Spin-Polarized Atomic Hydrogen

an experimental study

ACADEMISCH PROEFSCHRIFT

ter verkrijging van de graad van doctor aan de Universiteit van Amsterdam op gezag van de Rector Magnificus dr S.K. Thoden van Velzen, hoogleraar in de Faculteit der Tandheelkunde, in het openbaar te verdedigen in de Aula der Universiteit (Oude Lutherse Kerk, ingang Singel 411, hoek Spui) op woensdag 16 septembXl5.00 uur precies door 1q8f

AUGUSTINUS PETRUS MARIA MATTHEY geboren te Arnhem

1987

druk: krips repro meppel

Promotor:

Prof. Dr. I.F. Silvera Dr. J.T.M. Walraven

Co-promotor:

The work described "Natuurkundig "Stichting

in this thesis was carried of the University financially Onderzoek by the

out at the of Amsterdam (FOM).

Laboratorium" and supported

voor Fundamenteel

der Materie"

Aan Johanna

en Thomas

CONTENTS

Introduction

Chapter

I - Overview states and interatomic potentials surface

3 3

1.1 Atomic

1.2 Interactions

of H with the helium

1.3 Decay processes 1.3.2 Relaxation 1.3.3 Thermal 1.4 Rate equations References

in gaseous and adsorbed hydrogen

10
11

1.3.1 Recombination escape

16 18 19 22

Chapter

II - Apparatus system of the pressure gauge

25 25 32 38 42 47 48

2.1 Cryogenic

2.2 Design and construction 2.3 ESR Spectrometer 2.4 Trigger bolometers 2.5 Thermometry References

Chapter III - Experimental measurements 3.1 Pressure measurement of atomic hydrogen 3.1.1 Apparatus 3.1.2 Equation of state 3.1.3 Surface recombination rate 3.1.4 Results 3.1.5 Field dependence of the surface recombination rate 3.1.6 Implications for BEC

50 50 50 53 54 55 57 57

3.2 Temperature and field dependence of the surface recombination rate 59 3.2.1 Introduction 3.2.2 Temperature dependence results 3.2.3 Field dependence 3.2.4 Deuterium 3.3 Recombination heating 3.3.1 Introduction 3.3.2 Results 3.4 Search for spin waves References 59 59 64 68 69 69 71 72 74

Chapter IV - ESR pumping experiments in spin-polarized atomic hydrogen References

77 97

Summary

98

Samenvatting

100

Nawoord

103

Introduction

Spin to study.

polarized

atomic

hydrogen

is an attractive

substance bound

to create Fermions

and (the as a to

The hydrogen with spin

atom,

consisting

of two tightly with spin

electron composite display even

S = 1/2,

and the proton

I = 1/2) behaves atoms

boson.

A collection phenomena,

of spin polarized the most

hydrogen being

is expected

fascinating

prominent

its gaseous (BEG) and

nature, related hydrogen in a

at zero

temperature,

and Bose-Einstein Apart originate from

condensation nature mass,

superfluid atom, large these

behavior[l]. properties motion, in

the statistical its low atomic

of the

from

resulting

zero point

and the weak 1938 occurs that

interatomic in a

interactions. gas of the Boson lowest was the is

London[2] particles, momentum explored properties generally and BEG, a state

showed

non-interacting density, role

transition

at a critical populated. The

where

is macroscopically of papers

of the interactions dealing Although with it

in a series of

by Lee,

Huang

and Yang[3], spheres. between

a gas of weakly that

interacting

boson exists

believed

some connection

superfluid into

behavior a clear of 2.17 the of

it was never Liquid condenses

possible

to establish

this

connection below

picture. Kelvin, density the

helium-4, under

becoming

a superfluid pressure.

a temperature

its own vapor

In the gaseous phase BEG

phase,

is too low for BEG to occur, interactions low atomic atomic mask

and

in the liquid between

the effects and

interatomic The

the relation

superfluid make of

behavior.

mass

and the weak

interatomic to study

interactions the properties by

spin-polarized a weakly

hydrogen Bose

the ideal

substance

interacting state. atomic it is

gas as the density

is not

limited

condensation

into A

the liquid gas as H2· of

hydrogen highly pointed

of substantial due to

density the

is

not

provided

by

nature, H+H
-+

unstable

recombination this reaction field when

reaction may at be low

As first by His

out by Hecht atoms to await

in 1959[4],

surpressed temperature. and Walraven observable of research, gas by many maximum walls

polarizing suggestion

the had

in a strong realization hydrogen

magnetic until

1979,

Silvera time in

stabilized quantities showing orders

spin-polarized [5]. This

atomic

for the first

experiment

marks

the launching

of a new

field of the the the

that polarizing

the atoms

increased

the stability

of magnitude.

However,

this was

insufficient that

to bring covering

density

into

the BEG range. containing

Furthermore, the sample,

it showed with

of the chamber,

a thin coating

of liquid

helium-4 This

is

essential.

In the following years,

most of the

efforts

were Of

directed towards getting an understanding of the destabilization processes. thesis work may be considered as being a part of these efforts. particular the third liquid, interest was the study of the role of the liquid helium a small fraction of the gas condenses to the Furthermore, third body, form often surface surface, at low the large

surrounding the sample.

As hydrogen atoms are weakly bound to the surface of

temperatures. However, the collision frequency is high due to the lack of the dimension. of in a the the recombination of H towards H2 needs which is available at the Thus the intermediate surface may A in surface step. assistance quantities relaxation,

of helium atoms. appears as an

catalyse systematic as a

recombination of hydrogen atoms towards the molecular state or cause magnetic which has investigation This been carried out of the destabilisation is organized those as follows. In chapter I processes the

function of temperature and magnetic field. thesis theoretical with this to of concepts underlying the experiments will be reviewed. techniques, work. with their particularly Chapter Chapter II deals

that were developed in the course A separate chapter is

III gives a description of the various experiments,

together

results and interpretation.

devoted

microwave pumping experiments using ESR. References 1. I.F. Silvera, J.T.M. Walraven, Progress in Low Temperature Physics

vol. X, D.F. Brewer, Ed., North Holland Publishing Co., Amsterdam. 2. F. London, Phys. Rev. 54, 947; Nature 141, 663 (1938). 3. T.D. Lee, K. Huang and C.N. Yang, Phys. Rev. 106, 1135 (1957) and references therein. 4. C.E. Hecht, Physica 25, 1159 (1959). 5. I.F. Silvera and J.T.M. Walraven, Phys. Rev. Lett. 44, 107 (1980).

CHAPTER I OVERVIEW
Introduction

In this chapter, hydrogen will

the concepts Much

underlying

the stability covered in time later

of spin polarized of the the

be reviewed. investigations

of the material

is the subject chapters. and

experimental appearance size[1-3l, of

to be described at regular

With

review

articles was made

intervals

increasing

no attempt

to aim for completeness. static properties of atomic will hydrogen. be In

The first section 1.1,

two sections single between last

deal with

atom properties the atom

and pair

interactions helium-4

treated. in the in

Interactions section

and the liquid describe a review

surface processes,

are covered limiting and

1.2. The

two sections

dynamical

stability. section a form

Section

1.3 contains contained with

of the relevant

mechanisms, is brought

1.4 the material that may

in the preceding experimental

sections

into

be compared

results.

1.1 Atomic states

and interatomic

potentials

The state

magnetic

energy

levels

of an isolated

hydrogen

atom

in

its

ground

may be found

by diagonalizing

the spin Hamiltonian

(l.l)

The and the

first Iz are

two terms

stand

for the electronic of the electronic magnetic field

and nuclear and nuclear

Zeeman spin

energy,

Sz

the components

operators

along and PN by ge the

direction

of the applied magneton; are

B, PB is the Bohr magneton g factors The with are denoted term

the nuclear and gn'

the electronic both defined

and nuclear positive.

and

as being

last

in

Hamiltonian

represents

the hyperfine

interaction

coupling

constant labeled

(a/h
la>,

= 1420 MHz,
Ib>, Ic> and

a/kB =

68

mK).

The

eigenstates order

of this Hamiltonian, of energy

Id>, are

in increasing

la>

-cos8

1~1> + sin

Itr>

(l. 2 a) (l. 2 b)

[b>

In>

Ic> lei> where t (1)

cosO Itr> + sin 01+1>

(1.2

c)

Itt> state. The

(1. 2 d)

refers to an electron (proton) spin up

hyperfine

mixing parameter e = sinO is given by:

(1. 3)

where

~± = ge~B ± gN~N·

In

high

fields
ecr

(B ~

a/~+ = 506.2
e

Gauss) =

this

expression may be simplified giving and

a/(2~+B). In zero field,

~J2.

In strong magnetic fields and at sufficiently low temperatures, Ib> states are thermally populated. has a mixture referred other greatly enhanced stability against

only the la> and is

Such a spin-polarized two component recombination, The Ib> states and lei> states

to as spin polarized hydrogen (H+).

are pure Zeeman states.

Due to the hyperfine interaction, an amount e of the

Ic> states. The probability of finding a spin up la> state atom is given by e2. In a 10 Tesla field, e2 has a value of 6.4*10-6. Complete electron spin polarization in a strong plays a field key is therefore inhibited due to this admixture. role in the recombination of two hydrogen This admixture the atoms towards

electron spin orientation is mixed into the la> and

molecular state. The energies of the atomic states are given by

(1.4

a)

+(a/4) [1 - 2(~-B/a) 1

(1.4

b)

(1.4 c) +(a/4) [1 + 2(~-B/a)1 The energies are shown in fig.

(1.4 d)

1.1 as a function of magnetic field, together

with the allowed ESR and NMR transitions.

Id>=I1t> Ic>=cos e m>+sine Ib>=IJ.i> IOO>=~ lITt>-IU>J 10>=sine ITt>- cos e IH> In>

Fig.

1.1

Hyperfine

levels of H in a magnetic field.

Also

shown

are

the

possible electronic (S), nuclear (N) and longitudinal (F) magnetic resonance transitions (after Silvera and Walraven[lJ). The interaction between a pair of hydrogen atoms is represented by an

effective Hamiltonian of the form

(l. 5)

where Hn and Hex are the direct and exchange interactions between the and much Hexch Hdd is the interaction due to magnetic dipole forces. weaker were than the other terms. made by Kolos and Wolniewicz[4].
=

atoms, is + the the

The last term on in

Accurate ab initio calculations of Hn The potential depends two potentials;

total spin of the electrons S hyperfine interactions,

sl + s2' may

Leaving aside the weak intra-atomic

one

distinghuish

electronic ground state this leads to the l~~ or singlet potential VS' if the electronic spin state is antisymmetric (S = 0), and the 3~~ or triplet potential VT for the symmetric electronic spin state (S = 1). The singlet and triplet potentials are shown in fig. 1.2.a and 1.2.b. The singlet potential has a bound state with a dissociation energy of 36ll8.6±0.2cm-l [14,15] making a sample of atomic hydrogen highly unstable against molecule formation in zero magnetic field. at a bound state. The triplet potential has a shallow minimum of 6.46 Kelvin support an interatomic separation of 4.16 In a magnetic field,

A,

which

is

insufficient

to

the degeneracy of the

triplet

state is lifted, and one of the triplet potentials is partially shifted below the singlet potential. Polarization of the electronic spins in a strong field in combination with via sufficiently low temperatures, the triplet potential. This forces the atoms to interact pairwise strongly surpresses
5

t
~
c.!)

1
~2
E

3 B= 10 TESLA

®
Ms= 1

>2

a:
UJ Z UJ
...J

a:
,,-

> c.!) 0

0
/

;---

UJ Z UJ
...J

UJ l-

-c ~-2
I ,
'
\ ~I I

I
I

/U
9

<
~-1
UJ I-

Ms= 0

Q.-4

I\ I

1~+

Ms=-1

Q.-2

HI,r-: 1~+9
4

5 SEPARATION

NUCLEAR
Fig.

[AI

1.2

Interatomic line)

potentials

for a pair of hydrogen

atoms in the singlet (a) and in a

(dashed

and triplet

(solid line) states in zero field scale in fig. 1.2.b

field of 10 Tes1a (after Silvera

(b). Note the expanded

and Wa1raven[11).

recombination. In terms of the triplet and singlet potentials

VT and Vs one may write:

(1. 6)

where VD The magnetic dipole-dipole interactions are represented by the Hamiltonian:

(1. 7 a)

where

f(s,i) = s.i - 3(s.rl2).(i.rl2)

- '"

- '"

(1.7 b)

Here r12 is a unit vector.

The nucleon-nucleon term is small compared to the

electron-electron and electron-nucleon interactions; usually it is neglected. 1.2 Interactions The high gas. the role of of H with the helium the surface surface to be of critical importance in the

proved

stabilization experiments. In

If the atoms are strongly bound to the surface, a which severely limits the stability of 29] it was established that is essential. Liquid for covering helium atomic

surface density builds up, sample an and

the early experiments [5, 2B, almost ideal surface:

chamber walls with liquid helium

provides hydrogen

it has a low binding energy Hydrogen

is essentially free from contaminations.

molecules,

which are the product of the recombination reaction, bulk liquid [6], the liquid.

were found to enter the as it reveals is

eventually becoming a part of the solid substrate underneath

The adsorption isotherm is of primary importance,

the role of the surface in attaining BEG. The basic ingredient for the binding of a hydrogen atom to the liquid the H-He Van der Waals potential. nature of the problem, the known. follow Theoretical the A calculation of the binding energy has to Neither and a

take into account the surface density profile of the liquid and the many body in particular the H-He density correlations. surface density profile, nor the H-He correlation function is precisely Guyer used the and De Simone and Maraviglia [9] They Here we

estimates of the binding energy were made by

Miller [7]

Mantz and Edwards [B]

calculations of Mantz and Edwards in some detail. applied to

variational approach, succesfully

based upon Feynrnan [30] and Lekner [31], that had been predict the binding of helium-3 atoms to free In this method, knowledge of the ground state wave atoms with same is As the rl

surface of helium-4 [32]. function true He-H

wo~rl,r2' .~....,rN) is assumed of a system of N helium correlation function is unknown, as the helium atom.

coordinates rl' ....,rN' The hydrogen atom is treated as an impurity. correlations The helium atom with coordinate

it is assumed to have the

replaced by a hydrogen atom, and a trial wavefunction of the form (l.B) is used to minimize the energy by variation of yields a single f(rl)' The minimalization for the motion 7

procedure

particle SchrBdinger equation

Fig. 1.3.a Calculated profiles hydrogen

density on

for 3He and the isotopes adsorbed

the liquid 4He surface. Also shown is the 4He density distribution
20

p/po.

b H

Fig. 1.3.b Effective for Hand Edwards 3He [8]).

potentials

(after Mantz and

0 Veffiz) i~

~
-5 20 z~

perpendicular to the surface: (1.9) where m is the mass of the hydrogen atom, the bound in particle wavefunction; liquid surface the z the distance to the surface, an effective atoms Ez

state energy and ~(z) = f(z).(p(z)/po)~ is ground state.

single the the

p(z)/po is the normalized density distribution of the

The motion of the hydrogen along The effective potential Veff depends on

is free particle like.

one particle, ground state, hydrogen, hydrogen,

two particle and kinetic energy densities of the liquid in the and the H-He and He-He potentials. The resulting potential for with the density distributions for the hydrogen isotopes a binding in energy the of 1 of 0.63 Kelvin. the In view of the the 1.3 a and b. The potential yields a single bound state for calculation, agreement with

together having

is shown in figs. uncertainties experimentally

involved determined

value

Kelvin is reasonable. The adsorption isotherm is then

In the low density limit, the effects of Bose statistics and interactions between the hydrogen atoms may be neglected. given by 8

(1. 10)

where

ns

is

the density at the surface, is the thermal de

ny is the density
wavelength.

in The

the

bulk,

Ath ~ (2nn2/mkBT)~

Broglie

isotherm,

neglecting interactions between the hydrogen atoms yields unrealistic results at high surface densities. density also the diverges condition for Due to the effect of Bose statistics, the surface As ~ BEC in if the chemical potential ~ approaches zero. the onset of BEC in the bulk,

is this

noninteracting two phase system is impossible. the effect and

The inclusion of interactions Calculations on out by be linear density

between the hydrogen atoms changes this picture drastically. Silvera Goldman [10] and Edwards and Mantz[ll]. the In a

of interactions on the adsorption isotherm were carried

expansion, written

chemical potentials in the volume and at the surface may

~vO + 2 a3 ny, and ~s

(1.11)

Line of Critical Coverages for 2 Dim. Superfluidity

N I 2
b

?:
·iii

1012

c
GO

Adsorption Isotherms H+ on 4He Ea/k =O_96K

Bulk

15 10 Density

n (cm-31

10

20

Fig.

1.4 Adsorption

isotherms

of

on liquid

4He

(after

Silvera

and

Wa1raven[lJ).
9

where ~v and ~s are the chemical potentials in the volume and at the surface, ~vO and ~sO are the chemical potentials neglecting interactions and Q2 and Q3 are the linear The coefficients in the density expansion isotherms are shown in of the 1.4 interaction for several energy. resulting fig.

temperatures. cm-2 and below, surface is

Interactions may be neglected at surface densities of 5 depending on temperature; saturated at an estimated density of about 1014cm-2 for

*
a

1012

at BEG conditions in the gas

surface and 3.l0l3cm-2 for a 3He surface[ll.

1.3 Decay processes

in gaseous and adsorbed hydrogen

A sample of atomic hydrogen has a finite lifetime due to the formation of hydrogen is molecules. Recombination is the last stage in a reaction chain, which also may involve one or more relaxation steps. Sometimes the decay rate influenced by transport of atoms to a certain region in the experimental a system, where recombination is likely to occur. In the case that one of these steps is rate determining, the density decay rate becomes proportional to simple power (r) of the number density (n) of the gas: dn/dt Here K is the rate constant and r corresponds to the number of (1.12) hydrogen

atoms, that participate in the rate limiting reaction (r = 1, 2 or 3). Recombination literature reaction, stabilizing relaxation may for the of atomic hydrogen has been widely covered in the chemical Recent studies of find the best this elementary towards of wealth by the need to strategy during the past fifty years. encouraged high

densities at low temperatures,

has revealed a

possible new routes to the molecular state.

Thus far,

more than one hundred These processes is responsible the the in

and recombination channels have been identified.

be classified according to the type of interaction that that

reaction (exchange or dipolar interaction) or to the nature of catalyses the process (a helium atom or hydrogen atom or the liquid helium surface).

species

gaseous phase,

In the coming section a brief

description will be given, with some emphasis on processes that are effective at moderate densities (_1014_l0l6/cc).

10

1.3.1 Recombination In always practice needs the formation of a hydrogen molecule from a pair either in a single of atoms ternary

the assistance of a third body,

collision or two consecutive binary collisions. The two body process H + H ~ H2 + hv requires the release of the dissociation energy by the emission of radiation. This in however is prohibited, the molecule. to due to the absence of a permanent dipole moment of be momentum. It may The role of the third body is to provide extra degrees

freedom useful

obey the conservation laws of energy and

to spend a few sentences on standard room temperature

recombination.

In this case, the reaction proceeds in two steps. A possible mechanism is

H + X ~ (HX)*

where X F H. last stage

In the first step, is an exchange

a metastable molecule (HX)* is formed; The dominant process at

the

reaction.

ambient

temperatures is resonance recombination, also following a two step scheme:

where process the energy The as

X=H is

or some different species. an

In the first

stage, is

metastable This between energy Kelvin, a

molecule

having

energy above the dissociation

limit

formed.

resonant in the sense that it requires a close matching The quasimolecule may be stabilized by

incoming energy of the recombining atoms and the of the quasimolecule.

vibrational-rotational

transfer, if it collides with some particle X before it can dissociate. exchange mechanism is not effective at temperatures below 1 the helium isotopes, the only relevant candidates for X, do not form

quasi bound state with H. resonance

Several authors explored the possible relevance of The majority of the 11

recombination at temperatures below 1 Kelvin.

0 -20 -40

.=3

v=6

v=ll

H2

v=12

.=1·3

v=14 J=4

.=20

Dz .=21

.J=1 ':1=0

T:27
J=24

J=lO J=5

-60
~ -80

i;j

J =7

J=3 J=4

~ ~

""

--100

~-120 ~-140
CD

J=2

J=3

-160 -180 J=l -200 J=12 j;() J=l J=O J=2

Fig.

1.5

Vibrational-rotational limit (after Silvera

energy

levels

of the H2 molecule

near

the

dissociation

and Walraven

[1]).

quasi-molecular states, which are important at high temperatures, turn out to be thermally inaccessible in the low temperature regime. In fig 1.5 we depict some vibrational These rotational levels with an energy near the by dissociation applying by of a by the the that was is limit. magnetic Stwalley bound molecular states may become quasibound

field. This is due to triplet-singlet mixing. [12] some of these states may be tuned above of a magnetic field, those pairs, shifting the MS

As pointed out threshold asymptote

application be fulfilled

-1

triplet potential below the corresponding level. The resonance condition will for having a kinetic plus The Zeeman J~lO energy level however from coincides considered This state to expected of the with the energy of the bound state. this context by Kagan, v~12,

within be

Vartanyantz and reason. of

Shlyapnikov[13]. energy the

is accessible at fields>

14.6 Tesla.

This channel Apart mass

unimportant for the following After separating out the

constraints, also the existence of a centrifugal barrier governs the dynamics process. center motion, SchrBdinger equation for the pair takes the form

12

(l.13) where ~ = m/2 is the reduced mass of the pair, r = Ir1 - r21, and VCr) may be the which singlet is potential or the triplet potential. The effective The potential an effect term values, between provides the two atoms depends on the rotational quantum number J, characteristic for any central a rotational barrier, potential. which may be appreciable at high J

J-dependent

preventing an encounter, close enough to form a molecule. The v=14 , J=4 level was considered in detail recently [14] in relation to experimentally observed rates. From spectroscopic measurements it is known to be marginally bound, 1imit[15]. having The an energy of .26(46) Kelvin below fields the but of dissociation opens 0.52 level is inaccessible at zero fie1d[16],

a channel to recombination for pairs of la> state atoms at Tes1a and above.

The quasimo1ecu1e is calculated to have a lifetime of In the limit, where the loss of atoms

10 microseconds before it dissociates.

is bott1enecked by the formation of (H2)*, the rate constant is given by [14] (l.14) where ~ is the Zeeman energy of the pair and D14,4 the dissociation energy of the quasimo1ecu1e. The exponential threshold factor reflects the energy constraints, consistent a posed by the resonance condition. Experimentally observed rates with this relation were observed at relatively high temperatures, The experimental It is data also

field of 4 Tes1a and densities of order 1015/cm3. accurate than the

yield a value of 0.7 (0.1) Kelvin for the dissociation energy D14,4' which is more spectroscopically determined value. possible to give an improved value of the ground state dissociation energy DO,O of 36118.6 (0.2) cm-1, as the source of uncertainty in DO,O results from the uncertainty in D14,4' The dominant process at low temperatures is the direct process,

involving three bodies in a single encounter: H + H + X ~ H2 + X where X may be either H in the gaseous phase or in a surface bound state, a gaseous helium atom or the liquid surface itself. The rate constant or for 13

bulk

recombination and

in a background [16]. In this

gas of helium-4 case,

was

calculated decays

by

Greben, to

Thomas

Berlinsky

the density

according

(l. 15)

where

nH

and nHe

are the hydrogen state

and helium is enabled

densities

respectively. interaction.

The The

transition process

to the molecular is sometimes with the Van

by the exchange

referred

to as Van

der Waals

recombination,

due to the

interaction Both interaction electrons dependence the rate S

the third der Waals

body. interaction with atoms spin the third conserve I. As a body and the spin exchange of all the spin for be

between and the

the recombining total nuclear

the total

consequence,

may be factorized constant. correct:

out of the matrix simplistic

elements turns

in the expression therefore out to

A rather

picture

substantially in the presence character relatively without

the probability third

to recombine body

for two colliding related

atoms

of an appropriate into

is simply

to the singlet It is therefore this process, three-body

mixed easy going

the electronic

spin pair wavefunction. field full dependence quantum for

to derive into

the magnetic of a

the details

mechanical

h.h, h.h,

aa
E1j

ba ab

ea ae l(1j' - £')

da ad

bb

eb be

db bd

ee
-£'1

de cd ±ae

dd

S '0 0 0 0

M, 0 0 0 0

M/ 0 1 0

+-0:1]

+!
+Ctt:

+1

-,
+°11 ±a£ ae
±Ul1
a1j

0 0

HI
+!

'I ±~
-a1j

'1 '1 '1 0 0 0 0 0

+ + + + + + + + +

-,

=r»
«e
±0'11 'I
-a1j

ae
E1j a1j

_ j(1j' _ £2)

'1 '1 '1

0 1 0

+af
-F1j

a£

'1 '1 '1

0 1 -

Table

1.1

Decomposition spin

of hyperfine

pair

states

of hydrogen

with

respect [1J) .

to

the total

representation

\S HS I HI>

(after Silvera

and Walraven

14

approach.

One merely basis,

has

to decompose both

the hyperfine S and I are

pair

states The

with

respect

to a convenient of the pair out and may

in which respect

specified.

decomposition is worked to para-H2 to the rate

states 1.1.

with The

to the total

spin basis

IS MS I MI> correspond pairs

in table ortho-H2

10 0 0 0> and The

10 0 1 -1> states

respectively. via

contributions dependent

of the various fO and f1:

be expressed

the field

factors

where

na,

nb'

nc and nd denote rate

the fractional to the para

populations

of the to

hyperfine fO' and of

states. the ortho

The recombination rate

state

is proportional

is proportional

to fl'

We will

discuss

the characteristics

the process In (na = nb fl = 3/16. case

with

the aid of the above with equal

expressions. of a has the four lead hyperfine to states and

zero

field

occupation 1.16

nc = nd = ~) In a strong population

equations field

and b

fO = 1/16

the process

the following states Thus, 1/(2€2),

feat~res.

In the the of a 105 in

of equal

of the lower hyperfine to fO = fl = ~€2

(na = nb = 1/2) the application which amounts to 1/B2,

expressions strong field

for fO and fl lead will suppress

the rate by a factor the rate will A collision

a 10 Tesla

field.

Furthermore,

be proportional between two

which

may be verified atoms cannot

experimentally. lead

Ib> or two is of

Id> state particular states. state. be with

to the formation contains Ib>,

of a molecule. atoms

This

importance In a 50/50 Thus, lost, regard a in

if the sample mixture of

only

in the lower

hyperfine the la>

la> and

75 % of the rate all la> state which

depletes atoms is

the absence

of relaxation, Ib> state

will

eventually polarized

leaving to both

a gas of only electron

atoms,

doubly stable.

and nuclear always

spin

and completely due to dipolar rate. [17] This and

In practice, and the was by

small

la> density will

will decay

exist

relaxation,

sample

density by

at a reduced Berlinsky

relaxation later

bottleneck confirmed

predicted

Statt

and

experiments[26,27].

The

surface

analog

of the above as pointed

recombination Silvera have

process

was never [1]

studied the twoThe

theoretically. dimensional

However, process

out by

and Walraven a lot in

and

the

bulk

process

common.

15

interaction arguments

with we used

the third before

body

is still

of a nonmagnetic field

nature.

Thus, also

the apply we

to derive In default

the magnetic of a detailed model

dependence,

to the surface use data. radius a simple The

process.

theoretical to analyse gas

description, the

two-dimensional frequency by

gas kinetic

experimental with a

collision is given

in a two-dimensional

of disks

00

(l.17)

where density. section. same The

is crosslength (1.17)

the

average

velocity

of the

disks analog of a

the crossis the

200

is the two-dimensional assumed that

In equation

it has been

the cross-length order

for a-a

and a-b collisions. with a rate

The decay

is of second by

in the hydrogen

surface

density

constant

Ks given

(l. 18)

where leads event that atoms

f(B) = fO + fl is the field to the formation results there

dependent The

probability

that per

the

collision

of a molecule. two

loss

of two atoms for Ks'

recombination assumed if two

in a factor

in the expression in the form

It has been atom present

is always within

a third

body

of a helium

approach

a distance

200'

The [24] atom In

currently this

observed

maximum

density

in the gas body

is of

order with a

1018/cc hydrogen between three singlet will not the

density body

regime becomes

the direct dominant.

three

process of the

as the third relaxation The in

As a result

interplay involves

various

and recombination dipolar interaction

phenomena, with

this process atom

Ib>

atoms.

the third

introduces We

character discuss scope

the interaction in detail,

between

the recombining density

atoms. regime

this process thesis.

as the high

falls

outside

of this

1.3.2 Relaxation

Spin-exchange process were

scattering fields.

of hydrogen

atoms

is the

dominant

relaxation sections [21],

in low magnetic

Low-temperature

spin-exchange

cross

calculated a

by Berlinsky of about

and Shizgal 1

[20], and Morrow The field

and Berlinsky of

finding

value

A2.

dependence

exchange

16

depolarization was studied by Kagan, Vartanyantz and Shlyapnikov[13]. To find the may channels which contribute to the process and their field dependence, follow the same line of reasoning as in the case of Van der The allowed channels all involve a single or double we Waals and the state the at The

recombination. spin flip. process Zeeman factors very strong process is of

electron

In strong fields the dominant rates vanish as €2 - 1/B2, inelastic due to the difference in the initial and The rates are therefore further surpressed by

final

energies.

exponential

the form exp [2~BB/kBT]. An exception to these rules forms Normally however, the process is unimportant are completely empty.

ac~bd or bd~ac channel, fields, as

which is independent of field and elastic, and has a

large rate constant.

the upper hyperfine states

is of importance in the case of ESR experiments,

where a small non-

equilibrium population of the Ic> or Id> state is created. This point will be worked out in detail in chapter 4. In strong fields, relaxation in an almost pure Ib> gas, the rate is bottlenecked by Relaxation results either from

due to magnetic dipolar forces.

pair collisions or from the interaction with magnetic impurities at the walls of the sample container [25] Impurity rendered helium relaxation gives rise to a first order decay, and may be

unimportant by cleaning the cell of magnetic impurities or film. The their process was investigated by Statt experimental results with a et al.

coating who an

the interior of the cell with solid H2 in combination with a thick superfluid [22,23] interpreted model, predicting

exponential dependence on coating thickness. Two-body dipolar nuclear relaxation has been widely studied, cell is clean of magnetic impurities, this thesis work. [26,27] as it is of critical importance for the stability of the gas. It is only important if the a condition never met in the course of of the The surface variant of this process has been a source and theoretically predicted rates. Re-analysis of

confusion for several years, as there was a factor 50 discrepancy between the observed observed At The be rates[24,25] and an ESR study [34] then revealed that the the upper hyperfine states observed may be to

decay was due to a surface three body recombination process. high temperatures or low fields, thermally populated by electronic dipolar relaxation via two-body collisions. rate constant for this process grows exponentially with (~BB/kBT) as expected for a thermally activated process involving a single electronic 17

spin flip. 1.3.3 Thermal The placed effects, escape is

sample chamber used for the experiments described in this thesis in the center of a superconducting magnet. Due to fringe atoms in the la> and Ib> hyperfine states are driven to the

field field

center by a force of the form

F
where ~B is the magnetic moment of the atom and 8B/8z is the field assuming a uniform chemical interactions between the

(1.19) gradient In be

along the field axis. The density distribution may be found by evaluating the grand-canonical sum over states, the low density regime, where potential. atoms may

neglected, this leads to a density distribution of the form: n(z) where tail tube, zero. the

n(O) exp [~B(B(z) - B(O»/(kBT)]

(1.20) the

z is the coordinate along the field axis and z = 0 corresponds to In practice, our sample chamber of the density distribution may then extend into a zone of the

field center.

is an open ended system. The filling

where no protective helium coating exist and where the field is almost Atoms with sufficient velocity in the proper direction may escape from of the field to this zone in a single shot, and recombine

center

immediately.

The system will therefore behave as a volume V, connected via a unprotected by a helium coverage starts. Hence, even in the

tube to a perfect vacuum located at z > zd' where zd corresponds to the point where the zone, absence of recombination within the sample container, according to a first order process: dn/dt with a decay time the density will decay

-niT

(1.21)

= 4 CMV/KvA where V is the volume of the sample chamber, is the magnetic compression factor, For a K the typical

CM = exp[~B(B(zd) - B(O»/(kBT)] Clausing

factor and A the cross-sectional area of the tube.

geometry this leads to a time constant of about 100 seconds. Much longer time 18

constants of H+

were [5]

observed

in the first

successful study helium

attempt

to stabilize escape the

a sample [19] The tube,

and a following was attributed

systematic to refluxing

of thermal vapor in

difference retarding by time

filling

the motion

of the escaping compression a value who

atoms. factor

This

effect

may be

incorporated for the

including constant,

an additional which has

CHe in the expression

In view of the recent probabilities

of about

80.

results

of Berkhout wall

et al.

[33],

observed

substantial

for specular be even

reflection, than

the compression estimated.

efficiency

of the He-vapour

should

higher

originally

1.4 Rate equations

The

decay

of the densities coupled

of the four hyperfine first order differential temperatures states is

states

is governed

by a

set of four hyperfine a non

non-linear

equations.

The upper or if

states

are only occupied population Often

at high

or low fields, created by

equilibrium radiation.

of the upper upper

resonant neglected,

microwave leaving

these

states

may be completely

a set of two coupled we discuss

differential of these

equations. equations quantities for some such important cases, area and

Before we first volume, contribute by the

the form

introduce

the use of effective surface rate.

as effective the adsorbed are linked the

and effective

Both

the volume in both

and

phase together surface of the total sample

to the decay. adsorption

As the densities (1.10), rate

phases

isotherm

it is possible of volume only

to specify densities. small

contributions the atoms number sample is of

to the total chamber

in terms that
=

The volume of the the

is large

enough N

a very

fraction N is of

in the adsorbed atoms,

phase:

nyV + nsA '" nyV where

area the volume and

is the volume ns are the

and A the surface the

container, respectively. the volume

ny

and

surface a rate

densities Ks'

For a surface decays

decay

process to

of order

r with

constant

density

according

d(N/V)/dt

(1. 22)

Thus by

one may

translate

Ks into an effective

surface

rate

constant

K~ff,

given

(1. 23)

19

However, by the

for our experimental conditions,

the sample distributes itself and be an

over an inhomogeneous field region, the density distribution n(z) being given equation (1.20). expressions by for To account for this effect both the rate equations the rate constants for the various n by the central density n(O) processes and may introducing

modified effective

replacing

volume Veff and surface area Aeff.

The density distributes itself

over an effective volume Vo' defined by

N

(1.24)

Integrating the density distribution over the sample chamber yields

J~
where by the relation dN/dt where central r is

exp [~B(B(z) - B(O»/kBT]

dz

(1.25) dV/dz.

the real shape of the sample chamber is taken into account via

With a surface rate constant Ks an effective area Aeff is associated, defined

(1.26a)
rate constant at the

the order of the process and Ks is the Analogously,

field.

the effective volume associated with a

volume

rate constant Kv is defined by dN/dt or dn(O)/dt

(1.26b) (1.26c)

Integrating the rates over the volume or surface of the sample chamber yields the following expressions for Aeff and Veff:

J~
J~
temperature,

{Kv(B(z»/Kv(B(O») {Ks(B(z»/Ks(B(O»)

exp [r ~B(B(z) - B(O»/kBT] exp [r ~B(B(z) - B(O»/kBT]

dz dz

(1.27a) (1.27b)

Thus, every process has its own definition of Aeff or Veff' determined by the density and field dependence. If the field distribution has a

20

parabolic

shape,

the variation

of Aeff

and Veff

with

field

or temperature often

can has

be substantial sufficient

[27].

In practice

however,

the field

distribution are small,

homogeneity

that variations

in Aeff

and Veff

typically

a few percent. Note very that for some processes This is particularly factor the effective the case volume or surface constant area may be contains is a of in

large.

if the rate

magnetic second

threshold order, then

of the form exp[-2~BB/kBT]. factor cancel, and part giving

If the process

the threshold and (1.27b)

of the exponential

factor

equations density equal not a

(1.27a) tail

the contributions

from

the low center this will is be

of the density The above

distribution

and the density in this effective the center case. area

at the field In practice, or volume

weights. serious

integrals since

diverge

problem,

the actual from

bounded field and

due to transport region.

of particles

of the field then no longer

to the low applies, rates,

The equilibrium state distribution and drift

density

distribution

a steady

is set up,

determined gradient. these

by the various As an accurate are best

including of Aeff

diffusion or Veff

in the field in this case,

estimate studied

is difficult

processes

in a closed

geometry. this chapter, we discuss the form of the rate process equations (see for

To terminate the case that

the atoms

are lost by the direct This 10 form

surface

section for those

1.3.1)

and

thermal where low

escape.

is thought Furthermore

to be appropriate the is are densities unimportant.

experiments sufficiently shorthand

BIT
that n for

T/K.

should Using

be a

third-order

recombination equations

notation

ny(O) ,

the rate

(1.28a)

(1.28b)

where

K~~f

and K~£f G~£f

are the rate

constants

for recombination relaxation,

via

a-a or

a-b

collisions, surface

is the rate

constant

for dipolar constant

either

at the by

or in the volume,

Gimp

is the rate

for relaxation escape.

caused

magnetic'impurities,

and r is the time constant

for thermal

The rate

21

equations relaxation. from container. thermal

may be simplified Fast relaxation, In this

in two extreme compared of magnetic

cases,

fast relaxation

and

slow

to the recombination impurities of both hyperfine

rate, may result states reduce are to (l.29) become by the in

a large concentration equilibrium

at the walls of the sample

case the populations

(na = nb for T > 100 mK) and the equations dn at

= _

(1/2) (Keff + Keff) n2 aa ab

If

the

relaxation relaxation:

doubly polarized dipolar

is slow (Gimp = 0), then the gas eventually will (na/nb ~ 1). In this case, the rate is bottlenecked

at
This may be verified

dn [25]

(l. 30) dldt (na/nb) = 0 chapters. More complicated

by requiring

cases will be treated References l. 1.F. Silvera Physics", 2. T.J. Greytak Hardy,

in the experimental

and J. T .M. Walraven, and D. M. Kleppner,

in "Progress

in Low Temperature Amsterdam in Atomic (1986). Physics", Physica

Vol. X, D.F. Brewer,

ed., North-Holland, in "New Trends (1984).

North Holland

Publ. Co., Amsterdam Morrow,

3. W.N.
109

R. Jochemsen,

and A.J. Berlinsky,

&

110B, 1964 (1982). and L. Wolniewicz, and L. Wolniewicz, J. Chern. Phys. 49, 404 (1968). Chern. Phys. Lett. 24,457 J. Mol. Spectr. (1974). 54, 303 (1975).

4. W. Kolos W. Kolos,

W. Kolos and L. Wolniewicz, 5. I.F. Silvera 6. I.F. Silvera, 8. I.B. Mantz

and J.T.M. Walraven,

Phys. Rev. Lett, 44, 164 (1980).

Phys. Rev. B 29, 3899 (1984). Phys. Rev. Lett. 42, 1754 (1979). Phys. Rev. B 20, 4518 (1979). Chern. Phys. Lett. 60, 289 (1979). Phys. Rev. Lett. 45, 915 (1980).

7. R.A. Guyer and M.D. Miller, and D.O. Edwards,

9. C. De Simone and B. Maraviglia, 10. I.F. Silvera 11. D.O. Edwards 13. Yu. Kagan, 22 and V.V. Goldman, and I.B. Mantz,

J. de Phys. 41, C7-257 (1980). and G.V. Shlyapnikov,

12. W.C. Stwalley,

Phys. Rev. Lett. 37, 1628 (1976).

I.A. Vartanyantz

Sov. Phys. JETP 54, 590 (1981). 14. M.W. Reynolds, 15. I. Dabrowski, 16. J.M. Greben, I. Shinkoda, R.W. Cline and W.N. Hardy, (1984). Phys. Rev. B 34,4912 (1986). and A.J. Ber1insky, Phys. Rev. Lett, 45, 2105 (1980). by W.N. Hardy. Phys. Rev. Lett 44, 168 (1980).

Can. J. Phys. 62,1639 A.W. Thomas 945, (1981). suggested

Can. J. Phys, ~,

17. B.W. Statt and A.J. Ber1insky, This effect was originally 18. J.T.M. 19. J.T.M. Walraven Walraven,

and I.F. Silvera, I.F. Silvera, and B. Shizga1,

A.P.M. Matthey, Can. J. Phys. 58, 881 (1980). Can. J. Phys. 61, 1042 (1983). and E.J. Klein, (1985).

Phys. Rev. Lett. 45, 118 (1980). 20. A.J. Ber1insky 21. M. Morrow Journal and A.J. Ber1insky,

22. B.W. Statt, W.N. Hardy, A.J. Ber1insky of Low Temp. Phys. 61,471 23. B.W. Statt, A.J. Ber1insky

and W.N. Hardy, D. Kleppner and T.J. Greytak,

Phys. Rev. B 31, 3169 (1985). 24. H.F. Hess, D.A. Bell, G.P. Kochanski, Phys. Rev. Lett. 52, 1520 (1984). 25. R. Sprik, J.T.M. Walraven, 26. R.W. Cline, T.J. Greytak 1195 (1981). G.H. van Yperen and I.F. Silvera, 27. R. Sprik, J.T.M. Walraven, 28. W.N. Hardy, M. Morrow, R.M. Marso1ais, G.H. van Yperen and D. K1eppner, and I.F. Silvera, Phys. Rev. Lett. 47, Phys. Rev B 34, 6172 (1986).

Phys. Rev. Lett. 49, 153 (1982). R. Jochemsen, B.W. Statt, P.R. Kubik, A.J. Ber1insky and A. Landesman, and D.A. Smith,

Phys. Rev. Lett. 45, 453 (1980). 29. R.W. Cline, T.J. Greytak, J. de Phys. 41, C7-l5l 30. R.P. Feynman, 31. J.Lekner, D. K1eppner, (1980).

Phys. Rev. 94, 262 (1954). 1278 (1971). R. van Roijen, J.T.M. Walraven, F. Bridges

Philos. Mag. 22, 669 (1970). E.J. Wolters, I. Shinkoda,

32. W.F. Saam, Phys. Rev. ~, 33. J.J. Berkhout, 34. M.W. Reynolds,

Phys. Rev. Lett. 57, 2387 (1986). W.N. Hardy, A.J. Ber1insky, and B.W. Statt, Phys. Rev. B 31, 7503 (1985). 23

1 3He pumping 2 3He return 3 3He condenser 2 5

1

tube line

4 Flow impedance S 4He pumping 6 1.2 Kelvin 7 Film burner 8 Still 9 Continuous 10 Heatpipe 12 Copper 13 Cooling Thermal 14 Cooling lS Discrete (spiral) condensor heat exchanger 11 Cold plate sinter pole for Platform pole for far heat exchanger line 4He bath

3
7

8
4 9 11 14

10
12 13 15 16 17

21

infrared bolometer 16 Mixing chamber shield shield shield

22 23 33 24 32
31

17 Still temperature radiation 18 4.2 K radiation 19 77 K radiation 20 Outer case 21 FIR bolometer 22 HSC pinning 23 Mylar window 24 Pressure 2S Thermal gauge Platform

18 19

20

26 HEVAC with heatpipe evaporator 27 Accornrnodator 28~28 77 K pinning 29 Teflon valve

30

30 To microwave 32 Quadrupole

source coils

31 7.S T main coil 33 2S0 G + SO G coils

24

CHAPTER

II APPARATUS

Stabilization the development to be atoms. taken

of atomic

hydrogen

and the study

of its properties cooling

requires have

of non-standard the large

techniques. amounts such

Special

precautions

to handle

of heat, as pressure techniques

produced

by the recombining calorimetric to measure the

Suitable using

detection bolometers

schemes,

measurement, are useful

detection density This

and resonance

of the gas. chapter deals with experimental of this techniques, work. 2.1. The pressure bolometric with The emphasis on system those and

developments stabilization is devoted

that cell

were

part

thesis

cryogenic

are described detection resonance (section

in section techniques: (section

remainder

of the chapter (section (section

to various spin

measurement detection

2.2), 2.4)

electron and

2.3),

thermometry

2.5).

2.1 Cryogenic

system

An overview Hydrogen The (Oxford ~W

of the apparatus Cell were (HSC)

is shown

in figure line

2.1.

Figure

2.2 shows

the

Stabilization experiments Instruments).

and the fill out

in more

detail. refrigerator capacity rate with [1] of of 500 1000

carried

in a commercial has a maximum

dilution cooling

The dilution temperature cell

unit

at a mixing The

chamber sample and In

of 100 mK and a circulation to study atomic et hydrogen al.

~mol/sec. spin

is adapted

electron some

resonance,

is the one used by van Yperen geometry, It consists length: the cell is part tube

with

modifications. guiding outer mylar which the

this

of a light (inner sealed

pipe

system, 5 mm, a

the microwaves. 9 mm, which

of a copper and

diameter:

diameter: window, surrounds cell At

20 cm),

is vacuum

at the top with copper

is transparant part

to the microwaves. pipe,

A heavy

bar, on

the upper

of the light This bar this joint

transports

the heat

load

to the mixing low

chamber.

is soldered

to the cell with and has it becomes reduced The screw to copper

Woods a low

metal. thermal and This

temperatures,

is superconducting of 100 Gauss has been

conductivity.

In a field state

of the order after

normal zero. bar is

remains hysteresis

in the normal effect

the field

results

from

the freezing-in by a screw

of flux. This

is connected

to the mixing

chamber

contact.

contact

Fig. 2.1 (opposing page) Cryogenic

apparatus

25

(A)

(B)

TO CELL

INSERT

r+II>

___

MYLAR WINDOW

I----ACCOMM

ODATOR

:r:

"0

..
P

= ~.::

PRESSURE GAUGE

_3
~ .....
I

--~:i=:::=::
_ 0

TRIGGER

= "'''"''''''
f----l 1 em

=lr
(A) Hydrogen (B) Filling

ROOM TEMPERATURE H-SOURCE

Fig.

2.2

stabilization system

cell.

26

the

weakest

link in the thermal connection of the sample container With a heat load on the cell of 40

to

the the the

mixing

chamber.

microwatts,

temperature difference between The produces volume. Care All A cell a

the liquid inside the mixing chamber and

sample cell was 90 mK (cell temperature: 160 mK). is mounted inside the bore of a superconducting magnet maximum field of 7.5 Tesla with a 1:105 homogeneity in a main which 1 cc field. which and is

set of sweep coils is placed near the center of the

was taken to avoid magnetic impurities at the walls of the cell, copper parts were plated with gold to prevent and the mylar window. oxidation in

cause rapid relaxation of the nuclear spins of the sample of atomic hydrogen. air carefully etched in HCl and HN03. electrical field. tail of feedthroughs Stycast 1266[7] epoxy was used to seal the The geometry of the cell

such that the German-Silver parts of the filling line are practically in zero This prevents first order surface relaxation from taking place in the the density distribution. radiation The sample chamber is provided with a The a a

pressure gauge and a set of trigger bolometers[2] to destroy the sample. microwave Atomic a Teflon NiCr-plated sapphire sheet with a doped Germanium sensing element.

is detected with a composite bolometer consisting of

hydrogen is produced at room temperature by dissociating H2 in valve. To minimize recombination losses in the fill line and suitible wall materials have to be chosen. A Teflon

microwave discharge. sample chamber,

The atoms are allowed to enter the fill line by opening the tube

extends from room temperature to the accommodator, where the atoms are cooled down to about 5 Kelvin by wall collisions. Below 1 Kelvin, superfluid helium4 or a helium-3/helium-4 mixture is used as the coating material. The accornmodator is coated with solid hydrogen. More details of the production of cold atomic hydrogen can be found in ref. 3. The drives refluxes turns presence of a superfluid helium film in where it the stabilization The cell vapor

presents a cryogenic problem. The temperature gradient along the filling line the film towards the warmer parts, to the cold parts, the filling line into a heatpipe. evaporates. where it condenses. This convection mechanism

The heat load

due to this process

is found from: (2.1)

27

d being the tube diameter, Vc the critical velocity of the superfluid film, S the a film thickness and L the heat of condensation per unit volume. The and estimated heat load assuming a saturated film of thickness 300 angstrom

critical velocity of 30 centimeter/second amounts 0.5 milliWatts.

[4]

Surface the this also

roughness may increase this by a factor of four, as is known from beaker flow experiments temperature use a The To avoid such a heat load on the stabilization cell, At filling line should be pinned at a temperature of about 0.5 Kelvin. the helium vapor is effectively condensed. second thermal pinning point (to which we refer as thermal required cooling power for these pinnings is available

In practice we

platform, in at the a

see fig. 2.1) which normally operates at 300 mK. dilution unit. Several milliWatts may be extracted from the still. However, a disadvantage of the still as a heat sink is the fact that it operates temperature of 700 mK. much (T (T Another possibility is the continuous heat exchanger. flowing in the reverse condensed phase net H3C

The dilute stream, flowing from the mixing chamber to the still, represents a larger enthalpy flow than the condensed stream, At a temperature of 500 mK, 8.5 1.5 J/mole [5] The 500mK) 500mK) enthalpy of the direction. the enthalpy of the dilute phase H3D Thus there is a cooling a

J/mole at this temperature.

term on the still, which is normally wasted by electrical heating, but useful for our purpose. heat input Assuming that both phases have the same temperature (T),

(2.2) is needed to heat both streams from the mixing chamber temperature T, to the both

temperature phases effects. 300 our the

where TI3is the circulation rate and the enthalpies of

at the mixing chamber temperature have been neglected.

In the dilute osmotic for from is is in

stream the helium-3 concentration is a function of temperature due to osmotic Therefore the quantity H3D is sometimes referred to as the enthalpy[5]. From the above relation it follows that at a circulation rate of micromole/second continuous The situated thermal between heat (this corresponds to a normal operating condition about 2 mW cooling power may be at a temperature of 500 extracted mK, This which plate and is exchanger dilution refrigerator)

appropriate for our purpose. platform is connected to the cold plate. the continuous and first step heat exchanger,

28

contact with the dilute stream through a copper sinter. of electrolytic copper (length 50 cm,

An off the shelf bar

cross-sectional area 1 cm2) transports

the heat from the thermal platform to the cold plate. To increase the thermal conductivity, the copper has been annealed in a flame. The (helium device HEVAC apart several major part of the heat load due to the film is absorbed in the HEVAC vapor £ompressor). Apart from condensing the helium vapor, atoms. 10 this The cm also acts as a miniature diffusion pump for the hydrogen from the still. Here we are faced to another a distance of 60 cm, problem:

is connected to the continuous heat exchanger at a point about milliWatts over thereby producing a

transporting small

temperature gradient as possible. sample Figure chamber. the shows 2.3 Thus heat the we

In finding a remedy, we were guided by the a heat pipe, and filled the with heat helium-3, exanger. to the

cause of the problem itself: the heatpiping action of the filling tube of the utilized transferring continuous vapor load on the HEVAC to the HEVAC, the heat pipe continuous

connection

heat exchanger in detail.

The heat is transported by evaporation The difference in gas flow; the The heat

at the bottom of the heatpipe and condensation at the top. liquid, load Q, condensed at the top,

pressure between the warm and the cold end drives the drop ~T along a tube of length 1, radius a,

flows towards the bottom due to gravity. in case of a

temperature

is found from Poiseuille's law:

~T/l where mm i.d. bar gas, been of is the viscosity of the gas, R the gas constant, Pav the

(2.3) average

ry

pressure,

Pv the vapor pressure and L the molar heat of evaporation. For a 5 tube, equation 2.3 leads to a temperature gradient of 4 ~K/cm at a the same dimensions gives 10 mK/cm. Thus the transfer of heat by This principle cooling a

temperature of 0.5 Kelvin, and a heat load of 1 mW. A standard quality copper carrying its heat of vaporization is much more efficient than the by electrons or phonons in a conducting solid. applied in refrigation technology[6j, widely using heat has

transfer

agents

ranging from liquid hydrogen to liquid sodium. In practice the performance of the heatpipe is limited by the Kapitza resistance at the copper-liquid helium interface. distance, copper bar. 29 Therefore, if a large heat flux has to be transported over a long a heat pipe should be considered as a possible alternative for a

CONDENSER _ COPPER-----..:....:.~ SPIRAL

CONTINUOUS -HEAT EXCHANGER FLAT COPPER ----' WIRE

TO HSC

-EVAPORATOR M-;~7---COPPER
"'-----;7---

SPIRAL

p::s:::;s::s:::~

J~---HEVAC ---

SILVER SIN TER

THER M OMETER

---HEATER

Fig. from

2.3 Heatpipe cooling system, transporting the heat load on the HEVAC condensing vapor in the filling system to the continuous heat

exchanger in the dilution unit. 30

To are the having

overcome the Kapitza problem, a length of about 1 m.

both the evaporator and the 0.5 mm thick,

condensor and liquid,

equiped

with a spiralized copper plate,

15 mm wide

To enlarge the contact area with the

surface was deformed mechanically and sandblasted. Both spirals have a geometrical area of 400 cm2. Larger areas are more easily realized with however a temperature gradient is set up inside the sinter due interest. exists to An a blocking of convective flow and the poor thermal conductivity of and the sinter material in the temperature regime of of the heat transfer inside the sinter shows that there for a thin sinter, liquid

sinters, the helium-3 analysis

temperature dependent optimum sinter thickness: thermal factor. to conductivity

the heat limiting

transfer is limited by the Kapitza resistance, whereas for a thick sinter the of the liquid and the sinter material is the which is impractically thin. At a temperature of 500 rnK, the estimated optimum thickness amounts Another reason formation

25 microns for a silver sinter,

abandon the use of sinters inside the heatpipe is the possible A

of gas bubbles, particularly at the warm side of the heat pipe. Kapitza resistance also exists between the dilute stream and the walls continuous heat exchanger, the HEVAC. and between the helium and the copper 1 The continuous heat exchanger is a copper-nickel tube, of the

inside tube

meter long,

wound in the form of a spiral. To make contact to this tube, the flat area

was thermally shorted over a distance of 13 cm by soft soldering copper wire around the tube, resulting in a 15 cm2 effective contact

with the liquid. The copper connection to the condenser, the flat copper wire and the tube of the heat exchanger are soldered together with Woods metal. As the sample chamber and the filling line is normally covered with pure helium4, a 0.4 mm thick silver sinter was used inside the HEVAC. Test results of the heat pipe are shown in fig. resistance the dilute thermometers stream. rate The is were mounted at the HEVAC, HEVAC is provided with micromole/second, a 2.4. In this case the Calibrated carbon the temperature gradients inside the sinter are less of a problem. the condenser heater. is The and

continuous heat exchanger. This third thermometer measures the temperature of estimated for our input the circulation experiments. to heat, 300 which typical

These measurements were carried out with a constant heat Therefore, the circulation rate will vary somewhat with

the still.

applied to the HEVAC.

The estimated heat input by conduction from the 2.4. is displaced over

accommodator is 800 ~W. this amount.

The horizontal scale of fig.

At a heat load of about 1 mW the Kapitza resistance between the 31

800~--~-------------------------'
~

j 700
~
0::

o EVAPORATOR • CONDENSER o SPIRAL HEAT EXCHANGER

W600

::::> t-

~500
w
o,

~ w 400
t-

Fig 2.4. Test results of the heatpipe.

dilute behaviour

stream of

and the

the wall heat

of the heat at a heat

exchanger load

determines 4 mW

the the

overall Kapitza

pipe;

of about dominates.

resistance

between

the copper

and the helium-3

2.2 Design

and construction

of the pressure

transducer

A of tool

measurement

of the pressure its gaseous

of atomic

hydrogen

is the most

direct

way

demonstrating to study

nature. with

A pressure

transducer

is an attractive A sensitive in a

the density

decay

a non-destructive has been developed,

technique. which

diaphragm

type pressure

transducer and

operates

low temperature The diaphragm detected mm thick and

environment is shown

in strong

magnetic

fields. difference position, diaphragm across which the is

transducer results with and

in fig.

2.5. The pressure from the equilibrium streched

in a displacement technique.

a capacitive gold plated Piece

The

Kapton

is 0.012 (1

on one side.

It is sandwiched capacitor

between and

two pieces faces and pieces to provide copper

2) of copper. surface with the

1 is the fixed

plate,

the gold is in with an and

covered contact Stycast overload

of the diaphragm. HSC. [7]

r

Piece

2 contains

the sample copper choosen

The diaphragm The sandwiched 1266 has

is glued structure a good

to the was

1266 epoxy protection.

Stycast

adhesion

to both

32

Kapton.

The

seal is very strong and superleak-tight if the epoxy is allowed The space between the capacitor plates is A copper sinter gold strip from 1266

to cure under mechanical pressure.

in direct contact with the inner vacuum can of the cryostat. chamber. extending the Electrical

in the interconnecting tube prevents dust particles from entering this vacuum contact to the diaphragm is made through a from the gold plated central section. To isolate this strip

central section,

a small portion of copper is replaced by Stycast

before the vacuum chamber is machined into the fixed plate. The described displacement of the diaphragm from the equilibrium position with a simple model, where a uniform tension T can be the

provides

restoring force. The tension in terms of the relative strain e, the diaphragm thickness t and the Youngs modulus E is given by: Eet r results in

(2.4) the

Balancing

the

forces

on

a circular annulus of radius

following differential equation for the displacement u(r): (ljr) djdr (r dujdr)

- pjT

(2.5)

~~ STYCAST 1266

,
GOLD HEC TRODE
Fig. 2.5. The pressure showing

SINTER

gauge. Left: the gauge as seen from the top with the gold pattern on the membrane. Right: cross-

piece 2 removed, sectional

view of the gauge 33

where above

p is the pressure equation gives:

difference

across

the diaphragm.

Integration

of

the

u(r)

(2.6)

where

R is the radius
fringing

of the diaphragm. is:

The capacitance

G of the transducer

neglecting

fields

(2.7)

d being

the depth

of the vacuum is

chamber. because

Under

normal

operating

conditions, is the from

the non-linearity small order a

unimportant, However,

in our application is calibrated

the pressure of

lO-3Torr). where

the sensor

at pressures

of 1 Torr,

non-linearity

effects

are significant.

It follows

Taylor

expansion

of equation

2.7 that than

the reciprocal

capacitance Thus,

is a more for is the best

linear purpose obtained bridge,

function of

of the pressure the

the capacitance coefficient

itself.

calibration, a plot of

linear p.

in the expansion hand, the

from

I/G

versus

On the other the value reading, defined

capacitance

to be described

below,

measures

of G; to get an impression p should dG/dPI be plotted. from

of the non-linearities The eq. sensitivity

in the pressure of the transducer

G versus as I (l/G)

follows

(2.7):

I(l/G)

dG/dp I

(2.8)

Thus and

the sensitivity

can be optimized thickness

by maximizing

the diaphragm and the depth to the with

radius of the

minimizing chamber.

the strain, The bore

of the diaphragm sets an upper A vacuum

vacuum space depth

of the magnet

limit

available a uniform machining by the of sure fixed strain

and therefore of some by room

to the diaphragm of microns

radius.

chamber

tens

may be

realized

with

standard is determined

procedures strain copper that plate, at

a skilled

instrument

maker.

The strain

temperature between room

and the differential temperature between at room

thermal

contraction To make and the

and Kapton

and low temperature. the gold electrode

there

is no electrical

contact

the diaphragm

is prestressed is about

temperature.

The estimated

at low temperatures The transducer

2 percent. using the construction jig shown in fig. 2.6.

is assembled

34

l Fixed capacitor sample space 3 Brass ring 4 Brass ring 5 Membrane

plate

2 Copper piece 2 with

stretching

brass ring

6 Tightening motion plate

screws with springs for normal

7 Screw mechanism

of fixed capacitor for normal for lateral with

8 Screw mechanism motion 9 Screw mechanism respect Fig. 2.6. Construction jig, used to assemble

of piece 2

motion of membrane

to copper housing gauge.

the pressure

With this jig, the diaphragm and the copper pieces (1) and (2) can be aligned and brought into contact with each other. It allows for curing under mechanical pressure. oven, Because of its compactness, it can be placed in a small

thus allowing for a cure at elevated temperatures. Provisions are made

to monitor the capacitance between the fixed plate (1) and the gold electrode during assembly. The diaphragm is clamped between two flat brass rings (3) Before the gold is evaporated, and the whole assembly and
(4);

third ring with a smooth edge (5)

together with springs and tightening screws a mask (not is clean cleaned vacuum

(6) is used to stretch the foil.

shown in the figure) is ultrasonically in Freon.

added

The evaporation is carried out in a

system, to prevent the foil from contaminating with magnetic impurities. After the fixed plate (1) and the diaphragm are aligned, is temporarily removed to apply the epoxy. the 9) vacuum chamber to fill with epoxy. is used to bring the fixed plate Care has to be taken not to allow

To this end a very small quantity of with the foil other in a

epoxy is used, just enough to wet the copper surface. The screw mechanism (7the fixed plate into contact The controllable manner. same procedure is used to glue the copper 35

(a)

(b)

0
Fig.

LE

RE C,

C, Cz

-v,
-(Vj+OVj) ~

C2

2.7.

(a) Simplified

circuit

of transformer

bridge.

(b) Equivalent

circuit

of the bridge,

showing

the parasitic

impedances.

piece (2) to the Kapton. The hours epoxy at 60 °C.

Screws (7) and (8) are tightened to apply pressure. for a few the a brass spring is

is cured for 24 h at room temperature and post-cured After the superfluous Kapton is removed,

attached to make a clamped electric contact to the gold finger providing contact with the gold electrode. The in fig capacitance 2.7.a. most to of the transducer is measured with an equal secondary bridge ratio

arm to are

transformer bridge (General Radio, model 1615). A simplified circuit is shown By connecting the center tap of the capacitances use the We equivalent windings reading ground, shunted parasitic ground. that affect the

circuit [8]

of fig. 2.7.b to

analyse the errors and the noise performance of the detection electronics. Cl is the capacitance of the gauge, C2 the balancing capacitor of the bridge. CA and CB represent the total effective capacitances shunting the secondary windings of the transformer, RA and RB are their corresponding losses. RE, RF, LE A and LF represent the ohmic losses and the self inductances cables. connected the ZD to is the total impedance shunting represented the by of the secondary windings is of 8Vi. cables the The to vary a 1 interconnecting input. transducer in time. is In detector

mismatch

the bridge terminals with

coaxial use

suppress the influence of the capacitance between the leads, low temperature part of the apparatus we coaxial cable supplied by Lake Shore Cryotronics, large

which may

commercial

which unfortunately has

capacitance between the center conductor and the shielding of about

nF and a resistance of 10 O. These dominate CA and RE. A typical value of the self inductance of the leads is 1 ~H. approximately: The balance condition of the bridge is

36

32,---------------------------------,
o

Fig. 2.8. Calibration relation of the gauge.

of

the pressure-capacitance

o <! a:
<!

T = 1.5 K 0/

~31 u
(L

o / /

UJ

U Z <!

~30
<!
(L

<! U

29~

l_

----l---------~

0.5 PRESSURE

1.0 (TORR)

1.5

(2.9) At a frequency of 1 kHz all terms containing LE, LF, RE or RF are of the order of 10-5. The sensitivity of the bridge reading to variations in the corresponding affects long term components stability is quite small. The only important C2. by component stabilizing provides that has a its the the stability is the balancing of 10 ppm, capacitor

A good capacitor
capacity

which may be improved reference

temperature. by slow drifts If

In principle

a low temperature appeared gauge.

best result, but this solution in the pressure the bridge

not to be required

as we were limited signal

is in balance

at zero pressure,

the out of balance

may be used to monitor the variation

the pressure.

The out of balance

signal Ud in terms of

nCl neglecting

the parasitic

terms is given by: (2.10) by the input noise 37

AC - V i x D 1I[C 1 The smallest detectable capacitance

+ C 2 + (J·w Zn)-lj
is determined

change

of the detecto.ramplifier. source between bridge The Barocel corrected linear resistance is of

A common value of the noise figure at 1 kHz and a which is a typical value is 0.2dB. This results in a of the impedance of the detectable a are a 200 slope The excitation voltage minimum

100 kO,

the detector terminals,

of the order of 10 Volts. was calibrated at

relative capacitance change of 5xlO-7. transducer strain for 1.5 Kelvin with helium gas against fig. 2.8. The data Torr The Below a pressure of 1 gauge. The results are shown in effects [9]

thermomolecular

relation exists between the capacitance and the pressure.

gives a sensitivity of 1.65 pF/Torr. mK is about lxl014/cc. 2.3 ESR Spectrometer ESR atomic lower sample. Kelvin. A theory In magnetic detailed has a been has proved It

The smallest detectable density at

to be a valuable tool to study the

decay

kinetics

of the

hydrogen.

allows for a determination of the density of both The spectrometer is of a FIR

hyperfine states separately. Changes in the transmission

broadband are

transmission like design. It consists of a long tube, filled with a localized due to the presence of the sample detected by a semiconductor bolometer, which operates at a temperature of 0.5 description of the spectrometer and the given by van Yperen et al [1] ESR experiment a relevant only resonance a brief

Therefore

discussion of these items will be given here. conventional linearly polarized oscillating field. the In the case field is applied, in a direction perpendicular to a static

The transition rates may be found from a perturbation treatment. of atomic hydrogen, la> and Id> state and the Ib> and Ic> state.

the allowed transversal ESR transitions are between

If the oscillating field has an

amplitude Bl' the transition rates WBC and WAD are given by: (2.1la) (2.llb)

The resonance conditions are in the high field limit:

38

(2.l2a) (2.l2b) In The a 5.6 Tesla loss state field the resonance frequency is due to the absorption of radiation 157 of GHz. The dominant atoms is in the

broadening mechanism is due to the spatial inhomogeneity of the static field. power resonant

hyperfine

i is expressed via the ratio Pout/Pin'

where Pout

transmitted power and Pin the incoming power:

Pout/Pin = exp

[-J

dz' a(v,z')]

(2.13) (2.14)

a(v,z)
where sign
v

is the frequency of the radiation, to Both

vi(z) and ni(z)

the the

resonance plus Jb>-Jc>

frequency and the density of the Ji>-state atoms. applies transition.

In equation 2.14 the

the Ja>-Jd> transition and the minus sign to

vi and ni are a function of z via the field profile. It is To a first

assumed that there are only longitudinal field gradients present.

approximation the absorption as a function of frequency is given by:

The

divergences

of this expression at the field extrema may be From eq.

removed

by

incorporating the transverse field gradients. density The of the of 1016/cc,

2.15 it follows that if the

the lineshape reflects all the fine details of the static field profile. At a 10 % of the radiation is absorbed at resonance field has an inhomogeneity of 10-5 in a 1 cc volume. density of the two lower hyperfine states is measured as a line. The cell is provided with two coils for respectively. this

function purpose Gauss

time by alternately bringing the two states into resonance and monitoring resonance The 250

producing a maximum field of 250 and 50 Gauss,

coil is used to switch between the hyperfine states by reversing the current; the 50 Gauss coil is used to sweep the field through resonance. There are two problems associated with this approach: -The absolute inhomogeneity of the 250 Gauss coil is about four times better Reversing the current of this coil 39 than the inhomogeneity of the main field.

~~a~y~~o~:~tem

(a)

.0.
.<:._:._

_.--0CL
:--------

..

I I*

teflon. -vacuum

window

chopper Fabry- Perot Interferometer

Gunn - oscillator

I)j
~ 330Q

cut •• " fitter .

IT-transition Cl-+ 0 ---RG-137

crossed guide multiplier

(b) i"-r---.--c::::J...-+I ~

fromFtL

~
30V

to Gunn oscillator control

phase detector FM input

~HJ=11P:.
+15 V -15 V

2.
2k

\'-

i
~

lk

to reference crystal control

Fig. 2.9.
to

(a) Room temperature part of the ESR spectrometer. (b) Circuit used between different harmonics of the reference crystal of the

switch

frequency stabilizer. results in a slightly different field profile of the combined coils, but this is enough to produce a completely different line shape. -Because by the of the finite resistance of the persistent mode switch of the sweep coil, which results in a field pattern and thus a main coil, the field has a drift of 3 Gauss/hour. This drift has to be compensated lineshape varying in time, making it hard to calibrate the system. Part frequency 40 of these problems may be overcome by switching the microwave guarantees exactly over 1420 Mhz, the hyperfine frequency. This

Vl

Fig. A-D B =5.6 TESLA for

2.10. both

Resonance

1ineshapes at (a) and

>--

::::>

transitions frequency field

>-

constant constant that

cr: cr: >--

(b) , showing are field and

co

the shapes

cr:

«
>-

B-C
B=5.6TESLA +506.1 GAUSS

frequency

dependent.

-0

a:
-0

B-C

v=156.356 GHz B = 5.6 TESLA

A-D
V = 157.776 GHz

that the the

the

field

distribution

is the same

at resonance is short.

at both

transitions, version range.

if of A

time between spectrometer modification room

the two measurements a 1 % frequency

In the original the tuning

deviation

was beyond

simple The

of the spectrometer part

proved

to be effective. is shown operates in fig. 2.9a. and in a

temperature

of the spectrometer Oscillator, of 1%. with Higher

Microwaves is voltage

are generated tunable harmonic over

by a Cunn a range

which

at 39 CHz

harmonics

are produced diode

crossed nonlinear the wave frequency

guide

generator,

a CaAs (n=4)

Schottky-barrier is selected

as the of 3600

element. guide

The desired

harmonic

by the cut-off Systems MOS

and a Fabry-Perot locks 15 MHz

interferometer. Oscillator oscillator,

A Microwave to a high which (n

stabilizer stabilized crystal

the Cunn crystal limited overcomes

>

215)

harmonic range of The of

of an oven 0.3%. This

has

a tuning

oscillator 2.9b of tuning

the tunability this problem. (MOS APC

of the spectrometer. Normally, the output to the

circuit the

shown

in fig.

phase

detector voltage

the PLL system input. The

out)

is

fed

Cunn by

oscillator applying of fig.

frequency

of the system

may be tuned The

a voltage 2.9.b

to the crystal a switch

oscillator

(MOS Sweep to both

input).

circuit input of

adds

selectable

voltage

the tuning

41

the

crystal loop.

oscillator

and the output of the phase detector of

the

phase the

locked

Thus it allows for jumping between different harmonics of

crystal, keeping the loop in lock. This extends the effective tuning range of the stabilizer by about two orders of magnitude. Frequency modulation is used to employ Fig. phase 2.10 sensitive detection of the derivative of the resonance microwave to go signal. shows some typical line shapes. In fig 2.l0a the frequency is kept constant and the main field changed over 506 Gauss, from the a~b to the b~c transition. same. used, marked A In fig. the main

This ensures that the field distribution described above, is the led to field a the from both there is are still

is the same in these two situations. The two lineshapes in fig. 2.10a are the 2.10b the frequency switching technique, field is kept constant. the should be the same at the two transitions, two lineshapes. Thus we Despite the fact that

distribution

difference

between

conclusion that the microwave frequency plays some role in the lineshape too. possible explanation is the existence of standing waves, resulting reflections at irregularities in the sample chamber, the pressure gauge or the FIR bolometer. transitions at a high temperature, is slow and the relaxation is fast. equally populated. Comparision of such as the entrance of recombination gauge

The system is calibrated for

under conditions where the In this case, the signals with the

both hyperfine levels are pressure

provides an absolute calibration for the density of both hyperfine states.

2.4

Trigger

Bolometers

couple

of

bolometers

is

mounted inside

the

HSC

to

trigger

the in

recombination of the hydrogen atoms. the following situations: - at

They are used to destroy the sample

the end of a measurement of a pressure decay to establish the

baseline

of the pressure gauge. reading of the gauge.

Thus it is possible to correct for drift in the zero-

- by a rapid destruction of the sample,

the total number of atoms present in Furthermore, the cell. the

the HSC can be determined with a calorimetric measurement. helium film and the presence of helium-3 impurities in

current-voltage characeristics contain information about the thickness of the These

bolometer applications will be explored below.

42

2.4

2.0 _ 1.6

y-----........_
------

.! J

:/

.../

.. v'

./

----

'"" "- "-

T= 0.359 K

~
w
::::J

.......

::;: 900

.......

<,

..............

a:: 800

~ ffi
o,

700

~ ~
UJ

1.2 r

--------J

::;: ~ 600

a::

gem3 NTP

~08

'::i

~0.4

a
CD

~ 500 w ::;: 400

o 43em3 NTP D. 83em3 NTP

0.0 ;;--!:-~---;--!:-----!;;------!;;;:-7:----:;';;-' o 2 4 6 8 10 12 14 16
CU RRENT (MICRO AMPERE)

300~

~ 5

~ 10

FILM COOLING POWER (MICROWATT)

Fig.

2.11 Bolometer

current curve:

- voltage bare

Fig. film

2.12 Determination
critical cooling current

of powers voltage from

characteristics. bolometer. bolometer. covered mixture. Dashed Dotted

Full

curve: curve:

4He-covered bolometer

the bolometer characteristics. correspond at different

The arrows of Pcrit

with

a 1 ppm 3He-4He

to the values helium

coverages.

The material providing Fig. germanium follows helium) current, part

bolometers such

consist

of a small

(1

0.3

0.3 mm3)

piece

of resistive copper wires, cell. of a

as carbon thermal shows

or doped

Germanium,

and a pair body

of fine

a weak 2.11

link between typical helium-4

the bolometer current-voltage is introduced bolometer

and the sample

some When

characteristics in the (i.e. cell, the

bolometer. the at

curve with higher

characteristic low currents, a different heating

of the bare whereas shape.

uncovered a

the maximum In the case

is shifted of a helium

towards covered

having

bolometer, which is

of the Ohmic

is carried along

away by evaporation This extra

of helium, term

replenished to the cooling subtracted follows The amounts relate from

by superflow

the wires. film. the

cooling

is related to to the be

thickness of the from

of the helium bolometer, the total

To obtain carried The

the film by the

contribution leads from has the

heat

dissipation.

contribution bolometer. fig. of 2.12

leads

the I - V characteristic of this procedure in the cell

of the bare are shown in

results of the

for

different Kelvin. To was

helium

at a cell to its

temperature temperature,

0.359 the

bolometer against

resistance

bolometer

calibrated

the cell

thermometer.

43

Below 0.6 Kelvin, pressure rise

film cooling is ineffective because the helium-4 vapor The point, where the the steep current-

is too low in this temperature range.

at high flow rates sets in corresponds to the maximum in

voltage characteristic. of

At this point the superflow velocity along the wires Thus the critical velocity, the circumference power in from the This the

approaches the critical value.

the wires and the film thickness set an upper limit to the cooling A current that pulse through the bolometer beyond the maximum

of the film (see also equation 2.1). characteristic indicates To (arrows effective of the roughness. superflow this P/Po phase between We fig. use Strauss given by: (microwatt) dbeing Pcrit (2.16) triggers a rapid recombination of the hydrogen atoms.

part of the protecting helium coating

disappears

bolometer if the evaporating helium is not replenished by superflow. derive a value of the film thickness from the critical cooling powers in fig. 2.12), we need to know the critical velocity due to and the circumference of the bolometer leads. leads To differs from the geometrical calibrate some The effective circumference surface the At thickness helium-4 detector, on the the

circumference

the bolometer as a film data on the adsorption of

experiments have been repeated at a temperature of 1.3 Kelvin.

temperature

different adsorbed the relation [10] and in is

substrates are available. Thus the film thickness follows from the saturation where P is the pressure of the vapor in equilibrium with and Po film the on is the vapor pressure of the bulk liquid. cooling power at different saturations on By P/Po' measuring

critical

this power and the film thickness may be determined adsorption isotherms of Bowers et al. jewelers rouge [11] for this purpose. The

experimentally.

aluminium

The results are shown d

2.13.

critical power Pcrit as a function of film thickness

given in atomic layers, and do being number of immobile layers at the which zero cannot contribute to the cooling power. gives a value of 2 for do' Extrapolation with of to consistent other

substrate,

experiments. [lS] Below 1 Kelvin, we use a value of 1.S for do' [4] To check the validity of the above analysis, we calculate the film A

thickness 44

as a function of the amount of helium,

introduced in the cell.

~50
o
0:::

tJi [,0
o
Z w a: w ~ 1.32 w >a: ui >w

;30
V'J

<!

I I I I I I I I

3000 A

~ a:

:::>

LU

~20 :r: ~ 10
_J

1/
o
20 [,0
HELIUM

300A

30A

LL
a PIPe t>. PIPe "PIPe o PIPe =0.984 =0.946 = 0.827 = 0.696

o
ill

:E

<51.31

LOAD (cm3 NTP)

60

80

100

Fig.

2.14. Comparison

of film from

thicknesses
5 BOLOMETER 10 15 POWER {MICROWATT}

as determined
response values

the bolometer and calculated

(points) for various

Fig.

2.13. Bolometer
saturations

response PIPo'

at

sinter

pore

sizes

(curves).

different

complicating the large sinter). system rough The

factor

is the presence flat

of a silver (i.e.

sinter

in the HEVAC.

We model and r The a

as a completely surface, value in this

surface

the walls

of the cell) radius size.

having

many

dimples with

of characteristic the sinter por$

(the film and

of r is associated system is governed

distribution capillary If surface interfaces

by gravitational,

Van

der Waals,

forces. both of the the flat and the rough bulk by liquid, surface are at a potentials height at the h two above the

the chemical

film-gas

are given

J.!flat

J.!o mgh +

- V(dflat)

(2.l7a)

J.!sinter

J.!o mgh +

- V(dsinter)

- m1/pr

(2.l7b)

where mass,

J.!o is the chemical g is the

potential

of the liquid, V(d)

m is the is the

helium Van der

atomic Waals and p from

gravitational d from

acceleration, the substrate,

potential the

at a distance

1 is the surface surface tension term

tension arises

density

of the liquid.

The additional

45

the

decrease

in

surface free energy.

Because the

chemical

potential

is

constant through the system, the film thicknesses follow from

o
mgh - V(dsinter) - m~/pr For the Van

(2.l8a)

o
of

(2.l8b) Anderson and

der Waals potential we use the calculations

Sabisky [12]; we approximate their result by V(d) with (2.19)

a ~ 25 Kelvin (layers)3 as the appropriate Van der Waals constant for a Both film thicknesses may be calculated with the help of eqs (2.18 a) and

gold surface. (2.18 b), system area about is for different values of hand r. The total amount of helium in the the surface been has

follows

then from the film thickness in the sinter and

of the sinter.

A detailed study of submicron silver sinters

made by Robertson er al [13]

Our sinter is estimated to have a pore size of 0.5 micron and a total surface area of 1 m2. The model described here near saturation the shape of the liquid

valid for undersaturated films;

surface becomes strongly dependent upon the amount of helium in the system. Figure 2.14 shows the results of the calculations for three values of the sinter pore size, together with the results obtained by Van der Waals forces. from the bolometer is is by helium current-voltage completely the the film characteristics. At low coverages the film distribution When more

determined

introduced in the system, pore size.

the sinter becomes filled with bulk liquid whereas determined filled

thickness on the flat surface saturates at a value

This continues until the sinter is almost completely

and our simple model breaks down. The sinter is completely filled if about 30 cm3 NTP helium is loaded in the system, so we are safely below saturation. The bolometer pore results are quite well described by this model if we Note that the calculated use a is Van Thus of sinter rather diameter of .3 micron. thickness

insensitive to the choice of the relevant parameters (pore size,

der Waals constant) because of the form of the Van der Waals potential. we may have some confidence in the bolometer results. Addition 46

of a small quantity of 3He has a marked effect on the shape

W 0::: I-

::::>

«
0:::
l..LJ __J __J

T=0.367 KELVIN B= 7.9 TESL A 15 N=99x10

CL

wW
LJ

I-

0:::

I-w l..LJl!J

1'------

L«
__J

01__J

co>

PO

3 TIME

4 (SEC)

Fig. marks by

2.15. Destruction of hydrogen atoms by a bolometer current pulse. Lower bolometer voltage response. the evaporation of the The steep drop in the bolometer bolometer and the voltage of film on the onset

trace:

recombination. Upper trace: cell temperature response. Reproducing this trace a current pulse in the cell heater gives the total number of atoms in the cell.

the

current-voltage from This

characteristic.

A gradient body, known

in the drives

3He

concentration, along the

arising wires. A

evaporation results

on the bolometer effect,

the helium-3 flush

in a cooling

as the heat

effect

[16]. on the

1 ppm

3He impurity

may be easily

observed

in the form

of a shoulder (dotted line

bolometer 2.11).

current-voltage the bolometer

characteristic may be used

at low currents the purity

in fig. or

Thus

to check

of the helium-4,

to signal The energy

the presence recombination

of a leak. of two hydrogen K. Figure 2.15 atoms results how in the this release effect of an for

of E/kB

~ 52,000

illustrates number

allows

a destructive In the figure This wave order trace form 108

determination the temperature

of the total response

of atoms after

present

in the cell. is shown. the the a can

of the cell

triggering

is easily

converted with

into a recombination the cell heater. if the sensing

energy

by reproducing of as

electronically - 109/cm3

Very

low densities, is used A high both

may be detected

bolometer element.

trigger

device

and as a temperature

sensitivity

47

be

obtained

if

the flow in a short

impedance period

of the gas

to the

element

is

low,

to

destroy

the gas

of time,

of the order

of 10 milliseconds.

2.5 Thermometry

Several cryostat using a

200

1/8

W Matsushita below

[14] resistors 1 Kelvin.

were

mounted

in

the

to monitor bridge were

temperatures circuit

The resistance et al. [1].

was measured The following

described

by Van Yperen

standards -a NBS

used

to calibrate

the thermometers: fixed points device for temperatures below

SRM 768 superconducting

200 mK. - in situ helium-3 in scale section was vapor 2.2, pressure thermometry, 200 with the The pressure 1962 3He gauge vapor

described pressure -a mK

in the range

- 600 mK.

used[16]. resistor (Lake Shore Gryotronics) in the range 600

calibrated - 1000 mK. The vapor

Germanium

pressure The

may be used

to calibrate was

the thermometer measured to

in a a

nonzero field

magnetic dependent was

field.

magnetoresistance below after 200 mK.

obtain

calibration repeated has

The calibration temperature.

of the cell We believe below 300

thermometer that and our 5

always

cycling better

to room than

thermometry milliKelvin

an accuracy

3 milliKelvin

mK,

between

300 and 600 mK.

References

1. G.H. Phys. G.H.

van Yperen, Rev. B30,

J.T.M. 2386

Walraven,

I.F.

Silvera,

(1984). University Phys. Rev. of Amsterdam Rev. Sci. Lett. Instr. (1984). 44, 164 (1980). (1982).

van Yperen,

Thesis,

2. I.F. Silvera,
3. J.T.M.

J.T.M. I.F.

Walraven, Silvera,

Walraven,

53, 1167

4. For a recent

review

about

helium

films, Benneman

see The Physics and J.B.

of Liquid
eds.,

and Solid Helium, vol.

Wiley-Interscience 5. O.V. Lounasmaa, Press

II, K.H.

Ketterson

(1976).

Experimental
(New York, Treay, Inc.,

Principles

and Methods

below l K,

Academic 6. P. Dunn 7. Emerson

1974). Press, 24, Oxford 1978.

and D.A.

Heat Pipes, Pergamon

Nijverheidsstraat

and Guming

48

B-2431

Oeve1,

Be1gig. Trans. Inst. Radio Engrs. 1, 245 (1958). Phys. Rev. 102, 304 (1956). (1952). and S.G. Sydoriak,

8. A.M. Thompson, 9. T.R. Roberts 10. R. Bowers, 11. Strauss,

Phil. Mag. 44, 485 (1953). of Chicago andC.H. Anderson, Phys. Rev. A7, 790 (1973).

Thesis, University F. Gui110n

12. E.S. Sabisky

13. R.J. Robertson, 14. Sinhachiro

and J.P. Harrison, Sato, Rev. Sci. Instr. 46, 1226 (1975). below one Kelvin,

Can. J. Phys. 61, 164 (1983). Saito and Takashi 15. E. Long and L. Meyer, Sussex University Phys. Rev. 98, 6, 1616 (1955). and Thermometry

16. D.S. Betts, Refrigeration

Press, 1976.

49

CHAPTER III EXPERIMENTAL MEASUREMENTS The results, The of described in this chapter, motivation the of gaseous nature were obtained in two series of (section atomic decay 3.1), was and to the The and of hydrogen in an

experiments. rigorously feasibility second some of temperature studies the

the first series

demonstrate

pressure measurement in the study of

processes.

series

was devoted to the same type of experiments

extended

and field range and a better defined geometry (section 3.2) using a combination of pressure measurement and container; a search for the existence of

electron-spin spin waves is in

resonance. Section 3.3 covers the heat transfer between the gas and the walls sample described chapter 4. 3.1 Pressure The measurement of atomic hydrogen[1] the in section 3.4. ESR pumping experiments are covered in detail

first stable samples of atomic hydrogen immediately established

feasibility of pressure studies using a capacitive tranducer.

Deflections of

a flexible membrane due to applied pressure down to a fraction of an angstr5m may be observed experimentally, and correspond to a detection limit of = 10-6 Torr. This corresponds to a density of 3xl013 cm-3 at a temperature of 300 milliKelvin. Most of our experiments are carried out at densities of 1015_ 1016 cm-3. of Non-linearities about 1 Torr. in the pressure response are neglible up The capacitive transducer has become sensitivity, a accuracy to a and pressure detector common

in atomic hydrogen research due to its

convenience of operation. In this section we present the results of the first experimental study of atomic hydrogen using a strain gauge. Pressure measurement was used to demonstrate the gaseous nature of atomic hydrogen at low temperatures, and to study the temperature and field dependence of the surface recombination rate. We give a brief description of the apparatus, and emphasize used for our system first pressure in measurements chapter II. 3.1.1 Apparatus The first measurements with the pressure gauge were carried out in the the differences with the described

50

ELECTRICAL CONNECTOR -----!;;nJ • -HEATER -THERMOMETER

-co-

Fig. 3.1 Experimental with same made; chamber the pressure cryogenic

system, showing

the filling

line and the sample chamber

gauge. system where the early observations of atomic hydrogen evaporation were sample coolers. is shown 11 Tesla

only the sample chamber was replaced.

In this system both the

and the HEVAC were pinned to separate helium-3 being 270 mK. container The low temperature is positioned

The total running time was limited to about two hours, temperature in fig. 3.l. The 3.2, fills helium sample in the

the lowest attainable

part of the apparatus center of an

superconducting the film

coil of low homogeneity.

The pressure

gauge,

shown in fig. of the by and 51

is an integral part of the sample container. space between transducer. thickness. the deflection

In this design, the sample membrane was 11.1 to variations reference in the side of

the fixed plate and the flexible This makes the sensor sensitive In later designs, this of the membrane problem was

capacitive measuring

avoided pF/Torr

across the vacuum

the gauge. The measured

sensitivity

of this manometer

Fig. 3.2 Sample chamber with pressure relative pressure of 10-6 Torr. A outside its H2 1.3), value chamber, to the minimum capacitance

gauge. change was 1.7xlO-7, corresponding to a sample the the

valve was mounted in the filling line near the entrance of the close the cell at the bottom. the cryostat using a spring mechanism and a nylon thread

The valve could be operated from along

axis of the filling tube. The valve was necessary because in this system with short filling tube the atoms may escape from the cell and penetrate coated zone in a single shot. for the time constant This process, was studied in an earlier experiment in the same ~ystem(3).
T

thermal escape (see section A typical field was 8 Tesla

of 2000 seconds in an

observed. It was not possible, to close the valve completely, as a small leak remained. As such, the valve was effective as a controllable flow impedance. The was cell temperature was measured with a 270 Ohm Speer against a Germanium thermometer. in the cell, resistor, was which to calibrated present A bolometer with the used

trigger the recombination of the atoms. With this system, the total number of atoms, No could be determined calorimetric magnetic

technique, described in chapter 2. precautions were taken to keep the sample chamber free from impurities. At the time the system was prepared for the measurements, we were unaware of the relaxation limited recombination process (40)

52

3.1.2

Equation

of state

The

relation

between

the

pressure

and

the

density

for

weakly

interacting

gas of boson particles p

is given by[4)

(3.1) A value, representative by Friend and for atomic hydrogen Under typical to the is

where a is the hard core radius. a = 0.72 conditions pressure therefore determined measurement systematic part AngstrBm was calculated cm-3, (n = 1016 expected. between

Etters[5).

T = 300

mK) the relative contribution

due to the interactions

is of order 10-5, the pressure

and ideal gas behavior

A linear relationship of these

and the total number of atoms, by a simultaneous in fig. to 3.3. A result their in

by the calorimetric deviation

method,

could be established

quantities.

The results are shown

from the ideal gas law of 25 % is believed are blown out of the cell without Thus the values

from uncertainties full recombination

in the cell volume; we also believe energy.

that after triggering, releasing

of the hot molecules

for the total number of atoms

20
T·0.340 B·7.9 V.1.7 1.5
0

KELVIN TESLA CM-3

1.0
0

a: a:

~ ~

>-0,5
0 0 0

a:
::>
V>

c,

a:

3 NUMBER OF ATOMS

4
{x

5 10151

Fig. cell.

3.3 Relation

between

the pressure

and

the total number

of atoms

in

the

53

the

cell are a lower bound.

The data of fig.

3.3 provide the first

direct

evidence that atomic hydrogen is a gas at low temperatures. 3.1.3. Surface recombination rate

Under conditions,

typical for this experiment, recombination takes place The temperature dependence of the rate coverage. studied analyse function in the of is extensively 1.4, to

at the walls of the sample container. Although use the By second order surface

constant is mainly determined by the temperature dependent surface recombination experiments no detailed theory is available in the literature. described in section

We therefore

simple kinetic model,

experimental data. studying the temporal behavior of the gas density as a temperature, the surface binding energy of H on 4He could be determined. This method was successfully applied to the case of atomic deuterium adsorption on helium-4 to to by Silvera and Walraven[2]. In the limit of fast relaxation and equal populations of the lower hyperfine states, recombination and thermal escape, the only loss terms are due

and the total density decays according

dn/dt where we also included a term due to the filling flux~; volume of the density distribution (see eq. the density, we found to be small. experiments [6,7]. by 1.25). This is

(3.2) Vo is the effective by later

The terms third order in confirmed

The effective surface recombination rate constant is given

(3.3) where recombination and are the surface rate constants for a-a and a-b

respectively.

The solution of the differential equation (3.2)

0):

in terms of the pressure pet) is given by (~=

pet)

p(O) exp(t/r) (1 + [p(O) K~ff r (1 - exp(t/r»/kBT]J-l

(3.4)

54

5
4

~
P 10"4Torr]:

.-.B

1618
B = 7.9 TESLA T

3 2

K:"{T
(CM3K1/2 S-I)

0.46 KELVIN 10
19

. ,.,....------t[Min.).... 50

o
Fig.3.4

10

20

30

40

Typical

pressure

fill-decay

Fig. 3.5 Logarithmic vs liT. The adsorption determined

plot of K~ff energy is

plot. See text for explanation.

from the slope.

3.1.4 Results

typical pressure fill-decay plot is shown in fig. is then turned is off and the cell cools to the

3.4.

The C, D,

cell

is the the

filled for a period of 5-10 min. discharge temperature decay gauge. by but a is

with H+ starting at point A. At point B the point where where milliKelvin. The pressure

actively stabilized to within a

observed for periods up to one hour to point

remaining H+ is destroyed with a bolometer,

to establish the baseline of the chamber of the

The slow negative going pressure transient after triggering is caused small leak between the sample chamber and the vacuum This leak only contributed to the first order time constant, The data were sampled by a A non-

pressure gauge.

did not influence the second order constant.

Nicolet 1174 transient recorder and transmitted to a PDP 11 computer. of the form (3.4). gauge.

linear least squares fitting procedure was used to fit the data to a function The data were corrected for drifts in the baseline of the

To determine the adsorption energy EA, K~ff was measured as a function of temperature magnetic arises thermal in the range 0.34-0.54 Kelvin. All data were taken at a the the field from of 7.9 Tesla. To extract the adsorption energy EA from thermal velocity and

data we plot K~ffJT as a function of liT as shown in fig. the temperature dependence of the wavelength. From the slope of this plot we find

3.5. The factor JT

EA/kB = 0.89 ±0.07

55

Kelvin. table

Many

values of EA for H+ have been published in the literature; Most measurements use some as a probe, surface et zero time. or the is relaxation) the only exceptions Morrow being

in the

3-1 we compare them all.

process
a1.[8]

(recombination measurements investigated This quantity This

of Morrow et al. [8] and Reynolds et al. [9]. directly related to the surface

shift of the hyperfine frequency with NMR at residence

field. These above ESR

authors also determine EA from the recombination decay, result. mentioned is the only experimental evidence, iT is correct. Recently, factor

which gives the same in the

that the use of the

a satellite line

Table 3-1 Measured binding energies (H on 4He)

Method

KsB2/iT Measured by 2T2sec-1K-l/2) (cm UBC[8] UBC[8] 2.6xlO-7 4.0xlO-8 4.lxlO-8 4.0xlO-8 5.7xlO-8 5.9(5)xlO-8

1.15 1.15 0.89 1.01 1.06 1.15 1.10 0.96 1.00

Hyperfine shift Recombination Recombination Recombination Recombination Recombination Recombination Recombination ESR

Amsterdam[l] MIT[ll] Cornell [13]

Amsterdam, see section 3.2 UBC[14]

Harvard[lO] UBC[9]

56

spectrum resonance. adsorbed The

of

H~ was observed by Reynolds et al.[9], feeling

shifted from

the

main to

The shift was density dependent. atoms,

This feature was attributed

a slightly different field due to magnetic to

dipole 1.15

interactions [16]. This has enabled still another measurement of EA. values for the binding energy in table 3-1 range from 0.89 Kelvin. This spread in EA results in an unacceptably large uncertainty in the surface coverage, this leads to especially at low temperatures. At a temperature of 100 mK a factor 13 uncertainty in the unknown and surface coverage. two or Possible hyperfine film/wall thermometry, helium-3

errors in the determination of EA may result from[lO]: contamination, states, to

or varying polarization of the lower a Kapitza resistance at the gas/film

time varying surface area due to H2 build up, heating of the gas due

recombination

interface. We will return to this point in the next section. 3.1.5 Field dependence of the surface recombination rate

The field dependence was measured at a temperature of 0.385 Kelvin in the field range between 5 and 10 Tesla. For the case of strong fields, the model, described in chapter 1, The predicts a field dependence 3.6, of the we form: plot K~ff(B) = K~ff(0)x2€2. results are shown in fig. where

(K~ff)-l vs B2. A fairly linear relationship is observed, consistent with the value KsB2/JT larger Van der Waals recombination model. The dotted line in fig. 3.6 relates to a 22 F 2.6± 0.8*10-7 cm T sec-lK-1/2. This is about a factor of five than most of the measurements in table 3-1. This large difference is of related to the difference in binding energy. the results will be given in section 3.2. 3.1.6 Implications for BEG A more detailed comparision

These results have serious implications for observing BEC in this type of geometry. It has been shown, that in order to attain BEC, the surface must be 14 cm-2 (see section saturated with a surface coverage n~at EA/202 = 10 1.2). The maximum surface coverage in this experiment is about 1011 cm-2. Under BEC conditions, free surface recombination will therefore be five orders of If the sample container is by then the overall rate will be determined magnitude faster than in the present experiment. from magnetic impurities, dipolar relaxation, magnetic

as a result of the relaxation bottleneck [40] 57

(K:ff)-1 [10'8CM3S]
o' o o, '" o,

,'" '"
0

, ,, ,,

,' ,,

, ,
0

,'"

0 '"

"

, ,
0

, , '" 0 ,
20 40

, '"

S2[T2)

60

80

100

120

Fig. 3.6 Field dependence

of the surface

recombination

rate plotted

as

(K;ff)-l vs B2 for T = 0.385 K.

Other mechanisms found[6,71, A binding possible

which more severely means of reducing

limit the attainable

densities rate

have been working a lower were

but are beyond the scope of this thesis. the surface recombination density, is

with a surface,

that has a lower saturation

and therefore,

energy. To investigate

this aspect, some preliminary A reduction adsorption Therefore,

experiments

carried out with a 0.2 % 3He-4He mixture. of 6.1 was observed, is completely K EA/kB = 0.59 surface, 3He 3He was for attributed to a change in

of Kseff with a factor energy, of we the estimate helium-4 of to more

compared with the pure 4He case. If this change in K~ff the mixture. Helium-3 floats on top

especially

at low temperatures[151. since the fluctuating experiment[171,

a small addition data with contributes

may have a large effect. in the system, used In a subsequent

We were unable to get reliable 3He vapor pressure the binding

the noise.

in which a dilution

refrigerator saturated reduction

to achieve lower temperatures,

energy on a

3He_4He mixture was measured of the binding of the surface unchanged

to be 0.34 Kelvin.

Thus a substantial to BEG.

energy results from using a mixture, as being a major barrier

but this leaves the role

58

3.2 Temperature

and field dependence

of the surface recombination

rate

3.2.1 Introduction

The method field

measurements described

of the temperature 3.1, were

and extended

field

dependence,

using

the and is of of due at

in section out

to a larger geometry.

temperature This geometry profile parts rate

range,

and carried by a more rectangular line, which This

in a better

defined

characterised approximately the filling

homogeneous shape, results allowed field. ESR

field, and a high

leading

to a density

impedance

to the warm to the decay

in a low contribution for the measurement

to thermal low values

escape.

of the rate

constant

of the magnetic with

It was verified nuclear

that all data were

taken

under

conditions A

with

polarization using about ESR combined with a pressure This with

zero. measurement technique the walls from

density provides used to

determination information check that

the temperature in thermal

of the gas. equilibrium film was

was of the the

the gas was The thickness

sample current-

container. voltage were

of the helium of a helium

estimated

characteristics also of used to check

film bolometer;

the I-V characteristics contamination A (the

for the presence impurity used was

of a helium-3 easily

addition description

a 1 ppm helium-3 of the apparatus,

observable). is given

detailed 2.

in this

experiment,

in chapter

3.2.2 Temperature

dependence

results

Two

sets

of data were of 42

taken

at a field

of 5.6 Tesla,

and

at

estimated refer are

film-thicknesses to these as the

and 48 and the

A.

In the following film

discussion The

we will results

"thin" Both

"thick"

respectively. a two-dimensional the data explain

shown

in fig.

3.7. of 0.4

sets of data yield temperatures, to

recombination from the fits. thermal gaseous ended sinter order the

cross-section We considered

A.

At high

deviate this

several

possiblities, hyperfine body, states, and

effect: with open

population helium geometry.

of the upper

two body

recombination trough an

acting

as the third is a large

thermal

escape

As there this

surface may

area

of about

1 m2 in the silver to the second explanation

in the HEVAC, rate constant.

last process of these

effectvely

contribute

None

provides

a satisfactory

of

observed

increase

of the recombination

at high

temperatures.

59

Fig. 3.7. Temperature

o

dependence rate at

FILM-THICKNESS FILM'fHICKNESS

42 48

A A

of the recombination correspond of 42

B = 5.6 Tesla. The open circles

•
N

to a film-thickness

A A

and a surface binding

to a thickness

energy of 1.31 K. The closed circles correspond

10-18

~:z >
-' w
.,'>C.
LJ

of 48 Dotted and
\ \ \

and an energy of 1.15 K. line: resonance with b. = 0.7 K 7

w
V1

recombination
T
=

Jl-S.

\
\ \

•
\

\ .. .. •
0 \

.
r
\

3 11T (KELVIN-I)

similar

effect

was observed by Statt[44) and later

by

Reynolds

et

al.[20), show an

who measured Kab and Kaa as a function of temperature in a 4 upturn at temperatures above 0.5 Kelvin, recombination 1.14, yielding (see section 1.3.1). which was

Tesla to

field with ESR. Their sample chamber had a closed geometry. Their data of Kaa attributed fitted of resonance equation Kelvin, fig. rates, 3.7 The data were using

a quasi-molecule dissociation energy same faster

0.7(0.1) The

and a quasi-molecule lifetime of 7 microseconds. corresponds to the same mechanism with the in our experiment are somewhat can account for. However, observed

The dotted line in parameters. than resonance

recombination data.

it is difficult to obtain reliable likely of A the

data in this temperature regime, process at high temperatures. The 1.31 variation thin of

as is witnessed by the large scatter in the

We conclude therefore that resonance recombination is the most film results may be fitted with a surface binding and the the thick film results with a value of 1.15 result binding energy with film-thickness may

energy Kelvin. from

Kelvin,

penetration of the Van der Waals potential due to the solid substrate through 60

the

film,

which

affects

the binding in the Van

of the atoms der Waals

to the surface near and the the may by

in two ways: which atoms; to and of

the helium leads also the to

is compressed a stronger

field

substrate, adsorbed

attraction potential

between due

the liquid

the Van binding.

der Waals An

to the solid effect

itself was made

contribute Zimmerman

estimate

of the latter et al.[lO). is small;

Berlinsky[lB), the Van

and Godfried potential

At a thickness we estimate

of 40

the effect der

der Waals

that

the Van

Waals of 40

potential

for a metal However, may

is of the order the observed for. followed

of 10 milliKelvin change in binding

at a distance energy is much

angstrBm[19). than this value These of

larger

account were

measurements

by a systematic by Godfried

study

of the

dependence used a

and Ks on the film-thickness transducer surface to measure rate

et al.

[10)

They

capacitive effective

the film-thickness. temperature et al. have is their

In fig.

3.B we plot the

function of and

at constant of Godfried Also shown

and field been

as a

film-thickness. surface of

The data ratio.

scaled

to our field

to volume

calculated of the Van

thickness der Waals a

dependence potential

the rate, solid of

based substrate

on a model through

of penetration the liquid.

of the increase

The model 20

predicts below,

substantial whereas the

the rate data show

at thicknesses the influence a distance 1

of about

and

experimental of about 100 has

of substrate from

effects

up to a the Van

thickness der Waals

A.

At such

the substrate,

potential

a value

of roughly the role

mK.
roughness as a possible explanation we consider of spherical in contact is on to a the calculations, hills and valleys this

We have

explored

of surface

for the results, an eggbox-like shape, with provide rough an

mentioned surface:

above.

To facilitate covered with

a surface radius

all with

the same flat

of curvature.

We keep

system

atomically a reference,

surface;

the purpose thickness

of the flat

surface quantity surface

as the film specifying denoted the

is an ill defined on the flat

surface.

Thus,

a film-thickness sets

(in the

following the

discussion In

by d),

the chemical determined

potential thickness gauge

throughout from the

system.

fact,

experimentally

bolometer is related

current-voltage to the chemical

characteristic potential.

or the thickness In the case

capacitance this is

of the bolometer in section by Van der

the consequence The surface before film

of the calibration distribution forces. with in this We may sinters

procedure, system

outlined

2.4. Waals and

is governed lines

tension

follow

the same

of reasoning Within this

as we did context, it

in connection

(see section

2.4).

61

0 0
0;-

Vl m

u w

T= 300 MK B= 5.6 TESLA o THIS WORK o GODFRI ED ET A L.

~ 'Co
u, ~v>

L u S

""

\
\

' ......

::-::::-:=:-:o=-=~_-01

100

200

300 FILM-THICKNESS

.400 (AI

Fig.

3.B.

Effective

surface recombination rate at constant temperature and Squares: measurements by Godfried et curve gives the calculated

field as a function of film-thickness. al.[IOl. dependence Circles: for a flat surface.

this work. The dashed curve gives the predicted thickness The stepped

dependence for a rough surface. is useful to define a critical thickness do' magnitude. et al. value, uniform. situation a For a solid hydrogen substrate, [10) we find do = 3.5 rl/3, then in where both forces have the same using the potential of Godfried

where r is the radius of curvature (both is

do and r given in A). If the film-thickness on the flat surface is below this the Van der Waals forces dominate and the film distribution the valleys. In the valleys, unstable On the hills, the Van der Waals and If d > do' then the situation on top of the hills differs from the surface the as and tend to stabilize the thickness at d. As soon

tension forces counteract each other, value do' thickness becomes

both forces act in the same direction and with respect to variations in

d > do' the valleys become quickly filled with liquid. Surface ways. binding tension, the of roughness, Van in principle, affects the recombination rate in two enhances consider the a On top of the hills, where the film is locally thin due to the surface der Waals potential of the solid substrate To give a numerical example, the hydrogen atoms.

62

surface

covered

with will

spheres

with

a radius small

of 100 A. For potential wells

< 100A,
binding

about

10 %

of the surface

be covered

with

with

a depth

of 100 by the

mK

and a radius

of 100 A. We conclude

This will that this

increase effect

the overall is too small

energy for

about observed

10

mK.

to account

variation will

of the rate with also affect

thickness. the magnitude the rough of the surface

Roughness area. As more

the rate via

helium

is added

to the system,

structure

of the solid surface area.

substrate Due to

will

be smoothed filling

by the film, of the valleys area may result

leading in the

to a smaller surface, a

capillary of

considerable by

reduction some

the surface measuring a roughness

if the thickness, small

as measured amount.

thickness

device,

varies

by a relatively account et

We have rate

constructed variation consists calculated the surface area

distribution as observed

that may

for the observed al. The

with of four rate area

thickness length variation with

by Godfried

distribution

scales with

(r ~ 400 A, 4000 A, 40000 A and 400000 thickness, is shown The which follows 3.8. from

A). The
of the from

the variation that

thickness, 2.5.

in fig. shape

One notes

may vary

by a factor nature

step-like

of the curve

results

the discrete

of the distribution. as follows. We have and shown that that these substrate are much of a

We summarize effects larger flat are than

our considerations in both sets

important a model, can

of measurements, of the Van

based

on penetration for.

der Waals to

potential a

substrate,

account

It is possible the results

construct et al.

surface However, that

roughness it is not

distribution clear whether

that explains

of Godfried

such a distribution to explain

is realistic.

We also note, results. result the

the roughness Finally, formation especially necessary measurements low surface of

model we note a

is inadequate that

our own experimental roughness may

the presence which

of surface

in the

potential,

randomly

fluctuates of 20

along and below.

surface, is the the and are then

if the film ingredient described, densities usual transport

is thin, for Anderson

of the order

This

localization[38].

It does not

affect

but may be of importance of hydrogen atoms,

at very these

low temperatures localized surface one of is states is

if only motion

occupied. blocked; state ripplons electronic atomic to

The the

free particle-like of particles

along

the

occurs

by jumping

from

localized

another, in our

assisted case.

by the absortion via

or emission states

excitations, observed in in an

Transport

localized

systems; too.

it should,

at least

in principle,

be observable

system

63

3.2.3. Field The in 3.9.

dependence T = 381 mK

magnetic field dependence was measured at T = 300 mK and

the field range between 1.5 and 7.5 Tesla. At high fields (B>

The results are shown in fig.

3 Tesla), the rate is proportional to B-2. At lower We have calculated the rate as a function of field, hyperfine states, using a

fields, a deviation from the square law arises from thermal occupation of the upper hyperfine states. assuming binding a energy thermal distribution over the four

of 1.15 Kelvin and a two-dimensional cross-section of 0.4

for all possible pair collisions: (3.5) The contribution 1.3.1 for of the four hyperfine states is expressed the definition of f(B) and the other via symbols f(B) in (see this

section are from

equation). The solid lines in fig. 3.9 are the calculated results. Also shown the zero field values from Morrow et al.[8J; these values were derived good. The their data by interpolation and scaling to our surface-to-volume ratio.

The agreement between the calculations and the measured data is rather contribution of the upper hyperfine states to the rate constant is

At a field of about 2 Tesla the data are almost temperature independent.

roughly

proportional to exp[(2EA - 2~BB)/kBTJ. At a field of 1.7 Tesla both Boltzmann factors cancel. The implies assumption the To of thermal equilibrium of the process four and hyperfine deserves states further presence of a fast relaxation

investigation.

maintain

thermal equilibrium,

an electronic

relaxation

process is needed with a rate, at least comparable to the recombination rate. In search for a candidate, we considered several electronic dipolar and spinexchange relaxation processes. tail of the equilibrium. In fig. calculated lines 3.10 we show a compilation of all measured data, curves for some field values between 0 and 10 together with The solid Three Tesla. It turns out that aa-ac spin- exchange in the to maintain thermal density distribution is marginally able

have been calculated using a value EA = 1.15 K and

= 0.4

A.

body recombination with 4He vapor atoms as the third body has been studied by Jochemsen et al.[2lJ; they observed at high temperatures (=1 Kelvin) and zero field a density decay of the form 64

Fig. 3.9. Field dependence the surface recombination

~ THIS WORK,lOOMK oTHISWORK,l81MK o UBC

of rate.

The solid lines have been calculated using equation 3.5, of states. The by scaling the zero field assuming a thermal occupation the four hyperfine and interpolating measurements etal.fSJ. squares were obtained by Horrow

10-17

(3.6)

with K3 = 2.8*10-33 cm6s-l,

nH and nHe are the densities of the hydrogen and

helium gas atoms. The dotted lines in fig. 3.10 have been calculated assuming a field dependence of the form Kv(B) = Kv(0)*2€2. All 10 cm-l. data, factor measured We feel an since
s

data error and

have

been scaled to a surface to volume the the

ratio

of

that this is the best way of comparing in EA directly propagates into vice versa. This causes

experimental for 3-1),

pre-exponential

K B-2T-l/2

our initial value

K s B-2T-l/2

to deviate from the values,

obtained by others (see table

although the actual observed rates are in close agreement with one another. A possible explanation of the low value of EA obtained in our first is out. At the highest temperatures, the data might have been experiment carried by the limited temperature range in which the measurements could be

affected

resonance recombination in the tail of the density distribution. In part surface roughness may account for the large variation in Ks 65

-----_-

Kv
K~ff(A/v=10(ni1)

B=O

~ ,
,

/3 5
6,5

8 10

o o

°UBC • UBC =4 o AMSTERDAM = 5,6 • AMSTERDAM =7,9 oMIT =11 " CORNELL =8,3

10-21

234
1/T (KELVINI-1

Fig. 3.10. Compilation liT, dependence 66

of all measured

values of k~ff and Kv as a function calculated Kv.

of

for H on a 4He surface.

Solid lines:

field and temperature

of K~ff; dashed lines: calculated

10-12 10-13 ~,
\ 1\ j
I

,, \ \\ \,
\ \ \ \
\

Kv K:ff

(A/V = 10 cm-1)

~ \ \ 1\

10-14 10-15 10-16 10-17

tTl

I; \ \ ;\ \
I' \ I

,I
\ \

Q
I
\

I
\

,
I

B-P

,
I I
\

,
I
\

2 5

,
'~

,, ,

° 0°
,,
2 3 5

- 10u GI
II>

e u

18

x:: :::.
GIll>

>

8 10 20

x:: 10-19

10-20

° UBC
10-21 10-22

-OT • AMSTERDAM- 8 T

11T (KELVINr1

10

Fig. 3.11. Same as fig. 3.10 for H on a 3He or 3He-4He and Wa1raven[39J.

surface.

After Silvera

67

quoted results, films. In length

in the literature (see also table 3.1).

Leaving aside our own

early

the data in fig. 3.10 still show a variation of 50 % in Ks' indicating that surface roughness may be important, even in the case of thick

fig.

3.11 we give an analogous plot for the data, and a binding energy EA = 0.34 K.

observed in a 3He Both plots show at

covered sample chamber[17,22]. The curves have been calculated using a cross-

= 0.4 A

qualitatively

the same picture.

In the case of 3He,

bulk recombination

high temperatures is somewhat more important. 3.2.4 Deuterium Atomic deuterium has a binding energy on liquid helium-4 of 2.6 Kelvin[2] and a recombination cross-length of 400 A[23]. length obvious that the present Van der Waals Since the value of the crossit is not to model applies exceeds the hydrogen value by three orders of magnitude, recombination No experimental data at low fields are available.

deuterium. the that first may

In the present system, atomic

apparatus, no observable quantities of deuterium could be stabilized, whereas successfull stabilisation of deuterium took place in the be of help in the design of an apparatus for stabilizing described in section 3.1. deuterium. Both system, systems the differ with respect to their filling lines: in the "old" a A comparision of both systems provides information

atoms enter the cold part below 1 Kelvin of the apparatus in

region where the field is about half the value at the center; system, the atoms must travel a field. Assuming case of deuterium,

in the present in the

distance of about 10 cm along a tube in zero

that the present model of recombination also holds

the recombination losses in this part of the filling tube

may be estimated. Recombination losses in a tube were treated by Walraven and Silvera[24]. We find that a flux of 1013 s-l will pass the tube without appreciable losses in a 6 Tesla field at a temperature of .7 Kelvin. field however is zero, only a small fraction (=10-4) survives recombination. deuterium atoms down below 1 Kelvin in a reasonably strong field. If the without

These results indicate that it is of vital importance to cool

68

3.3 Recombination 3.3.1 Introduction Recombination dissociation between prevent the the low

heating

of

atomic

hydrogen

into

the

molecular

state

with of

energy

of 4.476 eV is accompanied by the production sample, thus or

large the

amounts of heat. As a result, substantial temperature differences may develop gas and the liquid that surrounds the temperatures in a high density gas, across liquid-solid interface near the metal wall of the sample container. providing This may another of

barrier on the route to BEG. In this section an experimental investigation of heat transfer from the gas to the thermal bath, Recombination energy atoms surface. the using a combination ESR and pressure measurement will be described. results in the formation of molecules in a highly involving several collisions with excited hydrogen helium vibrational-rotational state near the continuum level, releasing their excess in a multi-step process, Hence there in the gaseous phase or collisions with the surrounding liquid are two distinct paths along which the

recombination as by an the

heat can flow: one goes from the "hot" molecules directly into the liquid and underlying substrate, We may the other uses the gas of hydrogen atoms depict the flow of recombination heat intermediate step.

following diagram:

liquid helium cold H atoms

solid substrate

Fig. 3.12 Hodel of recombination Atoms exchange in

heat flow surface an amount of

a single collision with the

energy, characterized by the accommodation coefficient a, defined by: 69

(3.7)

where Ti is the temperature of the particles that hit the surface, temperature surface. of the reflected particles and Ts is the The net heat flux Qfrom the gas to the surface and the

Tr is the of the temperature

temperature

difference Tg - Ts are related by: (3.8) This linear relationship only holds for small Tg - Ts' The Kapitza resistance Rk may be expressed in terms of the accommodation coefficient: (3.9) where surface that and on O<n<l.
v is the gas to surface collision rate.

The atoms thermalize with the (sticking) or inelastic show Calculations [25-28]

either

by entering a surface bound state

scattering below

due to ripplon emission or absorption.

300 mK sticking is the most efficient energy transfer dependence:

that it has a relatively weak temperature a pure 4He surface, temperature. fluxes with They to heat

mechanism Rk = l/Tn, with [29,30] with steady and the The for a decreasing and gas

The accommodation coefficient was measured by Salonen et al. yielding a value of 0.3 at 200 mK, They employed ballistic heat pulses study the energy transfer between the [42] found an increasing a similar temperature using

increasing state liquid. sticking resonance. by

Recently,

Helffrich et al. increasing

accommodation technique. magnetic

cogfficient

cogfficient

s was measured by Jochemsen et al.[22] using Recently, for 3He and

found s~O.046 at T=200 mK for 4He surfaces and s~O.016

3He surfaces at T=lOO mK. yielding resistance s~0.33T between both

the sticking cogfficient was determined capillary, Kapitza is copper 4He surfaces. The measured

Berkhout et al.[41] from the flow of hydrogen atoms through a liquid helium and a metal susbstrate such as

given by RkT3 Little the cold accomodation obtained 70 is gas

0.05 k4W-lm2[31]. Third sound experiments yield a large value known about the energy transfer between hot H2 molecules atoms or the surrounding walls. an upper With known values of may bound of the gas temperature and the be

of Rk between a thin film and a solid substrate [32]

coefficient,

assuming that all recombination heat is transferred to the

surface

240 230
:z:

Vv

Fig. 3.13 Gas temperature function of recombination resistance. power. The slope gives Kapitza the

as a heating

220

>
_J

~ 210
UJ
_J

0 0

~ 200
cr:
IUJ

_J

~ 190
cr:
n,

-<

UJ IVl l!:J

~ 180
170 160 0

UJ

0
0

-<

HEAT LOAD

5 (MICROWATTJ

10

by the gas atoms. The only reliable therefore a direct measurement.

way of determining

the gas temperature

is

3.3.2 Results. Given the ideal gas law, Tg = p/(kB(na determined total by a simultaneous meausurement density (na + nb) using ESR. + nb))' the gas temperature of p using a pressure may be

gauge and the

The total heat production

Q follows from

the time derivative

of the total density: (3.10)

where Ed is the dissociation The derivative and time. d(na + nb)/dt data with respect the density The

energy per molecule is obtained

and Vo the effective

volume. ESR of

by differentiating

the smoothed as a function

to time. To obtain the data for Tg and in the lower hyperfine were carried out in a 3He-4He

0,

both the pressure chamber,

states are recorded

experiments

lined sample

with the wall kept at a temperature runs are plotted

of 164 milliKelvin. The results of four 16 cm-3, the gas heats up 70 in fig. 3.13. At a density of 10 resistance Rk = 9000 K/W. is The from the differentiation procedure to obtain

mK.
In

The

slope

of the data yields a Kapitza the contact area between

scatter

in the data results

O.
20 71

our geometry,

the sample and the surface

cm2. value

Using the

the relations,

given at the introduction of

this

section, have

we a is

estimate

Kapitza resistance between the gas

and the liquid to

of 2000 K/W;

the Kapitza resistance at the liquid metal interface

estimated to be 5700 K/W. resistance a silver has a

These results suggest, that the Kapitza resistance As this temperature dependence than the this gas-liquid interface.

at the liquid-metal interface is the most important in our geometry. stronger

resistance, we expect it to be more important at low temperatures. The use of sinter generally reduces the temperature drop at However, this reduction is always limited due to temperature gradients inside the sinter (see section 2.1 for a detailed discussion).
spin waves

3.4

Search

for

internal magnetic structure of the hydrogen atom has been predicted to have strong effects on its transport properties[33, 34,431. The so called identical between are This spin rotation effect originates from succesive binary transferring their transverse collisions identical particles, magnetization. Exchange effects of

The

This may lead to coherent propagation of spin oscillations. the interatomic potential, condition

important if the thermal de Broglie wavelength Ath exceeds the range is easily met at currently attainable densities.

even in a non-degenerate, weakly interacting gas. Standing driven by an a

waves may be observed in a closed geometry in a field gradient, RF field. Spin waves have been observed in NMR spectra[35,361

in the form of

density dependent modulation. lineshape and An is with waves in the ESR spectrum. illustrative always example even

We have undertaken an investigation of the ESR to explore the possible existence of these

in a gradient field,

Several derivative spectra at different densities technique. obtained The does by is shown in fig. 3.14, which is was

field gradients were recorded using the frequency modulation

integrating the derivative of the ESR spectrum. A reproducible fine structure observed, when the applied gradient zero. but field not gradient merely smears out the main resonance peak. The fine structure varies the applied field gradient and the microwave frequency, This leads to the conclusion, depend on the density. that the observed fine

structure cannot be interpreted in terms of spin waves. More probably, the observed fine structure in the ESR spectrum is related to an inhomogeneous distribution of the RF field, 72 resulting from reflections

Ie> -IC> TRANSITION B(O) = 5.6 TESLA

dB/dz =
Nb/Na=

8.5 GAUSS/CM
5

-5

+5

+10

+15

6B (GAUSS)
Fig.

3.14.
of

ESR

1ineshape

in a gradient scale may

magnetic

field. into

For

a given distance.

field The

gradient, wavelength

the horizontal the microwave

be translated is shown for

an axial

radiation

comparison.

at

wall irregularities such as the inlet of the pressure gauge. as a function of frequency and field. If we assume

This effect that the

was discussed in section 2.3, in relation to the observed behavior of the ESR lineshapes spatial density distribution is uniform and the field distribution is linear, then fig. 3.14 may be considered as a one dimensional mapping of the RF field distribution. scale pattern radiation the between is With probed the known value of the field over a distance of 1.5 cm. comparison. gradient, In the fig. horizontal 3.14, the the RF and of the more may be converted into a distance along the axis. is shown for

The wavelength A of a

The close correspondence between A

typical period of the oscillations supports the idea of

relationship

the fine structure in the spectra and a non-uniform distribution we note that Statt[37] also failed to observe spin waves in These results indicate that for some reason it is

the RF field. Finally ESR spectrum.

difficult to observe these waves with ESR than with NMR. We also note that we did not design our system with the possibility of these waves in mind.

73

References

1. A.P.M. Mattheij, J.T.M. Walraven and I.F. Silvera, Phys. Rev. Lett. 46, 668 (1981).
2. I.F. Silvera and J.T.M. Walraven, Phys. Rev. Lett. 45, 1268 (1980). 3. J.T.M. Walraven, I.F. Silvera and A.P.M. Mattheij,

Phys. Rev. Lett. 45, 449 (1980).

4. K.

Huang in

Studies

in Statistical

Mechanics

II,

J. de Boer and (1980).

G.E. Uh1enbeck eds., North-Holland Pub1. Co. Amsterdam (1964). 5. D.G. Friend and R.D. Etters, J. Low Temp. Phys. 39,409
6. R. Sprik, J.T.M. Walraven and I.F. Silvera,

Phys. Rev. Lett. 51, 479 (1983). R. Sprik, J.T.M. Walraven and I.F. Silvera, Phys. Rev. B32, 5668 (1985). 7. H.F. Hess, D.A. Bell, G.P. Kochanski, D. K1eppner and T. Greytak, Phys. Rev. Lett. 51, 483 (1983). H.F. Hess, D.A. Bell, G.P. Kochanski, D. K1eppner and T. Greytak, Phys. Rev. Lett. 22, 1520 (1984). D.A. Bell, H.F. Hess, G.P. Kochanski, S. Buchman, L. Pollack, Y.M. Xiao, D. K1eppner and T. Greytak, Phys. Rev. B34, 7670 (1986). 8. M. Morrow, R. Jochemsen, A.J. Ber1insky and W.N. Hardy, Phys. Rev Lett. 46, 195 (1981); erratum 47,455 and B.W. Statt, Phys. Rev. B31, 7503 (1985). 10. H.P. Godfried, E.R. E1ie1, J.G. Brisson, J.D. Gillaspy, C. Ma11ardeau, J.C. Mester, and I.F. Silvera, Phys. Rev. Lett. 55, 1311 (1985). 11. R.W. Cline, T.J. Greytak and D. K1eppner, Phys. Rev. Lett. 47, 1195 (1981). 12. R. Sprik, J.T.M. Walraven, G.H. van Yperen and I.F. Silvera, Phys. Rev. Lett. 49, 153 (1982). 13. B. Yurke, J.S. Denker, B.R. Johnson, N. Bigelow, L.P. Levy, D.M. Lee and J.H. Freed, Phys. Rev. Lett. 50, 1157 (1983). 14. B.W. Statt, W.N. Hardy, A.J. Ber1insky and E.J. Klein, Journal of Low Temp. Phys. 61,471 Phys. Rev. Lett. 46, 1461 (1981). 74 (1985). 15. F.M. Ellis, R.B. Hallock, M.D. Miller and R.A. Guyer, (1981). 9. M.W. Reynolds, I. Shinkoda, W.N. Hardy, A.J. Ber1insky, F.W. Bridges

16. A. Lagendijk,

Phys. Rev. B25, 2054 (1982). A.P.M. Mattheij, J .. T.M. Walraven and I.F. Silvera,

17. G.H. van Yperen, 18. D.S. Zimmerman 19. E. S. Sabisky 20. M.W. Reynolds,

Phys. Rev. Lett. 47, 800 (1981). and A.J. Berlinsky, and C.H. Anderson, I. Shinkoda, (1986). R. Jochemsen, B.W. Statt, P.R. Kubrik, Phys. Rev. Lett 45, 453 (1981). and W.N. Hardy, Can. J. Phys 61, 508 (1983). Phys. Rev. A7, 790 (1973).

R.W. Cline and W.N. Hardy,

Phys. Rev. B34, 4912 21. W.N. Hardy, M. Morrow, R.M. Marso1ais 22. R. Jochemsen, 23. R. Mayer 24. J.T.M.

and A.J. Ber1insky, M. Morrow,

A.J. Ber1insky

Phys. Rev. Lett. 47, 852 (1981). and G. Seidel, Proc. LT 17, 455 (1984). Rev. Sci. Inst. 53, 1167 (1982). (1983). (1985), J. Phys. Lett. 44, L-537 and N.A. G1ukhof Walraven and I.F. Silvera,

25. B. Castaing 26. Yu. Kagan,

and M. Papou1ar, G.V. Sh1yapnikov

JETP Lett. 40, 1072 (1984). 27. B. W. Statt, 28. V.V. Goldman, 29. K.T. 30. K.T. Sa1onen, Sa1onen, Phys. Rev. B32, 7160 (1985). Phys. Rev. Lett. I.F. Silvera, S. Jaakkola, 56, 612 (1986). Walraven and G.H. van Yperen, and T. Tommi1a, below 1 K', J.T.M.

Phys. Rev. B 25, 6002 (1982). M. Karhunen, Principles E. Tjukanov and Methods Proc. LT 17, Physica, 31. O.V. Lounasmaa, Academic Physica Press 32. J.E. Rutledge, 33. C. Lhui11ier 34. C. Lhui11ier 35. B.R. Johnson, p. 543.

'Experimental W.L. McMillan

(New York 1974). and J.M. Mochel,

107B, 401 (1981). and F. La1o~, J. de Phys. 43, 197 (1982). and F. La1o~, J. de Phys. 43, 225 (1982). J.S. Denker, N. Bigelow, L.P. Levy, J.H. Freed and and A. Rij11ard, and Techniques Bonn (1985).

D.M. Lee, Phys. Rev. Lett. 52, 1508 (1984). 36. T.O. Niinikoski, W. Meyer S. Penti11~, J.-M. Rieub1and Institut, Proc. 4th. Int. Workshop 37. B.W. Statt, Ph.D. Thesis, unpublished 39. I.F. Silvera (1984). Phys. Rev. 109, 1492 (1958). and J.T.M. Walraven, Progress in Low 75 38. P.W. Anderson, on Polarized University Target Materials Universitah"t of British

Ed., Physika1isches

Columbia,

Temperature

Physics,

Vol. X, Chapter (1986).

D, D. Brewer

Ed.,

North Holland 41. J.J. Berkhout, 42. J. Helffrich, 43. J.P. Bouchaud

Publ. Amsterdam E.J. Wolters, M.P. Maley,

40. ,B.W. Statt and A.J. Berlinsky,

Phys. Rev. Lett 45, 2105 (1980). R. van Roijen, J.T.M. Walraven,

Phys. Rev. Lett. 57, 2387 (1986). M. Krusius and J.C. Wheatley, Phys. Rev. B34, 6550 (1986). and C. Lhuillier, J. Phys. 46, 1781 (1985). 44. B.W. Statt, private communication.

76

CHAPTER

IV

ESR PUMPING

EXPERIMENTS

IN SPIN-POLARIZED

ATOMIC

HYDROGEN

A.P.M.

Mattheij+,

J. van

Zwol++,

J.T.M.

Walraven,

Natuurkundig Valckenierstraat

Laboratorium,

Universiteit

van Amsterdam, The Netherlands.

65, 1018 XE Amsterdam,

Isaac Lyman Laboratory

F. Silvera Harvard USA. University,

of Physics,

Cambridge,

MA 02138,

Pumping technique. of with a

experiments A model

were

carried

out

in atomic

hydrogen states

using

an

ESR

of the decay resonant

of the four hyperfine microwave field

in the presence and spin compared exchange of the

moderately experimental it

strong

is described that due to

results. is difficult

The results

indicate

relaxation, lower design estimate

to influence

the nuclear

polarization is applied

hyperfine

states

by means beam

of ESR pumping.

The model

to the We from

of an ESR operated that was 5xl013

source

of spin polarized

atomic

hydrogen. ejected

in our experiments sec-I.

the maximum

flux of particles,

the source

Present

address:

Hogeschool

West

Brabant,

PO Box

90.116,

4800 RA Breda,

The Netherlands. Laboratorium, The Netherlands. PO Box 80.000

++ Present

address:

Philips

Natuurkundig

5600 JA Eindhoven,

77

Int:roduct:ion Electron the been decay spin resonance has proven to be a valuable tool in the study of of spin applied polarized to study atomic various hydrogen. magnetic Using ESR, the and

kinetics

densities in the lower hyperfine states may be determined separately. ESR has succesfully relaxation recombination processes[1-4], and to observe atomic hydrogen in a surface

bound state[3]. Alternatively, instead of using ESR as a detection scheme, it may be applied at a high power level to directly influence the populations of the hyperfine states. of the atoms In a strong resonant microwave will be reversed, states, field, the the electron of in or a spin To spins the which results in creation The atoms regions,

electron spin-up atoms in a background gas of spin-down atoms. upper hyperfine created by ESR in an field, magnetic hydrogen explore was 4He made will be driven by magnetic forces to low magnetic field with spin-down atoms. field

inhomogeneous

magnetic

recombine

ESR-pumping and subsequent ejection from Conceivably,

has been proposed as a means of production of a doubly ESR of [6] a the

polarized atomic hydrogen beam [5].

very pure doubly polarized study lower of and

may be produced by removing the la> state atoms with of the effect of microwave pumping on the decay

these potential applications of the ESR pumping technique,

hyperfine state densities. The pumping experiments were carried out in a 3He_ lined sample chamber at a temperature of 160 mK and a magnetic field We shall present a model of ESR pumping and beam extraction, such (escape as recombination, from the spin-exchange relaxation, Our results thus giving show that 5.6 Tesla. processes, effusion states results

compare this model with experimental results. source).

Our model will include various diffusion and at moderate hyperfine different used to

densities spin exchange plays a crucial role in relaxing the upper before than they can be ejected, substantially Our model is predicted from a more naive picture.

analyze the limitations and the optimum operating conditions of an ESR pumped beam source. A model for the density First lea). some 78 we decay during microwave pumping sketched in fig.

briefly discuss our experimental geometry,

The apparatus is the same as the one used by Van Yperen et al.[l] with minor modifications. For a more detailed description of the apparatus,

Z =-L I

Z0

Z=+L ,

(a)

1.0
0

£D

0.8

......
N

£D
0

0.6 0.4
0.2 0 T=160mK 8(0) = 5.6 TESLA

......

( b)
-2

10

Z (CENTIMETER)

Fig. l.a. Experimental a background shown. Fig. l.b. Magnetic the lower hyperfine

geometry.

Electron

spin-up atoms are created by ESR in is of

gas of spin-down

atoms. A typical trace of a spin-up particle and equilibrium density distribution

field distribution states.

79

the reader is referred to this paper. design, Tesla. sample. Kelvin. in the spin-down the atoms operating It consists in Changes at

The spectrometer is of a FIR broadband and a central field filled with a of 5.6 are atoms of localized

a frequency of 157 Ghz,

of a long cylindrical tube,

the transmission due to the presence of the sample

detected by a semiconductor bolometer, which operates at a temperature of 0.5 ESR transitions create a small population of electron spin-up center of the magnetic field, atoms. in In fig. l(a) we also depict a possible path of a magnetic in the middle of a background gas

spin-up field

atom. In fig. l(b) we show a calculated typical axial density distribution of the lower hyperfine states together with the extending from z = -L to z profile. The density profile is approximately rectangular in shape, and has a length of 10 cm, L+. In our coordinate system, z = 0 corresponds to the center of the magnetic field and z = -L_ corresponds to the location of a mylar window, The width of where the density abruptly falls to zero. ESR transitions are for the zone at the field center where the

induced is determined by the field homogeneity. At resonance, this zone has a width of about 1 cm in our system. In the following treatment we shall
=

simplicity assume that all transitions take place at z

O. To escape without

recombination at the walls, the atoms with reversed spin have to traverse the zone to z = L+, a distance of about 6 centimeter, where the field gradient is practically phase may zero and where their motion is diffusion like due to result with in spin exchange, while wall collisions may collisions result the in to with the background gas and the sample container walls. Collisions in the gas recombination atoms in the lower hyperfine states. Once spin-up

atoms have left the background gas, the field gradient. The hyperfine involves and time. calculation states the We in

they will be rapidly accelerated due

of the temporal behavior of the densities of the of presence of a four coupled strong partial resonant differential

the

four field

microwave

solution

equations, The form of to

describing the occupation of the four hyperfine levels as a function of space use a one-dimensional treatment for simplicity. relaxation, ESR these equations is governed by conservation of mass, recombination, the field gradient. including terms due

and transport due to diffusion and drift in + d the equations are

Using a shorthand notation a,b,c,d for na(z,t), nb(z,t), a+b+

C

nc(z,t) and nd(z,t), and n

80

aa/at

- 2Kaa a2 - Kab ab - Kac ac - Kad ad + Gimp (b - a) (la)

ab/at

Gimp (b -a) - Gex (bd - ac) - W~~R (b - c) - aJb/az + Gbc~ nc

(lb)

ac/at
+ Gex (bd - ac) + W~~R (b - c) - aJc/az - Gbc~ nc (lc)

ad/at

- Kad ad - Kbd bd - Kcd cd - Gex (bd - ac) + w~~R (a - d) - aJd/az - Gad~ nd . (ld) As the lower

The all

K's refer to the two-body surface-recombination rate constants. we three-body terms.

densities in our experiments do not exceed a value of 1016 cm-3, hyperfine states is denoted by Gimp, dipolar relaxation is given of

neglect dipolar by

The impurity-relaxation rate constant for the Gad~ and Gbc~ are the two-body as it is

relaxation rate constants for d~a and c~b relaxation[lll. We neglect two-body the lower hyperfine states, dominated impurity relaxation in our experiments. relaxation exchange between fields, the case by created channel The rate constant for spin-exchange

by Gex. In a strong magnetic field, the only spin with a non-vanishing cross section involves collisions This process is normally unimportant in the at strong case

ac or bd pairs[71. of

if the occupation of the upper hyperfine states is zero. However, in a small population of the upper states, present

ESR, the spin-exchange rate may be very important. The rate ESR and WESR represent the transition rates due to ESR; as constants Wad bc outlined before, we assume that all transitions take place at z ~ 0 or WESR(z) ~ WESR generation relaxation rate terms 6(z), is with 6(z) the delta function. reduced. For but this reason rather we As spin-up atoms will them not in treat the net after these ESR generation by ESR may relax back to the lower hyperfine states, the effective explicitly, include

generation rates. Finally, Ja, Jb' Jc and Jd are the current densities due to diffusion or migration; their form will be discussed in the following 81

sections. In the following we shall solve these equations with a simple approximate model, rather than finding a rigorous solution, for instance by means of a Monte Carlo simulation. the pumping experiments. In our experiments, power The analysis yields a rather simple picture in terms and explains all of the salient features of We estimate the accuracy of our model to be of the to be described later, with a moderately strong to

of a few characteristic lengths, level of tens of percent. microwave of about 100 nW,

it takes at least a

hundred

seconds

With equal occupations of the four hyperfine states, the gas has a lifetime of a fraction of a second at a total density of 1016 cm-3. This the implies analysis. of that The the densities of the upper hyperfine recombination terms in equations states (la-ld) are small compared to the lower hyperfine state densities. This considerably simplifies containing these treat we products the such upper hyperfine state densities and the contributions of Furthermore, us to

remove the sample.

states to the ESR transition rates may be neglected. density distributions as quasi stationary, as switching on the microwave field. the

neglecting transient effects effectively To justify diffused diffusion a time

This allows

decouple this step, the over an

time and the spatial dependence of the densities. After a time t, <z> = (2Dt)1/2, spin-up

we estimate the time required to build up a stationary density in the spin-up atom has where D is the spends roughly average An distance electron

upper hyperfine states.

coefficient.
r = L+2/2D r = 0.1

atom therefore

in the background gas,

sec.

before it escapes to low field. We find at a total density of 1016 cm-3 (a more detailed discussion of After switching on the microwave we may we may solve spatial

diffusion constant r. regard the the

processes will be given later).

field, the density in the upper hyperfine states builds up with the same time As the total density decays on a much slower time scale, Thus first calculate the density distributions, for the densities, averaged over the spatial density distributions as quasi stationary.

proceed in two steps: rate the equations

neglecting

time derivatives of the densities in equations (la-ld),

and then the

distributions. losses,

Our final object will be the calculation of the recombination intensity of the emerging beam of spin-up atoms and the nuclear

polarization of the beam.

82

Density

distributions

Our central field zero. recombine center. also states having

system, part

sketched

in fig.

lb,

may be divided is vanishingly

into

three

parts:

the the

where

the field

gradient linearly, the

small,

a zone where

falls

off approximately atoms, as

and a zone where zero-field probability

the field will to

is almost eventually the field and

Spin-up there, We

entering they have

region,

a small

to return

assume

that

all densities in the density We approximate

are zero

in the zero-field of the lower

region

neglect beyond

the tail z

distribution the real

hyperfine by a field

L+.

field

distribution

a trapezoidal

shape:

B(z) B(z) B(z)

where zero. Tesla/cm

(2a)

(2b)

o
over which the field

z > L+ + ~z.

(2c)

~z is the distance In our geometry

the field gradient

drops

from

its maximum of

value

to 1

has

a value

approximately

and ~z = 5.6 cm.

In the central
diffusive, diffusion density (lc) and and

portion

of the density density

distribution,

the atomic where

motion

is

the current

is of the form J = -Dan/az, equations for the follow

D is the

coefficient. distributions (ld). Retaining

The

differential

quasi-stationary from equations

of the upper only

hyperfine

states

the leading

terms,

we get

K (a + b) c(z) + Gex

[ac(z)

- bd(z)]

(3a)

K (a + b) d(z)

- Gex

[ac(z)

- bd(z)]

(3b)

Gex

is

the

spin-exchange processes K. Using cm3/sec.

rate between

constant spin-up

and

we

have

assumed have

that a

all

recombination rate constant

and spin-down section

atoms

common

a spin-exchange The surface

cross

of 1 A2

[12] we find Gex measured by

8.2xlO-13 Hardy et al.

recombination

rate has been et al.

[8] for a 3He surface

and by Van Yperen

[9] for 3He_4He

83

mixtures. Using a binding energy of 0.39 K and a two-dimensional cross length of for all processes involving a spin-up and a spin-down particle we find K = 2.3xlO-ls cm3/sec at a field of 5.6 Tesla and a temperature of 0.16 Both diffusion particle-wall coefficient. and particle-particle collisions contribute The overall diffusion is estimated by simply to the O.lS

Kelvin. adding

the flow impedance due to both kinds of collisions; this gives

liD
where and Dk = (2/3)av is the Knudsen diffusion constant (a is the tube Here we neglect specularity in

(4)
radius no We

the

thermal velocity). as

atom-wall

collisions

observed by Berkhout et al.[13].

This seems justified as coefficient. (3b),

particular care was taken to obtain a flat substrate for the helium film. Our results are not too sensitive to the value of the diffusion use Do (a + b) = 1.SxlOlS cm-lsec-l. [10] Another set of two equations, for

similar to equation (3a) and

holds a-a 0, a

the lower hyperfine state distributions,

with additional terms for

=

and a-b recombination and b-a relaxation. To create spin-up atoms at z supply the lower axis Assuming of down-state atoms by diffusion is needed. hyperfine that four states are therefore of the same order of

The density gradients in magnitude. to the zstate lower

the densities of the upper states are small compared one expects that the presence of the upper the will the

state

densities,

distributions states. Then

cause a relatively small density variation along the We shall neglect this perturbation of distributions Ib, to z = +L+. length scales reduce to the in fig. spin-down

in the lower states. sketched

equilibrium

distributions, It is

which we approximate by a rectangular to

distribution, extending from z = -L

instructive to introduce three characteristic

express the relative importance of recombination and spin-exchange phenomena:

Lrec

[D/(K (a + b))]1/2

(Sa) (sb) (sc)

S4

100~--~~~------~------~-----'
Nb/Na=5

B =56 TESLA T=0.16 KELVIN

10
~
L
_j

~+

,<-

><1.0

(l)

_J

_j

0.113 10

DENSI TY (CW3)
of the characteristic

10

16

10

17

Fig. 2. Density experimental

dependence (L+).

lengths Lrec' L~x' L~x and length scale of the

the mean free path ~. Also shown is the characteristic geometry

In terms of these quantities, equations (3a) and (3b) may be rewritten: (6a) (6b) The physical meaning of Lrec'
r = (Kn)-l

L~x and L~x may be understood as with the recombination

follows.

lifetime thus axial is be

is

associated

process, in the

corresponding to an average diffusion length of (2Dr)Yz.

The quantity Lrec may down-state of the atom.

regarded as the typical distance a spin-up particle drifts

direction before it is lost by recombination with a clear that these lengths determine the contours

Similar considerations apply to L~x and L~x. distributions. The in fig.

From equations (6a) and (6b) it density

density dependence of all relevant characteristic lengths is plotted 2. At all densities a constant nuclear polarization nb/na = 5 At low densities «IOI4cm-3) has the particles principally collide 85

been assumed.

with the walls, and the mean free path is of the order of the diameter of the sample chamber; n-l/2. dominate and in this density regime Lrec' L~x and L~x are proportional to In the high density regime (>3xl015cm-3) collisions in the gas phase these to n-l. are lengths, In fig. together with the mean free path and

are high

proportional which the

2 the asymptotic behavior at low density. If the the

densities is also shown (dashed line). processes particle may characteristic length is longer than L+, escape,

From the figure it is easy to observe corresponding process under the width of the distribution, then

important at a given

without getting involved in

consideration (recombination or spin exchange). The experiments were carried out in the density regime of 1015_1016 cm-3, where Lrec ~ L+; from the behavior of L~x and L~x it is clear that a flipped spin is involved in several spin-exchange collisions before it is ejected. Adding equations (6a) and (6b) gives a simple differential equation for

the total spin-up distribution in the central field region: Lrec -2 [c(z) + d(z) 1 The solution of this equation is a sum of exponentials. In down the region z > L+, the spin-up particles are driven to low-field drift

(7)

regions by the density and field gradients. by wall collisions.

The diffusional motion is slowed the atoms move with a

In a field gradient,

velocity ~Igrad BI. Here the mobility ~ is given by

(8)

where Dk is the Knudsen diffusion constant and ~B the Bohr magneton. the contributions from the density gradient (diffusion) and

Summing field

the

gradient (drift) leads to the following expression for the current density:
(9)

where nt = nc + nd' Since there is no loss of particles due to recombination, the current density should be independent of z. This leads to the following differential equation for the density distribution:

86

T=160mK No + Nb= 5.10 15 CM3

;: 0.6

-2
Fig. 3. Normalized

8 10 Z (CENTIMETER)

spin-up density distribution. (10)

The

spin-up

density

falls

off

exponentially

in

this

region

with

characteristic length LB, where l/LB = -~B/kBT dB/dz. For our geometry LB has a value of 0.24 cm. nt(L+ + ~z)
=

The solution of (10),

satisfying the boundary condition

0 is given by nt(z)
(lla)

(llb)

The resulting particle flux ~ is given by (12) Here A is the cross sectional area of the tube. The last step in finding the spin-up density distribution is to match the solution gradient -no flux in the gradient-free region and the density The distributions must satisfy i.e. the distribution following in the zone. boundary 0 for 87

conditions: at the location of the mylar window, d(nc + nd)/dz

z = -1_;
-the equals -the sum of the fluxes at z = 0 in the positive rate; and the particle a discontinuity at the field A typical flux are continuous in the flux. by solving in fig. the 3. functions of and negative z-direction

the ESR generation density distributions

z, except The coupled

at z = 0, where spin-up density (la-ld).

exists center

may be found

equations

spin-up

distribution states

is shown

The estimated typical

density

of the upper experiments

hyperfine

at z = 0 under of

conditions

for our pumping the

is of the order Id> state

1012

cm-3. the coupled from our

To calculate differential analysis in The a

individual (6a) and

Ic> and

distributions,

equations

(6b) have

to be solved. differ from

It follows the sum

that region

the individual near

distributions

distribution terms. more state we the

z = 0 of width of the state

1~x or 1~x due to the spin-exchange is created by ESR shows a peak,

distribution than

which

pronounced shows may

the maximum location.

of the sum distribution, Since put in our experiment

whereas L~x and

the other

a dip at this with reasonable

L~x ~ L+, equal to

accuracy

the individual

distributions

sum distribution. Once levels, at the density distributions are known for the various flux hyperfine

the losses are

due to recombination calculated As

and the particle

due to ejection density Ic> fluxes gas of and ~c the Id> and

z = 1+

easily

as a function

of the total

lower states ~d

hyperfine are

states.

the density

distributions

of the

the same,

it is convenient
TV spent

to express

the ejection

in terms

of the time

by the atoms

in the background and V, the volume

between

generation with

by ESR and ejection density

by the field,

associated

the spin-up

distribution:

~c

(13a) (13b)

Here

and

d refer

to the upper

hyperfine

state

densities

at

z = O.

The

expression

for TV follows

from equation

(12):
(14)

88

1.0

_------r------,_----__,
T=160mK B= 5.6 TE SLA

I.....I
(l)

co
n:: o,

0.6 0.4

lLJ

c, w
l/)

-c

0.2

lLJ

o~-------~---------~------~ 2 3
DENSITY (1016 CM-3)
versus the escape probability states. In .fig. 4 we The for a spin-up particle

Fig. 4. Calculated density

of the lower hyperfine

The time constant be ejected (i.e.

TV

is a function of the total density only.

plot the probability that a spin-up particle,

after generation by ESR,

will

not lost by recombination with a spin-down particle).

escape probability is small if the total density exceeds a value of 1016 cm-3. If a high flux is wanted, the total density should remain below this value. The density density calculated ratio is flux of ejected particles as a function has been assumed. The effect of of 5. the total

at a constant microwave power level is shown in fig. nb/na = 5

A constant

relaxation

clearly visible in the fluxes at densities of 1015

spin-exchange cm-3 and below, spin

above, where ~c/~d = 5, the same ratio as the density ratio of the lower hyperfine states. At densities of 3xl013 cm-3 and exchange is of less importance, displayed sec-l. at spin-down atoms.

but still visible. The maximum in the fluxes

high densities results from recombination between spin-up

and Thus at our current power level, the maximum flux is 5xl013

89

T=150mK B

5.5 Teslo

Nb/No = 5

4>c
x => _J
1J._

1012

UJ
_J

,_

U 0:: <{

n,

10

11

14 10

15 10 DENSITY

15 10

(ew3,

17 10

Fig. 5. Calculated

flux of Ic> and Id> state atoms as a function power of 0.2 W. A density-independent is imperceptible

of total ratio on this

density and at a microwave transition) scale.

nb/na~ 5 has been assumed. Note that the difference and ~c (pumping the a-d transition)

in ~d (pumping the b-c

@
nd x1000

OL-

L_

___J

50

100

150
TIME (sec)

200

Fig. 6. Computer 90

simulation

of ESR pumping.

Solucion

of the rate equations equations (la-

Within the approximations discussed in the last sections, ld) may be rewritten: da/dt 2 -2Kaa a

Kab ab - K a(c +d)

ESR Gimp (b -W ad a + db/dt

a) + Gex (bd - ac)

(lSa)

Kab ab - K b(c + d) _W~~R b - Gimp (b

a) - Gex (bd - ac)

(lSb)

dc/dt

- K c(a + b) + Gex (bd - ac) (lSc)

C/TV

dd/dt +WESR a ad Here a, b,

- K d(a

b)

- Gex (bd - ac) - d/Tv

(lSd)

c and d denote the densities at z = O. In the above equations we We have ignored the

neglect the differences in the effective volume of the four hyperfine states. To our estimate this introduces an error of 20 percent. diffusion terms in equations lSa and lSb. were data. obtained by The values of Gimp, WESR and WESR ad be the experimental a

fitting the calculated decay curves to

A rather large value for Gimp of 3xlO-4 s-l was necessary to obtain For this

good fit. The hyperfine

reason, two-body dipolar b-a relaxation was not included Some four

in the analysis. above set of equations was solved with the help of a computer. 6 where we plot the densities of the states as a function of time. typical results are shown in fig. tuned to the Ib>-Ic> transition, switched off. atoms pumping is built up. Similar

At the point A the microwave power,

is switched on. At B the microwave power is found if pumping the la>-Id> ratio, 91

Soon after pumping has started, a density of Ic> and Id> state results are

transition is simulated numerically. One finds the surprising result that ESR of one of the lower hyperfine states does not change the b/a

i.e. to be

the

final nuclear polarization of the lower hyperfine states turns out irrespective of the transition being pumped consequence of the or if

always the same, type of

there is no pumping at all. This other behavior is a direct spin-exchange field relaxation rates, rates. which are at least two orders of magnitude faster than all

The transient behavior after switching on the microwave the other is created by spin exchange; Only a small fraction of the atoms

may be understood as follows. Soon after one of the upper hyperfine states is created by ESR, until by spin this process goes on is in the upper the net spin-exchange rate is small. exchange. The term bd-ac is driven to zero

hyperfine states; consequently, the populations of the lower hyperfine states do not change very much by pumping. mechanism, power is comparable so high, Spin exchange thus provides a feed-back becomes no which locks the ratio c/d to the ratio bfa, unless the microwave levels In our system this

that the density in the upper hyperfine

to the density in the lower hyperfine states.

occurs at power levels of 1 milliWatt and above. longer applies. Experimental results

Then the above analysis

The effects of ESR pumping were studied by measuring the density decay of the used lower two hyperfine states under different pumping conditions. both the to disturb the population of the four hyperfine in the lower hyperfine states. field was manually the magnetic tuned to states one density To pump ESR and of was to the the two time

measure

transitions, resonance

resonance;

densities were measured by switching the microwave frequency between the regions and sweeping the field through resonance at constant

intervals. Since it is not possible with our ESR equipment to pump one of the transitions and perform a density determination at the same time, transducer was used to monitor the total density. Pin = 0.2 every ~W was used. passage through resonance. a pressure density level A constant microwave power this

This power level destroyed 5% of the total No attempt was made to reduce

since our signals were quite noisy. The data were taken at a temperature of 160 milliKelvin, the walls of the sample fig. container 7, being covered by a 3He-4He mixture. Results are shown in where we plot the measured densities na and nb and the polarization Three situations are compared

P = (nb - na)/(nb + na) as a function of time. 92

rn
I

L
LJ

S
I

_- -

(a)

.6z .4~ .2 (bl

U"I

-+

I-

~2
Vl

«
0::.
_J

D,

UJ

Z
)

0 0...

.6 .4

0::.

w
_J

LJ

2
(el

2Z

:::>

6 4

·6 ·4

• c

nb

na p= (nb -na)/(nb 3000 TIME

+nal

.2

4000 (SEC)

Fig. 7. Experimental in this figure: The separate

results. See text for a discussion. 7a), pumping the Ia>-Id> transition (fig. 7a was taken for reference a to

no pumping (fig.

7b) and pumping the Ib>-Ic> transition (fig. 7c). measurement without pumping of fig. the effect of pumping from the unwanted destruction by density

determination with ESR. After the cell is filled with a density of about 1016 cm-3 a moderate nuclear polarization (nb/na = 5) builds up in the initial portion of the decay. Starting at a total density of 6xl015 cm-3 (point 1 in fig. 7) the by microwave power is switched off to allow for a free decay, undisturbed ESR (dotted lines in fig. 7a). When the total density has fallen to a value of 3xl015 cm-3 (point IIa in fig. 7) the microwave power is At point I in fig. 7b the power is not switched off as in fig. 7a but in 93

switched on again to resume measuring the densities na and nb'

contrast density increases

decay is perturbed by pumping the la>-Id> transition in the interval 6xl015 cm-3 - 3xl015 cm-3 (dotted line in fig. 7b). This the decay rate by a factor of four, In fig. the final polarization turns rapid 7c the decay is perturbed by pumping the Ib>-

the

out to be unchanged. decay, and 7c. Our experimental However, the

Ic> transition in the same density interval.

This results in the most

since the rate at which spin-up particles are created is proportional

to nb and nb/na = 5. The final polarization is about the same in figs. 7a, 7b results from our reflect the same behavior measurements as the can numerical be From drawn our

calculations. relaxation. regarding numerical

Therefore, we attribute the observed behavior to spin-exchange no conclusion precise value of the spin-exchange rate constant.

calculations we estimate that the maximum ejection flux was 5xl013

sec-I, containing c and d state atoms in a ratio 5 : 1. Possible The Id> spin applications model of the ESR pumping technique help in evaluating followed the by

developed

in this paper may be of

applications

,of ESR pumping mentioned in the introduction.

Pumping the la>-

transition exchange,

in a fairly pure Ib> state gas is likely to be

sufficiently

resulting in la>-Ic> pairs. This process is inhibited at low densities (below 3xl013 cm-3 in our geometry, see fig. 2).

Purifying a gas of Ib> state atoms with a small la> state impurity by pumping the la> state is therefore restricted to low densities unless the value of L+ is small. If we take 1 cm as a practically achievable lower bound for the width of the density distribution, then this technique will be effective up to densities of the order 1015 cm-3. This is easily seen from fig. 2. Generating factors emerging geometry choice an atomic the beam by ESR pumping seems more promising. of the field the The the and best affecting (field for particle flux and the nuclear microwave power, to In polarization magnetic obtain

beam are the total density, configuration).

this section we discuss how

each of these parameters may be found,

maximum

particle flux. The temperature and the field mainly affect Lrec' With a 3He_4He covered sample chamber the recombination rate will be minimum at a temperature of 200 mK[7]. At this temperature, the volume and the surface contributions to the recombination rate are equal. Our experiment was performed near this optimum, 94