Sec. 5.4. This cell behaves qualitatively similar to the aluminosilicate

cell (Fig. 5.3), with the T−11 approaching saturation at higher external fields and

local minima at coercive fields. Several transfers of gas were required to complete

the loop, and duplicate measurements at were made to check consistency before and

after the cell was refilled. We found that the wall rate was not completely consistent.

This is most evident for the −600 G measurements made on the return path from

−2000 G: the two points connected by the vertical line were made at the same field

and position on the loop but with different charges of gas. This offset in measured

71

0.06

0.08

0.10

0.12

0.14

0.16

0.18

0.20

-2000 -1000 0 1000 2000

T1-1

(h

ou

rs-1

)

Field (Gauss)

Figure 5.4. A T1 hysteresis loop of a bare (no Rb) Pyrex cell. The relaxation ratesare plotted vs. external field. The solid line with arrows was added to help guidethe eye. The data were taken at the fields of interest with no field cycling. Noticethe local minima at the coercive fields (≈ ±200 G). These data are qualitativelyconsistent with the aluminosilicate results and with the model in Chapter 3.

rates merely breaks symmetry in the plot but makes it difficult to state definitively

whether the data in Fig. 5.4 form a hysteretic loop. However, the bare Pyrex cell

shows evidence of the presence of ferromagnetism due to the local T−11 minima at

or near the coercive fields. This is apparently contrary to the results presented in

Chapter 3, where we showed that the relaxation rates of a bare cell measured at

30 G before and after exposure to a 1.0 T magnetic field did not change (see Fig.

3.3). Because both cells had about the same T1 when unmagnetized and measured

at low field, it is unlikely that T1 hysteresis was being masked by other relaxation

mechanisms in one cell and not the other. Further evidence is required for us to

draw any conclusions about the nature of bare-cell T1 hysteresis.

Similar experiments were conducted using three Pyrex spin-exchange cells (cells

containing Rb for optical pumping). The results show a very different behavior.

72

Figures 5.5, 5.6, and 5.7 show symmetric, closed loops of T−11 vs. external field, as

with the aluminosilicate and bare Pyrex cells. However, each cell shows a significant

field dependence that appears independent of the size of the magnetic moments:

as the magnitude of the field, thus the magnetization of the sites, decreases, T−11

increases dramatically until the external field passes through zero. Then the T−11

drops precipitously as the coercive field is approached, where each cell shows a local

T−11 minimum, as expected. This field-dependent effect is not accounted for in the

model, and is the only qualitative behavioral difference between different types of

glass that we have observed with 3He T−11 measurements. This is a very surprising

result, because a field dependence to the surface relaxation should only occur at fields

where the correlation time for the interaction is of the order of the Larmor period.

0

0.1

0.2

0.3

0.4

0.5

0.6

-2000 -1500 -1000 -500 0 500 1000 1500 2000

T1-1

(h

ou

rs-1

)

Field (Gauss)

Figure 5.5. A T1 hysteresis loop of Pyrex cell 9A. The relaxation rate is plotted vsthe external field. The data were taken at the fields of interest with no field cycling.The solid line with arrows is added to guide the eye. Local minima are presentat the coercive fields, as expected, but the strong field dependence of T−1

1 as theexternal field approaches zero from saturation is unexpected and unexplained.

73

0.0

0.2

0.4

0.6

0.8

1.0

-2000 -1500 -1000 -500 0 500 1000 1500 2000

T1-1

(h

ou

rs-1

)

Field (Gauss)

Figure 5.6. A T1 hysteresis loop of Pyrex cell 10A.

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

-2000 -1500 -1000 -500 0 500 1000 1500 2000

T1-1

(h

ou

rs-1

)

Field (Gauss)

Figure 5.7. A T1 hysteresis loop of Pyrex cell 18A. This particular cell was rinsedwith a 5% HF solution and a 37% HCl solution prior to Rb distillation, as discussedin Sec. 8.5.3.

74

Thus, we currently have no satisfactory explanation for the observed phenomenon

that is consistent with previous models.

We would expect the correlation time to be the collision time with the wall, which

should be much shorter than the Larmor period at any of our fields of interest. If

the field of a relaxation site is dipolar, then the general form of relaxation would

have the form [30]:

1

T1

∝ τ

1 + ω2τ 2, (5.1)

where ω is the Larmor frequency, which is proportional to the external field, and τ is

the correlation time. The relaxation models we develop and verify in Chapter 3 and

Chapter 6 assume ω2τ 2 ¿ 1, which is easily the case with 100 kHz Larmor frequency

(at ≈ 30 G) and a correlation time on the order of 10−12 s [17]. The dramatic field

dependence shown in Figs. 5.5, 5.6, and 5.7 implies that the same assumption is not

valid here. This is troubling, because pressure-dependence measurements supported

Eq. (3.1). A long correlation time would mean that the relaxation mechanism in

the presence of a magnetic site would be due to 3He diffusion through the field

gradient caused by that site. As given in Eq. (1.6), T−11 due to field gradients is

directly proportional to the diffusion coefficient which varies inversely with pressure.

Thus, the long correlation-time limit would result in opposite pressure dependence

than was observed in Chapter 3. Also troubling is the fact that the strange field

dependence was observed only in Rb-coated Pyrex, and not bare Pyrex, Rb-coated

quartz, or Rb-coated aluminosilicate.

One difference between Pyrex and aluminosilicate glasses is the permeability to

helium: Pyrex is about 104 to 105 times more permeable at room temperature [77].

If the 3He is able to permeate the glass and remain in the vicinity of a magnetic

75

site for much longer than a typical adsorption interaction (≈ 10−12 s [17]) then

a field dependence such as that observed would be explained. However, there

are several problems. First, temperature-dependence studies detailed in Chapter

7 suggest that the presence of Rb in the cells reduces permeability significantly

when compared to that of bare Pyrex. Second, the bare Pyrex cell did not show

the same field dependence as seen in the Rb-coated Pyrex. Third, quartz is more

permeable than Pyrex by an order of magnitude at room temperature [77]; thus we

expected to see the effect even more strongly than in Pyrex. However, preliminary

measurements strongly indicated that quartz spin-exchange cells do not show the

same field dependence as the Pyrex spin-exchange cells. Fourth, we can think of no

way of explaining how the 3He might remain in the vicinity of a relaxation site long

enough to cause the effect. Since the effect has thus far been observed only in Pyrex

cells containing Rb, then it is likely that a chemical interaction between the Pyrex

and Rb is somehow responsible. Clearly, this remains an open question warranting

further study.

5.5.2 Low-field Hysteresis

Other researchers have reported seeing changes in T−11 with variations of cell

orientation in low field (tens of Gauss) for unmagnetized or degaussed cells [73, 78].

We measured the relaxation rates for two cells at various fields in the range of +30 G

to −30 G (see Fig. 5.8). Cell 20B was never exposed to a high magnetic field, so its

T1 hysteresis properties are not known. Cell 14A was used as a control, since it is

not affected by T1 hysteresis to a significant degree (<10%, see Fig. 4.6). Both cells

had the same size, shape, and gas pressure. An incomplete loop was acquired for

each cell: cell 20B was measured at various fields from +30 G to −30 G and back

up to −5 G, and cell 14A was measured at +30 G, +5 G, −5 G, and −30 G. The

76

0.01

0.1

-30 -20 -10 0 10 20 30

14A

20B

T1-1

(h

ou

rs-1

)

Field (Gauss)

Figure 5.8. Low-field T1 hysteresis. A plot of T−11 vs applied magnetic field for

two cells. Cell 20B has never been exposed to a large magnetic field, and cell 14Bwas degaussed. The data were corrected for bulk He–He relaxation. Field cyclingwas employed with all data taken at +30 G. Cell 20B was measured from +30 Gto −30 G and back up to −5 G; multiple data points were acquired at +5 G toassure consistency in measurements due to intervening polarization of the gas. Cell20B shows strong field dependence while cell 14A shows almost no field dependence.Cell 14A was used as a control because it was known to not exhibit significant T1

hysteresis effects.

large increase in T−11 seen in cell 20B cannot be attributed to gradient relaxation,

because cell 14B did not show similar behavior. If gradient relaxation dominated at

±5 G, then both cells would be expected to have similar T−11 ’s at those fields.

There are three characteristics of Fig. 5.8 that we wish to address. First, there

is a significant difference between the measured rates at +30 G and −30 G for cell

20B. Second, cell 14B does not show the same strong field dependence that 20B

shows. Third, the low-field data lack evidence of a coercive field.

First, we seek an explanation for why cell 20B had such different measured rates

at +30 G and −30 G. If the cell was thoroughly degaussed and if no measurable

77

hysteresis was present, then there should be no difference between the magnitude

of M at ±30 G, so the measured rates should be the same. However, the rate

increase from −30 G and +30 G was about a factor of 2, which implies that the

cell was not completely degaussed. Reasons for this are not clear. Either we did

not do a thorough job of degaussing (although cell 20B was never exposed to a

static field above 30 G) or there are magnetic sites that have some slight permanent

magnetization. We have looked at other unmagnetized and degaussed cells at ±30 G

with mixed results. Some have shown a difference between the two fields and others

have not. This wide variation between different cells makes it difficult to draw any

conclusions.

Second, the large increase in rate seen in cell 20B (Fig. 5.8) as the magnitude of

the field was lowered toward 0 G is similar to the rate increases seen in the high-field

measurements of Rb-coated Pyrex (see Figs. 5.5, 5.6, and 5.7). In contrast, cell

14B showed very little change. Since cell 14B showed only slight T1 hysteresis, we

hypothesize that cell 14B simply lacks a sufficient number of magnetic sites to make

the effect measurable.

Third, it is clear that there is a field-dependent relaxation mechanism at low fields

in some cells, although the results of low-field hysteresis are qualitatively different

from the high-field results. The most significant difference is the lack of local T−11

minima at coercive fields. This may be due to two things: the corresponding

hysteresis loop does not cross the H axis, or the lack of any hysteresis. First, we

assumed that at low fields the magnetization in the cells would have hysteresis loops

centered about the origin of the M vs. H plot. However, it is possible that even a

degaussed cell has some degree of remanent magnetization, the origins of which are

not clear. This would place the hysteresis loop above, or below, the horizontal (H)

axis so that the loop would never cross the axis for a sufficiently low range of fields.

78

Second, M may be a nearly single-valued function of H between ± 30 G, meaning

that the hysteresis loop may be very narrow or even linear, resulting in virtually

no change in T−11 at fields of a given magnitude. Because T1 measurements of cell

20B showed only a field dependence with a lack of hysteresis, both possibilities may

hold.

We note that if the M vs. H dependence for degaussed cells in the range of ±30 G

is linear or nearly linear, then we should see a T1 field dependence that is parabolic,

according to the relationship T−11 ∝ µ2 given in Eq. (3.1). The very strong field

dependence observed is opposite this, implying that the actual field dependence is

much stronger.

5.6 Conclusion

We observed a very strong field dependence of T1 values measured in Rb-coated

Pyrex cells at fields between ±2000 G that is independent of the size of the moments

of the magnetic sites. This dependence is not explained by the model developed in

Chapter 3. We did, however, observe field dependence in Rb-coated aluminosilicate

and bare Pyrex that was qualitatively consistent with the model. This is the first

qualitative difference between different types of glass that we have observed using

T1 measurements of 3He relaxation. We also observed field-dependent effects in

Rb-coated Pyrex at low fields similar to those observed at high fields, with the

exception that no coercive fields or definite hysteresis were observed. The cause of

the unexpected field dependence in the Rb-coated Pyrex is indicative of very long

interaction times between the 3He atoms and the relaxation sites. These times would

have to be many orders of magnitude longer than the ballistic collision time. The

cause of such long interaction times is currently a mystery.

CHAPTER 6

FUNDAMENTAL MECHANISMS OF3He RELAXATION ON GLASS

6.1 Preface

This chapter is a manuscript submitted to Chemical Physics Letters in September

2002. This paper presents the first experimentally verified model of 3He relaxation

in bare (containing no Rb) glass, an important step in understanding relaxation

mechanisms in alkali-coated glass. Two significant points make this paper strong.

First, the theoretical prediction of T−11 temperature dependence is confirmed inde-

pendently by two different research groups using cells of different surface to volume

ratio, different 3He pressures, and measurement fields of different strength. These

aspects support the universality of the results for all bare Pyrex cells. Second, we

show that physical properties of the glass alone determine the relaxation rate of the

gas. Follow-up work has shown that the theory is general enough to extend to other

types of glass, and, more importantly, that the presence of an alkali metal in the

cells changes the fundamental relaxation mechanisms, a fact that was not previously

appreciated (see Chapter 7).

My coauthors are B. Driehuys, of Amersham Health, and B. Saam, my advisor.

Much of the theory in this paper was based on derivations by B. Driehuys, with my

refinements and explanatory text. I also compiled the numerical predictions and

analyzed the data.

80

6.2 Abstract

We present a model of 3He relaxation on the surface of borosilicate glass which

accurately predicts observed relaxation rates and their temperature dependence.

Above room temperature 3He dissolves into Pyrex, where interactions with Fe3+ ions

result in a relaxation time of ≈ 1 ms. Gas exchange across the glass surface of an

enclosed vessel leads to T−11 = A/V × (3.9± 1.4)× 10−2 cm/h at room temperature,

where A/V is the surface-to-volume ratio. The activation energy for relaxation is

13.7 ± 0.7 kJ/mol and is dominated by the activation energy of 3He diffusion in

glass. This is the first successful confirmation of predicted 3He relaxation rates in

glass vessels.

6.3 Introduction

Spin-exchange optical pumping (SEOP) [1] and metastability-exchange optical

pumping (MEOP) [2] are common methods of producing very high, nonequilibrium

nuclear polarization in certain noble gas nuclei. The gas is typically polarized

and/or stored in glass vessels, or cells. Workers in the field have long attempted to

determine a quantitative and predictive model of 3He surface relaxation on glass.

Since 3He surface relaxation has proven to be a very complex problem, understanding

even a single model system would be critical progress. The ultimate goal is a

better understanding of 3He relaxation in spin-exchange cells (cells containing an

alkali metal), where magnetic inclusions in the glass can dominate relaxation [45].

Researchers who use bare glass cells of all types, and Pyrex in particular, as storage

cells for polarized gas research may find these results especially pertinent.

Previous measurements of 3He relaxation as a function of temperature on glass

surfaces have been made in bare (containing no Rb or surface coatings), sealed

Pyrex, aluminosilicate, and quartz cells [19, 76]. For Pyrex, Fitzsimmons et al.

81

provided significant insight into 3He relaxation mechanisms by showing that adsorp-

tion dominates relaxation below about 130 K and absorption dominates at higher

temperatures. They derived and verified a model for adsorption-based relaxation.

However, a quantitative understanding of the absorption regime, which is relevant

for most practical experiments, has eluded researchers. In this paper we provide

a theory valid for all bare Pyrex cells which accurately predicts the measured

rates for dissolution-dominated relaxation. We show that such relaxation can be

characterized by an Arrhenius relation with a relaxivity %0 and the appropriate

activation energy EA:

1

T1

=A

V%0 exp

(− EA

R T

), (6.1)

where A/V is the surface to volume ratio.

6.4 Theory

6.4.1 T > Room Temperature

Our model for relaxation of polarized 3He is based on the solubility, diffusivity,

and intrinsic relaxation of 3He in the glass. We assume that all relaxation is due to

interactions of 3He with paramagnetic impurities in the glass, and that the number

of 3He atoms in the gas is much greater than the number of dissolved atoms. The

net flow of magnetization is from the gas to the glass in the −ξ direction, while

ξ = 0 represents the glass-gas interface. In the limit of weakly-relaxing walls [79],

the polarization may be assumed uniform in the gas and continuous across the

glass–gas interface.

The diffusion equation in the glass is

82

∂

∂tM(ξ) = Db(T )

∂2

∂ξ2M(ξ) + Q(ξ), (6.2)

where Db(T ) is the temperature-dependent diffusion coefficient of the helium in

the bulk glass, M(ξ) is the 3He magnetization, and Q(ξ) is a source term. The

magnetization loss is

∂

∂tM(ξ) = − 1

T1

M(ξ), (6.3)

where T1 is the measured relaxation time. The source term represents the magneti-

zation destroyed while in the dissolved phase:

Q(ξ) = − 1

T1 b(T )M(ξ), (6.4)

where T1 b(T ) is the temperature-dependent relaxation time of the dissolved gas.

From the above assumptions, T1 À T1 b. Equation (6.2) becomes:

Db(T )∂2

∂ξ2M(ξ) − 1

T1 b(T )M(ξ) ≈ 0. (6.5)

The general solution to Eq. (6.5) is

M(ξ) = S(T ) M0 exp(ξ/λ), (6.6)

83

where M0 is the gas-phase magnetization, S(T ) is the Ostwald solubility (Suckow [80]

refers to the Bunsen solubility1, which we will use in later calculations), and λ =√

Db(T ) T1 b(T ) represents a characteristic penetration depth of magnetization in

the glass. We note that solubility is usually calculated from measurements of

permeability K and diffusivity, since S = K/D.

The observed rate I at which total magnetic moment leaves the gas phase and

enters the dissolved phase is

I = −M0 V

T1

, (6.7)

where V is the cell volume. This rate is also the flux of total magnetic moment at

the interface multiplied by the interface area A:

I = −A Db∂

∂ξM(ξ) = −A

√√√√ Db(T )

T1 b(T )M0 S(T ). (6.8)

Equating (6.7) and (6.8) gives a prediction for the relaxation rate of polarized 3He in

a bare glass cell entirely in terms of the cell geometry and the bulk glass properties:

1

T1

=AS(T )

V

√√√√ Db(T )

T1 b(T ). (6.9)

A similar equation was partially derived by Deaton et al. in their study of 3He

relaxation on polymer surfaces [81].

1The Ostwald solubility is the volume of gas dissolved in a unit volume of a liquid at a specifiedtemperature and pressure. Bunsen solubility is the Ostwald solubility measured at STP.

84

The general form of relaxation due to dipolar interactions in the bulk [30] is

1

T1 b

=6

15

Mr6

τc

1 + ω2τ 2c

, (6.10)

where r is the separation, τc is the correlation time of the interaction, ω is the

3He Larmor frequency, and M = γ2gas γ2

e h2 S(S+1), where γgas and γe are the

gyromagnetic ratios for 3He and electrons, respectively. Mazitov et al. showed that

3He relaxation in bulk borosilicate glass depends most strongly on interactions with

Fe3+ ions (spin S = 5/2), which have a correlation time τFe for electron spin flips

of approximately 8× 10−9 s at room temperature [18]. To consider the effect of all

Fe3+ ions on a 3He nucleus, the expression in Eq. (6.10) must be integrated from

the distance of closest approach a through all space:

1

T1 b

=∫ ∞

a

6

15

Mr6

τc

1 + ω2τ 2c

N 4πr2 dr, (6.11)

where N is the density of Fe3+ ions in the glass. Since τc ≈ τFe, Eq. (6.11) becomes:

1

T1 b

=24π

45

N Ma3

τFe

1 + ω2 τ 2Fe

. (6.12)

Our measurements are made at low fields (see Sec. 6.5). Since ω2 τ 2Fe ¿ 1 we can

simplify Eq. (6.12) to:

1

T1 b

≈ 24π

45

N Ma3

τFe. (6.13)

85

Combining Eqs. (6.9) and (6.13) gives:

1

T1

=A

V

√24π

45

N Ma3

S(T )√

τFe(T ) D(T ). (6.14)

The temperature dependence of S(T ), Db(T ), and τFe(T ) can be characterized by

Arrhenius relations [18, 80]:

S(T ) = S0 exp(− ES

R T

), (6.15)

Db(T ) = D0 exp(−ED

R T

), (6.16)

τFe(T ) = τ0 exp(−EFe

R T

), (6.17)

where ES, ED, and EFe are molar activation energies for solubility, diffusion, and

Fe3+ electron spin flips, respectively, R is the universal gas constant, and T is the

absolute temperature. The subscript 0 indicates an asymptotic (T →∞) value. By

substituting Eqs. (6.15), (6.16), and (6.17) into Eq. (6.14) we have:

1

T1

=A

V

√24 π

45

N Ma3

S0

√τ0 D0 exp

(− EA

R T

)=

A

V%0 exp

(− EA

R T

), (6.18)

where we have defined %0 as the relaxivity and the total activation energy as

EA = ES +1

2(EFe + ED). (6.19)

86

We are primarily interested in relaxation in Pyrex, a borosilicate glass made by

Corning, because it is commonly used for spin-exchange cells. To calculate the

relaxation rate predicted by (6.18) for Pyrex we use the values obtained from bulk

glass measurements shown in Table 6.1. τ0 was calculated using Eq. (6.17) and

values for EFe and τFe. Based on [18] and a discussion with Mazitov, we assume

a 10% uncertainty for τ0 and N . As Shelby [82] points out, there is generally

poor agreement in reported activation energies for permeation of He in Pyrex,

ranging from 21.8 kJ/mol to 31.4 kJ/mol, and similar discrepancies exist for diffusion

measurements. Bulk-glass activation energies reported in [80] were used in Table 6.1

rather than measurements by several other workers (see, for example, [77, 82, 83])

because the former measurements were made on Duran, a borosilicate glass made

Table 6.1.Important values for Eq. (6.18) for Pyrex glass.

Variable Value Uncertainty Units Ref

D0 7.0× 10−4 0.6× 10−4 cm2/s [80]

ED 27.8 0.5 kJ/mol [80]

S0 6.3× 10−3 0.6× 10−3 cm3 STP/cm3 [80]

ES 1.5 0.6 kJ/mol [80]

τFe(295K) 0.77× 10−8 s [18]

τ0 1.9× 10−9 10% s

EFe −3.4 0.3 kJ/mol [18]

N 8× 1018 10% cm−3 [18]

a 5 A [18]

87

by Schott which is very similar to Pyrex, and because an uncertainty was given

with each of the measured values. From Eq. (6.13) we estimate the relaxation time

in bulk glass as T1 b ≈ 1 ms, therefore the magnetization penetration depth λ is

≈ 30 nm.

Inserting relevant values into Eq. (6.18) gives the relaxivity:

%0 = (10 ± 2) cm/h. (6.20)

Equation (6.19) gives the activation energy:

EA = 13.7± 0.7 kJ/mol, (6.21)

which is dominated by the activation energy of 3He diffusion in glass. We then

calculate an expected room temperature relaxation rate:

1

T1

=[A

V(3.9 ± 1.4)× 10−2

]h−1. (6.22)

Equation (6.22) predicts that bare Pyrex is of marginal utility for polarized 3He

storage, a fact that has been verified by several investigators [17, 19, 20]. As

discussed in Sec. 6.6.1, this situation changes drastically for cells containing alkali

metals.

6.4.2 T < Room Temperature

At lower temperatures the 3He lacks sufficient kinetic energy to overcome the

potential barrier for dissolution. For example, 13.7 kJ/mol of kinetic energy is

88

required for the 3He to overcome the potential barrier of dissolution relaxation,

whereas only 1.7 kJ/mol is available at 200 K. Relaxation mechanisms with negative

activation energies, such as adsorption to the cell wall, will begin to dominate the

measured T−11 as the temperature decreases. Although typical sticking times are only

≈ 10−13 s at room temperature [17], there is no potential barrier to overcome, since

the interaction is slightly attractive. Fitzsimmons et al. [19] derive an expression

for relaxation in a cell where adsorption dominates:

T1 =N

n(tad + Tad), (6.23)

where N is the total number of gas atoms in the cell, n is the total number of

gas atoms adsorbed to the surface at any instant, tad is the average adsorbed-atom

sticking time, and Tad is the relaxation time of an adsorbed atom. The number of

adsorbed atoms is assumed much less than N , and is given by n = N v A tad/(4 V ),

where v is the mean thermal velocity of the 3He atoms. They show that Tad À tad

and that Tad = tad/2 W , where W is the probability of an adsorbed atom relaxing.

W is proportional to t2ad, and tad follows an Arrhenius relation with activation energy

Ead. Thus Eq. (6.23) becomes:

1

T1 ad

=A

Vκ0 exp

(−2Ead

R T

) √T , (6.24)

where κ0 = W√

2 R/NA mπ, the 3He mass is m and NA is Avogadro’s number.

Muller [84] reports an activation energy of He adsorption on glass of

Ead = −0.96 ± 0.19 kJ/mol. (6.25)

89

This was independently confirmed by Fitzsimmons et al. [19] by observing nuclear

spin-relaxation of polarized 3He in low-pressure, sealed cells at various temperatures

below room temperature.

By assuming that a 3He atom will only relax in a collision with a Fe3+ ion at the

surface, we can approximate W and find κ0. The Fe3+ concentration is about one

part in 104 by volume [18], and Timsit et al. estimate that an average of 106 collisions

are required to relax a 3He atom [17]. Therefore, κ0 ≈ 3× 10−2 cm h−1 K−1/2.

6.5 Experimental

All measurements were made on spherical, valved, bare Pyrex cells. Cells pre-

pared at Utah were ≈ 50 cm3 and contained ≈ 4 amagats of 3He, and the cell

prepared at Amersham Health (AH) was ≈ 180 cm3 and contained ≈ 1 amagat of

3He. The Utah cells were prepared by baking under vacuum for ≈ 48 hours at up

to 400C. (Procedures used at Utah for cell fabrication and a detailed description

of the cells can be found in Chapter 4.) Polarized gas was transferred into an

evacuated cell from a similar, higher-pressure spin-exchange cell by connecting the

cells, opening the valves, and allowing the pressure to equilibrate. The gas could

not be polarized in the cells directly because bare cells, as we define them, do

not contain Rb. All Utah T−11 measurements were made at ≈ 30 G using 100

kHz pulse NMR (see Chapter 2) and very small flip angles to excite only a small

fraction of the gas. The AH measurements were made at ≈ 7 G (24 kHz). The

initial heights of the free-induction decays acquired at appropriate time intervals

were fit to a single exponential to extract T−11 . Several T−1

1 measurements could

be made on a single charge of gas with intervening changes in temperature. The

above-room-temperature measurements were done in a forced-air oven typically used

for SEOP. The temperature was maintained to a few tenths of a degree by a resistive

90

temperature detector and controller. The measurements below room temperature

were done in an insulated cylinder connected to a liquid nitrogen dewar. The desired

temperature was reached by boiling off the liquid nitrogen at a specific rate with

submerged heating tape powered by a variable AC transformer. The temperature

was monitored with a thermocouple and maintained to within a few degrees. In all

cases the cell valve was kept at room temperature to prevent o-ring failure.

6.6 Results and Discussion

6.6.1 T > Room Temperature

Relaxation rates of three bare Pyrex cells, labelled 19A, 19B, and PXX05, were

measured at various temperatures between 298 K and 473 K (see Fig. 6.1). By

comparing the average curve fit results of the three cells to Eq. (6.18) we find that

EA = 14.7 ± 0.3 kJ/mol and %0 = (36 ± 4) cm/h (represented by the solid line

in Fig. 6.1). This equates to a room-temperature relaxation rate of

1

T1

=[A

V× (9.6 ± 1.6)× 10−2

]h−1. (6.26)

Our results are in excellent agreement with the predicted value of EA = (13.7

± 0.7) kJ/mol and in good agreement with %0 = (10 ± 2) cm/h (represented in

Fig. 6.1 by the dashed line), providing strong evidence that the model, Eq. (6.18),

accounts for the majority of relaxation in this temperature range. The discrepancy

in the intercepts in Fig. 6.1 is directly related to the discrepancy in %0. We note,

however, that the experimental value of %0 was obtained by assuming a perfectly

smooth cells surface (minimum value of A/V ). It is not difficult to imagine that the

actual A/V is larger by a factor of two or more, which would bring the measured

91

0.01

0.1

1

2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4

19A

19B

PXX05

theory

V/A

T1-1

(cm

ho

urs

-1)

1000/T (K-1)

Figure 6.1. Temperature dependent relaxation rates for three bare (no Rb) Pyrexcells above room temperature. Equation (6.18) is also plotted using (6.20) and(6.21). Cells 19A and 19B were prepared and measured by the group at Utah,and cell PXX05 was prepared and measured at Amersham Health. Cells 19Aand 19B were ≈ 50 cm3 and ≈ 4 amagats, and cell PXX05 was ≈ 180 cm3 and≈ 1 amagat. Rates were measured at temperatures between 298 K and 473 K.Bulk He–He relaxation was subtracted from the data. The data were fit to anArrhenius relation with % and EA as free parameters. The resulting activationenergy is EA = 14.7 ± 0.3 kJ/mol. Differences in the slopes represent a differencein EA, whereas differences in the intercept could be due to an underestimation inA/V , since we assumed a smooth, spherical surface.

%0 into much closer agreement with theory. In addition, slight differences in glass

composition [85] or thermal history [86] could lead to variations in N or D0 beyond

the quoted errors we assumed. Differences in the relaxivity might lead to nonlinear

behavior in Fig. 6.1 or to a slope different from that predicted. The theory accounts

only for absorption relaxation, thus Eq. (6.22) represents a lower limit for T−11 at

room temperature.

We have also studied the temperature-dependent relaxation of 3He in several

Pyrex cells containing Rb metal. In stark contrast to bare cells, we found that

92

diffusion-based relaxation is absent but that an adsorption-based relaxation model

seems much more appropriate for the longest-lifetime cells. This suggests that the

Rb strongly inhibits dissolution in spin-exchange cells leaving other relatively weak

mechanisms to dominate relaxation.

6.6.2 T < Room Temperature

To investigate adsorption-based relaxation, we measured 3He relaxation in a

temperature range of 95 K to 175 K. Figure 6.2 shows T−11 vs. 1000/T for two

of the cells discussed in Section 6.6.1. Results of a global fit of the data to Eq.

(6.24) give Ead = −0.63± 0.03 kJ/mol and κ0 = (1.6± 0.1)× 10−3 cm h−1 K(−1/2).

0.01

0.1

1

10

5 6 7 8 9 10 11

19A rate

19B rate

theory

T1-1

(h

ou

rs-1

)

1000/T (K-1)

Figure 6.2. Relaxation rate vs 1000/T for two bare Pyrex cells at ≈4 amagats.These are two of the same cells shown in Fig. 6.1 but with a different charge of gas.The dashed line is a plot of Eq. (6.24) using (6.25) and κ0. Temperatures rangedfrom about 95 K to 175 K. The data, with bulk He–He relaxation subtracted,were fit to Eq. (6.24). The results from a global fit give an activation energyof Ead = −0.63 ± 0.03 kJ/mol, somewhat weaker than reported in [19, 84] of−0.96± 0.19 kJ/mol.

93

There is good agreement between our value of κ0 and the predicted value of 3× 10−2

cm h−1 K(−1/2), although our value is somewhat lower. This could be a result of

fewer Fe3+ ions on the glass surface than anticipated, and suggests the possibility

of using polarized 3He to measure Fe3+ ion surface concentration in various types of

glass. The activation energy we found was somewhat weaker (closer to zero) than

that reported by Fitzsimmons et al. [19] and Muller [84] of −0.96 ± 0.19 kJ/mol.

We note that Fitzsimmons et al. found a local minimum in T−11 of 3He relaxation

in bare Pyrex at about 120 K, reflecting the transition between adsorption- and

absorption-dominated relaxation. We clearly observed adsorption behavior up to

about 170 K and a local minimum at about 200 K. We have no direct explanation

for the discrepancy, but we point out that the relevant data in Ref. [19] carry large

error bars, particularly near the T−11 minimum; in addition the curves used to fit

their data are at least somewhat speculative.

6.7 Conclusion

This work represents the first successful quantitative verification of predicted

3He relaxation phenomena in bare Pyrex glass. We conclude that we have identified

the correct relaxation mechanism for bare Pyrex, that our theoretical calculation

represents a lower bound on T−11 , and that experimental values will be larger as A/V

departs from an ideally smooth surface. The relaxation is dominated by interactions

of dissolved 3He with Fe3+ ions in the glass. We have experimentally verified the

predicted activation energy of dissolution-based relaxation, which depends on the

activation energies of 3He solution, diffusion, and Fe3+ electron spin flips. By

comparing these results to the results of similar studies of vessels containing Rb,

we will gain further insight to the relaxation mechanisms in spin-exchange cells.

94

This will lead to more consistent production of quality vessels and more efficient use

of the spin-polarized gas.

We acknowledge helpful discussions with R. K. Mazitov, M. Conradi, and W.

Happer, and we are grateful for the expert glassblowing of J. Kyle. This work was

funded in part by Amersham Health.

CHAPTER 7

3He RELAXATION IN BARE AND

Rb-COATED GLASS

7.1 Abstract

By studying the temperature dependence of 3He relaxation in various cells, we

show that 3He relaxes in cells containing an alkali metal for spin-exchange optical

pumping by a different mechanism than in bare vessels, containing no alkali metal.

Dissolution into the glass and adsorption to the glass provide dominant relaxation

mechanisms in bare Pyrex vessels, but dissolution-dominated relaxation is virtually

nonexistent in the presence of Rb. Thus, lower than expected polarizations (50% vs.

80%) cannot be attributed to increased wall relaxation rates at high temperatures

required for spin-exchange. In addition, we observe that adsorption relaxation is

dominate in bare aluminosilicate cells at our temperatures of interest. By contrast,

adsorption relaxation is dominate in bare quartz vessels only at temperatures above

about 120 K.

7.2 Introduction

3He wall relaxation times in spin-exchange cells can vary from 10s of minutes to

100s of hours, and very little is understood about the wall relaxation mechanisms.

By studying the temperature dependence of relaxation rates in bare glass (no alkali

metal or surface coatings) we hope to better understand wall relaxation mechanisms

96

and the role that alkali metals play in inhibiting relaxation. Bare Pyrex has already

been investigated (see Chapter 6), so in this chapter we show results of similar

studies on bare aluminosilicate, bare quartz, and Rb-coated Pyrex. The ultimate

goal is to apply an understanding of the basic physics of 3He wall relaxation to the

consistent fabrication of long-lifetime spin-exchange cells.

The 3He polarization during optical pumping is given by Eq. (1.1). We estimate

that 〈PA〉 can be maintained at near 100% under our SEOP conditions [14]. Using

a measured spin-exchange rate at 180C [26] and the saturated vapor pressure curve

for Rb [13], the maximum attainable 3He polarization in a typical 40 h cell is about

80%. There are no reports in the literature of 3He polarizations achieved with

SEOP over about 50%. This polarization deficit remains unexplained at present,

but it has been speculated that it may be due to increased wall relaxation times at

spin-exchange temperatures. We show in Chapter 6 that relaxation rates in bare

Pyrex do, in fact, increase exponentially with increasing temperature. In this paper

we show that the same phenomenon does not occur in spin-exchange cells and cannot

account for the polarization deficit.

7.3 Theory

7.3.1 Aluminosilicate Glass

In the quest to find a suitable glass for the consistent and reliable production of

spin-exchange cells, much attention has been paid to aluminosilicate glass because

of its low permeability. Aluminosilicate glass is several orders of magnitude less

permeable to He than Pyrex [77] but suffers from lower availability and workability.

In spite of the low permeability, only limited success in making long-lifetime cells has

been realized with little evidence that it can be used to consistently produce spin-

exchange cells with lower relaxation rates than Pyrex. Aluminosilicate is greatly

97

advantageous in some applications that require low-boron, low-permeability glass

[87].

We can apply the relaxation model derived for bare glass in Chapter 6 to bare

aluminosilicate cells. The very high activation energy of diffusion (see Table 7.1)

should cause the room-temperature relaxation rate to be quite small. Applying

permeation and diffusion measurements made on Supremax glass (an aluminosilicate

glass) shown in Table 7.1 to Eq. (6.19), we calculate an activation energy EA =

29.6 ± 1.1 kJ/mol. Assuming an iron content similar to Pyrex [18, 88] and a cell

with A/V = 1, we approximate a room-temperature relaxation rate of 5× 10−5 h−1,

about 103 times smaller than bare Pyrex. Such relaxation rates have not been

observed in aluminosilicate cells, although the rates are generally somewhat lower

than bare Pyrex. Gentile et al. reported a relaxation time in a bare, 40 cm3, sealed

GE-180 cell of 6 h and in 1720 (Corning) as high as 72 h [44]. Other workers have

Table 7.1.Important values for Eqn. (6.18) for aluminosilicate glass.

Variable Value Uncertainty Units Ref

D0 1.8× 10−3 0.3× 10−3 cm2/s [80]

ED 56.9 1.0 kJ/mol [80]

S0 3.1× 10−3 0.5× 10−3 cm3 STP/cm3 [80]

ES 2.8 1.0 kJ/mol [80]

τFe(295K) 0.77× 10−8 s [18]

τ0 1.9× 10−9 10% s

EFe −3.4 0.3 kJ/mol [18]

98

reported times in 1720 ranging from about 6 h to about 30 h [17, 19, 76]. It is likely

that another, undetermined relaxation mechanism masks any absorption relaxation

that occurs at room temperature, thus causing the lower than expected relaxation

times. However, it has been shown that GE-180 cells that contain Rb can have very

long lifetimes: a 0.85 atm cell had a measured lifetime of 840 h [87].

The activation energy of adsorption of He on alumonosilicate should be similar

to that of Pyrex, as the principal constituent of both glasses is SiO2. Because of

the much higher EA for aluminosilicate, we anticipate that the transition tempera-

ture between absorption and adsorption relaxation, reflected as a minimum of the

relaxation rate in a plot of T−11 vs. temperature, should occur at a somewhat higher

temperature than the ≈ 200 K for Pyrex. By adding Eqs. (6.18) and (6.24) we have

a general equation for relaxation at all temperatures:

1

T1

(T ) =A

V%0 exp

(− EA

R T

)+ κ0 exp

(−2Ead

R T

)√T . (7.1)

By setting the first derivative of Eq. (7.1) with respect to T equal to zero then solving

for T , we can predict the transition temperature between absorption and adsorption

relaxation. By using the same value of %0 for aluminosilicate that was calculated

for Pyrex, since S0, D0 and N are approximately the same (GE-180 contains 0.02%

iron oxide by weight [88], which is approximately the same concentration as Pyrex,

if not slightly lower [18]), and using the measured values of Ead and κ0 in Pyrex,

we estimate that the transition temperature for aluminosilicate to be ≈ 300 K.

This temperature may be too high for us to clearly see absorption relaxation in our

temperature range, but adsorption relaxation should be easily observed.

99

7.3.2 Quartz Glass

Quartz glass is about an order of magnitude more permeable to He than Pyrex

at room temperature [83], but the iron content is far lower. To predict relaxation in

quartz, we can assume two limits: no contribution to relaxation from iron (iron-free

quartz) and iron-dominated relaxation (enough iron in the glass to dominate relax-

ation). In the zero-iron-content limit, the dominant bulk-glass relaxation mechanism

T1 b will be from interactions with unpaired electrons from broken SiO2 bonds, which

have a spin-flip correlation time of about 1.6× 10−7 s [18]. The correlation time

for diffusion is τD = r2/6D(T ) ≈ 4× 10−9 s at room temperature, where r is the

average jump distance of≈ 2.5 A [18] and D(T ) was measured to be about 2.2× 10−8

cm2/s at room temperature [89]. The 3He relaxation would be dominated by the

mechanism with the shorter correlation time: 3He diffusion in the glass. Using Eq.

(6.9), it can be shown that T−11 would depend only on the temperature-dependent

3He solubility S(T ). The activation energy of solution in quartz is negative (see

Table 7.2), implying that the relaxation rate of 3He in quartz would decrease with

increasing temperature for all temperatures. In the second limit, the correlation time

for Fe3+ electron spin flips would dominate relaxation and the expression derived for

Pyrex glass would be applicable. Using values shown in Table 7.2, we calculate the

activation energy EA = 7.2 ± 0.4 kJ/mol and a room-temperature relaxation rate

of approximately 3× 10−2 h−1 for a cell with A/V = 1. The resulting temperature

dependence for absorption relaxation would be qualitatively similar to that of Pyrex.

GE fused silica, the glass we used, has an iron content of 0.2 ppm by weight [90]. If,

due to the low, but nonzero, iron content, we are in a regime where both mechanisms

contribute to T1 b then we would expect to see an activation energy somewhere

between ES and EA.

Assuming that Ead will be similar for quartz and Pyrex and that the limit in

100

Table 7.2.Important values for Eqn. (6.18) for fused silica.

Variable Value Uncertainty Units Ref

D0 3.0× 10−4 cm2/s [89]

ED 23.4 0.2 kJ/mol [89]

S0 7.3× 10−3 cm3 STP/cm3 [89]

ES -2.9 0.3 kJ/mol [89]

τFe(295K) 0.77× 10−8 s [18]

τ0 1.9× 10−9 10% s

EFe −3.4 0.3 kJ/mol [18]

which iron dominates relaxation is correct, then we would expect to see a transition

between absorption and adsorption relaxation at a temperature somewhat lower

than the 200 K for Pyrex, due to the much lower EA for quartz. Since the iron

content of fused silica is ≈ 103 times smaller than for Pyrex, we estimate that %0

will be lower by a factor of about 30 and κ0 will be lower by a factor of about 103.

Using Eq. (7.1) we predict a transition temperature of ≈ 90 K, possibly too low for

us to see. If the broken-bond model is correct, then there may be no discernable

transition since both regimes would have small, negative activation energies. If there

are contributions to relaxation from both models, then a transition temperature

would be very difficult to predict, but would still be expected to be below that of

Pyrex due to the much lower EA.

101

7.3.3 Rb-coated Pyrex Glass

The effects of the presence of an alkali metal on 3He relaxation are not well

understood but are critical to wall relaxation times in spin-exchange cells. It has

been shown that the presence of an alkali greatly decreases the room temperature re-

laxation rate over bare cells of several types of glass [17, 19, 20, 45], and that a visible

coating of other pure metals can be very polarization friendly (i.e., nonrelaxive) [20].

Alkali metals are especially favorable because the loosely bound conduction electrons

prevent He, and other noble gases, from readily adsorbing [91]. Thus, alkali metals

have a very weak attractive potential for He adsorption [92]. More specifically, Heil

et al. show that the wall relaxation rate is proportional to the cube of the work

function of the metallic surface [6], thus alkali-metal surfaces, especially Rb and Cs

with relatively low work functions of 2.26 eV and 1.95 eV, respectively [75], result

in very long relaxation times. However, in spin-exchange cells it is not expected

that the alkali metal forms a perfect metallic surface in the entire cell. Therefore,

surface chemistry between the glass and alkali may also affect 3He relaxation by

altering the ability of the 3He to interact with iron ions. It is possible, for example,

that the alkali metal atoms block interstitial holes into which 3He could otherwise

diffuse [77], thereby inhibiting absorption relaxation. Unfortunately, relatively little

is known about the chemical effects of alkalis on Pyrex, other than that Corning,

the makers of Pyrex, warn consumers that hot alkalis will etch the surface [93].

7.4 Experimental

All of the cells tested were spherical, valved, and approximately 50 cm3. The

Pyrex spin-exchange cells and the quartz cells were fabricated by the University of

Utah glass blower, and the aluminosilicate cells were fabricated by the glass blower

at Princeton University. All cells were prepared and measured at Utah (see Chapter

102

4 for our cell fabrication and preparation techniques). No Rb was distilled into the

quartz and aluminosilicate cells, so they were prepared on a truncated manifold.

We note that the stem of the quartz cells consisted of a ≈ 2.5 cm length of quartz,

a graded seal, and a ≈ 2.5 cm length of Pyrex for the pull-off. The spin-exchange

cells contained ≈ 8 atm of 3He, and the bare cells contained ≈ 4 atm. Polarized gas

was introduced into the bare aluminosilicate and quartz cells using a gas-transfer

manifold (see Fig. 5.2) attached to our cell-filling system, which provided a nitrogen-

gas purge and vacuum. Gas was transferred by attaching an empty cell and a

spin-exchange cell, opening the cell valves, allowing the pressure to equilibrate, and

closing the valves. Several T1 measurements could be made on a single charge of gas.

All measurements were made at 30 G using the 100 kHz pulse NMR spectrometer

described in Chapter 2 and an RF coil on the cell stem (see Fig. 4.1). The initial

heights of the free-induction decays acquired at appropriate time intervals were fit

to a single exponential to extract T−11 . Very small flip angles were used to excite a

fraction of the gas in the stem for minimal global polarization destruction. Above-

room-temperature measurements were made in the oven normally used for heating

spin-exchange cells for SEOP, and the temperature was maintained to within a few

tenths of a degree by a resistive temperature detector and controller. Below-room-

temperature measurements were made in an insulated aluminum cylinder attached

to a liquid nitrogen dewar. The cell was cooled by boiling off the nitrogen with

heating tape submerged in the dewar controlled by a variable AC transformer. The

temperature was monitored using a thermocouple and maintained to within a few

degrees. During all measurements the cell valve was kept at room temperature to

prevent o-ring failure.

103

7.5 Results and Discussion

7.5.1 Aluminosilicate Glass

7.5.1.1 T > Room Remperature

Figure 7.1 shows a plot of T−11 vs. 1000/T for two bare aluminosilicate (GE-180)

cells for temperatures from room temperature up to about 460 K. The data do

not follow the form of Eq. (6.18), suggesting that diffusion through the glass may

not be a dominant mechanism of relaxation in this temperature range, consistent

with expectations. Perhaps these data represent the transitional regime between

adsorption- and absorption-based relaxation predicted in Sec. 7.3.1 to be around

300 K. In addition, the room temperature relaxation rate is comparable to bare

Pyrex, contrary to the prediction but consistent with the findings of other workers

0.10

0.15

0.20

0.25

2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4

Al1B

Al1A

T1-1

(h

ou

rs-1

)

1000/T (K-1)

Figure 7.1. Relaxation rate vs. 1000/T for two bare aluminosilicate (GE-180) cellsfor T ≤ 300 K. The cells were at ≈ 4 atm, and polarized gas was transferred in froma spin-exchange cell. The data show an apparent lack of dissolution-dominatedrelaxation, as expected for this very impermeable glass.

104

using various types of aluminosilicate glass [17, 19, 20, 44]. The cause of the much

lower than predicted relaxation rates in bare aluminosilicate is not understood.

7.5.1.2 T < Room Temperature

Figure 7.2 shows T−11 vs. 1000/T data for the same two cells shown in Fig.

7.1 for temperatures from about 100 K to room temperature. The solid line is a

best global fit of Eq. (6.24) to all the data and the dashed line represents the bare

Pyrex results from Chapter 6. The measured activation energy for adsorption on the

aluminsilicate is Ead = −0.69±0.05 kJ/mol. By comparison, the result for Pyrex was

Ead = −0.63 ± 0.03 kJ/mol. The excellent agreement between the aluminosilicate

10-2

10-1

100

3 4 5 6 7 8 9 10 11

Al1B

Al1A

Pyrex

T1-1

(h

ou

rs-1

)

1000/T (K-1)

Figure 7.2. Relaxation rate vs. 1000/T for two bare aluminosilicate (GE-180) cellsfor T ≥ 300 K. The cells were at ≈ 4 atm, and polarized gas was transferred fromspin-exchange cells. A global fit (solid line) of Eq. (6.24) gives Ead = −0.69± 0.05kJ/mol. For comparison, bare Pyrex data from Chapter 6 is plotted with a fit tothe same equation (dashed line). The similarity in the activation energy, reflectedby the slope of the lines, indicates that the relaxation mechanism in these cells issimilar to that in the bare Pyrex cells.

105

and Pyrex results strongly suggests that the same relaxation mechanisms dominate

in the different types of glass in this temperature regime. The difference in intercept

could be due to different concentrations of Fe3+ ions at the surface [although GE-180

has somewhat less iron than Pyrex (see Sec. 7.3.1)] or to differences in surface to

volume ratio. Little is known about microscopic surface smoothness of the different

types of reblown glass, although atomic force microscopy studies are currently being

done [78].

We note that the room-temperature T−11 measurements in Fig. 7.2 are about a

factor of two different from those in Fig. 7.1. The sets of data were taken with two

different charges of gas for each cell (polarized gas was transferred into the cells from

spin-exchange cells at room temperature), with the above-room-temperature data

taken first. The first charge of 3He was evacuated prior to the second introduction

of polarized gas. The dramatic drop in relaxation rate after the second charge

of gas is a mystery, but could be due to trace amounts of Rb atoms that were

transferred in with the polarized 3He (≈ 1018/cm3 at room temperature [13]). These

Rb atoms may have neutralized relaxive surface sites by reducing Fe3+ ions to Fe.

We showed in Chapter 3 that the 3He relaxation rates are certainly very sensitive

to the concentration of relaxation sites, and 3He may be sensitive enough to detect

such slight changes in Fe3+ surface concentrations.

7.5.2 Quartz

Figure 7.3 shows T−11 vs. 1000/T for two bare quartz cells for temperatures

ranging from about 100 K to 460 K. Neither cell shows relaxation characteristic

of the iron-free model. The data appear to support the iron relaxation model, as

the activation energy of relaxation is positive in most of the temperature range. For

these cells, the average activation energy is EA = 2.5±0.2 kJ/mol. This is somewhat

106

10-1

100

2 4 6 8 10 12

Q4A

Q4B

T1-1

(h

ou

rs-1

)

1000/T (K-1)

Figure 7.3. Relaxation rate vs. 1000/T for two bare quartz (GE fused silica) cells at≈ 4 atm. Polarized gas was transferred into these cells from spin-exchange cells. Thesolid lines represent fits of Eq. (6.18) to the data for which absorption dominates,above about 130 K. The data were taken in two segments: above and below roomtemperature. For reasons that are not clear, the behavior of Cell Q4B above roomtemperature is inconsistent, and these data were not included in the curve fit. Theaverage activation energy EA for both cells is 2.5 ± 0.2 kJ/mol. A minimum inT−1

1 is apparent at ≈ 120 K, where there is a transition between absorption andadsorption relaxation, close to the predicted value of ≈ 90 K.

lower than the predicted value of 7.2 kJ/mol in the iron-dominating limit. Perhaps

the low iron concentration is allowing the two regimes to compete. Figure 7.3 shows

a minimum in T−11 occurring at about 120 K, where there is a transition between

absorption and adsorption relaxation. This temperature is lower than we saw in

Pyrex, as anticipated, and is very close to the transition temperature of about 90 K

predicted in Sec. 7.3.2. It may be interesting to repeat this experiment with a very

high purity quartz, such as Suprasil, to see if the broken-bond model can be verified.

As an aside, we note that the room temperature relaxation rates for bare quartz

are rather short. We have seen relaxation times as long as about 30 hours in Rb-

107

coated quartz, and Heil et al. have seen similar times in a quartz cell containing Cs.

These relaxation times are not as long as those for most alkali-coated Pyrex cells,

indicating that the glass still plays an important relaxation role in alkali-coated cells.

Section 7.5.3 discusses relaxation in Rb-coated cells in greater detail.

7.5.3 Pyrex Glass

We measured the relaxation rates for several Rb-coated Pyrex cells in the temper-

ature range 393 K to 295 K (see Fig. 7.4). It is difficult to measure accurately T−11

at higher temperatures, because spin-exchange interactions with the alkali metal

will dominate 3He relaxation. Spin-exchange rates are difficult to calculate since

10-3

10-2

2.4 2.6 2.8 3.0 3.2 3.4

K1B

9B

15B

18A deg'd

1000/T (K-1)

T1-1

(ho

urs

-1)

Figure 7.4. Wall T−11 measurements of several spin-exchange cells for T > 295 K.

All of the cells contain Rb except for cell K1B, which contains potassium. He–Hedipole relaxation and spin-exchange relaxation rates have been subtracted from thedata. The cells with the smallest wall rates show some temperature dependence, butthis dependence becomes less apparent in shorter T1 cells. The cells do not displaybehavior modelled by Eq. (6.18), which was derived for bare Pyrex.

108

Rb number densities are not always understood and we currently have no method

of direct measurement. We approximated the Rb number density using the vapor

pressure curve presented by Killian [13]. As shown in Fig. 7.4, no single, consistent

relaxation mechanism seems to dominate spin-exchange cells. Cell K1B, which

contains potassium (all other cells contain Rb), has a strong increase in T−11 with de-

creasing temperature, an effect directly opposite to that predicted by the bare-Pyrex

dissolution model. Cell 9B has a weaker temperature dependence, but also opposite

to that of the bare-Pyrex model. Cell 18A has a slight temperature dependence

which is absent when the cell is magnetized (with a room temperature relaxation

rate of about 0.35 h−1). Cell 15B behaves differently from all the other cells with an

apparent local minimum in T−11 somewhere between about 350 K and 300 K, similar

to what is predicted for aluminosilicate glass (see Sec. 7.3.1). Cells 6A and 18A when

magnetized (not shown in Fig. 7.4) were also measured in the same temperature

range, and they showed no temperature dependence whatsoever; each had a room

temperature relaxation rate of about 0.35 h−1. These two cells and the bare cells

discussed in Chapter 6 had similar relaxation rates at ≈ 380 K but very different

behaviors with changing temperature: the bare cells’ T−11 ’s followed Eq. (6.18) while

6A and 18A (magnetized) had T−11 s independent of temperature. It appears from

Fig. 7.4 and cells 6A and 18A (magnetized) that the shortest-lifetime cells show

the least temperature dependence, suggesting that temperature dependence can be

masked by faster wall relaxation mechanisms.

The data in Fig. 7.4, when compared to Fig. 6.1, suggest that the alkali metal

strongly inhibits dissolution of 3He into the glass. It appears that adsorption-like

relaxation is especially significant in cell K1B whose T−11 increases dramatically as

the temperature decreases. To be consistent with the adsorption model in Eq. (6.24),

the activation energy of adsorption for cell K1B would have to be much lower than

109

that of bare Pyrex. A fit of the K1B data to Eq. (6.24) gives Ead ≈ −6.7 kJ/mol,

which is significantly more attractive than the −0.63 kJ/mol we measured for bare

Pyrex. Data for cell 9B gives Ead ≈ −1.3 kJ/mol. This is much weaker than Ead for

K1B, but still significantly stronger than for the bare Pyrex. Because He is so weakly

adsorptive to alkali metals, we hesitate to attribute wall relaxation in alkali-coated

cells to a mechanism that appears so strongly adsorptive and easily masked by faster

mechanisms. We have presented models for two potential relaxation mechanisms

that can be temperature-dependent through the gas diffusion coefficient D. It can

be shown that D ∝ T 3/2 for a dilute ideal gas [94]. The two models are T1 hysteresis

[see Eq. (3.1)] and magnetic field gradient relaxation [see Eq. (1.6)]. Magnetic field

gradients cause T1 ∝ 1/D, which results in behavior opposite to what is observed

in Rb-coated cells. However, the T1 hysteresis model predicts relaxation due to

magnetic sites with T1 ∝ D, which gives a temperature dependence qualitatively

consistent to what is observed. Magnetic sites that are not thoroughly degaussed or

are slightly aligned by the 30 G SEOP field could, therefore, be the major cause of

3He relaxation in the longest-lifetime cells. Whatever the cause of relaxation, our

results strongly indicate that increases in wall relaxation rates at high temperatures

are not responsible for the low (≤ 50%) 3He polarizations that we achieve.

We, along with other researchers [14, 73, 78], desire to know how the Rb interacts

with the glass in order to gain a better understanding of relaxation mechanisms in

spin-exchange cells. It was shown in Chapter 3 that Rb has the effect of reducing

the relaxation rate by a factor of 10 or more vs. the T−11 measured in the same cell

before Rb was introduced. Other workers have demonstrated a similar phenomenon

[19, 20]. By comparing the temperature-dependence data from Chapter 6 and Secs.

7.5.1 and 7.5.2 to alkali-coated cells in Fig. 7.4, it appears that the presence of

the Rb inhibits dissolution relaxation significantly, if not completely. We suggest

110

five possible explanations for this: (1) The Rb acts to block sites into which the

3He might dissolve by fitting into the “holes” of the irregular glass lattice; (2) The

Rb reduces the Fe3+ ions at or near the surface to less-relaxive Fe; (3) The Rb,

which is known to etch Pyrex, creates a new, amorphous layer on top of the glass

substrate; (4) Rb oxides could form at the surface from residual oxygen in the glass;

(5) The Rb forms a thin, metallic coating over most of the glass, which would be

very nonrelaxive to polarized 3He.

First, if the Rb blocks diffusion of 3He into the glass, disrupting the otherwise

dominant relaxation mechanism of interactions with Fe3+ ions in the bulk of the

glass, then a significant decrease in T−11 would be expected. The one cell containing

K in Fig. 7.4 had a much stronger absorption-like temperature dependence then

the cells with Rb, suggesting even less dissolution is occurring in this cell than the

others. Perhaps the smaller K atoms are more effective than the Rb at entering

the glass matrix and, therefore, are more effective at blocking dissolution of 3He.

Altemose [85] demonstrated that addition of any alkali oxide to a simple borosilicate

glass increased the activation energy of permeation and diffusion to He significantly,

while the activation energy of solubility was only weakly affected. Not all alkalis had

equivalent effects: he showed that as the diameter of the alkali ion was increased, the

activation energy of permeation at room temperature decreased until the alkali was

as large as Rb+. The addition of Rb or Cs had the effect of increasing the activation

energy when compared to K (or Na, but not Li). He proposed that K+ fits very well

into the holes of the irregular glass network, while Rb+ and Cs+ are large enough

to actually spread the glass network slightly. Although we added the alkali after

the glass was fabricated (Altemose added an alkali oxide prior to fabrication), it is

still reasonable to assume that the alkali was able to enter the glass network to the

depth of perhaps a few nm. The result would not be a totally impermeable surface,

111

so this possibility could only be a contributing factor.

Second, if the Rb reduces the relaxive Fe3+ ions to Fe at the surface and some

depth into the glass, then a significant relaxation mechanism will be diminished.

This possibility is effectively equivalent to the first since they both inhibit the ability

of the 3He to interact with relaxive Fe3+ ions. In Chapter 6 we showed that the

magnetization penetration depth in Pyrex is about 30 nm. It is unlikely that the Rb

is able to reduce Fe3+ ions to such a depth. The radius of a Rb+ ion is about 1.7 A

and the average hole diameter in the glass lattice is about 3 A [18], thus making

it difficult for the large Rb atoms to diffuse deeply into the glass. Therefore, this

possibility could also only be a contributing factor.

Third, it is possible that the result of the alkali reacting with (etching) the glass

is the formation of a thin, relatively impermeable, amorphous layer coating the cell.

As stated above, the insertion of an alkali during glass formation may fill random

holes in an irregular glass structure, reducing the He permeation rate significantly. If

a Rb-rich amorphous layer is formed during the initial interaction of the Rb with the

glass, then the layer would be significantly less permeable than the Pyrex substrate,

resulting in a longer-lifetime cell. This thin layer would also form a protective layer

to inhibit further etching by the Rb. In a separate experiment, we measured the

relaxation rate in each of several cells three times: first bare, then with Rb, finally

with the Rb rinsed out (see Sec. 8.5.2). We found that the rinsed cells had a marginal

improvement in T−11 vs. the original bare cells (typically a factor of ≈ 2 improvement

compared to the the factor of ≈ 10 when the Rb was still present). This suggests

that, if such a layer is formed, it is not solely responsible for the observed decrease

in relaxation rates in Rb-coated cells. We note that the cells showed T1 hysteresis

both after the Rb was introduced and after the Rb was rinsed out, whereas before

the introduction of Rb they did not. In addition, two of these cells were rinsed

112

with HCl after the Rb was rinsed out, then additional Rb was distilled in. The cell

relaxation rates exhibited only slight changes before and after the HCl rinse, possibly

suggesting that the acid was not effective in affecting the Rb-exposed glass. Further

experiments to reveal chemical changes to the glass surface would be necessary to

form any conclusions.

Fourth, we have already shown that Rb oxides are very nonrelaxive (see Sec.

4.12). The presence of such oxides in regular spin-exchange cells has been observed,

since the dark, nonmetallic oxides are often seen in cells that have been sitting on the

shelf for several weeks or months. Such oxides at the surface would act as a physical

barrier which might inhibit the permeation or adsorption of 3He thus reducing the

measured relaxation rate somewhat.

Fifth, a metallic coating of Rb on the glass surface would inhibit permeation and

adsorption. A metallic layer should be very nonrelaxive because they are weakly

adsorptive to 3He; an alkali metal surface is the most inert to noble gas adsorption

[91]. Heil et al. showed that a coating of Cs, the alkali metal with the largest atomic

radius, is very effective at inhibiting 3He relaxation [20]. A metallic layer can be as

thin as <1 monolayer, as long as the electron wave functions overlap. The formation

of a thin, continuous metallic layer would require a smooth glass surface and for the

adsorbed alkali be in equilibrium with the vapor. As the vapor pressure increases,

surface wetting occurs and may enhance the metallic layer thereby improving the

measured T−11 . A spin-exchange cell with a perfectly smooth surface would have very

favorable relaxation characteristics, but slight surface imperfections or impurities

would cause breaks in the metallic layer allowing for interactions of the 3He with

the glass [95]. Also, large magnetic sites may protrude through the layer and cause

relaxation.

It is certainly possible that all five phenomena occur in a cell to varying degrees.

As the first four phenomena are independent of cell geometry and probably occur

113

relatively equally in all cells, the best spin-exchange cells are probably those with

the smoothest, cleanest surface and therefore the best metallic coating. The number

of magnetic sites also plays a key role, and we believe that a very few sites can make

the difference between a long- and short-lifetime cell. In Chapter 3 we predicted

that ≈ 4× 104 sites occupy a typical cell, and cell-to-cell variation of a factor of two

is easy to imagine. Since that number is beyond our control during cell fabrication

and cell smoothness may or may not be a function of how the cells are blown, the

overall wall relaxation properties may be beyond control. One can only maximize

the chances of making a quality call through good preparation practices, which we

have detailed in Chapter 4.

7.6 Conclusion

We have shown that the presence of Rb in spin-exchange cells significantly inhibits

relaxive interactions of 3He with the glass surface and bulk glass by preventing 3He

from dissolving into the glass or adsorbing to the glass, contributing significantly to

long relaxation times observed in Rb-coated cells. Importantly, temperature depen-

dent changes to wall relaxation rates do not contribute to low polarizations achieved

during SEOP. We have also shown that interactions of dissolved 3He with Fe3+

ions in fused silica is a significant relaxation mechanism and that dissolution-based

relaxation is not significant in GE-180 cells at room temperature. The low observed

relaxation time in GE-180 cells indicates that mechanisms other than absorption

dominate the relaxation at the temperatures studied. The results of these studies

will help in the search for a quantitative understanding of relaxation in spin-exchange

cells, hopefully resulting in a better understanding of effective cell fabrication and

preparation protocols.

CHAPTER 8

CELL RINSING

8.1 Abstract

We show that rinsing cells with acids is not effective in eliminating T1 hysteresis

or in improving cell relaxation times. In fact, cells rinsed with acid tend to have

lower than average relaxation times and exhibit T1 hysteresis more strongly than

unrinsed cells. Also, we provide evidence in separate experiments that the magnetic

sites responsible for T1 hysteresis originate in the glass and are not brought in by

the Rb. In one experiment we rinsed cells with a chemical reducing agent, and in

another experiment we removed the Rb from cells. In both cases, T1 hysteresis was

observed.

8.2 Introduction

In Chapter 3 we showed that ferromagnetic sites at or near the surface of the

cells can be a major source of relaxation in spin-exchange cells. A key goal in our

work and the work of others [78] has been the elimination of such sites, which would

be a significant step towards consistent production of very long lifetime cells that

never require degaussing. However, since degaussing has proven to be reliable and

effective, a reasonable alternative is producing very long lifetime cells that may need

periodic degaussing. This chapter addresses several of our efforts to either eliminate

the sites or improve our cell preparation protocol.

115

The production of long-lifetime cells requires an understanding of physical phe-

nomena of 3He interactions at the cell surface. We have repeatedly shown in this

thesis that the basic model of relaxation, Eq. (1.9), does not accurately describe

3He relaxation. While other chapters of this thesis have addressed the dependence

of T1 on external parameters, such as temperature or magnetic field, this chapter

addresses effects of altering the interior surface of cells with three specific goals:

first, to investigate the origin of the magnetic sites, which should make it easier to

eliminate them; second, to eliminate T1 hysteresis by eliminating the magnetic sites;

third, to develop a method of producing consistently long-lifetime cells in spite of

T1 hysteresis effects.

It has often been suggested that we rinse cells with acid to accomplish the second

goal. We have generally had poor results using cells rinsed with acids. Specifically,

cells we rinsed with HF were particularly poor and had magnetized T−11 ’s higher

than their unmagnetized or degaussed T−11 ’s by factors of 20 to 100, far greater

than the typical factor of ≈ 2 for unrinsed cells. Other acids, such as nitric acid,

have been tried by B. Saam and T. Gentile [73], with mixed results. We have found

that acids do not remove T1 hysteresis-causing iron sites, but often enhance the

effect, possibly by exposing additional sites by etching the glass.

8.3 Theory

Iron and its oxides are known impurities in Pyrex at a concentration of ≈ 0.04%

by weight [18]. Our hypothesis for the origination of T1 hysteresis is that small

clusters of iron oxides (Fe2O3) may be distributed homogeneously throughout the

glass. Clusters that end up at the surface of a cell and are exposed to Rb may be

reduced to bulk iron, forming multidomain magnetic sites. This is plausible because

the energy required to reduce Fe3+ to Fe2+ is 0.77 eV and to reduce Fe2+ to Fe is

116

−0.45 eV, while the Rb oxidation potential is 2.93 eV [96]. That is, Fe3+ accepts an

electron if 0.77 eV is donated with it, whereas Fe2+ accepts two electrons and returns

0.45 eV. These sites, when magnetized, have a dramatic effect on wall relaxation of

polarized 3He. To achieve the goal of consistently producing robust, long-lifetime

spin-exchange cells for magnetic resonance imaging, it is desirable to remove the

iron clusters without any adverse effects on the wall relaxation rates. In theory,

the use of a gentle acid rinse to dissolve a significant number of sites would result

in long-lifetime cells that do not exhibit T1 hysteresis. If the surface of the glass is

etched, however, the resulting increase in surface area and/or exposure of additional

sites may result in very short lifetime cells.

Other rinsing techniques employed in this chapter were done to study the cell

surface by observing 3He relaxation rates. As stated above, we believe that the Rb

acts as a reducing agent on Fe3+ ions already in the glass. The use of a chemical

reducing agent can provide evidence both that the iron originates in the glass and

that the Rb reduces Fe3+ to Fe. A bare cell can be rinsed with a reducing agent such

as decamethylcobaltocene (DMC) [97] and tested for T1 hysteresis using polarized

gas that is transferred into the cell. The oxidation potential of DMC is 1.47 eV

[98], more than sufficient to reduce Fe3+. A second technique is to rinse Rb out of a

cell and measure the relaxation characteristics of HP 3He in the resulting bare cell.

Thus, effects that the Rb has on the glass surface may be detected. For example,

if the magnetic sites are brought in with the Rb, then it is reasonable to assume

that they should be rinsed out with the Rb, and the T1 hysteresis would vanish. Or,

if Rb etches the glass in a manner that exposes additional surface area, then the

relaxation rates in a rinsed cell should be faster than in the same cell before Rb is

introduced.

Another approach is to ignore the magnetic sites, since cells can be degaussed

117

if needed, and concentrate on techniques for making the longest lifetime cells.

Compared to many other types of glass, Pyrex is quite permeable to He [77], so

limiting diffusion of 3He into the bulk of the glass will limit relaxation due to

interactions with paramagnetic impurities such as Fe3+. It has been shown that

alkali-rich borosilicate glasses are less permeable to He than Pyrex by an order

of magnitude or more [83]. Further, it was demonstrated that potassium, among

all alkalis, fit best into the interstitial gaps of the amorphous glass structure, thus

inhibiting He diffusion most effectively [85]. Two of the best spin-exchange cells that

we ever produced contained K instead of Rb. Therefore, it is possible that treating

the cells with K prior to introducing Rb will result in very long-lifetime cells.

8.4 Experimental

We tested Pyrex cells, both Rb-coated and bare (not containing Rb). Cells that

contained Rb were prepared according to our usual methods outlined in Chapter

4. Bare cells were prepared as usual, but the Rb distillation steps were skipped.

Polarized gas was transferred into the bare cells using the gas transfer manifold

discussed in Chapter 5 (see Fig. 5.2). All T−11 measurements were made using

100 kHz pulse NMR detection (described in Chapter 2) at ≈ 30 G. Very small

flip angles, delivered through a coil on the cell stem (see Fig. 4.1), were used

to minimize magnetization destruction. Relaxation rates were typically measured

in three situations: unmagnetized, magnetized, and degaussed. As in Chapter 3,

unmagnetized refers to a newly fabricated cell that has not previously been exposed

to a high magnetic field. A magnetized cell has been exposed to a high magnetic

field (typically ≈ 104 G) of an electromagnet, and a cell is degaussed by exposing

it to an oscillating, decreasing magnetic field. The set of three measurements was

usually done with no intervening heating of the cell or exposure to laser light. The

118

cells were transported to and from the electromagnet in a battery-powered solenoid

to prevent polarization loss.

Acid and DMC rinses were done after cell fabrication but prior to attaching

the cell to a manifold. With the valve removed, the rinse could be easily added

and drained through the stem, after which the cells were thoroughly rinsed with

deionized water. The three acid solutions used were 10% HF, 37% HCl, and Aqua

Regia (3 parts HCl, 1 part HNO3).

Some cells were tested bare, but after containing Rb. To rinse Rb out of a cell,

the pressure was relieved, the valve was removed, and the tip of the cell stem cut

off. The Rb was allowed to react with the room air until it was no longer volatile.

Ethanol was then added through the stem and the cell thoroughly rinsed several

times. The cells were reattached to a manifold for evacuation and baking prior to

the introduction of HP 3He. We assumed that the reactions of the Rb with air or

the ethanol did not affect the glass surface.

8.5 Results/Discussion

8.5.1 Reducing-agent Rinse

To test the hypothesis that Rb acts as a reducing agent in the cells by reducing

existing iron oxide clusters to bulk iron, we rinsed two cells, labelled 13A and 13B,

with DMC. The unmagnetized, magnetized, and degaussed relaxation rates are

shown in Fig. 8.1. Both cells show T1 hysteresis to a slight degree, but more so

than any of the several other bare, unrinsed cells that we have tested. For example,

when cell 11A was tested bare with no rinses and prior to introduction of Rb, it

showed no T1 hysteresis even though its relaxation rates were similar to those of cells

13A and 13B (see Fig. 3.3). Cell 13B had an especially fast relaxation rate, much

faster than predicted for bare Pyrex in Chapter 6 and faster than other bare cells

119

0.10

0.15

0.20

0.25

0.30

0.35

0.40

un

mag

mag

deg

13A

13B

T1-1

(h

ou

rs-1

)

Figure 8.1. Cells rinsed with a chemical reducing agent. Cells 13A and 13B wererinsed with DMC for several minutes after fabrication but prior to baking. Bothcells show T1 hysteresis, representing the only T1 hysteresis that we have observedin six bare Pyrex cells tested. Error bars are too small to see. These data supportthe hypotheses that magnetic sites are intrinsic to the glass and are not introducedby the Rb, and that the Rb acts as a reducing agent.

we have observed. If T1 hysteresis effects were masked in other bare cells by faster

wall relaxation mechanisms, then we expect that no hysteresis would be detected

in such a short-lifetime cell. These data support our hypothesis that the magnetic

sites originate in the glass, and that existing iron oxide sites are being reduced to

give rise to T1 hysteresis.

8.5.2 Rb Rinses

One method of investigating the effects of the presence of Rb on 3He relaxation

is to measure T−11 before and after Rb is rinsed from a cell. This may indicate

whether effects of the Rb (i.e., long wall lifetimes and T1 hysteresis) depend on the

presence of Rb. Figure 8.2 shows relaxation rates of three cells before and after

120

0.01

0.1

1

10

5B 11A StL

deg with Rbmag with Rbdeg rinsedmag rinsed

T1-1

(h

ou

rs-1

)

Figure 8.2. Relaxation rates for three different cells before and after Rb is rinsedout. Error bars are to small to see. The cell labelled StL was made by collaboratorsat Washington University. The three cells were measured both magnetized anddegaussed when they contained Rb and again after the Rb was rinsed out. Polarizedgas was transferred into the rinsed cells to make the measurements. The cellsgenerally have longer relaxation times when the Rb is present, but they still showT1 hysteresis when rinsed.

the Rb was rinsed out. Two important features are shown in the figure. First, the

degaussed relaxation times of the cells, represented by circles, are generally better

when the cells contained Rb (cell StL showed little change). Second, the rinsed cells

still magnetize, verifying that the Rb itself is not responsible for T1 hysteresis, and

suggesting that the magnetic sites that are responsible are permanently attached to

the cell. This is evidence against the sites having been introduced by the Rb during

distillation, and supports the idea that reduced iron oxide causes T1 hysteresis.

Figure 8.3 shows data from cell 11A only. The cell was measured prior to any

Rb distillation (bare), after Rb was distilled in, and after the Rb was rinsed out.

This figure emphasizes both the lack of T1 hysteresis before the addition of Rb and

121

10-2

10-1

100

un

mag

mag

deg

11A bare11A11A rinsed

T1-1

(h

ou

rs-1

)

Figure 8.3. Relaxation rates for cell 11A when bare (new, no Rb), with Rb, and withthe Rb rinsed out. Error bars are too small to see. Polarized gas was transferredinto the cell for measurements without Rb. When rinsed, the relaxation rate isbetter than the bare cell, but not as good when it contained Rb. This indicates thatchanges to the glass by the Rb are both friendly to polarized gas and permanent.

the dramatic reduction in T−11 after Rb was introduced. The significant difference

in T−11 between the bare and rinsed measurements indicates that rinsing a cell with

Rb alters the surface in such a way to make the cell less relaxive. This may include

filling the holes of the amorphous glass into which 3He would otherwise diffuse and

relax. As noted in Sec. 8.5.1, T1 hysteresis was measured in bare cell 13B in spite

of the very fast relaxation rate. The fact that cell 11A, when bare (open circles in

Fig. 8.3), did not show any measurable hysteresis indicates that magnetic sites were

not present or were not significant contributors to 3He relaxation. This is further

evidence that Rb acts as a reducing agent in creating magnetic sites.

122

8.5.3 Acid Rinses

8.5.3.1 HF Rinse

The first cells rinsed with acid, labelled 5A and 5B, were cylindrical with rounded

ends with a volume of≈ 35 cm3. These cells were rinsed with a 10% hydrofluoric acid

(HF) solution for several minutes after initial fabrication and prior to attaching them

to the manifold for preparation. As shown in Fig. 8.4, these cells had unmagnetized

relaxation rates of about 0.025 h−1 for 5A and 0.091 h−1 for 5B. We note that the

unmaganetized rate of 5B is uncharacteristically high compared to that of most other

cells we have made (see Fig. 4.6). We have no explanation for the large difference in

unmagnetized relaxation rates between the two cells, since the cells were fabricated

and prepared in exactly the same way and at the same time, but we have observed

similar discrepancies in other cell pairs. These cells were not particularly “good”

10-2

10-1

100

101

un

mag

mag

deg

5A

5B

T1-1

(h

ou

rs-1

)

Figure 8.4. T1 hysteresis of HF-rinsed cells. Error bars are too small to see. Cells5A and 5B were rinsed with a 10% solution of HF for a few minutes prior toRb distillation. Both cells have magnetized T−1

1 s significantly higher than whendegaussed, indicating that a higher than average number of magnetic sites werepresent. It is apparent that rinsing with HF is not beneficial.

123

(we loosely define a good cell as one with T−11 ≤ 0.025 h−1), and they magnetized

abnormally strongly, by factors of 20 to 100 compared to a more typical factor of

≈ 2. The magnetized T−11 of 5B, 10 h−1, is by far the highest relaxation rate we

have measured for any spin-exchange cell.

To further understand the relaxation of 3He in Pyrex exposed to HF, we prepared

several samples of flat, ≈ 1 mm thick Pyrex glass for atomic force microscopy (AFM)

studies. The samples differed from glass used to make our cells because they were

not “reblown” (see Sec. 4.5). Some samples were untreated and others were rinsed

with HF. The HF-rinsed samples were rinsed in a 10% solution for several minutes,

then rinsed thoroughly with deionized water. The unrinsed samples were gently

cleansed with ethanol and a lint-free wipe. The AFM images indicate that HF

makes the surface very rough leaving many small pits and crevices, increasing the

surface area significantly (compare Figs. 8.5 and 8.6). In addition, it is possible

that HF simply exposed more potential magnetic sites than it removed. Although

0µm

12µm

24µm0µm

12µm

24µm

65nm

32nm

0nm

Figure 8.5. An AFM image of an untreated Pyrex sample. The ≈ 1 cm3 sample offlat Pyrex was rinsed with ethanol and a lint-free wipe prior to imaging. The imageshows a random section of the sample 24 µm on a side with a vertical scale of 65nm.

124

0µm

12µm

24µm0µm

12µm

24µm

90nm

0nm

45nm

Figure 8.6. An AFM image of a Pyrex sample rinsed with a 10% HF solution forseveral minutes. The image shows a random section of the sample 24 µm on a sidewith a vertical scale of 90 nm. This sample is much more rough than the sampleshown in Fig. 8.5, which likely explains why cells rinsed with HF had relatively highrelaxation rates and magnetized very strongly (see Fig. 8.4).

several researchers have observed that the presence of Rb dramatically improves

wall relaxation times over bare glass [17, 19, 20, 45], the damage done by the HF

rinse may be beyond repair.

8.5.3.2 Aqua Regia Rinse

Next we rinsed cells 16A and 16B with aqua regia (AR), a solution of 3 parts

HCl and 1 part HNO3. We used this acid in an effort to dissolve surface iron sites

because the acid does not aggressively attack glass. These cells were spherical with

a volume of ≈ 50 cm3. The bodies of the cells were submerged in a beaker of the

AR solution overnight with the capillaries sticking out; the cells were found floating

partially submerged in the morning. After flushing thoroughly with deionized water,

the cells were attached to a manifold for baking and Rb distillation. These cells had

unmagnetized T−11 ’s of about 0.028 h−1 for 16A and 0.083 h−1 for 16B (see Fig.

125

8.7). Again, we have no explanation for the large discrepancy in the rates between

the two cells, since they were prepared identically. We note that these cells did not

fully degauss to the original T−11 , nor were they considered good. However, these

cells magnetized by a factor of ≈ 3 to 10, which is less than the factor of ≈ 20 to

100 for cells 5A and 5B rinsed with HF but more than the typical factor of ≈ 2.

Sufficient etching of the glass by the AR may have taken place to expose at least as

many potential sites as may have been dissolved while increasing the surface area

somewhat.

8.5.3.3 HF and HCl Rinse

The next experiment involved cells 18A and 18B, both spherical 50 cm3 Pyrex

cells. We chose to initially rinse with an aggressive acid to expose as many sites

10-2

10-1

100

un

mag

mag

deg

16A

16B

T1-1

(h

ou

rs-1

)

Figure 8.7. T1 hysteresis of cells rinsed with aqua regia. Both cells magnetizesignificantly, but not to the extent of the cells rinsed with HF (see Fig. 8.4). It isapparent that rinsing with AR is not beneficial to cell quality, nor does it eliminateT1 hysteresis.

126

near the surface as possible, then to rinse with an acid that does not etch glass

aggressively. The idea of the second rinse was to dissolve the sites that were exposed

by the first rinse while avoiding exposure of additional sites. We rinsed the cells first

with a 10% HF solution for several minutes and then with a 37% HCl solution for

several minutes. The cells were then attached to a manifold and prepared as normal.

The unmagnetized T−11 s of these cells were about 0.022 h−1 for 18A and 0.037 h−1

for 18B (see Fig. 8.8). (We note that in this chapter the “B” cells are consistently of

shorter T1. We feel this is just a coincidence, since this behavior is reversed or totally

absent in other cell pairs. Cell placement on the manifold is random, and labels are

assigned only after the cells are attached.) Similar to other acid-rinsed cells, the

original T−11 ’s were not fully recovered by degaussing. These cells magnetized by

a factor of ≈ 10 to 30, stronger than typical cells that have not been rinsed, and

stronger than the cells rinsed with AR. To better understand the effects of rinsing

10-2

10-1

100

un

mag

mag

deg

18A

18B

T1-1

(h

ou

rs-1

)

Figure 8.8. T1 hysteresis of cells rinsed with HF and HCl. Error bars are too smallto see. Cells 18A and 18B were rinsed with a 10% HF sloution then a 37% HClsolution prior to Rb distillation.

127

with HCl, we prepared samples of flat Pyrex rinsed with a 37% HCl solution for

AFM; see Fig. 8.9. It is difficult to discern surface characteristics from Fig. 8.9, but

the results of T−11 measurements indicate that HCl has similar qualitative effects as

the other acids we used. Specifically, it may increase the surface area and expose

additional relaxation sites.

8.5.3.4 Intervening HCl Rinse

Finally, two other cells, 5B′ and 11A′, were rinsed with acids. The prime denotes

that the original Rb was rinsed out and Rb distilled in a second time. 5B′, originally

labelled 5B, was previously rinsed with HF prior to initial Rb distilation (see Sec.

8.5.3.1); its relaxation properties are described above (see Fig. 8.4). 11A′ was

initially prepared as a normal, unrinsed cell labelled 11A (see Fig. 8.3), which

reflected a more typical factor of two change in rate when magnetized. The Rb was

rinsed from both cells with ethanol, then the cells were rinsed with an HCl solution

for several minutes. The idea was to eliminate magnetic sites whose existence

0µm

12µm

24µm0µm

12µm

24µm0nm

10nm

20nm

Figure 8.9. Atomic force microscopy of a bare Pyrex sample rinsed with HCl.Compare to Figs. 8.5 and 8.6, noting the difference in vertical scale.

128

had been verified; this is the key difference between this experiment and previous

acid-rinse experiments. After the HCl rinse, the cells were reattached to a manifold,

baked, and Rb distilled in. Figure 8.10 shows T1 hysteresis results before and after

the intervening HCl rinse. Surprisingly, the unmagnetized and degaussed relaxation

rates for each cell before and after the rinse are similar, implying that wall relaxation

properties were not affected much by the intervening acid rinse. This suggests that

the cell wall relaxation properties are largely set during the fabrication process or

the initial introduction of Rb.

10-2

10-1

100

101unm

ag

mag

deg

5B5B'11A11A'

T1-1

(ho

urs

-1)

Figure 8.10. T1 hysteresis of Rb and HCl rinsed cells. Error bars are too small tosee. Cell 5B′ (originally cell 5B, see Figure 8.4) was rinsed with HF prior to initialRb distillation. The Rb was rinsed out, the cell was rinsed with HCl, then the cellwas attached to a new manifold Rb distilled in again. Cell 11A′ (originally 11A) wasinitially prepared as a standard spin-exchange cell. The Rb was rinsed out and thecell prepared again with 5B′. The data show that the unmagnetized and degaussedrates were similar before and after the acid rinse, possibly indicating that the aciddoes not have a dramatic effect on cells that have contained Rb.

129

8.5.4 Potassium Rinse

It has been shown that K is the most effective of the alkali metals in filling

interstitial gaps in a borosilicate glass matrix [85]. In fact, two of the longest-lifetime

spin-exchange cells we have made contained K instead of Rb (cells 12A and 12B in

Fig. 4.6). We prepared two standard spin-exchange cells normally except that the

manifold had a second retort for a K ampoule [33] positioned between the liquid

nitrogen trap and the “B” cell (see Sec. 4.5 for a detailed description of a standard

manifold). After baking but prior to Rb distillation, a small amount of K (≈ 30

mg) was distilled into each cell. The cells were then heated to ≈ 200 C for several

hours to imitate the heating during optical pumping that the previous K-coated

cells experienced. This period of heating effectively drove all visible quantities of

K from the cells. However, we think that K was present long enough to chemically

interact with the entire cell surface. Finally, Rb was distilled into the cells. It is

certain that some K was distilled in also, since the two metals became mixed in the

manifold. The relaxation rates of these cells, labeled 20A and 20B, were measured.

The unmagnetized rates of the cells were 0.035 h−1 and 0.029 h−1, respectively. (The

magnetized rates of these cells were not measured since they were being used for

low-field hysteresis measurements; see Sec. 5.5.2.) It seems that pretreating with K

is not any better at inhibiting relaxation than simply introducing Rb.

8.6 Conclusion

We have demonstrated that rinsing cells with acid prior to Rb distillation is an

ineffective treatment. The acid-rinsed cells had higher than average relaxation rates

when unmagnetized or degaussed, and they tended to magnetize more strongly than

unrinsed cells (that is, for the acid rinsed cells, the ratio of the magnetized rate to

the unmagnetized rate was larger than average). This behavior suggests that the

130

acids etched the cell surface to expose additional surface area and potential relaxive

magnetic sites. However, cells that were rinsed with Rb prior to an acid rinse did

not show a significant change in relaxation rate before and after the acid rinse.

This indicates that the cell properties may be largely established upon initial Rb

distillation.

We have also demonstrated that bare Pyrex cells can demonstrate T1 hysteresis

when rinsed with a chemical reducing agent. This is evidence in favor of the

hypothesis that the magnetic sites originate in the glass, and that Rb behaves as a

reducing agent in the cells. Further evidence is seen in cells that have had their Rb

rinsed out. These cells have somewhat higher rates than when the Rb was present,

but they continued to show T1 hysteresis even in the absence of Rb.

We acknowledge J. Miller of the University of Utah Chemistry Department for

his assistance with the DMC rinsing. We also acknowledge J. Kyle and M. Delong

for assistance with the acid rinses. Our thanks to B. Anger and C. Inglefield of

Weber State University for the AFM images.

CHAPTER 9

MRI OF FLOWING POLARIZED 3He

9.1 Abstract

To demonstrate the feasibility of gas flow imaging in physiologically relevant

phantoms, we imaged flowing HP 3He using both a spin tagging, time-of-flight

technique and velocity-sensitive phase encoding. The images were made using

straight and curved tube phantoms with an inner diameter approximately the same

as an adult rat trachea. We imaged flow and diffusion with a resolution of 625 µm,

high enough to see a clear velocity and diffusion distribution in the tube. The spin

tagging images were made with a resolution of 315 µm, and rapid signal wash-out

due to diffusion is clearly visible. This chapter is a discussion of the results of the

first velocity and diffusion maps made on flowing HP 3He. These experiments were

feasible because of our success at making long-lifetime spin-exchange cells.

9.2 Introduction

Magnetic resonance imaging (MRI) is typically a spin-density imaging technique

that takes advantage of the unequal energy-level populations of an ensemble of

nuclei in an external magnetic field. The thermal equilibrium population difference

is simply a Boltzman distribution. The energy difference is ∆E = hγB0, where γ

is the nuclear gyromagnetic ratio and B0 is the external magnetic field strength,

conventionally along the z-axis. Each nuclear spin precesses about the external field

132

with a frequency ω0 = γB0, called the Larmor frequency, and tends to align itself

with the field, resulting in a net magnetization of the sample. The application of

a radio frequency (RF) magnetic field at ω0 and orthogonal to B0 causes energy

absorption by the nuclei and tips the net magnetization away from B0. When the

RF field is turned off, the precession of the nuclear spins about B0 induces a voltage

in a pick-up coil, which is often the same coil that delivers the RF field. Magnetic

field gradients can be used to spatially encode the spins. For example, a gradient

G = ∂Bz/∂xi applied in an arbitrary direction xi gives a field B(xi) = B0 + Gxi.

This encodes the nuclear spins along the xi axis with a unique position-dependent

Larmor frequency ω(xi) = γB(xi). Thus, the intensity of the response at a given

frequency is proportional to the spin density at the corresponding position. By

appropriate use of gradients, an image can be produced by sampling a sufficient

number of positions in the sample.

The last several years has witnessed rapid development of the use of hyperpo-

larized (HP) gas as the signal source for MRI, especially of pulmonary air spaces.

Advances in MRI strategies have allowed researchers to visualize 6th generation

airway branches in live rats [99]. Other studies of 3He diffusion coefficients in

lung tissues have revealed significant physiological differences between healthy and

diseased tissue [100]. Clinical use of HP gas would provide spatial and temporal

resolution not possible with any other pulmonary imaging modalities.

The work described in this chapter uses HP gas to map flow velocities, a technique

that has drawn little attention, since most images are maps of spin density. HP

gas allows for very rapid imaging because the enormous nonequilibrium nuclear

polarization eliminates the need to wait for the magnetization to relax back to

thermal equilibrium between pulses. Thus, images of gas flowing, for example into

lung air spaces, can be made in virtual “real time,” that is, images can be made in

133

sufficiently rapid succession to visualize flow dynamics. This is especially important

when imaging 3He because the diffusion coefficient is very high. Diffusion through

field gradients causes signal attenuation due to dephasing of the magnetization.

The data we collected was preliminary for the purpose of demonstrating the

feasibility of flow measurements using HP gas in physiologically relevant phantoms.

The Virtual Lung project at Pacific Northwest National Laboratory (PNNL) uses

computational fluid dynamics (CFD) models to predict long-term disease progres-

sion in pulmonary tissue. Model validation requires flow data of gas in the bronchial

tree and diffusion data in alveolar air spaces. In March of 2002 I transported the

polarizing equipment and several spin-exchange cells to Richland, WA to perform

the preliminary imaging experiments in collaboration with Dr. Kevin Minard. This

work was largely possible due to our success at making long-lifetime cells, since we

usually required cells to sit for many hours after polarizing and prior to imaging

while relying on minimal polarization loss. This was because we could polarize a

sample of gas every 12 hours, but we could only image during regular working hours.

9.3 Theory

The effects of diffusion on bulk flow and vice versa in a rapidly diffusing gas

like He are not well understood, particularly at physiologically relevant flow rates.

In flow measurements, rapid diffusion may cause significant signal attenuation, due

to the magnetization dephasing that occurs in the presence of a magnetic field

gradient, thereby setting a lower limit on the measurable flow rate. If laminar flow

is established, then random sampling of gas atoms between streamlines will affect

the measured diffusion coefficient by making it appear even higher, an effect called

Taylor dispersion [101].

134

A bipolar gradient sequence, or an equivalent sequence with unipolar gradients

separated by a π pulse (a π pulse is an RF pulse that inverts the magnetization by

rotating it 180) is sensitive to both velocity and diffusion by changing the phase

and attenuation of the signal, respectively. Such a pulse sequence is shown in Fig.

9.1. If flow is at a constant velocity V , then the phase change Φ for a moving spin

is [102]:

Φ = γV∫

tG(t) dt , (9.1)

where γ is the gyromagnetic ratio and G(t) is the applied gradient parallel to the

velocity component being measured. The sequence shown in Fig. 9.1 has (idealized)

rectangular gradients of duration T separated by a time Td. From Eq. (9.1) the

resulting phase change experienced by constantly moving spins is:

Φ = γV GTTd. (9.2)

RF

Td

Signalπ pulseπ/2 pulse

G

Figure 9.1. The velocity and diffusion sensitive gradient sequence used to make im-ages in Fig. 9.3. The sequence is run twice, once with the gradients off (indicated bythe dashed lines) and once with the gradients on. The phase change is proportionalto the velocity, and the signal attenuation determines the diffusion coefficient. Thegradients are applied in a direction parallel to the measured velocity and diffusion.

135

Using the same pulse sequence, the signal attenuation due to diffusion is [102]:

exp[−γ2G2DT 2 (Td − T/3)

], (9.3)

where D is the diffusion coefficient. Thus, two measurements of phase and attenua-

tion, one with the bipolar gradients off and the other with the gradients on, can be

used to extract velocity and diffusion coefficients.

An effective method of visualizing flow in a phantom is called multistripe tagging

[103]. A stripe is a region of the sample in which all of the magnetization has been

destroyed. Stripes can be “tagged” using an RF pulse generated by modulating

a series of evenly spaced, uniform, rectangular pulses with a sinc function, where

sinc(t) ≡ sin(πt)/πt. The effective region of striping is determined by the pulse

width, the profile and width of each stripe are determined by the sinc envelope, and

the stripe spacing is determined by the applied gradient and pulse spacing. The

striping pulse puts the magnetization of the stripe regions into the transverse plane,

after which a “crusher” gradient destroys all transverse magnetization so that no net

magnetization remains in the stripes. By incrementally increasing the delay time

between stripe tagging and imaging, a temporal progression of the stripes can be

seen as the fluid moves.

9.4 Experimental

These experiments were conducted at PNNL using the polarization apparatus and

spin-exchange cells transported from Utah. Polarized gas was delivered to various

phantoms using a home-built, small-animal ventilator similar to others [104]. The

ventilator controlled delivery and mixing of the gases. The experimental set-up is

136

schematically shown in Fig. 9.2. HP 3He was dispensed from a high-pressure cell into

an evacuated Tedlar (DuPont) bag, which is contained in a sealed box. The gas was

then pneumatically forced out of the bag with dry N2 gas and through a tube to a

pneumatic valve at a controlled rate by pressurizing the box. The pneumatic valve,

placed in the magnet core, allows mixing of the 3He with up to three other gases.

We used a mixture of approximately 10% 3He and 90% N2 gas with a total flow rate,

monitored with separate flow meters, of approximately 200 cm3/min, physiologically

relevant for a rat. The phantoms used were elastic tubes with an inner diameter of

≈ 2.3 mm, comparable to an adult rat trachea. Each experiment used ≈ 350 cm3

of 3He at an initial polarization of ≈ 30%. The measured polarization of the gas at

the phantom was 7 ± 2 %. The images were made in a 30 cm diameter, horizontal

bore, 1.5 T superconducting magnet with a tuned RF coil. No slice selection was

used in the images.

X

X

ventilator

Tedlar

bag

magnet

pneumatic valve

N

N2

He

N2

source

regulator valves

flow meters

3He

to imaging phantom

Figure 9.2. Experimental set-up for MRI flow imaging.

137

9.5 Results and Discussion

Figure 9.3 shows an image of measured velocity vs. position, or velocity map

(green background), and an apparent diffusion coefficient (ADC) map (black back-

ground). The direction of gas flow is indicated by the dashed arrow. No slice

selection was used, so the measured velocity and diffusion at any position in the

image is a projection through an axis perpendicular to the direction of flow. For

the velocity map, the color scale represents velocities between +84 cm/s in violet

through 0 cm/s in green to −84 cm/s in red. For the ADC map, violet represents

2.3 cm2/s, green is 1.1 cm2/s, and red is 0 cm2/s. For both images the field of view

is about 6 cm × 6 cm, and the spatial resolution is 625 µm.

The velocity map shows flow velocities parallel to the motion probing gradient

in the range of ± 84 cm/s in the vertical sections, parallel (or antiparallel) with

the direction of the applied motion-probing gradients (indicated by the bold arrow),

Figure 9.3. Velocity map (left) and ADC map (right) for flowing HP 3He througha tube. The bold arrow shows the direction of applied motion probing gradientsand the dashed arrow indicates direction of gas flow. The color scale on the far leftapplies differently to both maps. For the velocity map, the scale represents velocitiesfrom +84 cm/s (violet) to −84 cm/s (red). For the ADC map, the scale representsADC values from 2.3 cm2/s (violet) to 1.1 cm2/s (green) to the limiting case of 0cm2/s (red). The field of view is 6 cm × 6 cm, and the spatial resolution is 625 µm.

138

which agrees well with the known flow rate of approximately ± 80 cm/s. In the turn

at the top, where the bulk flow is perpendicular to the motion-probing direction,

the measured velocity transitions through zero from negative to positive values, as

expected.

The ADC map shows diffusion coefficient values in a range of 2.3 cm2/s to 0

cm2/s. The value of 1.1 cm2/s measured in the turn is close to ADC values of dilute

3He in nitrogen with no bulk flow [100]. The somewhat higher values measured in

the straight segments suggest the presence of Taylor dispersion [101]: as 3He atoms

diffuse between different flow stream lines, the mean square displacement in the

direction of the flow increases and is reflected as a higher ADC.

Figure 9.4 shows a sequence of images taken after stripes, with separation of

≈ 1 cm, were burned into the flowing gas, and corresponding 1-d projections on

the right. These images were made by creating ≈ 3.5 mm wide stripes, then

imaging the remaining magnetization after increasing time delays. The 15 images

are temporally spaced by 1 ms with an initial 1 ms delay between the RF pulse

and the first image, thus showing flow progression over 16 ms. Spatial resolution is

315 µm. The stripes are clearly visible, with more intense signal in blue and less

intense in red. Motion of the gas can be seen as the stripes progress to the left

(the vertical dashed lines were added to guide the eye). The stripes blurred almost

immediately as diffusion caused mixing between the stripes and the regions with

magnetization, with nearly thorough mixing having taken place after only 16 ms.

This is reasonable if we assume D ≈ 1.1 cm2/s for dilute 3He in nitrogen, which was

measured in Fig. 9.3. This corresponds to a mean linear distance traveled for each

atom of about 1.3 mm after 16 ms. Since magnetization was diffusing into stripes

from both ends, the 3.5 mm stripes should be nearly filled after 16 ms, as observed.

139

TIME

≈1 cm

Figure 9.4. Visualization of real-time MRI of 3He flowing through a tube. The 3Hewas diluted to a ≈ 10% concentration then flowed through a 2.3 mm i.d. tube at ≈200 ml/min. Stripes were burned in (less intense signal is red, more intense signal isblue), and images were taken at 1 ms intervals. The rapid diffusion of the 3He (≈ 1cm2/s) caused immediate mixing of the polarized gas with the stripes until the signalwas almost evenly distributed after only 16 ms. Corresponding 1-d projections ofthe signal are on the right. The dashed lines are added to guide the eye.

140

9.6 Conclusion

We have demonstrated the feasibility of MRI velocity and diffusion maps of

flowing HP 3He. The very high diffusion coefficient of 3He, which may be enhanced

by Taylor dispersion, places a lower limit on the velocity that can be measured. The

high diffusion coefficient also complicated efforts to visualize flow of 3He through

a tube by rapidly washing-out the striping pattern created by the rf pulse. The

results of these imaging experiments are difficult to interpret without a knowledge

of the actual flow dynamics. In order to make this imaging technique useful, MRI

results will have to be compared to computational models of flow in the simple

tube phantoms. One can then proceed with model validation in more complicated

phantoms.

We acknowledge K. Minard at PNNL for his MRI expertise used in setting up and

running the pulse sequences. Thanks to G. Samuelson for designing and building

the ventilator.

APPENDIX

INTERMEDIATE-FIELD

SPECTROMETER

The NMR spectrometer used in Chapter 5 was built especially to handle NMR

frequencies from about 300 kHz to about 3 MHz. It does a very adequate job, but

suffers in signal-to-noise at the lower frequencies because we use a single, ≈ 0.8 µH

probe in series with a 50 Ω resistor for all frequencies. This spectrometer is very

similar in principle to the spectrometer described in Chapter 2 but consists of

RF components designed to handle the higher frequencies. It was built in three

13”×17”×3” chassis boxes. Each of the boxes is described below.

Box 1 (Fig. A.1) contains the power supply and RF-pulse power amplifier. The

RF from the synthesizer is split by a ZSC-2-1 power splitter [105], which is mounted

outside the box for additional shielding for the internal components. The two ZAD-1

frequency mixers [105] act as gates to prevent RF leakage to the coil. The ZAD-1

has three ports labeled “L,” “I,” and “R.” “L” is the local oscillator input, and “I”

and “R” are the intermediate frequency (IF) and RF ports, respecitvely, both of

which can act as an input or an output. The variable attenuator [106] can be used

to switch in as much as 65 dB in the RF pulse line. The ZHL-32A broadband power

amplifier [105] amplifies the RF pulse which is then sent to the duplexer in Box 2.

Box 2 (Fig. A.2) contains the receiver and duplexer. A ZSC-2-1 power splitter

acts as the duplexer. The ZSC-2-1 has three ports labeled “S,” “1,” and “2.” Ports

142

ZAD-1 ZAD-1

L I R L I R1

S

20-65 dB

ATTENUATOR

ZSC-2-1

POWER

AMP

GATE

INPUTS

SYNTH IN

PSD

BOX 3

+15V DC

BOX 2

POWER

AMP

BOX 2

POWER SUPPLY

+24 V DC

ZHL-

32A

Figure A.1. Box 1 of the intermediate-field spectrometer. This box contains thepower supply, RF-pulse gates, and the RF-pulse power amplifier.

“1” and “2” are isolated from each other, so the splitter acts as a switch in this ap-

plication. Although not ideal, it does an adequate job at these frequencies. Initially,

a switchable, tuned LC circuit, similar to that used in the 100 kHz spectrometer,

was used. However, we found that such a duplexer added too much noise to the

NMR signal. The preamp [107] is powered by an isolated +15 VDC input from Box

1 for additional noise reduction, and it has been modified by the manufacturer to

143

PREAMP

+15 V IN FROM BOX 1

PREAMP OUT TO

BOX 3

ZSC-2-1

2 S 1

TO NMR PROBE

POWER AMP IN FROM BOX 1

DUPLEXER

DIODE GATES

MITEQ

Figure A.2. Box 2 of the intermediate-field spectrometer. This box contains theNMR signal receiver and the duplexer.

144

have a < 50 µs recovery time. Diode gates are contained in Bud boxes (indicated

by the dashed squares) for additional shielding.

Box 3 (Fig. A.3) contains both the pulse generation circuitry and phase detector

(separated from each other by the dashed line in the figure). Figure A.4 shows the

details of the pulse generator and low-pass filter section. The pulse length can be

continuously varied from 10 µs to 10 ms. The variable attenuator can be used to

switch in as much as 65 dB in the receiver line to avoid saturating the ZFL-500

amplifier [105]. The pulse generator circuit requires +5 VDC and the ZFL-500

requires +15 VDC, both provided by the power supply in Box 1.

This spectrometer was designed and built with heavy reliance on the expertise

and advise of M. Conradi. B. Anger and S. Morgan did most of the actual assembly.

145

ZAD-1

LOW-PASS

SECTION

0-65 dB

ATTEN-

UA

SYNTH IN

FROM BOX 1

PREAMP

IN FROM

BOX 2

L I R

PULSE

GENERA

CIRCUIT

SIGNAL

OU

SCOPE

SCOPE

TRIGGER

Z500

AMP

PHASE

SENSITIVE

DETECTOR

GATE

OUTPUTS BOX 1

Figure A.3. Box 3 of the intermediate-field spectrometer. This box contains thepulse generator circuitry, NMR signal amplifier, and the phase detector. See Fig.A.4 for pulse generator and low-pass filter details.

146

74121Q

+5

+5

+5

+5

SCOPE

TRIGGER

GATES

330

330 330

3306.8k

6.8k

PULSE GENERA CIRCUIT

+5 143

4

7

11 10

74121 DETAILS

2.0k

+5

25k 0.0068

0.068

0.68

SWITCH

center: 10-100 µs

left: 100-1000 µs

right: 1-10 ms

OU

SCOPE

3.3k1.0k

270 .001

IN

FROM

PSD ROTARY SWITCH

.001.0033 .01 .022 .05 .1 .18 .47

LOW-PASS FILTER DETAILS

Figure A.4. Intermediate field spectrometer pulse generator and low-pass filtersection details. See Fig. A.3.

REFERENCES

[1] T. G. Walker and W. Happer, Rev. Mod. Phys. 69, 629 (1997).

[2] T. R. Gentile and R. D. McKeown, Phys. Rev. A 47, 456 (1993).

[3] P. Anthony, Phys. Rev. Lett. 71, 959 (1993).

[4] N. R. Newbury, A. S. Barton, P. Bogorad, G. D. Cates, M. Gatzke, B. Saam,L. Han, R. Holmes, P. A. Souder, J. Xu, and D. Benton, Phys. Rev. Lett. 67,3219 (1991).

[5] T. E. Chupp, R. J. Hoare, R. L. Walsworth, and B. Wu, Phys. Rev. Lett. 72,2363 (1994).

[6] W. Heil, J. Dreyer, D. Hofmann, H. Humblot, E. Lelievre-Berna, and F. Tasset,Physica B 267,268, 328 (1999).

[7] D. Raftery, H. Long, T. Meersmann, P. J. Graninetti, L. Reven, and A. Pines,Phys. Rev. Lett. 66, 584 (1991).

[8] M. S. Albert, G. D. Cates, B. Driehuys, W. Happer, C. S. Springer Jr., B.Saam, and A. Wishnia, Nature 379, 199 (1994).

[9] J. C. Leawoods, D. A. Yablonskiy, B. Saam, D. S. Gierada, and M. S. Conradi,Concepts Magn. Reson. 13, 277 (2001).

[10] A. Kastler, Phys. Radium 11, 225 (1950).

[11] M. A. Bouchiat, T. R. Carver, and C. M. Varnum, Phys. Rev. Lett. 5, 373(1961).

[12] W. Happer, Rev. Mod. Phys. 44, 169 (1972).

[13] T. Killian, Phys. Rev. 27, 578 (1926).

[14] T. Walker, University of Wisconsin, private communication.

[15] N. R. Newbury, A. S. Barton, G. D. Cates, W. Happer, and H. Middleton,Phys. Rev. A 48, 4411 (1993).

[16] G. D. Cates, S. R. Schaefer, and W. Happer, Phys. Rev. A 37, 2877 (1988).

148

[17] R. S. Timsit, J. M. Daniels, and A. D. May, Can. J. Phys. 49, 560 (1971).

[18] R. Mazitov, P. Diehl, and R. Seydoux, Chem. Phys. Lett. 210, 543 (1993).

[19] W. A. Fitzsimmons, L. L. Tankersley, and G. K. Walters, Phys. Rev. 179, 156(1969).

[20] W. Heil, H. Humblot, E. Otten, M. Schafer, R. Sarkau, and M. Leduc, Phys.Lett. A 201, 337 (1995).

[21] M. F. Hsu, G. D. Cates, I. Kominis, I. A. Aksay, and D. M. Dabbs, Appl.Phys. Lett. 77, 2069 (2000).

[22] B. T. Saam and M. S. Conradi, J. Magn. Reson. 134, 67 (1998).

[23] M. A. Rosenberry and T. E. Chupp, Phys. Rev. Lett. 86, 22 (1997).

[24] X. Wu et al., Phys. Rev. Lett. 85, 2900 (2000).

[25] E. Brunner, M. Haake, A. Pines, J. A. Reimer, and R. Seydoux, Chem. Phys.Lett. 290, 112 (1998).

[26] A. B. Baranga, S. Appelt, M. V. Romalis, C. J. Erickson, A. R. Young, G. D.Cates, and W. Happer, Phys. Rev. Lett. 80, 2801 (1998).

[27] T. B. Smith, T. E. Chupp, K. P. Coulter, and R. C. Welsh, Nucl. Instrum.Methods Phys. Res. A 402, 247 (2000).

[28] G. L. Jones, T. R. Gentile, A. K. Thompson, Z. Chowdhuri, M. S. Dewey,W. M. Snow, and F. E. Weitfeldt, Nucl. Instrum. Methods Phys. Res. A 440,772 (2000).

[29] R. E. Jacob, S. W. Morgan, and B. Saam, J. Appl. Phys. 92, 1588 (2002).

[30] N. Bloembergen, E. M. Purcell, and R. V. Pound, Phys. Rev. 73, 679 (1948).

[31] J. O. Hirschfelder, C. F. Curtiss, and R. B. Bird, Moleculal Theory of Gasesand Liquids (Wiley, New York, NY, 1954), p. 15.

[32] R. Barbe, M. Leduc, and F. Laloe, J. Phys. (Paris) 35, 935 (1974).

[33] Atlantic Metals and Alloys, Stratford, CT.

[34] Z. Q. Qiu, Rev. Sci. Instrum. 71, 1243 (2000).

[35] M. V. Romalis and M. P. Ledbetter, Phys. Rev. Lett. 87, 067601 (2001).

[36] H. J. Jansch, T. Hof, U. Ruth, J. Schmidt, D. Stahl, and D. Fick, Chem. Phys.Lett. 296, 146 (1998).

149

[37] T. Piertaß, R. Seydoux, and A. Pines, J. Magn. Reson. 133, 299 (1998).

[38] E. E. de Lange, J. P. Mugler III, J. R. Brookman, J. Knight-Scott, J. D. Truwit,C. D. Teates, T. M. Daniel, P. L. Bogorad, and G. D. Cates, Radiology 210,851 (1999).

[39] F. D. Colegrove, L. D. Schearer, and G. K. Walters, Phys. Rev. 132, 2561(1963).

[40] M. V. Romalis, E. Miron, and G. D. Cates, Phys. Rev. A 56, 4569 (1997).

[41] L. D. Shearer and G. K. Walters, Phys. Rev. 139, A1398 (1965).

[42] F. J. Norton, J. Appl. Phys. 28, 34 (1957).

[43] J. R. Johnson, A. K. Thompson, T. E. Chupp, T. B. Smith, G. D. Cates, B.Driehuys, H. Middleton, N. R. Newbury, E. W. Hughes, and W. Meyer, Nucl.Instrum. Methods Phys. Rev. A 148, 356 (1995).

[44] T. R. Gentile, D. R. Rich, A. K. Thompson, W. M. Snow, and G. L. Jones, J.Res. Natl. Inst. Stand. Technol. 106, 709 (2001).

[45] R. E. Jacob, S. W. Morgan, B. Saam, and J. C. Leawoods, Phys. Rev. Lett.87, 143004 (2001).

[46] Part no. 826460-0004, Kimble-Kontes, Vineland, NJ.

[47] Part no. CG-351-01, Chemglass, Inc., Vineland, NJ.

[48] Part no. SS-8BK-K5, Nupro Co., Willoughby, OH.

[49] Model no. TSU071E, Pfeiffer Vacuum Inc., Nashua, NH.

[50] Model no. PKR-251, Pfeiffer Vacuum Inc., Nashua, NH.

[51] Model no. RGA200, Stanford Research Systems, Sunnyvale, CA.

[52] Model no. MVP015-T, Pfeiffer Vacuum Inc., Nashua, NH.

[53] Part no. VCMT-1000-BA, Varian Vacuum, Lexington, MA.

[54] Part no. NPC-410, Lucas Novasensor, Fremont, CA.

[55] Part nos. CG-303-108 and CG-303-101, Chemglass Inc., Vineland, NJ.

[56] Part no. 370-5, Cotronics Inc., Brooklyn, NY.

[57] Part no. 18S-36, Thermostatic Industries, Los Angeles, CA.

[58] Omega Heater, Inc., Ronkonkoma, NY.

150

[59] Part no. SS-14DKS4, Whitey Co., Highland Heights, OH.

[60] Part no. MMA-077-2QF, Kurt J. Lesker Co., Clairton, PA.

[61] Part no. 00040-VCR-N2, Nupure Inc., Manotick, Ontario, Canada.

[62] Part no. AH75-6MF, Hotwatt Inc., Danvers, MA.

[63] Part nos. CN76000 and F3105, Omega Engineering Inc., Samford, CT.

[64] D. K. Walter, W. M. Griffith, and W. Happer, Phys. Rev. Lett. 86, 3264(2001).

[65] Model no. B1-79-40C-19-30-4, Coherent Semiconductor, Santa Clara, CA.

[66] Model no. Affinity RAA-003A-CB01CB, Lydall Inc., Ossipee, NH.

[67] Model no. MWPS-795-30-4, CVI Laser Inc., Albuquerque, NM.

[68] Model no. PC2000, Ocean Optics Inc., Dunedin, FL.

[69] Part no. XA46-780NB2, Omega Optical Inc., Brattleboro, VT.

[70] B. Larson, O. Hausser, P. P. J. Delheij, D. M. Whittal, and D. Thiessen, Phys.Rev. A 44, 3108 (1991).

[71] T. A. Altes, P. L. Powers, J. Knight-Scott, G. Rakes, T. A. E. Platts-Mills,E. E. de Lange, B. A. Alford, J. P. M. III, and J. R. Brookeman, J. Magn.Reson. Imaging 13, 378 (2001).

[72] B. Saam, W. Happer, and H. Middleton, Phys. Rev. A 52, 862 (1995).

[73] T. Gentile, NIST,private communication.

[74] O. V. Sorokin, Radio Eng. Electron. 14, 1335 (1969).

[75] Handbook of Chemistry and Physics, edited by D. R. Lide (CRC Press, BocaRaton, FL, 2002-2003), Vol. 83, p. 12-130.

[76] J. G. Ganiere, Helv. Phys. Acta 46, 147 (1973).

[77] F. J. Norton, J. Am. Ceram. Soc. 36, 90 (1953).

[78] B. Dreihuys, Amersham Health, private communication.

[79] M. A. Bouchiat and J. Brossel, Phys. Rev. 147, 41 (1966).

[80] A. Suckow, P. Schlosser, H. Rup, and R. Bayer, Glass Technology 31, 160(1990).

151

[81] D. M. Deaton, P. A. Cella, K. C. Hasson, D. Zollinger, and B. Driehuys,Containers for Hyperpolarized Gas and Associated Methods, U.S. Patent6,128,918, 2000.

[82] J. E. Shelby, J. Appl. Phys. 45, 2146 (1974).

[83] V. O. Altemose, J. Appl. Phys. 32, 1309 (1961).

[84] D. Muller, Z. Phys. 188, 326 (1965).

[85] V. O. Altemose, J. Amer. Cer. Soc. 56, 1 (1973).

[86] L. C. Walters, J. Amer. Ceram. Soc. 53, 288 (1970).

[87] D. R. Rich, T. R. Gentile, T. B. Smith, A. K. Thompson, and G. L. Jones,Appl. Phys. Lett. 80, 2210 (2002).

[88] D. Warrington, GE Lighting, private communication.

[89] D. E. Swets, R. W. Lee, and R. C. Frank, J. Chem. Phys. 34, 17 (1961).

[90] http://www.gequartz.com/en/tab11.htm.

[91] L. W. Burch, M. W. Cole, and E. Zaremba, Physical Adsorption: Forces andPhenomena (Oxford University Press, New York, 1997).

[92] E. Cheng, M. W. Cole, W. F. Saam, and J. Treiner, Phys. Rev. Lett. 67, 1007(1991).

[93] http://www.latoscientific.com/borosilicate.htm.

[94] F. Reif, Fundamentals of Statistical and Thermal Physics (McGraw-Hill, NewYork, NY, 1965), p. 485.

[95] J. Shi, University of Utah, private communication.

[96] Handbook of Chemistry and Physics, edited by R. C. Weast (CRC Press, BocaRaton, FL, 1979-1980), Vol. 60, p. D-155.

[97] Product no. 40,178-1, Sigma-Aldrich, St. Louis, MO.

[98] J. L. Robbins, N. Edelstein, B. Spencer, and J. C. Smart, J. Am. Chem. Soc.104, 1882 (1982).

[99] G. A. Johnson, G. P. Cofer, L. W. Hedlund, R. R. Maronpot, and S. A.Suddarth, Magn. Reson. Med. 45, 356 (2001).

[100] B. Saam, D. A. Yablonskiy, V. D. Kodibagkar, J. C. Leawoods, D. S. Gierada,J. D. Cooper, S. S. Lefrak, and M. S. Conradi, Magn. Reson. Med. 44, 174(2000).

152

[101] L. G. Kaiser, J. W. Logan, T. Meersmann, and A. Pines, J. Magn. Reson.149, 144 (2001).

[102] A. Caprihan and E. Fukushima, Physics Reports 198, 195 (1990).

[103] M. V. Icenogle, A. Caprihan, and E. Fukushima, J. Magn. Reson. 100, 376(1992).

[104] L. W. Hedlund, G. P. Cofer, S. J. Owen, and G. A. Johnson, Magn. Reson.Imag. 18, 753 (2000).

[105] Minicircuits, Brooklyn, NY.

[106] Model no. 50B-001, JFW Industries, Inc., Indianapolis, IN.

[107] Model no. AU-1467-100, Miteq, Inc., Hauppauge, NY.