Sputtering of Uranium - CaltechTHESIS

Sputtering of Uranium - CaltechTHESIS
CaltechTHESIS
A Caltech Library Service
About
Browse
Deposit an Item
Instructions for Students
Sputtering of Uranium
Citation
Gregg, Ron
(1977)
Sputtering of Uranium.
Dissertation (Ph.D.), California Institute of Technology.
doi:10.7907/RQYH-V370.
Abstract
Sputtering yields for uranium metal under bombardment by 13 - 120
keV protons and by 20 - 120 keV He+ are presented. Angular distributions
of the material sputtered by these ions are also given. Sputtering yields
for 40 and 80 keV Ar+ were measured as well.
The technique employed to make these measurements was the detection
of fission tracks in mica produced by ^(235)U sputtered onto collector foils
which were subsequently exposed to a high fluence of thermal neutrons.
The technique is extremely sensitive and allowed the measurement of
sputtering yields less than 10^(-4) atoms per ion. It also made possible
a detailed study of the emission of chunks from the uranium targets
during sputtering. Mass distributions of chunks emitted during bombardment
by 40 - 120 keV protons and by 80 keV argon are presented.
Comparisons are made between the experimental results and those
predicted by the Sigmund theory of sputtering.
Item Type:
Thesis (Dissertation (Ph.D.))
Subject Keywords:
(Applied Physics)
Degree Grantor:
California Institute of Technology
Division:
Physics, Mathematics and Astronomy
Major Option:
Applied Physics
Thesis Availability:
Public (worldwide access)
Research Advisor(s):
Tombrello, Thomas A.
Thesis Committee:
Unknown, Unknown
Defense Date:
13 May 1977
Record Number:
CaltechTHESIS:07182014-082315611
Persistent URL:
DOI:
10.7907/RQYH-V370
Default Usage Policy:
No commercial reproduction, distribution, display or performance rights in this work are provided.
ID Code:
8558
Collection:
CaltechTHESIS
Deposited By:
INVALID USER
Deposited On:
18 Jul 2014 16:04
Last Modified:
18 Sep 2024 22:53
Thesis Files
Preview
PDF
- Final Version
See Usage Policy.
18MB
Repository Staff Only:
item control page
CaltechTHESIS is powered by
EPrints 3.3
which is developed by the
School of Electronics and Computer Science
at the University of Southampton.
More information and software credits
SPUTTERING OF URANIUM

Thesis by
Ron Gregg

In Partial Fulfillment of the Requirements
for the Degree of
Doctor of Philosophy

California Institute of Technology
Pasadena, California

1977
(Submitted May 13, 1977)

ii
ACKNOWLEDGEMENTS
For motivating this experiment in the first place, for his undiminished interest in it, and especially for allowing me to pursue the
research in my own style, my greatest thanks to Tom Tombrello.

I also

wish to thank Don Burnett and his crew in Mudd for initiating me into
the mysteries (if not the joy~ of the fission track technique.

Ziggy

Switkowski is also deserving of a special mention for his constant
enthusiasm and for good advice on many occasions.

Conversations with

Bob Weller, Joe Griffith, and others in Kellogg proved quite useful at
times.

The helpfulness and general friendliness of the many staff mem-

bers in Kellogg was greatly appreciated.
This section would not be complete without acknowledging the denizens of Page House, Chester's Jesters, and the many other friends here
at Caltech (who shall remain nameless for their own protection) who have
made the last few years so enjoyable.
The person deserving of the most gratitude of all is Sue, who gave
up so much that I might pursue graduate studies at Caltech.
I would also like to acknowledge the efforts of the scanners, Sharon
Streight, Robert Thornton, Dave Walker, Norm Murray and Joel Okazaki, who
spent many hours at the microscope counting tracks,

iii
ABSTRACT
Sputtering yields for uranium metal under bombardment by 13 - 120
keV protons and by 20 - 120 keV He+ are presented.

Angular distributions

of the material sputtered by these ionsarealso given.

Sputtering yields

for 40 and 80 keV Ar+ were measured as well.
The technique employed to make these measurements was the detection
of fission tracks in mica produced by 235u sputtered onto collector foils
which were subsequently exposed to a high fluence of thermal neutrons.
The technique is extremely sensitive and allowed the measurement of
sputtering yields less than 10- 4 atoms per ion. It also made possible
a detailed study of the emission of chunks from the uranium targets
during sputtering.

Mass distributions of chunks emitted during bombard-

ment by 40 - 120 keV protons and by 80 keV argon are presented.
Comparisons are made between the experimental results and those
predicted by the Sigmund theory of sputtering.

iv
TABLE OF CONTENTS
I. Introduction

II. Experimental Procedures

A. An Overview

B. Sputtering

The Sputtering Assembly

The Ultra-High Vacuum System

The Ion Beams

Target Preparation

The Sputtering Runs

C. Track Detection

The Mica

Neutron Irradiation

Etching

Scanning

III. Experimental Results

A. Target Surfaces

B. Angular Distributions of Sputtered Atoms

C. Total Sputtering Yields

D. Chunk Emission

E. Error Analysis

IV. Theory and Speculation

A. The Collision Cascade Theory

B. Comparison to Results

C. Chunk Emission

Summary

51a

VI. Appendices

A. Sensitivity of the Fission Track Detection Technique

B. Ultra-High Vacuum Technique

C. Effect of Adsorption on the Sputtering Yields

D. Neutron Fluence Determination Using NBS
Neutron-Irradiated Standards

VII. References

VIII. Tables

IX. Figures

-1-

INTRODUCTION

Sputtering, to put it as succinctly as I can, is the ejection
by a flux of particles of some of the constituent atoms of an object
from one of the surfaces of that object.

The principal quantity which

characterizes sputtering is the sputtering yield "S" defined as the
number of atoms ejected per incident particle .

Scientific investiga-

tion of sputtering extends back many decades and ~as grown into a
vigorous and multifaceted field, as one may quickly determine by consulting one of the many recent books and review articles on the subject.2•30

Design considerations for thermonuclear reactors provide some
of the strongest motivation for studying sputtering today. 20 This is
due to the fact that bombardment of the first wall of the reactor ves sel
by energetic particles will erode the wall and may sputter enough material to poison the plasma. Sputtering of the lunar surface by the solar
wind 31 • 47 and sputtering of such diverse astrophysical surfaces as those
found on Io 33 and on interstellar grains 51 are also topics stimulating
current interest in the field.
My own involvement in this sputtering business developed out of
another experiment--an attempt to measure the cross-section at energies
near 100 keV
for 10 B(p,a) 7Be using very thin 10s targets. It was
realized by Ziggy Switkowski that it was probable that the large proton
bombardments contemplated would sputter away a large fraction of the
target.

It would therefore be necessary to measure the amount of sput-

tering, which was jus t fine as far as Ziggy was concerned since the
measurement would be perfectly suited to the use of a very sensitive

-2technique--namely, the use of plastic track detectors and the 10 B(n, a )
reaction to determine the amount of sputtered 10 B collected on a foil.
I must digress for a moment to comment that the interest in sensitive
techniques is largely motivated by the low sputtering yields typi cally
produced by light ions.

The interest in sputtering by light ions

(such as neutrons, protons, and a particles) is largely motivated by
the fact that they are the primary constituents of the solar wind and
contemplated first-generation CTR plasmas.
In any event, Ziggy took his enthusiasm for this sensitive technique to a typical gathering of the minds in Lauritsen Library in
Kellogg, where Tom Tombrello proposed that sputtering of 235 u would be
even more sensitive, and easier.

One would still use the nuclear track

detection method to measure the amount of sputtered material collected
on a foil, but in this case exposure of the collector foils to a high
neutron fluence would result in 235 U-fission fragments which would
register in pieces of mica placed against the collector foils.

The

tracks would be made visible by etching in HF and could be counted
with the aid of a microscope.

The attractiveness of this proposal was

enhanced by the presence in the Caltech community of one of the leading
experts on the use of nuclear track detectors--Don Burnett. It also
just so happened that Don had some 99% pure 235 u metal foils left over
from the lunar neutron probe experiment. 6 My fate was sealed. Having
been sidetracked from 10 B(n,a) 7Be by proton sputtering of boron, I now
found myself pursuing proton sputtering of uranium.

-3-

The feasibility of the experiment depended on the magnitudes of
a few physical and empirical quantities which are connected by a
couple of simple equations.

The fission track density on the micas

placed next to the collector foils during neutron irradiation is the
quantity which must ultimately be determined by microscopic scrutiny
of said pieces of mica.

The number of tracks per square centimeter

is given by
(1-l}

where

Nu = number of 235 u atoms/cm2 on the collector foil
N = neutron fluence from irradiation in the reactor
n (=thermal neutrons per cm 2 )
a = 235 u fission cross section at 2200 m/sec = 580 bn

The density of 235 u atoms collected is given approximately by
Nu --~
2nr
where

(1-2)

S = sputtering yield (atoms/ion)
NP = total number of incident ions
r = distance from target to collector foil

Equations (1-1) and (1-2) can be combined to yield
( 1-3)

For the experiment to work, this equation must yield a track density
which is greater than the background track density due to 235 u in the

-4-

mica, in the collector foils, or acquired on the surface of either
from the environment.

Using information supplied by Don Burnett's
group in Caltech's Mudd Laboratory, I estimated that NT= 10 5cm- 2

would be sufficiently above background and would provide a convenient
track density for the microscopic scanning. (Both of these conjectures
proved to be correct).
Now for the right-hand side of equation (1-3). Based on the
work of Finfgeld et a1. 10 on the sputtering of several metals by
0.5-8 keV protons, I guessed that the yield for 100 keV protons on
uranium would be on the order of 10- 4 (which turned out to be true). It
was also known that a fluence of 1014 neutrons/cm2 was readily obtainable in the thermal colurm at the UCLA reactor.

A practical distance r

from target to collector foil is about 3 em, determined from considerations of ion-beam size and sputtering chamber geometry.

Plugging these

numbers into (1-3) then implies that a total incident proton dose of
~ 10 18 would be required to produce a final track density NT= 105cm- 2 .
This in turn implies bombarding the target with 10 ~A of beam for
5 hours, which is possible with the ion source available in Kellogg.
The conclusion was, therefore, that the experiment would be feasible.
In actual practice 10 ~A of 100 keV ions was found to deposit
more energy than the . 001" foil targets could to 1erate without severe
thermal crinkling.

Beam currents of about 1 ~A were employed instead.

It was also found that collector foils and mica track detectors were
just small enough to fit in a small plastic container which could be
lowered into the core of the UCLA reactor, affording neutron fluences

-5of up to 1o 16 cm- 2 .

This compensated for the reduced ion-beam current

and permitted hydrogen sputtering runs of between 15 minutes and
5 hours (depending on choice of NP and Nn). Due to the much larger
sputtering yield from helium (S ~ 10- 2 ), a typical helium run was 1 ~A
for 60 seconds.

For argon the sputtering yield is on the orrler of 2,

which meant runs with argon lasted less than a minute with ~10 nA of
beam to avoid too large a track density* in the mica detectors.
Jt was realized in the beginning (by Tom and Don at least) that
the experiment held promise in a number of ways.

First of all, it

ought to be sensitive enough to measure the sputtering yield for protons
with energies near 100 keV.

It was also expected that angular distribu-

tions for the sputtered material could be obtained.
expectations have been fulfilled.

Both of these

Sputtering yields and angular dis-

tributions have been measured for 13-120 keV protons.

Additionally,
angular distributions and sputtering yields for 40-120 keV 4He
have been obtained.

It was further hoped that the technique

would prove to be sensitive enough to imply the feasibility of
doping various materials with 235 u to allow low-sputtering-yield
studies of other materials utilizing the same technique.

The technique

has in fact proved to be sensitive enough, and this is discussed in
some detail in Appendix A.

Another area in which results were looked

for was the area of chunk emission, since chunks collected on the foils
would show up as star-bursts of fission tracks in the micas.
would pennit one to actually determine the chunk size.

This

In fact many

chunks have been seen on the collector foils, and data on the amount
*for convenient sc~nntng in ~n opticql microscope

-6-

of chunk emission and distributions of chunk size are presented . One
final bit of information that is presented is some limited data on
molecular sputtering by H2 and H3 .
Inasmuch as there are a number of parameters that one may vary
in a sputtering experiment, I will summarize here what was and what
was not varied.
1. The effect of beam energy on sputtering yield was inves tigated for 10-120 keV 1H and 40-120 keV 4He.
2. The angular distribution of the sputtered uranium was detenni ned.
3. The angle of incidence of the beam was not varied; all sputtering was done with perpendicular incidence of the beam on
target.
4. A comparison between the sputtering yield from slightly oxidized target surfaces and sputter-cleaned target surfaces is
presented, but there was no investigation of yield as a
function of surface finish.

All the targets were more or

less uniformly rough at the 1~ level.
5. There was no systematic study of dose effects; new target
areas were used for each run to eliminate possible effects
from non-un i form dose.
6. No effect from varying target temperature was looked for;
beam flux was always kept low enough that heating of the
target was negligible. (See p.l9)

-7Before we plunge ahead into the experimental procedures, a few
words should be said about why the sputtering is performed in ultrahigh
vacuum.

The primary concern is that the target surface be as clean as

possible during the sputtering runs.

Cleanliness of the surface is im-

portant because the sputtered atoms originate from within only a very
few monolayers at the surface.

If the surface is oxidized or has

various substances adsorbed on it, then the sputtering yield of uranium
will decrease as energy is expended to sputter atoms other than uranium.
Provided that the surface is clean at the atomic level at the beginning
of a sputtering run, it will remain so if the rate of adsorption of
gases is small compared with the rate at which they are sputtered away
by the ion beam.

Since the intrinsic sputtering rates for keV protons

are quite low, and since the beam current density must be kept low to
prevent overheating of the targets, the vacuum must be quite good to
avoid accumulation of surface contaminants during hydrogen sputtering
runs.

The vacuum requirements for sputtering by keV helium and argon

ions are less stringent because of their much greater sputtering
yields.

This is all discussed quantitatively in Appendix C. The conclusion is that the vacuum employed (on the order of l0- 9 torr) is good

enough to prevent decrease of the sputtering yield from

100 keV

protons by more than a few percent, as an upper limit.
Initial cleanliness of the targets is obtained by washing them
in HN0 3 and then sputter-cleaning them in vacuo with an argon ion beam.
(See Chapter II for details).

The nitric acid bath is necessitated by

the fact that uranium oxidizes progressively in air; which is to say

-8-

that an acquired oxide layer affords no protection against further ox i dation.

This means that over a period of weeks uranium targets stored

at atmospheric pressure acquire an oxide layer which is too deep to be
removed conveniently by presputtering, but which dissolves in HN0 3 in
a few minutes.

The uranium foils used in this experiment had been

stored in atmospheric conditions for an extended period and were given,
therefore, the complete cleaning treatment.

-9-

II.
A.

EXPERIMENTAL PROCEDURES

An Overview
The basic apparatus in which the uranium sputtering was conducted

is pictured in Fig. 1, and the important elements are sketched in Fig. 2.
The positions of the bands of sputtered uranium as they are collected on
the catcher foils are displayed in Fig. 3.
Approximately 40 sputtering runs were made to gather the data for
this thesis.

The primary steps required for each run are presented

here:
(1)

A metallic 235 u target roughly 1.5 em x 1.0 em x .0025 em is

cleaned by dipping it in 35% HN0 3 for a few minutes.
(2)

The target and others like it are mounted on a movable target

plate which in turn is mounted on the sputtering assembly (Fig. 6).
(3)

A cylinder designed to collimate the sputtered uranium is

attached to the assembly (Fig. 7).

Clean collector foils (5.0 em x

3.75 em x .005 em) are installed in a cylinder which is attached to a
manipulator (Fig. 8) so that a number of sputtering runs may be made
each time the assembly is placed in the ultrahigh vacuum (UHV) chamber
(Fig. 9).
(4)

down to

The assembly is placed in the UHV chamber which is then pumped
10- 9torr.
At this point most of the targets are sputtered by 80 keV Ar

to remove any lingering surface contamination.
by a 135 keV ion source.

Ion beams are provided

-10(5)

The sputtering runs themselves are now made using beams of

H1,
H+2 , H+3 , ..Her or Ar+ ; the collector fo1ls
are moved between runs so

that the sputtered uranium is collected in bands (Fig. 3).
(6)

The collector foils are removed from the assembly and inter-

leaved with mica foils of the same dimensions in a plastic container.
NBS standards containing known quantities of 235 u are also placed between mica sheets in the container.
(7)

The container is exposed to a high fluence of neutrons
(typically 10 15 -1o 16 ;cm2 ) in the R-1 nuclear reactor at UCLA to fission
some fraction (typically 10- 6 to l0- 5 ) of the collected 235 u.
(8) The mica foils are then etched (typically for 15-60 min) in
48% HF to make the fission tracks visible.
(9)

The micas are scanned with the aid of a microscope (at 450X

usually) to obtain the raw data.
(10)

The track density in the micas next to the NBS standards is

used to compute the neutron fluence which in turn is used in conjunction with track densities to compute the amount of 235 u collected at
various places on the collector foil.
and sputtering yield are determined.

Finally, the angular distribution

-11B.

Sputtering
In this section I will discuss the relevant aspects of
the sputtering assembly, the ultrahigh vacuum system, the ion source

and associated beam line, beam integration, target prP.paration, and the
actual sputtering.
The Sputtering Assembly
This appears in various stages of assembly in Figs. 4-8.

Figure 4

shows the portion of the assembly which remains in the atmosphere.

consists of an 8" OD UHV flange on which are mounted two precision
manipulators.

The left-hand one is a bellows-sealed 6-inch linear

travel feed-through designed by Jon Melvin and constructed in the Kellogg
shop.

It is used to position the cylinder carrying the collector foils.

The right-hand one is a VF-172-2 rotating/translating manipulator with
2" travel manufactured by Huntington Mechanical Laboratories.
used to position the target plate.
is pictured in Fig. 5.

It is

The vacuum side of the 8" OD flange

The end of the 6" travel manipulator protrudes

through the hole on the right.

Mounted on the end of the other manipu-

lator is a stainless steel block to which the target plate is mounted.
The block is electrically isolated from the manipulator by the boron
nitride piece seen directly below it.
In Fig. 6 the target plate which was used for most of the sputtering runs is shown attached.

It was machined from l/8" OFHC copper to

provide a good heat sink for the targets.

The targets were mounted by

clamping with pairs of copper bars, two of which are shown in position.
A piece of quartz glass for viewing the beam spot is also shown attached.

-12This target plate supplanted one made from l/8" aluminum, on which tl-)e
targets were clamped behind 8 nm holes by single l/8" bars of aluminum.
The sputtering assembly is next shown in Fig. 7 with the collimation cylinder in place on its insulating glass ring.

This stainless

steel cylinder has a 3/8" wide slot extending 180° around to limit the
collection area on the collector foils for a given run.

It also has

two pairs of vertical side bars to ensure that the target plate is kept
perpendicular to the beam.
Figures 1 and 8 display the entire sputtering assembly, which has
been completed by the addition of the collector-foil holder.

This

cylinder (which is also stainless steel) holds the collector foils with
the edges nearest the beam butted against the inwardly-protruding frame
around the vertical slot, and with the outer edges clamped by vertical
bars (see Fig. 2).

The collector-foil holder is moved by the 6-inch

travel feed-through to position the foils for different sputtering runs.
The resulting arrangement of bands of sputtered uranium is shown in
Fig. 3.
The Ultrahigh Vacuum System
A photograph of the UHV system appears in Fig. 9 and a sketch of
the important components comprises Fig. 10.

All connections between

components are sealed with OFHC copper gaskets.

The only materials used

in the system (i.e., stainless steel, copper, aluminum, glass, mica,
and boron nitride) are low vapor pressure materials which may be baked
at high temperature to drive adsorbed gases from the inner walls. Except
for the ion pump, the entire system may be baked at 450°C; the ferrites

-13-

in the ion pump's magnets limit its bake-out temperature to 250°C.
Proper cleaning of the components (see Appendix B for cleaning procedures) and baking enable a vacuum of <10-9 torr to be routinely
achieved.
The sputtering charri:>er is a 6" ID stainless steel tee, 11" high,
which was manufactured by Ion Equipment Corp.

It has three 8" flanges

to which the sputtering assembly, a large viewport, and the UHV pump
attach.

The pump is a COV-500 by Ion Equipment which is attached

directly below the sputtering assembly for maximum pumping speed in the
vicinity of the targets.

It contains both a titanium sublimation

getter-ion pumping assembly and a 25 £/sec differential sputter-ion
pump. The combination is rated at 500 £/sec for N2 at l0- 8 torr and
6£/sec for argon at the same pressure. When the system is reasonably
clean the pump is capable of bringing the pressure from 10- 3torr to
l0- 9torr in 12 hours.
The pressure rises to about 4x 10- 9 torr when a 1 vA H+ ion beam
is brought into the chamber and to about 2 x 10 -?torr for a 1 vA argon
beam.

A vA of beam implies 1 £/sec of particles entering the system at
2 xl0- 7torr and 50 £/sec at 4x 10- 9torr. It is quite likely that most

of both the argon and hydrogen diffuse back out of the targets within
seconds and contribute to the gas load.

In addition, an unknown quan-

tity of gas is desorbed from the target surface while being bombarded
by the beams and must also be pumped.

In the case of argon bombardment

the amount of desorbed gas is likely to be several times the amount of
argon.

It appears, therefore, that the observed pressure rises are con-

sistent with the expected pumping speed of the system.

-14The system is roughed to 10 -3 torr by an Ion Equipment Corp. SP-11
molecular sieve sorption pump.

The SP-11 consists basically of a fancy

bucket of zeolite which is immersed in a simple dewar of LN 2 to promote
adsorption of gases.

It is capable of pumping the vacuum system down
from atmospheric pressure to l0- 3torr in less than 30 minutes. When
the SP-11 is not in use it is isolated from the ultrahigh vacuum portion
of the system by an Ion Equipment Corp BVV-152 copper-sealed right-angle
valve.
The UHV system is connected to the rest of the beam line through
a 40-cm-long cold trap with a 1 em ID central beam tube. A Huntington
MS-075 5/8" ID copper-sealed in-line valve serves to isolate the sputtering system from the low vacuum (typically 3 xl0- 6 torr) of the beam
line when beam is not being run.

With the cold trap filled with LN 2
and the UHV system pumped down to l0- 9torr there is essentially no
change in pressure in the sputtering chamber when the MS-075 valve 'is
opened.

Thus, contamination of the system by oil vapor from the diffu-

sion-pumped beam line is virtually eliminated.
Another component of the vacuum system which should be mentioned
is the Faraday cup at the end of the beam line (see Figs. 9 and 10).
This was made by Ion Equipment Corp. and has a very short section of
glass in the middle to electrically isolate the end flange.

Checks of

beam integration accuracy in the sputtering chamber can be made by
bringing the beam through to the Faraday cup.

In addition, the end

flange normally has a window mounted on it so that the beam spot may be
viewed on a piece of quartz glass installed on the target plate.

This

-15was very important in this experiment, since the beam optics are such
that a wide variety of beam spot shapes and sizes is available at the
target plate, and an appropriate spot must be selected for each run.
The Ion Beams
Views of the ion source and beam line appear in Figs. ll and 12.
Beams of positive ions of hydrogen, helium, argon, and other gases
were extracted from the 135 kV duoplasmatron ion source located in the
Kellogg Radiation Laboratory at Caltech.
by a 31° deflection magnet.

The beams were energy analyzed

The beam energy was calibrated by direct

measurement of the ion source voltage using a precision resistive divider.

Beam focusing was provided by a doublet quadrupole magnet which

was located between two sets of x-y steering magnets.
Final collimation of the ion beam was provided by a pair of 3 mm
apertures in .020" stainless steel disks mounted in the beam tube between the cold trap and the MS-075 in-line valve.

The downstream aper-

ture was 12 em from the other and 30 em from the target.

The beam op-

tics were such that the focus position of the beam could be moved a
considerable distance to either side of the apertures, hence the
apertures did not completely determine the beam spot, but only limited
it to between 3 and 6 mm in diameter.
this experiment:

This actually proved useful in

for sputter-cleaning of the targets a large (6 mm

diameter) argon beam spot was used, while a smaller 3-4 mm spot employed during the data runs ensured that the portion of the target being
sputtered was well within the sputter-cleaned area.

As mentioned

earlier, selection of beam spot size was facilitated by viewing the beam

-16-

on a piece of quartz glass attached to the target plate.
Beam integration was performed using an Ortec model 439 digitizing current integrator.

The target plate, collimation cylinder, and

catcher-foil carrier were all connected together and isolated from
ground so that an effective Faraday cup was formed. Secondary electrons
knocked out of the target by the ion beams could only escape from the
sputtering assembly through the 1 cm2 beam entrance aperture, which
amounted to <1% of the 2n solid angle available to them.

This meant

that the beam integration was very insensitive to application of a bias
voltage.

There was no change in hydrogen beam current at the most sen-

sitive level that could be read on the beam current monitor (2%) when a
+300V bias was applied to or removed from the sputtering assembly .
There was also no change at the level of repeatability for argon beam
current readings (4%) when a +300V bias was applied or removed.

addition, there was no detectable change ( < 2%) in hydrogen beam currents when the target plate was moved so that the beam impinged on a
uranium target rather than the imaging quartz.

There appeared to be an

increase in measured beam current of about 5% when an argon beam was
moved from quartz to uranium target.
Secondary electrons emitted from the edges of the final collimation apertures were suppressed by small magnets placed near the beam
tube just before the sputtering chamber.

The magnets were brought almost

close enough to the beam being run to just barely deflect it. Here also
the beam currents were very insensitive to placement or removal of a
suppressor magnet, implying that very few secondary electrons were making

-17it from the final collimator to the sputtering assembly even without
suppression.

For argon beams the current reading changed by only 1%

when a magnet was brought close enough to deflect the beam a bit, and
for hydrogen beams there was no detectable variation in beam current .
Target Preparation
Basic cleaning of the uranium targets was accomplished by dipping
them in 35% HN0 3 for approximately 5 minutes.

The acid bath was fol-

lowed by a rinse in distilled water and an acetone bath.

Targets which

had been stored in atmospheric conditions were covered with a macroscopic layer of black oxide which presumably is mostly uo 2 .

Nitric acid

has a much faster attack rate for the oxide than for the metal.

Most

of the oxide dissolved within two minutes, leaving a semi-shiny silvery
surface. Occasionally a few tiny black streaks aligned with the coldrolling grain of the foils were visible after cleaning.
After cleaning, the targets were typically at atmospheric pressure for 30-60 minutes before pump-down began, and another few hours was
required for the vacuum to reach l0- 9torr. During this period the foils
acquired a definite yellowish cast indicative of the formation of an
oxide layer on the order of 100~ thick. 14 Just prior to most of the
sputtering runs this oxide layer was removed by sputtering with an
80 keV argon beam.

For this presputtering the argon beam spot was a

uniform diffuse blob approximately 7 mm in diameter.

For sputtering

runs in series 2 and 3 the targets received a presputter fluence of
nearly 1.5 x 10 16 argon ions, and series 4-7 targets received a fluence
of 5 x 1016 .

The sputtering yield for 80 keV argon is ~2 (as measured

-18in this experiment), and the average interatomic distance in uranium
is 2.75~. 24

Thus, roughly 140~ (50 monolayers) of the target surface

was removed from series 2 and 3 targets, and about 400~ (150 monolayers)
was removed for series 4-7.

That should have been more than enough to

ensure removal of all surface oxidation and other contamination.

In all

cases the appearance of the beam-spot area of the target foils was very
s i 1very after sputter cleaning.
Scanning electron micrographs of typical target surfaces appear
in Figs. 15-22; they are discussed in Section IliA .

In Appendix C

the possibility of some re-oxidation of the foils after sputter cleaning
is discussed.
The Sputtering Runs
Figure 3 displays the positions of the catcher foils and their
data bands as seen from the position of a target.

The foils were

1}" x2" pieces of .005" thick 5052 aluminum for runs in series 1-3 and
1}" x2" pieces of .002" Indian ruby muscovite mica for runs in series
4-7. Mica was substituted for aluminum to take advantage of its lower
235 u content after it was discovered that mica could be placed in the
sputtering chamber without degrading the vacuum. The data bands resulted from collimation of sputtered 235 u by the 3/8" wide slot in the
collimation cylinder.

During presputtering cleaning runs the foils

labeled GL and GR were in front of the slot.
Before each sputtering run the appropriate ion beam was carefully
focused and centered on a piece of quartz mounted below the targets.
The beam spot was 5 mm maximum width to ensure that it was well within

-19-

the sputter-cleaned area.

Moving the target plate with the precision

Huntington manipulator allowed the accurate positioning of the beam
on target.
During all sputtering runs the beam current was kept low enough
(< 3 vA) so that no overheating of the target foils occurred.

A trial

run with 20 vA of 100 keV H~ on target had produced thermal crinkling
of the foil, and a run at 5 vA had produced some discoloration (but no
crinkling), but 3 vA produced no visible effects.

Hydrogen ion beam

currents were usually 1-2 vA which meant 3-5 hours to accumulate the
fluence of 1017 ions in series 1-3 runs and less than 15 minutes to
accumulate 5 x 10 15 ions for series 4-7.

Due to the much higher sput-

tering yield from argon, a typical argon sputtering run was 10 nA of
beam for 15 seconds.

Helium sputtering was done with 1 vA of beam on

target for ~60 seconds.
The pressure in the sputtering chamber was always ~l0- 9 torr
before any sputtering began. It rose to as much as 5 x 10- 7 during argon
5 x 10-9 immediately afterward.
The pressure was always a few times 10-9 torr during hydrogen sputtering
beam sputter-cleaning and dropped to

runs and dropped back down to l0- 9 torr shortly after completion of the
1 as t run i n a s e ri es .
When a series of 5 or 6 runs was finished the catcher foils were
removed from the sputtering assembly and placed in individual plastic
boxes to await exposure to neutrons at the UCLA reactor.

-20C.

Track Detection
This section will be concerned with the details of the nuclear

track detection technique employed to ascertain the surface density of
sputtered 235 u on the collector foils. The basis of the technique is
that mica in contact with the collector foils during neutron irradiation
will record a fission-fragment damage track for each 235 u atom that
fissions.

These tracks are later etched in HF to render them visible

under transmitted-light microscopy and are counted to determine the
amount of 235 u on the collector foils.
The Mica
The muscovite mica employed in this experiment is an ideal track
detector for fission fragments for a number of reasons.

In the first
place, it has 100% registration efficiency for fission fragments. 11 In
addition, it does not register tracks at all for elements with Z < 10, 12
which avoids any possible background of tracks from a-particles, protons,
etc.

Further, the counting efficiency for the tracks produced by a sur-

face source of fission fragments is virtually 100%.

This is due pri-

marily to the distinctive shape of the etched tracks (see Fig. 13):
long

(15~

full length), thin tubes with diamond-shaped cross sections.

With a moderate etch (> 15 minutes in 48% HF), even tracks seen end-on
are unmistakable under the microscope.

This distinctive shape makes

possible accurate track counting even in the presence of a substantial
background of random pits, bubbles, cracks and tiny dirt particles.
important virtue of muscovite is that, since the surface is dissolved

-21-

away by the etchant at a rate which is completely negligible compared
with the etching rate for tracks, it can be etched for very long times
(order of hours) to make even short fission tracks unmistakable without
erasing tracks having a very low angle to the surface.

(See Fleischer,

Price, and Walker, Ch. 2 for a detailed discussion of the etching
process.)
A significant attribute of the high quality Indian ruby mica used
in this experiment is an effective surface density of 235 u which was
negligible compared with the amounts collected on the catcher foils.
· 1ow 235 u concen t rat1on,
· d w1t
· h great sens1· t 1v1
· · t y f or 235 u
Th 1s
comb1ne
through induced fission makes possible measurement of some extremely low
sputtering yields, which is discussed in Appendix A.
Additional advantages of mica are that it suffers little radiation
damage and, quite significantly, no track fading during neutron irradiation. 41
The only disadvantage of mica is its susceptibility to neutron activation. At a fluence of 10 16 n/cm 2 the resultant activity decays to a
negligible level in a period of a few days, hence it is not a serious
drawback.
Neutron Irradiation
When a group of collector foils was accumulated they were interleaved with 1-1/2" x 2" pieces of muscovite mica and placed in a small
plastic container for transport to UCLA and subsequent neutron irradiation.

All pieces of mica had the surface layer on each sirle stripped

off with adhesive tape to minimize possible environmental uranium

-22contamination.

The mica placed up against each foil recorded the

uranium distribution on the foil by registering one of the fission
fragments from each 235 u that was fissioned. In the cases where
sputtered 235 u was collected directly on mica foils, two copies of
each data band resulted .
Neutron irradiations were performed at the R-1 nuclear reactor
at UCLA's Nuclear Energy Laboratory . This reactor produces a peak
flux of 2 x l0 7n/(cm2 -sec-watt) and is therefore capable of delivering
a fluence of lo 16 n/cm2 in a period of 90 min at its full operating
power of 100 kW.

For runs in series 1-3 a fluence of ~ lo 15 n/cm2 was

employed, and for series 4-7 the fluence was ~lo 16 n/cm2 .
The fluences were determined with the aid of NBS glass standards
containing known quantities of 235 u. These standards are available in
four nominal uranium concentrations (500, 50, 1, and 0.07 ppm), the
higher concentrations being appropriate for use with lower neutron
doses and vice-versa.

The primary technique employed to ascertain the

neutron fluence ~1as simply to place a standard between two sheets of
mica in the package of catcher foils being irradiated and count the
number of fission tracks per cm2 that were produced in the mica. The
fluence is then calculated from

where

Nt = number of tracks per cm2
235
Nu = effective number of
u atoms 1 cm2 in the standard
of = fission cross section = 582 bn

-23Nu in turn is given by

where

nu = number density of 235 u in standard

average single fission-fragment range in the standard glass

and the factor of~ accounts for the fact that only half the fissioned
235 u atoms within a di stance Rf of the surface of the standard glass
produce a fission fragment which leaves the surface (given that the
thickness of the glass is greater than Rf).
The accuracy of this technique depends on two things: 1) careful
counting of all fission tracks produced in the mica, which is difficult
because the actual track lengths are distributed all the way down to
zero due to the fact that the source of the fission fragments is thicker
than Rf' and 2) knowledge of the value of Rf itself.

For each reactor

run I personally counted a minimum of 1000 tracks (under direct microscopy at 450X) spread over a large area on the micas which \'/ere in contact with the glass standards.

I feel confident that the counting

efficiency was at least 95% and that the counting accuracy (involving
resolution of short tracks and discrimination against tiny features which
are not tracks) was about ~5% . It has been reported, 41 however, that
discrepancies in neutron flux determination due to differences in track
counting criteria may be expected to be as large as 20%.
A range Rf appropriate to the soda-lime glass composing the NBS
standards was calculated by Haines 18 using the range and energy-loss

-24relations of Lindhard. 25 •26

The calculation was normalized to the

range of n-induced fission tracks in fluorapatite as measured by
Bhandari et a1. 4 Their measurement of the mean etchable track length
yielded a value of 15.3 ± 1~ (2.37 ± 0.15 mg/cm2 ). The range derived
by Haines for soda-lime glass is 2.24 mg/cm2 . Other calculated values
(including 2.22 mg/cm2 for pure Si0 2 and 2.29 mg/cm2 for phlogopite
mica) make it clear that the range is not strongly sensitive to the
exact atomic composition of such silicates.

In addition, the measured

etchable range of fission fragments in muscovite mica is very close to

15~. 11

It is apparent, therefore, that calculational errors in deriving

Rf are smaller than the error quoted by Bhandari, and a value of
2.24 ± 0.15 mg/cm2 is thus employed. Combining this error (± 6.5%) with
the stated counting error (± 5%) yields an overall uncertainty in the
value of the neutron fluence of less than ± 10%.
As a check on this neutron fluence determination method, the
included glass standard wafer from one of the neutron exposures was compared with an identical pair of standards which had received prior
irradiations in the NBS reactor and been certified by the NBS as to
neutron dose.

These are sold by the NBS as "calibrated glass standards

for fission track use" in the same four uranium concentrations as the
non-irradiated standards.

The pair of standards used contain
0.823 ~ 0.002 ppm (by weight)uranium with 0.2792 atom% 235 u and is
designated "SRM 963" by the NBS.

The two wafers are separately desig-

nated RT-3 and RT-4 corresponding to the different positions in which
they were irradiated in the NBS reactor.

The RT-4 position is farther

-25from the reactor core center, and the neutron flux is better thermalized there than in position RT-3.

This is indicated by the cadmium

ratios of 10.2 for gold foils and 65 for copper foils at RT-3 and
ratios of 87 for gold foils and 536 for copper foils at position RT-4. 8
The neutron fluxes measured by the NBS for SRM 963 are shown in Table
1; total neutron fluence is found by multiplying by the exposure times
given in footnote b.

The wafer identification numbers for the stand-

ards I used are 614016 for RT-4 and 614141 for RT-3.
After irradiation of the blank wafer, which I designated SRM-A,
all three wafers were ground, polished, etched, and counted as detailed
in Appendix D.

Results are tabulated in Table 2.

The neutron fluence

values determined here for the UCLA reactor run are a bit lower than
the value of 1.08 t .10 x lo 16 n/cm2 found by counting fission tracks in
mica sheets placed against SRM-A during irradiation.

However, the pro-

portion of fast neutron flux in the center of the UCLA reactor is somewhat higher than found even in the RT-3 position of the NBS reactor.
This is indicated by a cadmium ratio for 40 mgm/cm2 gold foils (as
measured by Bruce Taylor, private communication) of 3.3 in the UCLA
reactor center as compared with 10.2 for gold foil{ at RT-3 and 87 at
RT-4.

The values determined by comparing SRM-A with RT-3 thus actually

underestimate the neutron dose received by SRM-A by a few percent. The
conclusion, therefore, is that agreement between the two methods of
fluence determination is within the error quoted for the primary method
due to counting error and uncertainty in Rf, and no correction is
applied to fluences derived by the primary technique.

*also 40 mgm/cm2

-26-

Etching
After the neutron irradiation, and after the resultant radioactivity decayed to a reasonable level, the micas 1~ere etched in 48% HF
at room temperature to render the fission tracks visible.

Etching

times of 15 to 60 min were employed; a short etch was used for micas
where it was desired to make star bursts of tracks easy to count, and a
long etch was given to the standards to enhance the visibility of the
many short tracks.

The etching action of the HF was stopped after the

desired time by immersing the micas in full strength NH 40H for 10-15 min.
This was followed by rinsing for a few hours in running water to remove
all traces of HF.
Results of etching a typical mica detector appear in Fig. 13. The
etching time in this case was 15 min.

The magnification is ~lOOOX and

the actual track lengths are close to 15~.

This particular photo con-

tains a small "star" of about 30 tracks resulting from fissions of 10- 6
of the 235 u atoms in a chunk of approximately 3 x 10 7 atoms. Figure 14
displays a large fission star at the same magnification.
Scanning
Once a mica has been etched, all that remains to extract the raw
data from it is to count the tracks in given areas.
~15~

Since the tracks are

in length it is necessary to use a microscope to do this.

The

microscopes utilized in this experiment were equipped with precision xandy-deflection gauges on the stage to enable given points on the data
bands to be accurately located.

They also possessed precision grids in

the eyepieces which permitted accurate determination of track densities.

-27The areas of the grids were calibrated using a Zeiss objective micrometer engraved with a scale 1 mm long markerl off every 10~.
Scanning was often facilitated by viewing the areas to be counted
on a closed-circuit TV attached to one of the microscopes.

Track s ize

and contrast were sufficient that the TV could be employed for routine
work with perfectly adequate results.

The resolution was not quite

enough for counting the standards (with many short tracks) or counting
the number of tracks in a given star-burst, and in these cases direct
viewing through the microscope was always used.
A small number of scanners were employed at one time or another
in the tedious counting operation.

Their results were checked fre-

quently and found to be consistent and reproducible to within 10%,which
is smaller than the scatter in the data.

Generally, a minimum of

400 tracks were counted at each data area, which was typically an area
1 mm high and. 15 mm wide in the long direction of a data band.

Each

data area supplied one data point on an angular distribution of sputtering yield curve.

Thus the error from counting statistics was ±5%

for the angular distributions.

Since the sputtering yield for each ion-

target combination summed over 7 data areas, the statistical error on
the values of S(E) due to counting was <2%.

-28III. EXPERIMENTAL RESULTS
A.

Target Surfaces
Before looking at the results of the sputtering measurement, it

would be useful to have an idea of the condition of the surfaces beinq
bombarded.

To this end a number of the targets were examined with a

small scanning electron microscope (SEM).

A series of photographs of

a typical target were taken and v-1ill be discussed here.
Figure 15 shows the target at a magnification of 240X.

Striations

due to cold rolling of the uranium foil are clearly visible.

They are

also visible at 2400X (Fig. 16).

It is seen here that gross surface

features on the target are on the order of a few microns in size.

The

dark pits presumably result from nitric acid dissolution of oxides
during the surface-cleaning baths.
Fig. 17.

This same area is seen at 8400X in

Here it becomes evident that the targets are quite rough at

the 0. 1~ level, in addition to being lumpy at the 1~ level.

This is

significant because the projected range of 120 keV protons is on the
order of 0. 1~ 40 and the other ions have even shorter ranges. This
implies that even though the ion beams always impinged perpendicularly
to the target surface, the sputtering yields are actually averaged
over a very wide range of incident angles.
An area of this same target which has been sputter-cleaned with
a fluence of"'l0 17 80 keV argon ions appears in Fig. 18 at a magnification of 8400X.

In this case the surface roughness is very great and

fairly uniform on the scale of 1~.

Here it is quite clear that the

ion beam impinges on the target at effectively all angles of incidence.
This same spot is seen in Fig. 19 at 2400X.

A similar area which was

-29-

also sputter-cleaned in the same fashion is shown in Fig. 20 at the
same magnification.

Also at 2400X is another photo of an unsputtered

area of this target {Fig. 21).
A couple of sputtering runs were made on a target which was not
cleaned at all.

This target had been s t ored in a small plastic box

inside a larger plastic box at atmospheric pressure for more than a
year.

It was fully oxidized to a depth of about 10~ (as revealed by

thickness measurements before and after washing it in HN0 3 ).

An SEM

photo of it at 800X appears in Fig. 22.
B.

Angular Distributions of Sputtered Atoms

Although the basic Sig~und theory predicts an isotropic distribution for the sputtered flux, 43 experimentally determined angular distributions tend to be near cos 8, 2 for perpendicular incidence and 8
measured relative to the beam direction. Results on proton sputtering
of various metals reported by Finfgeld 10 include distributions that
are peaked more sharply than cosine.

Having seen in the previous sec-

tion that the uranium surfaces bombarded in this experiment are very
rough on the scale of the collision-cascade size, one miqht expect
that the angular distributions would be smeared out considerably from
a cosine form.

Such &s the case, as can be seen in Figs. 23-26 .

The angular distributions are not quite isotropic, which may well
be due to self-shadowing (i.e., atoms sputtered from a low area of the
target cannot leave the surface at a large angle without impacting
high areas nearby), but they are clearly not cosines.
Cos 114 e gives

-30-

a fit whose accuracy is commensurate with the scatter in the angular
distribution data and with the errors on the individual data points.
Figures 23 and 24 display respectively the best and the worst fits of
cos 114e to the proton sputtering angular yields. Figure 25 gives a
typical fit to the molecular hydrogen results, and Fig. 26 is a typical
fit for the helium sputtering case. In all cases cos 114e was chosen
for purposes of calculating total sputtering yields and is not intended
to represent any underlying physics.

It is possible, of course, that

one might be able to show that such a distribution arises naturally,
given an intrinsic isotropic distribution and some reasonable model of
a rough surface.

In any case it would be interesting to study the

angular distribution as a function of surface finish.

This would be

difficult for light-ion sputtering of uranium, however, because the
sizes of the collision cascades are so small.

120 keV protons have a

projected range in uranium of approximately 1000~ (see Schi~tt 1966),
implying collision cascades of that extent.

This implies a surface

finish on the order of 100~ would be required for the surface to appear planar to the cascade.
C.

Total Sputtering Yields
The total sputtering yield results are summarized in Tables 3

and 4 and Figs. 27-29.

The yields were calculated with the formula
( 3-1)

where

number of fission tracks per cm2 on the collector foil

-31extrapolated to 8 = 0°

number of incident atoms (not ions; e.g., an

H; ion has

3 atoms)
N =

r =

neutron fluence to which the collector foils were exposed
cross section for fission of 235 u by thermal neutrons
distance from target to catcher foil.

The factor 8n/5 arises from fitting the angular distribution results
by Nt(8) = N~cos 1 1 4 8, as discussed in the previous subsection.

It may

be instructive to observe that this factor would be 2n if an isotropic
distribution was fitted.

This would increase the calculated values of

S by 25% if the same values of N~ were employed.

If one attempted in-

stead to fit a cos 8curve to the differential yield results, the factor
would ben and the total yields calculated would be reduced by 37.5%
using the same N~ values.

To obtain even a rough fit to the angular

distributions with cos 8 it would be necessary to use a substantially
higher N~ value, so the reduction in calculated values of S would be
more like 15%.
Looking now at Fig. 27, three sets of hydrogen sputtering data
can be seen.

The upper set of points was

obtained by sputtering tar-

gets which had just been sputter-cleaned in vacuo with an 80 keV argon
beam.

The total proton fluences in these runs were on the order of
~0 x1o 15 . The middle set of points are from bombarding sputter-cleaned
targets with 10 17 protons.

The reduction in yield may be due to accumu-

lation of surface oxides during the runs (3-5 hours here as opposed to
15 minutes in the previous case).

This point is discussed in detail

-32in Appendix C.

It may also be that the increased proton dose has

some effect, since strong dose effects have been observed elsewhere. 2
The lowest yields were obtained with the same proton dose on targets
which had not been sputter-cleaned.

These targets were cleaned in

HN0

immediately prior to placement in the UHV system, but acquired
many monolayers of oxide before being sputtered .
It could be argued that the increased surface roughnes s of the
sputter-cleaned targets 1-1as responsible for the increased yields.

There are two arguments against that, however.

First of all, the

electron micrograph of Fig. 17 demonstrates that all of the targets
are quite rough at the lOOOg level, which is the relevant scale here
because the collision cascades (for the energy and mass-ratio regimes
we are concerned with) are of this extent.

Second, the theoretical

prediction is that the sputtering yield will not vary much with incident angle when M2 >> M1 (see Sigmund 1969, p. 404).

It is true,

though, that the theory is not sophisticated enough to handle largeangle incidence by ions accurately, particularly for light ions on
heavy targets.
To try to get a better idea of the effect of surface oxidation
on the sputtering yield, a pair of runs were made on targets th at
were heavily oxidized.
22.

An average area of one of them appears in Fig.

As discussed in Section IliA, the oxidation extended to a depth

of 10 ~ or so.

Due to the storage conditions it is expected that the

surface consists primarily of uo , but the precise stoichiometry and
the amounts of any contaminants are unknown. The sputtering yields

-33(presented in Table 3) are approximately 10 times lower than those
of sputter-cleaned targets also bombarded by 10 17 protons. This sets
an approximate lower limit to the sputtering yield of uo 2 due to protons near 50 keV; if the surface is completely uo

to a depth of ~10~
then these numbers are sputtering yields for uo 2 (assuming the sputtering proceeds stoichiometrically) .

If, however, there are other

materials present, the sputtering yield has been reduced accordingly,
since the fission track technique is insensitive to them.
The energy dependence of each of the three sets of hydrogen sputtering data follows the trend predicted by the Sigmund theory and
indicated by the curve (see Section IV).

Exceptions are the two points

at 13 keV and 20 keV which were due to sputtering with 40 keV
40 keV H~ respectively.

H; and

No cause to doubt the validity of these meas-

urements was discovered, and the reason for the disagreement with
theory is unknown.

The other proton yields derived from H~ and

bombardment agree well with the H~ yields with equal energies per
proton.

These results are shown in Fig. 28.

results of H~ and

They agree with previous

H; sputtering of several metals obtained by Kenkniqht

and Wehner (1964), who found that the sputtering yield per proton of a
given energy was the s arne for H2 and H .
Sputtering yields due to helium ion beams are displayed in Fig. 29
and tabulated in Table 4.

It is immediately apparent that the yields

are about 30 times those from proton sputtering on similar sputtercleaned surfaces.

The energy dependence appears to be somewhat flat-

ter than Weissman and Sigmund's prediction (solid curve; see Section IV).

-34Here, too, the magnitudes of the yields are much lower (by a factor
of about 10) than predicted by the theory.
A couple of data points for uranium sputtering by argon were
also obtained, as seen in Tables 3 and 4.

In this case the theory comes

much closer, being high by only a factor of 3.
D.

Chunk Emission
It has been discovered in recent years that sputtered material does

not always emerge as single atoms.

There are instances in which some of

the material is ejected as agglomerates of many atoms, commonly referred
to as "chunks".
Chunk emission from niobium during sputtering by 14 MeV neutrons
has been reported by Kaminsky 19 and Kaminsky and Das? 1 •22
Biersack,
et al: have studied chunk ejection during neutron irradiation of uo 2
utilizing the same mica detection of fission tracks as employed in this
experiment.

The technique is particularly appropriate to i nves ti gati on
of chunk emission from materials containing 235 u because of the striking
"star-burst" signature left in the mica by a chunk during neutron irradiation.

Two examples appear in Figs. 13 and 14.

The number of fission

tracks in more than 1000 such stars was counted to yield the chunk emission data summarized in Tables 5 and 6 and Figs. 30-32. The number of
235 u atoms per chunk is obtained by dividing the number of tracks in
the corresponding star by the fraction of 235 u atoms fissioned:
8.5xl0- 7 for runs in series 1-3; 8.0xlo-6 for runs in series 4-7.
The mass distribution of chunks emitted during the first five proton sputtering runs of Table 5 is shown in Fig. 30.

The lower cutoff

-35-

at 5 x 106 atoms per chunk is due to the fact that smaller chunks
normally produced less than 5 tracks in the mica and could not be
recognized.

The distribution displays a long tail, indicated by the
two bars at the right, extending out to 109 atoms per chunk. The

curves represent visual fits to the data of the forms N(n) = a/n 2 and
b/n.

The former fit represents the histogram fairly well for larger

n but diverges badly from it at small n.

The latter fit, on the other

hand, is not as bad for small n but is quite a bit high at larger n.
An exponential actually fits the histogram much better, as seen in
Fig. 31.

More data would be required to ascertain whether or not the

exponential adequately represents the tail of the distribution, which
has been omitted from Fig. 31.

Fits to the mass distribution will be

discussed further in Section IVC.
The mass distribution of chunks ejected during sputtering by
80 keV argon is shown in Fig. 32.

The low number of stars with 5-10

tracks is rather surprising but evidently quite real.

No data on

chunks of smaller size is yet available to confirm the existence of
the apparent maximum in the distribution.

It is also worth noting

that the distribution is somewhat flatter here than in the H1 case and
the tail is longer--several chunks with a few times 109 atoms were
counted.
This size ( ~10 9 atoms) is typical of the many chunks of uo 2 seen
by Biersack, et al~ but I do not know what sort of size distribution
their experiment produced.

The average chunk of niobium sputtered by
neutrons, as reported by Kaminsky and Das, 21 contained 2 x 10 12 atoms,

-36and chunks as small as 10 7 atoms were seen. The median chunk size seen
here was ~2.5 x 10 7 atoms, for 80 keV argon sputtering, and ~2.0 x 10 7
atoms for proton sputtering; the mean sizes were 6.6 x 10 7 atoms and
4.3 x 10 7 atoms respectively. These values are of course influenced by
the arbitrary cutoff at 5 x 10 6 atoms/chunk and by the low statistics
for large chunks. Assuming a spherical shape, a chunk containing
5 x 10 7 235 u atoms would be ~o. l~ in diameter.
The observed chunk size for protons at reduced dose (5 x 10 15 instead of 10 17 incident ions per target) was only ~10 6 (see Table 5).
The statistics here are very poor, however, since a reduction in dose
by a factor of 20 resulted in a reduction in chunk yield by a similar
amount.

This reduction is probably significant in itself, since it

indicates that the chunk emission is not due to release by the first
wave of H~ ions of energy stored in the target surface.

A great deal

more information on dose dependence is necessary before anything
definitive can be said.
The

observed median size of chunks emitted during helium sputtering was ~2 x 10 6 . Here again the He+ dose was down by an order of
magnitude from what it would have been in conjunction with a neutron
fluence of 10 15 cm- 2 (lo 16 cm- 2 was the neutron fluence for the helium
sputtering runs), but the dose effect is unknown.
Looking now at Table 6, it is apparent that the efficacy of chunk
emission correlates roughly with the efficacy of single atom sputterThe He+ sputtering yield is ~so times greater than for H~, and
· d t o eJec
· t an amount of 235 u
roug h l y 50 t1mes
t he H+ fl uence was requ1re

ing.

-37in chunks comparable to that ejected by He+ (taking into account the
fact that the chunks were smaller in the He+ case).

S for argon

sputtering is two orders of magnitude greater than S for helium, and
two orders of magnitude more He+ fluence yielded a number of chunks
comparable to that ejected by argon ions.

Since the He+-ejected

chunks were smaller, the amount of uranium ejected as chunks was considerably greater for the argon case.

This agrees with the observation

that chunks composed a much greater fraction of the total sputtering
yield forargon ("' 10%) than for hydrogen or helium( "' .3% and'V.l %
respectively).

This fraction is determined in each case by multiplying
the average chunk size by the observed number of chunks per cm2 and
dividing by the number of 235 u atoms/cm 2 (see Table 6), with the assumption that all single tracks represent single atoms.

If the size dis-

tribution of chunks does in fact follow an exponential dependence, then
this assumption has very little effect on the calculated fractions (see
Section IVC).
Another observation of possible importance is that the amount of
chunk emission under H~ bombardment seems to increase with ion energy,
which is the reverse of the energy dependence of the total sputtering
yield.

Here again

more data is needed to establish this trend.

Before closing this topic, a few words are in order concerning
possible spurious contributions to the chunk distribution on the micas.
As a test of environmental contamination,a 3 cm2 area on the back of
data band LL2-1 was scanned very carefully at 450X.

A total of two

stars were found (see "Blank" entry in Table 5), while the sar.1e area

-38-

on the front contained >300 stars.

This argues strongly against gen-

eral environmental contributions or spurious stars internal to the mica.
Actually, on rare occasions a uranium-rich inclusion near enough to the
mica surface to be etched was encountered.

An internal star is readily

told apart from a star due to a surface deposit by focusing the microscope up and down, so none of those were counted.
Scans at low power (25X and lOOX) were made of the relevant micas
to see if any patterns that would suggest contamination could be found.
None were, except for broad bands of stars along the sides of micas in
series 6, (He+ sputtering) where enough uranium had accumulated (during
the sputter-cleaning runs) on the clamping bars to scrape off on the
micas when they were installed.

The scanning area for series 6 was ac-

cordingly reduced to avoid the contaminated area by more than l em.
A final argument that spurious stars do not make a large contribution is that data bands on micas which have been through identical
processes, and even data bands on the same mica, display very different
chunk densities, but which are uniform within each data band.
All things considered, there is undoubtedly some contamination
included in the data, but it is expected to be only a small portion.
E.

Error Analysis
The primary sources of error in determination of values for the

yieldS were uncertainty in Nn (the neutron fluence) and error in determining the track density N~ (see eq. (3-l)).
As discussed in Section IIC the error in Nn was due to counting
error (±5%) in determining the number of tracks in the standard micas

-39-

and to uncertainty (of ±6.5%) in the established fission-fragment range
Rf.

The overall error in Nn is then ±8.5%.

The error inN~ had three sources: l) countinq statistics (±2%)
(see Section IIC); 2) error in fitting the cos 114e curve to the angular distribution data due to scatter in the data (varied from ~±5% to
±10%); and 3) background subtraction.
traction varied considerably.

The amount of background sub-

For most runs in series 4-7 it was zero,

due to collecting the sputtered uranium directly on mica (see Section
IIC).

In series l-3 the sputtered uranium was collected on aluminum

which contributed ~30 background tracks per field of view, which had
typically 100-300 tracks total.

Uncertainty in the background deter-

mination was ~±10%, which contributed typically ±l% to ±4% to N~.
Another source of background on some data bands was 235 u atoms
bouncing or drifting up onto them during the sputter-cleaning runs.
Although the uranium being sputtered during the cleaning runs was well
collimated, the amount was so great that a small fraction findinq its
way up toLL and LR foils (see Fig. 3) instead of sticking to GLand GR
foils gave rise to a non-negligible background "fog" of tracks on some
of the low-numbered data bands.

This background subtraction \'las handled

by scanning vertically across the data bands and plotting the fall-off
of the track density with distance away from the lower edges of LL and
LR micas.

In the worst cases the subtraction amounted to

25% and is

responsible for the relatively large error bars on the Ar+ sputtering
yields.

-40-

The result of the foregoing is that the error in N~ had to be
determined individually for each sputtering run.
cally ±10% .

Values were typi-

Once the error on N~ was obtained for a given sputtering

run it was added (in quadrature) to the error on Nn to produce the
quoted error on the value of S for that run.

Quoted errors are typi-

cally ±15%.
All other sources of error inS were determined to be negligible
compared with those discussed, but a couple of them deserve mention.
One of those is beam integration error, which was found to be <1% for
hydrogen ion beams and perhaps as large as 5% (systematic error) for
the argon beams (see Section liB).

Another possible source of systematic error is the sticking fraction for the sputtered 235 u atoms on
the catcher foils, since a sticking fraction less than 1 would reduce
the measured values of S accordingly. The sticking fractions for sputtered 235 u atoms on both aluminum and mica have been measured in the
Kellogg Laboratory by Bob Weller and Joe Griffith (private communication) and found to be >.97 in both cases.

-41-

IV.
A.

THEORY AND SPECULATION

The Collision Cascade Theory
In view of the complexity of the atomic processes involved, it is

not surprising that a definitive theory of sputtering has not yet
emerged. The collision cascade theory due to Sigmund 43 •49 is the most
satisfactory theory to date; it makes definite predictions concerning
the important quantities measured in sputtering experiments and has
proved to be quite accurate in many cases.

It also serves to bring

into focus the primary physical interactions at the atomic level which
occur during sputtering.

Before examining the quantitative predictions

of the Sigmund theory, a brief qualitative look at the sputtering
process is in order.
An impinging ion suffers a series of collisions with target atoms,
creating a number of primary recoil atoms which in turn collide with
other target atoms, and so on.

This collision cascade continues until

none of the recoiling atoms has sufficient energy (on the order of 20 eV)
to displace another atom from its position in the lattice.

Typically,

the cascade proceeds through many steps, and a large number of atoms are
set in motion with very low energies.

The fact that the cascade pro-

gresses through many generations of recoiling atoms implies that momentum information about the original projectile is lost, and the cloud of
low energy atoms becomes essentially isotropic.

This makes the theoret-

ical solution of the sputtering problem much more tractable.

In any

event, atoms in the cascade which arrive at a surface of the target with
enough energy to overcome the surface binding (typically a few eV) are

-42sputtered.

So many atoms are set in motion with very low energies

that they account for the bulk of the sputtering yield, even though
energetic recoils and backscattered ions account for most of the
sputtered energy. 42 Experimental energy spectra of sputtered atoms
peak strongly at a few eV (see, for example, Thompson 1968).

The low

energy of most sputtered atoms also implies that they originate within
a few ~ngstroms of the surface.
In addition to losing energy via collisions with target atoms
("nuclear stopping") an ion loses energy to the electrons in the target
("electronic stopping").

Due to the enonnous mass difference between

atoms and electrons, energy once lost to the system of electrons is
never returned to the collision cascade.

In calculations of sputtering

it is necessary, therefore, to detennine and subtract out this energy
loss.

For heavy ions on moderate atomic-weight targets, nuclear stop-

ping of the ion predominates and electronic stopping is neglected, but
for light ions on heavy targets, electronic stopping is greater and
must be taken into account.

As a pertinent example, the electronic

stopping power for 100 keV protons in uranium is 330 times the nuclear
stop pi ng power.
Sigmund's basic approach to the calculation of sputtering yields
is to make use of transport theory to detennine the distribution of
energy deposited in atomic motion in the target.

This energy is then

converted into a distribution of low energy recoils, and it is determined how many of them arrive at the surface with sufficient energy to
overcome the surface binding forces.

Knowledge of certain quantities

-43is essential to the procedure: the cross sections for scattering of
high-energy ions and atoms (Rutherford cross-sections for higher energies and Thomas-Fenni cross-sections for lm'ler energies), cross sections for low energy target-atoms scattering from each other (ThomasFermi for most energies and Born-Mayer cross sections for the lowest
energies), electronic stopping powers (Lindhard's expression 25 •26
SefE} = KE 112 is used except when the ion velocity is high enough to
warrant use of an S (E) ~ 1/E approximation to the Bethe-Block
formula) , and the surface binding energy U (sublimation energy is

generally used for lack of better infonnation).

The procedure is ap-

plicable only to amorphous targets, but should provide a reasonable
approximation to sputtering from polycrystalline targets.

It also ap-

plies only to bombarding ions with energy much greater than the surface
binding energy, i.e., to ion energies of at least 100-200 eV.

addition, the energy threshold Ed ( ~ 20 eV} for displacement of an atom
from its lattice position and binding energy lost to the lattice hy
recoiling atoms are neglected.

One final approximation that may be

significant is that the target surface is a smooth plane.
The depth distribution F(l)(x,E,n) of energy deposited in atomic
motion satisfies the equation44,45,49,53

where

E = initial ion energy
n = cos 8 and 8 is the angle between the ion beam and the
target surface-normal (the x-direction )

-44n' = cos 8' and 8' is the polar angle for a scattered ion
n"

cos 8" and 8" is the polar angle for a scattered target
atom

F(x,T,n") is the deposited-energy profile due to a target
atom with energy T; it is computed from a similar transport equa t 1. on. 45
do

= do

{E,T)dT is the differential cross section for energy

loss T in an elastic ion-target-atom collision
Se = Se(E) is the electronic stopping cross section

N = target atom number density
To solve equation (4-l) the F functions are expanded in terms of
Legendre polynomials in the angular dependencies Tl• n', and n"· Equation
(4-l) is then multiplied by xn (n=O,l ,2,···) and integrated over x to
generate a series of moment equations.

These integrodifferential equa-

tions are solved numerically for n ~ 3 or expanded to first order in
U /E and the first few moments solved for analytically. 43
Finally, the

energy deposition distribution is constructed from these moments by use
of the Edgeworth expansion in terms of Gaussians and derivatives of
Gaussians. 43 •49 The sputtering yield for a given ion-target combination
may then be written in terms of the energy deposited at the surface:
S(E.n)
where

(4-2)

A= - 2 NC IJ

41T

0 0

(4-3)

-45and C0 is a constant related to the Born-Mayer interaction between
target atoms (see Sigmund 1969, p. 390).
The energy dependence of sputtering yields is essentially that
of the nuclear stopping cross sectionS n (E).

It is customary, there-

fore, to introduce the function a (E,n) such that

(4-4)
so that (4-2) may be rewritten as

(4-5)

S(E,n) =

It is also customary in calculations of a and of nuclear stopping
powers to work with the dimensionless energy £ and dimensionless distance p given by

£ -

p -

(4-6)

4MlM2
(M +M )Z Nna . X
1 2

(4-7)

z1z2e 2 (Ml+M2)

where subscript 1 refers to the projectile and 2 refers to the target
and

is the screening parameter in the Thomas-Fermi potential :
. a

a 0 is the Bohr radius.

(4-8)

-46Equation {4-5) can then be rewritten as
Ml 0.042 1
)(d £)
S(E,n) = 4'1Taz z 2e 2M +M
-U
"""""02 a £ , n crp n

(4- 9)

[A]

It is this form of the yield equation which has
proved most useful for calculations to compare with experimental res ults
in Section IVB.
B.

Comparison to Results
Equation (4-9) was used to calculate theoretical values of S vs.

energy for H1 , He and Ar.

The results are tabulated in Tables 7-9 and

graphed for H1 and He in Figs. 27 and 29.

The values for the reduced

nuclear stopping cross sections n (£) = (d£/dp) n are from Schi~tt (1966).
The values a(£) = a( £ ,n = 0) for perpendicular incidence are taken from
Weissman and Sigmund (1973) for H and He. For argon sputtering the
data presented in Andersen (1974, p. 392) were consulted to arrive at
an (energy-independent) value a( £) = 0.5.

A value U0 = 5.4 eV for the

surface binding energy was taken from Gschneidner (1964).
The energy dependence of H1 sputtering as predicted by the theory
and graphed in Fig. 27 follows the data except for the measured yields
at 13 and 20 keV.

These two points were determined using 40 keV H; and
40 keV H; ion beams. According to Fig. 28 and other sources, 23 • 50 an H;
ion with energy E gives the same sputtering yield as two H~ ions with
energy l/2E, and similarly for H;.

The points at 13 and 20 keV should
be reasonable yield measurements for H+ , and the reason for disagreement
with theory is not known.

-47-

For He+ sputtering the measured sputtering yield also begins to
diverge from the theoretical curve at lower energies (Fig. 29).

This is

tantalizing, but more measurements at lower energy are needed to estab1ish the dependence firmly.
The magnitudes of S as predicted by (4-9) do not agree with the
experimentally determined values very well.

The theory is high by a

factor of about 25 for H+1 on sputter-cleaned targets

(the curve in Fig.
27 is the prediction by the Sigmund theory multiplied by l0- 2 ). In the
case of He+ ions on sputter-cleaned targets the theoretical curve i s high
by a factor of ~ 10 (see Fig. 29).

The calculated values of S(E) appro-

priate to sputtering by argon (Table 9) are higher by a factor of ~ 3
than the measured values at 40 and 80 keV (Tables 3 and 4).

The theor-

etical predictions appear to approach the empirical values of S as the
target-to-ion mass ratio {M ;M 1 ) decreases. This is not too surprising
as it is a well-known feature of the original Sigmund (1969) theory. Why
it should be true of the Weissman and Sigmund (1973) version, in which
the approximations appropriate to light-ion sputtering have been made,
is unclear.
One can imagine several reasons why the experimentally-determined
values of the yield might be reduced.

For one thing, surface contamina-

tion or oxidation of the targets would reduce the yield.

This effect is

considered in detail in Appendix C and is expected to be unimportant for
the sputter-cleaned targets.

The surface roughness would also be ex-

pected to influence the yield relative to the predicted yield, since the
theory assumes a planar surface.

The expectation, and the usual res ult,

however, is that the sputtering yield increases with increasing s urface

-48-

roughness, since more ions impinge at oblique angles and deposit more
energy near the surface.

The ion dose probably has some effect as

well, but the doses delivered in this experiment are not unusually
high or low compared with experiments with heavy ions which the
theory fits quite well.

The conclusion is that the lack of agreement

between theory and this experiment is probably due to inaccuracy of
the theory at large M2/M 1 .
C.

Chunk Emission
The study of chunk emission from solids under ion bombardment

is of recent genesis, and a cogent theory of the emission mechanisms
has not yet been fonnul ated.

I do not intend to fonnul ate one, but

will confine myself to a few brief remarks on the subject.
It is clear that none of the chunks observed in this experiment
could have been ejected directly by a single ion, since the energy required to create the necessary free surface is greater than the total
ion energy available.

To put it in rough quantitative terms, imagine

a small, cubic chunk of 106 atoms perched on one face on the target
surface.

The energy required to release the chunk is given roughly by
the number of atoms on the attached face (10 4 ) times the cohesive energy
per atom (~ 5 eV) or 50 keV.

Especially when it is considered that the
average energy t rans f erre d t o a pnmary
reco1· 1 2 35 u a t om by any of th e
incident ioni is near 100 eV, it is evident that the ejection mechanism
must involve some variety of energy accumulation.
Guinan (in Kaminsky et al, 1974, p. 171) suggests a process
whereby energy deposited in collision cascades serves as a trigger to

-49-

release much larger reserves of energy stored as internal stress es in
the material.

Cold rolling of the metal, for example, stores cons id-

erable energy in various types of dislocations.

It appears that ene rgy

deposited by cascades in the vicinity of certain dislocations is sufficient to initiate a crack.

Continued energy deposition contributes

to propagation of such microcracks and leads to chunk emi ss i vt~ when
cracks intersect the surface.

A prediction based on this mechanism is

that chunk emission should be enhanced by increased surface roughnes s .
This effect has been observed by Kaminsky 21 and is certainly not con tradicted by the amount of chunk emission from the very rough surfaces
sputtered in this experiment.
Returning for a moment to the idealized cubic chunk perched on
the surface, we observe that as we increase the number

n of atoms in

such a chunk the energy binding it to the target surface increases as
n213 •

If we naively expect the probability of emitting a chunk to be

inversely proportional to the energy binding it to the surface, we
would then look for an n- 213 dependence in the mass distribution of
A look at Fig. 30 reveals that an n -1 fit to the distribution
is too flat--deviating markedly from the data at larger n. An n-2/3
chunks.

dependence deviates even more, so our naive expectation does not appear
to have been very fruitful.
Suppose we now look at the mass distributions and see if any information can be gleaned from them.

The distribution of chunk size in

the case of argon sputtering (Fig. 32), with a maximum at small n, is
suggestive of the sort of distribution of pieces obtained when an object

-50(such as a glass) is smashed by a sudden blow: lots of small pieces,
less tiny ones, and a few large ones.

It is conceivable in the case

of chunk emission that the "object" is a blister which suddenly ruptures when the internal pressure of gas (accumulating as ion implanta20
tion continues) becomes too great. A number of groups have reported
blistering (and blister rupturing) of various metal s under bombardment
by hydrogen and helium ions.

No~ priori

ess should not also take place in uranium.

reason is known why the procParticularly for argon ,

which has a much lower mobility in uranium th an hydrogen, and whi ch
does not form compounds with uranium as hydrogen does, gas bubble
formation and subsequent blistering may be an important mechanism for
chunk emission.
If we do not learn anything immediately from the exponential fit
to the chunk size distribution, at least we can use it to calculate a
total sputtering yield in chunks.

For

N(n) = N e-An

(4-10)

where N is the number of chunks consisting of n 235 u atoms (in units of
106 atoms), the total number of 235 u atoms contained in chunks with size
between n1 and n2 is simply
n2
Nu = J
nl

n2
N( n) n dn =

N n e -An dn

(4:-11)

Nu = N [ (),n+ l)e -An]"2
n1

(4-12)

-51Using the values N0 = 180 and A = 0.05 for the fit to the distribution
of hydrogen-sputtered chunks in Fig. 31, and integrating from n1 = 10
to n2 = 100, equation (4-12) gives Nu = 6.2 x 10 3 This is the number
of 235 u atoms (in units of 106 ) contained in all chunks having sizes
between 10 7 and 10 8 atoms found on the 30 cm2 of mica that was scanned;

1 .e.,

Nu = 2.1 x 10 8 em -2

This is a bit less than 0.1% of the total

single-track yield per cm2 (see Table 6).

Integrating instead from
n 1 = 0 to n2 = oo , we obtain Nu = 2.4 x 10 8 atoms cm- 2 . This exponential
fit says, therefore, that most of the total yield of atoms contained in
chunks is found in chunks with sizes from 10 7-10 8 atoms. The fact that
this estimate of chunk contribution to total sputtering yield (0. 1%) is
lower than the estimate of 0.3% in Table 6 is due presumably to the exponential fit underestimating the quantity of chunks in the tail of the
distribution.
Performing the same calculation for the argon sputtering case of
Fig. 32, equation (4-12) yields (with N0 = 60, A = 0.03, n1 = 0, n2 = oo ,
and a scanned area of 4 cm2 ): NU = 1.6 x 10 10 This is actually 30% of
the single-track yield, implying that this fit overestimates the quantity
of chunks in the tail of the distribution.

-51 aV.

SUMMARY

One of the most significant results of this experiment is that
it demonstrated the great sensitivity of the technique of detecting
fission-fragment tracks in mica for measuring sputtering yields of
materials containing 235 u. Values of the sputtering yield S as low as
3x 10-S were readily measured for proton bombardment of uranium. and it
is shown in Appendix A that yields as low as 10- 6 could be measured for
bombardment by keV ion beams of materials doped with 0.1 atomic % 235 u.
One of the primary limits on the technique is simply the available beam
fluence. and the possibility of measuring S = 10-ll for such things as
photon sputtering of uranium is also discussed in Appendix A.

It should

be mentioned that the sensitivity of the technique implies that very
little of a given target surface is actually sputtered away (less than
one monolayer was removed during any sputtering run in this experiment).
This in turn implies that extraordinary precautions may be required to
avoid reduction of the sputtering yield by surface contamination and
that. as is discussed in Appendix c. even a vacuum of 10- 9 torr may be
insufficient to maintain the cleanliness of the target surface during
a given sputtering experiment.
The fission-track detection method possesses another important
virtue: it facilitates detailed study of chunk emission from targets
containing 235 u since a neutron-irradiated chunk produces a star-burst
pattern in the mica collector sheet (see Figs. 13 & 14). and the number
of tracks in the star is proportional to the number of 235 u atoms in
the chunk.

Some of the most exciting results of this experiment concern

-5lbthe emission of chunks during sputtering of uranium by 1 H+ , 4 He+ , and
40Ar+ ion beams with energies in the vicinity of 80 keV. These results
are summarized in Tables 5 & 6 and Figs. 30-32 .

The uranium targets

consisted of cold-rolled foils which had been pre-sputtered with 80 keV
Ar+ ions to a dose of approximately 1017 cm- 2 and which had very rough
surfaces (see Figs. 15-21).
Mass distributions for chunks emitted during sputtering by 1H+
and 40Ar+ were found to have approximately exponential shapes over the
range of sizes seen - from the scanning cutoff at 5 x 106 atoms/chunk
to a few times 10 9 atoms/chunk (see Figs. 31 & 32). The average size of
the chunks seen was close to 5x 107 235u atoms. The sputtering yield in
chunks was about .3% of the total sputtering yield for proton bombardment
and about 10% in the argon case. By way of comparison, Kaminsky and
Das report 22 emission of chunks from niobium during irradiation with
14 MeV neutrons and observed chunk sizes ranging from -s x 1o6 to -5 x 1012
atoms.

They also found chunk contributions to the total sputtering

yield ranging from zero to about 60% for different areas of targets
having various surface finishes.

One should not attach too much signifi-

cance to similarities and differences between the chunk emission results
reported here and those reported by Kaminsky since the bombarding particles are quite dissimilar, the doses are 3 - 7 orders of magnitude
apart, the energy deposition per primary recoil and per volume differ by
several orders of magnitude, and a considerable amount of the chunk
emission here may be due to bursting of small blisters containing gas
bubbles of the implanted ions.
It was also found that the amount of chunk emmission due to

-5lcl + 4 +
40 +
bombardment by H , He , and Ar correlated roughly with the sputtering yield in each case. For example, the yield S for argon sputtering
of uranium is about 104 times that for H+, and approximately 10 5 times
as many protons as argon ions were required to eject a given amount of
uranium in chunks.
Angular distributions for the sputtered uranium are plotted in
Pigs. 23 ~ 26,

The cos 114 e shapes are due presumably to smearing out,

by the very rough target surfaces, of much sharper distributions .
Total sputtering yields are summarized in Tables 3 & 4 and Figs.
27- 29. Typical values of S, for sputter-cleaned targets, are 3x 10- 4
for 80 keV 1H+, 10- 2 for 80 keV 4He+, and 2 for 80 keV 40Ar+ ion beams.
Calculations of S based on the theory due to Sigmund 43 and Weissman and
Sigmund 49 overestimate the results by approximately a factor of 3 in the
argon case, by a factor of -10 in the helium case, and by a factor of
nearly 25 in the case of proton bombardment.

There are several things

which could contribute to this discrepancy .

For one, the theory assumes

a planar surface, but the surfaces being sputtered in this experiment
are very rough (see Section IIIA).

While roughening a smooth surface
generally increases the yield, it has been demonstrated 55 that an

extremely rough surface can reduce the yield considerably.

This may

well account for a factor of 3 reduction in the yield as compared to
theory.

Another important consideration is that the theory tends to

overestimate the sputtering yield for large ratios of target atomic
weight to projectile atomic weight, which probably explains much of the
discrepancy in the hydrogen and helium cases. For comparison, Furr and
Finfgeld 15 reported yields for H+ on gold approximately 5 times lower

-Sldthan the theoretical values calculated by Weissman and Sigmund 49 , and
Summers, et a1~ 6 found values of S for H+ on niobium that were 8 to 9
times lower than theory predicts 49 . One other possible source of discrepancy is that the real target surfaces incorporate a certain amount
of contamination, due to adsorption and due to implantation of the ions
themselves, which the theoretical calculations ignore.
One last result from this experiment derives from sputteri ng runs
made with molecular hydrogen beams (see Fig. 28). Consistent with the
findings of others 15 •16 •23 , it was found that, for the few energies
investigated, the sputtering yield from an H~ molecule having energy E
is the same as that from n 1H+ ions having energy E/n.

-52APPENDIX A
Sensiti vity of the Fission Track Detection Technique
In this appendix I would like to discuss the intrinsic sensitivity
of the fission-track detection method for measuring the sputtering yield
of 235 u and to suggest possible extensions of the technique. The primary limit to the sensitivity of the technique is the 235 u background in
the mica itself.

Good (but readily available) mica such as was used in
235
this experiment has an effective surface contamination of ~10 8
u atoms/
cm2 (equivalent to ~2. 5 ppb). This permitted keeping the level of background tracks below 1% while measuring a sputtering yield S of ~10- 4 for
100 keV H~ on 235 u with a fluence of 5 x 10 15 protons and an irradiation
of the mica with lo 16 neutrons/cm2 .

This sensitivity implies that the
technique could be used to measure sputtering yields down to S = 10- for
materials doped with 0.1% (atomic) 235 u and bombarded with~lo 17 beam particles.

In that case the background level of tracks in the mica would be

at about the 50% level.

It would of course be necessary to allow for non-

stoichiometric effects in such a sputtering of very small amounts of
material from a surface doped with something as heavy as 235 u.
If one wishes to push the technique to measure smaller sputtering
yields, the most obvious approach is to increase the fluence of beam
particles, either by running for long periods of time or by increasing
the beam current.

In the latter case one may run into the problem of

excessive heating of the target and begin to evaporate, rather than
sputter, material.

We found it possible to run as much as 10 ~A/cm

100 keV H~ on 0.025 mm uranium foil without excessive hea ti ng, which

-53implied 5 hours to accumulate 10 17 beam particles. Metals more conductive than uranium would allow higher beam currents, as would ope r ation
at lower beam energies. Finfgeld et a1, 10 for example, have sputtered
thin metal foils with H+ and D+ beams of a few keV at beam current densities around 1 mA/cm 2 for beam currents of ~100 ~A. Such a beam current allows accumulat i on of 10 19 projectil es in 5 hours and makes pos sible the measurement of S = 10-6 for materials doped with 0. 1% (atomi c )
23su.

One can also gain a bit in the density of s puttered material collected by decreasing the distance from target to collector foil.

For

backsputtering it is difficult to gain a great deal due to practi cal
limits set by the aperture required in the foil holder to admit the beam
and by the finite beam size itself.

For transmittance sputtering these

practical limits do not apply, since the catcher foil may be placed very
close to the target . In that case the density of collected 235 u atoms is
simply S xbeam fluence x doping fraction, and S = 10- 8 could easily be
measured for a beam fluence of 1o 20 cm- 2 and 0. 1% (atomic) doping. This
would also be true for backsputtering by particles with enough energy to
pass through a catcher foil placed immediately in front of the target . In
that case one would use a foil with low 235 u content to collect the sputte red rna te ri a 1 and then p1ace the foi 1 on the s urface of a mica s heet
during the neutron irradiation.

Good results have been achieved using

high purity aluminum for such catcher foils.
The great sensitivity of this technique may also prove useful in
the investigation of such exotic forms of sputtering as electron and

-54photon sputtering, which are expected to have very low yields. As an
example, photon sputtering of 235 u is being looked for in Kellogg
Laboratory.

The UV light source employed emits 3 x 1015 4.88 eV

photons cm- 2 sec-l at a distance of 4 em (Barbara Cooper, private communication). Requiring that 10 8 235 u per cm- 2 be col lected on the
mica catcher foils (to be above intrinsic 235 u background) in a period
of 106 seconds implies, via the formula
NnS
N = __r:__
2nr2

(A-1)

that S < 10-ll could be measured. In fact, substituting Nt = 10 8cm- 2 ,
r = 4 em, and Np = 3 x 10 15 x 106 = 3 x 10 21 in (A-1) yields a measurable S value of 3 x lo- 12 .
It is evident that, in the cases of photon and electron beams at
least, the possibility of very large beam fluences implies that extremely low values of S are subject to measurement.

Ultimate limits

may be set by such competing processes as self-sputtering of the uranium
target by fission fragments or by a 's from decay of 234 u.

-55-

APPENDIX B
Ultra-High Vacuum Technique
The aim of this appendix is to pass on to the next generation of
Caltech graduate students some of the practical knm-11 edge I have 1earned
concerning ultra-high vacuum technique.

That which I have not learned

fills many volumes of the extensive literature on the subject.

Two ex-

cellent gateways to that literature, which are very valuable sources of
information in their own right, are High Vacuum Technology (Roth 1976) 37
and the chapter on vacuum in the Handbook of Thin Film Technology (Maisel
and Glang, ed, 1970). 27 The first chapter of Roth contains a list of the
primary journals dealing with vacuum, the transactions of the important
vacuum conferences and most (if not all) of the books that have ever been
written about vacuum.

I would recommend taking a close look at this book

before undertaking any extensive work involving ultra-high vacuum.
The technology necessary for producing systems possessing considerable experimental utility and operating in the range of l0- 9 -lo- 10 torr is
well established.

Several firms sell the type of components incorporated

in the system utilized in this experiment.

To be useful in ultra-high

vacuum the materials used to construct these components must have low intrinsic outgassing characteristics.

In addition, they must have low vapor

pressures at the baking temperature (400°-500°C) required to effectively
drive off gases adsorbed when the system is at atmospheric pressure. This
requirement eliminates such metals as lead, zinc, and cadmium from consideration and implies that brass should never be admitted to an UHV system.
Stainless steel is the primary material employed, and other materials

-56commonly used include oxygen-free high-conductivity (OFHC) copper
(primarily for metal-seal gaskets), dense glasses like Pyrex (primarily
for viewports), and certain ceramics (primarily for electrical insulation).

Common alloys of aluminum in rolled or extruded form can be

used in ultra-high vacuum if they are well-cleaned and baked (see Roth
1976, p. 142, 332 and Power and Robson 1962).

For a while both the

target plate and collimation cylinder in the sputtering assembly were
6061 aluminum, and no difficulty in attaining 10-gtorr was experienced.
In fact, one can place small quantities of any material in the vacuum
provided the pumping speed is adequate to compensate for the outgassing.
Roth (1976, p. 142) gives a nomogram for matching pumping speed and
outgassing properties to achieve the desired pressure.

Pragmatically

speaking most materials are still prohibited in a system of the size used
in this experiment, but it has proved feasible to introduce considerable
amounts of materials not normally recommended for UHV work. It was possible, for example, to reach l0- 9torr in the normal pumpdown time with
225 cm2 of unbaked muscovite mica catcher foils installed.
Selection of suitable materials is not in itself sufficient to
guarantee attainment of ultra-high vacuum.

It is also necessary to re-

duce the outgassing rates of these materials.
such a process.

There are two steps in

First, the materials are carefully cleaned to minimize

the amount of gas brought into the system.

Second, after bein9

installed, they are baked at low pressure to drive off gas adsorbed on
the surface and to decrease the concentration of dissolved gases which
diffuse into the vacuum from the bulk material.

-57Rapid pumpdown is greatly facilitated by thorough cleani ng of
pieces before introducing them into the UHV system.

This is parti cu-

larly important if one hopes to add parts to the system and pump down
to ~lo- 9 torr quickly without baking.

It has proved possible to do s o

with this system provided the parts constitute a fairly small fra ction
of the system's surface area.

A set of cleaning procedures that have

been used successfully in Kellogg for a few years appears in Tabl e 10.
For most materials the procedures consist of a degreasing step, an
etching bath to remove oxides and other porous surface material wh ·lch
adsorb gases well, rinses to remove the various solvents, and a final
methanol rinse to reduce the amount of water adsorbed on the surface
and to remove any lingering grease.

More information on cleaning (in-

cluding other recipes for etch baths, more detail on degreasing, and
so forth) can be found in section 9.4 of Espe (1970) and section 7. 2
of Roth (1976).
A chemical cleaning process that is commercially available is
DS-9 sold by the Diversey Co. of Chicago.
one of which is an etching solution.

It consists of three baths,
Milleron 32 reports an outgassing

rate for stainless steel of ~lo- 12 torr- 1 cm- 2sec-l after treatment
with DS-9 and 12 hr of pumping, but no baking.
gassing rate without baking.

This is a very low out-

I would be surprised to find that the

treatment for s tainless steel specified in Table 10 results in quite as
low an outgassing rate.
If it is necessary (as it usually is) to reduce the outgassing
rate below that obtained after cleaning, by far the most effective

-58-

method is heating in vacuum. Power and Robson 34 obtained outgassinq
rates of 3 x 10- 14 torr .R. cm- 2sec-l for stainless steel and 2 x 10- 14 f or
99% aluminum shim after 16 hr of baking at 400°C and P ~ l0- 9 torr . Outgassing of the same samples was about 2 x lo- 10 torr- .R.-cm- 2sec-l after
24 hr of pumping at l0- 9 torr without baking.

The rate of degassing

increases very rapidly with temperature, being about ten times greater
for outgassing stainless steel at 400°C than at 300°C (as shown in
Ca 1der and Lewin 1967; this is an excellent reference on the theory and
results of baking stainless steel).

Most of the UHV system used in thi s

experiment is bakeable to at least 400°C.

As a practical matter the

rate of degassing must be kept low enough so the ion pump does not stall,
which means that 400°C must normally be approached gradually.

The most

efficient way to do this is to employ a controller (such as the one
built by Bob Weller for Kellogg Lab) which senses the ion pump current
and keeps it below an acceptable maximum by regulating the current to the
bake-out heaters.
Baking of new components is especially important since a great deal
of gas is dissolved in the materials during manufacture.

Since gas dif-

fusion constants are much larger at elevated temperature than at room
temperature, UHV materials may be effectively degassed irrevers ibly by
baking at 300-400°C.

It often requires a few days of baking at such

temperature to achieve the lowest possible outgassing rate.

Again, the

higher the temperature the quicker the results, 1 hour at 1000C 0 being
equivalent to 2500 hours at 300 °C. 7 The gas that is the worst offender
in this regard is hydrogen.

A typical concentration of H2 in stainless

-59steel is 0.3 torrQ..cm- 3 ,7 and long after all water vapor and other
gases have been desorbed from the surface during bake-out H2 will continue to evolve, comprising 99% of the gas output. 7 Since the diffusive release of hydrogen is much less at room temperature, it is
-16

theoretically possible to achieve an outgassinq rate of~ 10
torr Q..
cm- 2sec-l for a typical stainless steel component. 7 I t is not poss ible
to go much lower \'lith a thin-walled component because of penneation
through the walls by hydrogen in the atmosphP.re.
Once the system has been thoroughly degassed there are a couple of
things that can be done to minimize outgassing during subsequent pumpdowns.

First, the system should be vented with dry nitrogen when it is

brought back up to atmospheric pressure.

This is to eliminate the con-

densation of water that occurs \'/hen room air is all owed to expand into
the vacuum. Adsorbed water vapor is difficult to get rid of without
baking. 28 The second thing is to make sure that none of the surfaces
which will be in vacuum comes in contact with skin--the grease deposited
in a fingerprint acts as a vast gas source in ultra-high vacuum.
When pumping down, the ion pump will start readily at pressures
below l0- 2torr provided the outgassing from the system is small. In
practice, the outgassing load is frequently great enough that the ion
pump will not start without difficulty.

If the gas load is large the

high pump current required causes the pump to heat up, which in turn
causes the pump itself to outgas.

The increased gas load leads to still

higher pump current, and the cycle may run away resulting in shut-down
of the pump by the thermal safety relay.

Forced air coolinq of the pump

-60helps avoid overheating, but the most effective technique is simply
to plug the pump into a Variac and use that to maintain the pump current below some safe level. It is also very helpful to pump the
system down below l0- 4torr by opening up to beam- line vacuum (with the
cold trap filled) before attempting to start the ion pump.

This pro-

cedure has the additional benefit of prolonging the life of the pump
element by decreasing the volume of gas to be pumped.

-61-

APPENDIX C
Effect of Adsorption on the Sputtering Yields
I will consider here the question of whether or not the vacuum
employed was sufficient to prevent reduction of the sputtering yi elds
due to adsorption of gases onto the target surfaces.

The factors which

determine the amount of gas adsorbed per time are the partial pressures
of the reactive gases in the vacuum system's atmosphere, their sticking
fractions on impact with the target surface, and the residence times
before adsorbed atoms are vibrated back off the surface.
The sticking fraction is somewhat dependent on temperature and
the amount of gas already adsorbed, but for most gases on most metals
it is between 0.1 and 1. 29 For a reactive metal like uranium we can
assume it is close to 1.
The average residence time is given by 29

( C-1)

where 'o is the adsorbed atom's period of thermal vibration normal to
the surface, and Ed is the energy required to desorb the adatom. 'o is
approximately l0- 13sec. 29 • 38 An Ed of 25 kcal/mole (which is typi cal
of H2o on metal surfaces) thus gives T = 105sec at room temperature.
Since uranium is a good getter, 39 and since adsorptio.n energies of the
common reactive gases (N 2 , 02 , co , etc.) on other getters are normally
greater than 25 kcal/mole, we conclude T > 10 5sec for all relevant noninert gases.

The maximum time between sputter-cleaning of a target and

-62the end of the sputtering data run is a few hours, hence it is apparent
that most of the atoms adsorbed on the uranium surface remain during
the sputtering run unless sputtered away.
An upper limit to the amount of gas adsorbed on the uran ium su rface by the end of a data run can be obtained by assuming all of the
residual gas in the system at its base pressure is reactive. The base
pressure is 510- 9 torr, and the gas density n ~ 3.5 x 10 7cm- 3 at this
pressure.

The average velocity

v of the gas particles is given by 36
(C -2 )

For a molecular weight Mmol = 28 and T = 300° K, v = 4.5 x l0 4cm/sec. The
flux of gas particles impacting the target surface is

nv

= -

(C-3)

in this case.

This implies a time of 2 x 10 3sec (~ 30 min) to adsorb

a monolayer on the target surface.

In the worst case several mono-

layers would have been adsorbed between sputter-cleaning and sputtering
of a target.
In reality, it is expected that at least 90% of the residual gas
in the UHV system is hydrogen and helium. 7 , 37 This is due primarily
to the low pumping speed for helium and to the high emission rate of H2
from the system's stainless steel walls and from the titanium sublimator
filaments.

Helium is not adsorbed.

Hydrogen is, but the amount that

-63-

might be adsorbed is considerably less than the quantity being delivered
to the target surface by the ion beam anyway.

It is probable, therefore,

that less than a monolayer of gas which might affect the hydrogen sputtering yields was adsorbed before the end of the short (15 min) sputtering runs. The masses of the adsorbed gases are so much less than that
of 235 u that a monolayer of gas added to the top few monolayers of
uranium actually being sputtered cannot be expected to reduce the yield
of 235 u by more than a few percent.
It is possible that a few monolayers may have been adsorbed by the
end of the longest (5 hour) runs.

This may be responsible for the fact

that the long runs gave sputtering yields about 25% lower than the short
runs.

An idea of the maximum effect adsorption may have can be gained

from the observation that the runs on targets which had not been sputtercleaned (and were thus known to have surface oxidation many monolayers
deep) produced sputtering yields roughly a factor of 2 lower than the
sputter-cleaned targets.
It has been assumed thus far that the sputter-cleaning of the
targets during proton bombardment was not sufficient to ensure an atomically clean surface.

Let us look briefly at that assumption.

The fact
that relevant adsorption energies are on the order of 1 eV/atom37 and
the knowledge of proton sputtering yields for a variety of materials 10 • 31

make it clear that the desorption yield in this case for such gases as
nitrogen and oxygen must be at least 10- 2 . With a typical ion beam current density of 6 x lo 13 cm- 2sec-l, the desorption rate would then be at
least 6 x lo 11 cm- 2sec-l, which is approximately the adsorption rate for

-64-

air at l0- 9 torr (eq. C-3).

Given that most of the residual gas is

hydrogen and helium, it appears that the targets actually became substantially cleaner during hydrogen ion bombardment.
All things considered, it is very unlikely that the hydrogen
sputtering yields in the short runs were reduced by more than a few
percent, as an upper limit, by effects due to gas adsorbed on the target surface.
Since the argon and helium runs were very short (< 60 sec) and
took place immediately after sputter-cleaning, and since the sputtering
rates are so much higher than those of hydrogen, there was no effect on
these yields due to adsorbed gas.

-65APPENDIX D
Neutron Fluence Detennination Using NBS Neutron-Irradiated Standards
The National Bureau of Standards has available glass wafers containing precisely known quantities of uranium and thorium which have
been irradiated with carefully determined fluences of neutrons.

Iden-

tical non-irradiated standards are also available which the user may
include with samples being neutron irradiated .

The user's neutron

fluence may subsequently be determined by comparison, after etching
both standards and counting the etch pits due to fission tracks in each.
The basics of this technique are covered in NBS Special Publication
260-49: "Standard Reference Materials: Calibrated Glass Standards for
Fission Track Use". 8 I will simply summarize them and add some infonnation of a practical nature.
The first step is to decide which of the four available uranium
concentrations (nominally 500 ppm, 50 ppm, 1 ppm and .07 ppm) is appropriate for the anticipated neutron dose, bearing in mind that the
uranium is <0. 7 atom % 235 u. It was found that SRM 963 (1 ppm) gave a
convenient density of tracks when exposed to 1o16 -1o 17 n/cm2 and etched
according to NBS recommendations (75 sec in 16% HF at 20°C).
Once a blank wafer has been irradiated, both it and compari son
standards must either be broken or ground to reveal an internal surface
free of fission track contamination from the environment.

Fracturing

immediately yields a smooth internal surface which can be etched without
further treatment.

Unfortunately, the fractured surface is not flat and

is rather inconvenient to s can under the microscope.

Grinding and

-66polishing requires more time and effort but, done properly, yields a
smooth, flat surface which is conducive to easy scanning.

The details

of grinding and polishing proceduresdeveloped by the author are presented next.
The initial step, and it is an important one, is to round off
the edge (about l/64" will do) of the face of the wafer to be ground.
Next, at least 30~ of the face is removed by hand grindinq directly on
a smooth glass surface using a slurry of water and 5~ alumina.

With

firm pressure from one finger and 3-4 small circular motions per
second, 30~ can easily be removed in less than 5 minutes.

30~

is re-

moved because it is greater than the longest fission-fragment range.
The first reason for rounding off the edge of the wafer is to avoid
big scratches on the surface made by tiny fragments of glass which chip
off a sharp edge during the grinding operation.

The reason for grinding

directly on a sheet of glass is that if one attempts to grind on a piece
of polishing paper, it will take close to forever to remove 30~.
After a brief ultrasonic cleaning, polishing can now begin.

Low

speed polishing is necessary to avoid heat buildup, which causes track
fading. 8 I found the Mini-Met polisher (manufactured by Buehler, Ltd.
of Evanston, Ill.) to be quite useful for this step.

The polishing may

also be done by hand i f one has the patience and endurance.

If one

employs the Mini-Met, it is necessary at this juncture to either fabricate a sample holder sized to the NBS wafers or to glue them to 1" glass
wafers, which fit the standard Mini-Met holder.

The latter option was

chosen by the author and offered the additional advantage of a convenient

-67-

handle for manipulating the wafer without worryinq about marring its
surface.

It was found that 1" glass wafers, l/16 11 thick (which are

commonly used for mounting petrographic thin sections) are readily
available and are ideal.

Cyanoacrylate cement gave a quick, strong

bond between the glass wafers.
The first of two polishing operations utilizes a slurry of water
and ll-1 alumina on (Texmet®) polishing cloth.

One hour of polishing

with the Mini-Met set at its lowest speed and third-from-lowest pressure
gave an adequate surface finish without noticeable heating of the standards.

The second reason for rounding off the edge of the wafer will

now become obvious if it has not been done, for the sharp edge will dig
into the polishing cloth and cause the standard to roll around on its
edge rather than riding around with its full surface in contact with the
polishing surface.
If the polished surface is viewed with phase-contrast microscopy

at this point, many llJ-Wide scratches and a fairly small population of
llJ pits will be evident.

If the standard is etched at this point, these

tiny pits will etch to about the size of typical fission-fragment
track etch pits and will make accurate track counting difficult. It is
standard practice, 11 therefore, to give the surface a final polishing
using 0.3lJ alumina.

Another brief ultrasonic bath precedes the 0.3lJ

polishing, which is performed on very soft polishing cloth (Microcloth®
is used with the Mini-Met).

With the same settings as above, the Mini-

Met produced a surface virtually devoid of ll-1 pits in 30 minutes.
A final brief ultrasonic bath, distilled water rinse, and filtered-air blow drying leaves the sample clean and ready to be etched.

-68-

It is imperative that care be taken not to mar the polished surface
until after it is etched.

Wiping it with a Kimwipe, for example, pro-

duces enormous numbers of 1~ pits which will interfere with track
counting.
The etch recommended by the NBS {75 seconds in 16% HF at room
temperature) gave excellent results.
3-4~ across

Track etch pits were typically

and were easily counted in reflected light using phase-

contrast microscopy at 400X.

This technique is recommended for the

great clarity it imparts to etch pits and other surface features . Also,
the colors are very beautiful.

More importantly, it allows accurate

discrimination against etch pits which are not due to fission tracks.
This is based on the fact that the 75-second etch is short enough that
most of the fission tracks have not been etched all the way to the end
and are thus sharp-pointed cones. 11 •13 Pits which were present before
etching were not sharp-bottomed to begin with and are nicely-rounded
pits afterwards.

These rounded bottoms reflect light in a characteris-

tic manner which is different from the characteristic reflection from
sharp-bottomed pits.

The difference between the two types of reflection

is most apparent when the microscope is focused up and down through the
depth of the pits.

As a rounded bottom is brought into focus, the re-

flection coalesces symmetrically and uniformly into a bright but somewhat diffuse spot which is essentially the same shape as the pit and
whose size is a considerable fraction of the pit's.

When a sharp-

bottomed pit is brought into focus, the reflections move around in a
nonuniform manner and coalesce into a spot which is more intense and

-69-

better defined, at least at one end, than in the previous case.

For

oblong track pits a criss-crossing of reflections (due to second order
reflections from the cone walls, I suppose) as the focus is moved up and
down is generally observed.

For large cones seen end on there is very

little reflection at all.
In studying the difference between track pits and spurious pits,
I found it useful to grind and partially 1~ polish an un-irradiated
standard.

After giving it the standard etch many pits were visible

which were not due to tracks and a graphic comparison to a well-polished,
irradiated, and etched standard, where most of the pits were due to
tracks, could be made.
Since tracks crossing the surface have lengths anywhere from zero
to full range, tracks shorter than some length will be etched all the way
to the end and become round-bottom pits which are not counted.

This

technique cannot therefore yield directly an absolute neutron fluence
from a single user-irradiated glass standard.

However, if both user- and

NBS-irradiated wafers receive identical etches and are counted carefully
following identical counting criteria, very accurate fluences relative
to the NBS values may be determined.

Identical etches are required be-

cause the density of revealed tracks is a strong function of the etching
time and etchant dilution.
One final detail is that one must choose between the two NBSirradiated standards supplied in each package of standards if one desires neutron fluence values better than ~10%.

This is because one of

them (RT-3) was exposed to neutrons in a position in the NBS reactor

-70having a higher proportion of fast neutrons than the other (RT-4). This
is indicated by different cadmium ratios for the gold and copper fluxmonitor foils in the two positions. 8 If the neutron flux is well thermalized during irradiation of the user's blank standard, then RT-4 would
be the wafer to compare to.

Wafers from position RT-3 would give a

better approximation to neutron energy distributions nearer to core.

To

give a quantitative example, the cadmium ratio for the gold foil monitors
is 10.2 in position RT-3 and 87 in position RT-4 and is 3.3 (Bruce Taylor,
private communication) in the center of the UCLA reactor core.

A stand-

ard exposed in the center of the UCLA reactor and compared to an RT-3
wafer would slightly underestimate the effective flux.

-71VII.

REFERENCES

1.

Almen, 0. and G. Bruce, 1961, NIM l.]_, 257,279.

2.

Andersen, H., 1974, in V. Vujnovit {ed.) 7th Yugoslav Symposium
and Summer School on the Physics of Ionized Gases (Institute of
Physics of the University of Zagreb), p. 361.

3.

Bethe, H. and J. Askin, 1953, in 0. Segne (ed . ) Experimental
Nuclear Physics (John Wiley & Sons, New York), p. 166.

4.

Bhandari, N. et al., 1971, Earth Planet . Sci. Lett . .J.l, 191.

5.

Biersack, J., D. Fink and P. Mertens, 1974, J. Nucl. Mat. Met.

~.

194.
6.

Burnett, D. and D. Woolum, 1974, Earth Planet. Sci. Lett.£!._, 153.

7.

Calder, R. and G. Lewin, 1967, Brit. J. Appl. Phys. ~. 1459.

8.

Carpenter, B. and G. Reimer, 1974, NBS 260-49: "Calibrated Glass
Standards for Fission Track Use."

9.

Espe, W., 1966, Materials of High Vacuum Technology (Pergamon Press,
New York) Vol. 1: "Metals and Metalloids".

10.

Finfgeld, C. et al., 1974, AEC Report OR0-3557-15.

ll.

Fleischer, R., P. Price and R. Walker, 1975, Nuclear Tracks in
Solids (University of California Press, Berkeley).

12.

Ibid, p. 318.

13.

Fleischer, R. and P. Price, 1964, J. Geophys. Res. 69, 331.

14.

Flint, 0., J. Polling and A. Charlesby, 1954, Acta Metall. £, 696 .

15.

Furr, A. and C. Finfgeld, 1970, J. Appl. Phys. !!.._, 1739.

16.

GrtSnland, F. and W. Moore, 1960, J. Chern. Phys. 32, 1540 .

-7217.

Gschneidner, K., 1964, Solid State Phys. ]_§_, 275.

18.

Haines, E., private communication.

19.

Kaminsky, M., 1974, Proc. 5th Int. Conf. on Plasma Physics and
Controlled Nucl. Fusion Res., IAEA No. CN-33, p. 287.

20.

Kaminsky, M., H. Wiedersich and K. Zwilsky (eds), 1974, Proc.
Conf. Surface Effects in Controlled Thermonuclear Fusion Devices
and Reactors in J. Nuc. Mat. Met. 53.

21.

Kaminsky, M. and S. Das, ibid, p. 162.

22.

Kaminsky, ~~. and S. Das, 1975, Proc. 6th Symp. on Eng. Prob.
Fus. Res., IEEE No. 75CH1097-5-NPS, p. 1141

23.

Kenknight, C. and G. Wehner, 1964, J. Appl. Phys. ~. 322.

24.

Kittel, C., 1971, Introduction to Solid State Physics (Wiley &
Sons, New York), p. 39.

25.

Lindhard, J., M. Scharff and H. Schi~tt, 1963, Kgl. Danske
Videnskab, Selskab Mat.-Phys. Medd. ]1, No. 14.

26.

Lindhard, J., V. Nielsen, M. Scharff and P. Thomsen, 1963, Kql.
Danske Videnskab. Selskab Mat.-Phys. ~1edd. ]1, No. 10.

27.

Maisel, L. and R. Glang (eds), 1970, Handbook of Thin Film Technology (~1cGraw-Hill, New York), Ch. 2: "High Vacuum Technology".

28.

Ibid, p. 2-44.

29.

Ibid, p. 2-41.

30.

Mayer, J. and J. Ziegler (eds), 1973, Ion Beam Surface-Layer
Analysis, Thin Solid Films 19.

31.

tkDonnell, J. and D. Ashworth, 1972, Space Res. XII, 333.

32.

Milleron, N., 1967, IEEE Trans. Nucl. Sci. NS-14 (3), 794.

-7333.

Nash, D., D. Matson, T. Johnson and F. Fanale, 1975, J. Geophys.
Res. 80, 1875.

34.

Power, B. and F. Robson, 1962, Transactions of the Second International Vacuum Congress (Pergamon Press, Oxford), p. 1175.

35.

Price, P. and R. Walker, 1963, J. Geophys. Res. 68, 4847.

36.

Reif, F., 1965, Fundamentals of Statistical and Thermal Physics
(McGraw-Hill, New York), p. 268.

37.

Roth, A., 1976, Vacuum Technology (North-Holland, New York).

38.

Ibid, p. 174.

39.

Ibid, p. 262.

40.

Schi¢tt, H., 1966, Kgl. Danske Videnskab. Selskab Mat.-Fys. Medd.
35, No. 9.

41.

Seitz, M., R. Walker and B. Carpenter, 1973, J. Appl. Phys. 44, 510.

42.

Sigmund, p. ' 1968, Can. J. Phys. 46, 731.

43.

Sigmund, p .• 1969. Phys. Rev. 184' 383.

44.

Sigmund, p .• 1972. Rev. Roum. Phys. ll_, 823,969,1079.

45.

Sigmund, p .• M. fv1atthes and D. Phi 11 ips, 1971, Rad. Effects ll_, 39.

46.

Summers, A.' N. Freeman and N. Daly, 19 71 , J. App 1 . Phys. !S 4774

47.

Switkows ki, Z. , P. Haff, T. Tombre 11 o and D. Burnett, 1977, J.
Geophys. Res. (in press)

48.

Thompson, M., 1968, Phil. Mag. ~. 377.

49.

Weissman, R. and P. Sigmund, 1973, Rad. Effects }2_, 7.

50.

Weissman, R. and R. Beh ri sch, 1973, Rad. Effects .!2_, 69.

51.

Wickramasinghe, N., 1972, Mon. Not. Roy. As tro. Soc. 159, 269.

-7452.

Winterbon, K., 1972, Rad. Effects .J.l, 215.

53.

l~interbon,

K., P. Sigmund and J. Sanders, Kgl. Danske Videnskab .

Selskab Mat.-Fys. Medd. 12· No. 14.
54.

Yonts, 0., C. Normand and D. Harrison, 1960, J. Appl. Phys. I!_, 447.

55 .

Ziegler, J., J. Cuomo and J. Roth, 1977, Appl. Phys. Lett. 30, 268

-75-

TABLE 1
Neutron fluxes measured by the National Bureau of Standards for
fission track glass standards SRM963 (seep. 25).

. . .. 4

-76Table 1
SR..\1 963 Fission Track Glass Standard
Wafer Number
NBS Reactor
Identification Position

Neutron Flux Mean Valug
and Standard Deviat~on
(xl0 11 n·cm- 2 ·sec- 1 2 >

Tolerance Int er vals

(95%)

~ 9 9 '6

±0.09
±0.09
±0 .09

Cu Foil
614001-614040
614041-614090
614091-614125

RT - 4
RT-4
RT-4

1. 26 ± 0.026
1. 26 ± 0.026

±0.06
±0.06
±0.06

614126-614165
614166-614215
614216-614251

RT-3
RT-3
RT-3

5.38 ± 0.17
5.10 ± 0.17
5.14 ± 0.17

±0.42
±0.42
±0.42

±0.60
±0.59
±0.60

±0.09
±0 .08
±0.09
±0.21
±0.21
±0.21

1. 24 ± 0.026

Au Foil
1. 41 ± 0.024

614001-614040
614041-614090
614091-614125

RT-4
RT-4
RT-4

1. 44 ± 0.024

±0.06
±0 .06
±0.06

614126-614165

RT-3
RT-3
RT-3

6.03 ± 0.059
5.97 ± 0.059
6.13 ± 0.059

±0. 1 5
±0.15
±0.15

614166-614215
614216-614251

1. 36 ± 0.024

a Standard deviations refer to individual metal foils.
blrradiation was performed at a power of 10 megawatts;
80 seconds in RT-3, or 120 seconds in RT-4.
cA 95 percent tolerance interval is estimated to include the
measurement of approximately 95 percent of all individual
wafers of the population of wafers. Thus, the probability is
approximately 95 percent that any individual wafer measurement
will lie inside the 95 percent tolerance interval. A similar
definition holds for 99 percent tolerance interval.

12

-77TABLE 2
Determination of total fluence received by NBS glass standard SRM-A
during neutron irradiation at UCLA reactor hy comparison to known
fluences received by NBS fission track glass standards SRM963 (RT-3
and RT-4) (see p. 25).
Fields of view were areas 260~ x 260~ on the standard glass wafers as
seen under phase-contrast microscopy at 400X.
Neutron fluences for SRM-A are obtained by multiplying the NBS-determined fluences for RT-3 and RT-4 by the ratio of tracks counted on
SRM-A to the number counted on the corresponding RT-standard.
Neutron fluences as determined by the NBS for glass standards RT-3 and
RT-4 are obtained from Table 1 by multiplying the neutron flux mean
values for the appropriate wafers by the times they were irradiated in
the NBS reactor (from footnote b).

The serial numbers of the particular

wafers used in this experiment are 614016 for RT-4 and 614141 for RT-3 .

-78-

TABLE 2
SRM-A

RT -3

RT-4

Fields of View Counted

20

40

80

Total Number of Tracks Counted

750

732

556

Tracks per Field of View

37.5

18.3

6.95

2.05

5.40

Track Density Ratio: SRM-A/RTNeutron Fluence (n/cm2 ) as
determined by NBS with
Au foils:

4. 82 X 10

Cu foils:

4. 30 X 10

Neutron Fluence for UCLA Reactor
Run by comparison of
SRM-A to RTAu foils:

9. 88 X 10

Cu foils:

8.82 X 10

15
15

15
15

1. 69 X 10
1.51

10

15
15

9.13 X 10
8.15 X 10

15
15

-79TABLE 3

Sputtering yields of 235 u due to bombardment by ion beams at normal
incidence to the targets. Yields "S" are atoms of 235 u emitted per
incident atom.

With a cos 114e fit to the angular distribution data,

S is calculated from

N°

S = 8nr
-5- N N of

p n

where N~

nurmer of fission tracks per cm2 on collector foil at

e = oa
Np

number of incident atoms

Nn

neutron fl uence = l. 46 X 1o15 cm- 2

of

582 bn

3.7 em

Most targets contained 99.71 % 235 u but a few (marked 93 in the % 235 u
column) contained 93.32% 235 u, which necessitated a correction when
computing S.
Quoted errors on the values of S are overall errors including uncertainty due to counting and error in determination of Nn.

-80TABLE 3
Run

% 235u

I on Beam

Np

No

3.9 X 10 5
2.7 X 10 5

3.2 ± 0.5 X 10-4
2. 7 ± 0.4 X 10- 4

2.5 X 105
2.0 X 105

2.0 ± 0.3 X 10- 4
1.6 ± 0. 25 X 10- 4

1.5 ± 0.25 X l0 - 4

2. l X 10 5
1.3 X 105

1.8± 0.2 X 10- 4

S(atoms/atom)

Sputter-cleaned targets:
3-3

40 keV Hl

99

1017

3-4

60 keV Hl

2-4

80 keV H

II

2-6

100 keV Hl

II

2-5

120 keV Hl

II

II

l. 8 X 10

Targets cleaned only HN0 3:
l-4

40 keV H1

93

II

l-2

60 keV H1

99

II

l-3

80 keV H1

II

II

1-l

100 keV H1

93

II

l-5

120 keV H1

99

II

l.l X 105
8.9 X 10 4

1.0± 0.15 X 10-4
8.9 ± 0.15 X 10- 5
7.7 ± 0.10 X 10- 5

8.3 X 10 4

6. 7 ± 0.07 X 10-5

3.6 X 104
5.4 X 10 4

2.9 ± 0.6 X 10- 5
3.9 ± 0.6 X 10-

4.7 X 10 4

l . 9 ± 0.4

Targets not cleaned at all:
3-6

40 keV Hl

99

II

3-5

60 keV Hl

99

II

Sputter-cleaned target:
2-3

80 keV Ar

99

2.0x10 12

-81-

TABLE 4
Sputtering yields of 235 u due to bombardment by ion beams at normal
incidence to the targets. Yields "S" are atoms of 235 u emitted per
incident atom (rather than per incident ion).

With a cos 114e fit

to the angular distribution data, S is calculated from

where N~ =number of fission tracks per cm 2 on collector foil at
8 = 0°

Np

= number of incident atoms

Nn

= neutron fl uence = l. 37 x 10 16 cm-2

of

= 582 bn

3. 7 em

All targets in runs in series 4-7 contained 99.7% 235 u.
Errors on the values of S are overall errors including uncertainty due
to counting and error in determination of N .

-82TABLE 4
Run

Ion Beam

Np

5-5

40 keV H3

1.5 X 1016

5-6

40 keV H2

l.Q X 1016

5-4

40 keV H1

5.0 X 1015

7-6
4-6

50 keV H1
80 keV H1

2.5 X 1015
5.0 X 1015

4-4

100 keV H1

It

4-5

120 keV H1

It

7-2

80 keV H2

It

7-3

100 keV H2

It

7-4

120 keV H2

It

7-5

120 keV H3

7.5 X 1015

6-6

40 keV He

3.0 X 10 14

6-5

60 keV He

It

6-2

80

keV He

It

6-3

100 keV He

It

6-4

120 keV He

It

4-3

40 keV Ar

4-2

80 keV Ar

1.0 X 10 12
It

N~(cm- 2 )

S(atoms/atom)

7.0 X 10 5
4.8 X 10 5

4.0 ± 0.5 X 10- 4
4.2 ± 0.5 X 10- 4

3.0 X 10 5
1.5 X 105
1.6 X 10 5

5.2 ± 0.7 X 10-4

1. 5 X 10 5
1.4 X 10 5

2.6 ± 0.4 X 10-4

3.2 X 105
2.9 X 10 5

5.5 ± 0.8 X 10-4

2.7 X 105
3.6 X 10 5

5.2 ± 0.7 X 10-4
2.8 ± 0.3 X 10-4

2.4 ± 0.3 X 10-4

5.0 ± 0.6 X 10-4
4.7 ± 0.6 X .10 -4
4.2 ± 0. 5 X 10- 4

4. 7 X 10 5
5.0 X 10 5

1.3 ± 0.15 X 10- 2
1.4 ± 0.15 X 10- 2

4.2 X 105
3.3 X 10 5

1.2 ± 0.15 X 10-2
9.4 ± 0.10 X 10- 3

3. 6 X 105

1.0 ± 0.10 X 10- 2

2.0 X 105
2. 3 X 105

1.7 ± 0.5
2.0 ± 0.5

-83-

TABLE 5
The number of fission stars consisting of a number of fission tracks
Nt, in four ranges, produced by total doses NP of various ion beams
bombarding uranium metal targets.

In each case the numbers were deter-

mined by careful scanning with a transmitted-light microscope of the
specified area on mica detectors which recorded the 235 u distribution
on the sputtering collector foils.
Nt is the median star size in each case, and Nu is the number of atoms
in the corresponding chunk.

(See Section IIID).

The entry marked "Blank" gives the results of scanning the back of the
mica with the 80 keV Ar+ data bands on it.

Blank

80 keV Ar

80 keV He
100 keV He
120 keV He

60 keV H1
80 keV H1
100 keV H1
120 keV H1

II

II

II

II

II

II

2 X 1012

II

II

1.8

II

II

3 X 1014

II

II

II

II

5 X 1015

II

II

lO 17

40 keV H
60 keV H1
80 keV H1
100 keV H1
120 keV H1
II

Ion Beam

II

Area Scanned
(cm2)

178

27

13

126
133

7l

27
17

76

12

35
37
30

39

17
12

66

26
30
21

Number of Stars with Nt =
> 75
5-24
25-49 50-74

TABLE 5

25

20
25

10

15
20
25
20
15

Nt

2.5 X 10 7

2.0 X 106
2.5 X 106

106

1.5 X 10 7
2.0 X 10 7
2.5 X 10 7
2.0 X 10 7
1.5 X 10 7

Nu

00

-85-

TABLE 6
A comparison of chunk emission data for H;, He+, and Ar+ ions on
uranium.

The total number of chunks counted in each case is given,

as well as the number/cm 2 on the micas.

The ion beam fluence is the

total fluence delivered to a given target, and which produced the
stated number of chunks/cm2 . The mean number of 235 u atoms per
chunk and the number of single atoms/cm2 are averages over all the
data bands counted to produce chunk data for each of the three ion
species (e.g., ten data bands for 40-120 keV H; were scanned).
See p.51

-86TABLE 6

He

Ar

Chunks counted

612

67

359

Ion beam fluence

1017

3 X 10 14

Chunks/cm2

20

12

90

4. 3 X 10 7

5.3 X 106

6.6 X 10 7

Single atoms/cm2

3.2 X 10 11

4.6 X 10 10

5.5 X 1010

"" · 3%

""· 1%

'V l0%

Chunk contribution to
sputtering yield

2 X 10

12

-87-

TABLE 7
Sputtering yields of 235 u under bombardment at normal incidence by

H~ ions with energy E, as calculated using equation (4-9).

£ and p

are the dimensionless energy and range respectively (see Section IVA).
a(£) values are from Weissman and Sigmund (1973).

The dimensionless

nuclear stopping cross-section s n (E) = (d£/dp) n values are from
Schi~tt

(1966).

ing energy.
See p.46

A value of U0 = 5.4 eV was used for the surface bind-

-88TABLE 7

H~ Sputtering of 235 u
E ( keV)

a(£)

sn( E: )

S(E)

.38

.06

2.8

.411

6.4 X 10- 2

10

. 76

.09

2. 35

.378

15

1. 14

.11

2.0

. 345

4.9 X 10-2
3. 8 X 10- 2

20

1. 52

. 13

1. 75

. 313

40

3.04

. 18

1.3

. 236

3.1 X 10 -2
1. 7 X 10 -2

60

4.56

. 22

1.0

. 193

1. 1 X 10-2

80

6.09

.25

0.8

. 165

100

7.62

. 29

0.7

. 144

7.3 X 10-3
5.9 X 10- 3

120

9.14

. 31

0.65

. 137

4.7 X 10- 3

E: = 7.62 x 10- 2 E (keV)

-89-

TABLE 8
Sputtering yields of 235 u under bombardment at normal incidence by
He+ ions with energy E, as calculated using equation (4-9).

£ and

p are the dimensionless energy and range respectively (see Section IVA).
a(E) values are from Weissman and Sigmund (1973).

The dimensionless

nuclear stopping cross-section s n (£) = (d£/dp) n values are from
Schi¢tt (1966).
ing energy.
See p.46

A value of U0 = 5.4 eV was used for the surface bind-

-90TABLE 8
He+ Sputtering of 235 u
E ( keV)

a(£)

sn(£)

S(E)

20

. 742

.358

2.35

.38

. 39

40

l. 48

.535

1.7

. 32

.23

60

2.23

.675

1.5

.27

. 18

80

2. 97

. 795

1.3

.24

. 14

100

3.71

.901

1.2

.22

.11

120

4.45

. 998

1.0

.20

.09

= 3.71 x 10-2 E (keV)

-91TABLE 9
Sputtering yields of 235 u under bombardment at normal incidence by
Ar+ ions with energy E, as calculated using equation (4-9).

£ and

p are the dimensionless energy and range respectively (see Section
IVA).

a( £) value of 0.5 is from Andersen (1974) p. 392.

The dimen-

sionless nuclear stopping cross-section sn( £) = (d£/dp)n values are
from Schi~tt (1966).
binding energy.
See p.46

A value of U0 = 5.4 eV was used for the surface

-92TABLE 9
Ar+ Sputtering of 235 u
E ( keV)

a(e:)

sn(e:)

S(E)

20

.064

.20

0.5

.34

5.2

40

. 129

. 32

II

. 39

5.9

60

. 193

.43

II

.40

6.2

80

.257

. 52

II

. 41

6.3

100

. 322

.61

II

.41

6.3

120

. 386

. 70

II

. 41

6.2

e: = 3.22 x 10 -3 E (keV)
p =

3.22 x 10 -3 s (~gm em-2)

-93TABLE 10

Cleaning procedures used to prepare materials for use in ultrahigh
vacuum (see Appendix B).
in the next table.

Recipes for the acid dip solutions are given

94TABLE 10
Cleaning Procedures for Parts to be Used in Ultrahigh Vacuum
Ceramics:

Ultrasonic rinse in clean methanol; hot air dry

Copper:

Vapor degrease in trike,* detergent wash, acid dip (50% HCl
at room temperature for 1-3 min}, water rinse, methanol
rinse, acetone rinse, warm air dry

Copper Wire: Vapor degrease in trike, detergent wash (Labtone}, acid
dip (50% HCl at room temperature for l-3 min), water, methanol,
and acetone rinses while brushing with SST brush, warm air dry
Gold:

Acid dip in aqua regia (3HC1 + HN0 3}

Aluminum:

Vapor degrease in trike, dip in 10% NaOH solution saturated
with common salt for 15-50 sec, rinse; i f discolored \~ash in
20-30% HN0 3 , rinse in running water 3-5 min, dip in 12% H2so 4 ,
HaOH solution again for 1 min, rinse in running water, methanol
rinse, warm air dry

Glass Parts: Vapor degrease in trike, potassium dichromate saturated
solution 35 cc in 1 liter concentrated H2so 4 , or Cr0 3 saturated
solution instead of potassium dichromate; n.b. slowly stir the
acid into the chromate or trioxide solution; use at ll0°C
(solution should be red), deionized water rinse, ultrasonic
methanol bath, warm air dry
Stainless Steel: Vapor degrease in trichloroethylene (10-15 min), rinse
in warm tap water, dip in stainless steel cleaning solution no.2
for 8 min, rinse in cold deionized water, dip in stainless
steel cleaning solution no.3 for 10-15 min, rinse in tap water,
rinse in deionized water, dip in methanol (ultrasonic cleaner),
hot air dry
Titanium:

Dip in titanium cleaning solution (30 sec, inspect, and repeat
until surface is smooth and clean), rinse in deionized water,
rinse in methanol, hot air dry

*trichloroethylene

-95TABLE 11

Acid dip solutions called for in table of cleaning procedures

-96TABLE 11
Stainless Steel Solution No. 2
% by volume

minimum

maximum

specific
gravity

HN0 3-nitric acid

19.0

18.0

20.0

1. 4078

HCl-hydrochloric acid

1.1

1.0

1.2

1. 16

HF-hydrofluoric acid

2.2

2.0

2.4

1.258

H20-deionized water

77.7

balance

Makeup:

Fill tank somewhat less than half full with deionized water,
add acids in above order, while stirring. Fill up with water.

Stainless Steel Solution No. 3
HN0 3-nitric acid

50.0

H20-deionized water

balance

Makeup:

46.0

50.0

1. 4078

Fi 11 tank somewhat less than half full with deionized water,
slowly pour in HN0 3 , while stirring. Fill up with water.

Titanium Cleaning Solution
HN0 3-nitric acid

25.0

23.0

30.0

1. 4078

HF-hydrofluoric acid

2.0

1.5

2.1

1.258

H20-deionized water

ba 1ance

-97-

FIGURE l
The sputtering assembly.(see p.ll)
is sketched in Fig. 2.
flange.

A cross-section of the apparatus

The assembly is mounted on an 8" 00 UHV

-98-

-99-

FIGURE 2
Cross-sectional view of the sputtering apparatus pictured in Fig. 1 .
(See p.9)
Bars bracketing edges of target plate ensure that it remains perpendicular to beam at all times .

-100-

0:::

_J_J
_j_

00

ULL

(.)

0:::1.!1

<(

CD

Wl'0 II
.....J.....Jo
- 0 ·0-r-o

lL ..L - -

_J
_J

-101FIGURE 3
The arrangement of the bands of sputtered uranium resulting from
moving catcher foils to different postions behind 1 em horizontal
slot in collimator cylinder. (see Figs. 1 & 2 and p.9)

Viewpoint

is that of target, so labels on the backs of catcher foils appear
backwards.

Labels are drawn as for runs in series 1.

Beam enters

between left and right foils.
Uranium sputtered during sputter-cleaning of targets is collected
on foils labelled GL & GR.

All dimensions in em.

-102-

-,~.or

IJU

5.1
l~U

00
('()

IJJ

I~J

_l
-.

IJC)

-103-

FIGURE 4
The manipulators for positioning the targets and catcher foils.
(seep.ll)

-104-

-lOS-

FIGURE 5
The vacuum side of the sputtering assembly mounting flange with the
assembly removed.

(see p. ll)

-106-

-107FIGURE 6
Target plate. (see p.ll)

-108-

-109-

FIGURE 7
Collimation cylinder added to assembly.

(see p.l2)

-110-

-111FIGURE 8

Complete sputtering assembly.

(see p.l2)

-112-

-113FIGURE 9
The ultra-high vacuum system.
sketched in Fig . 10.

(see p. l2)

Essential components are

The entire assembly may be adjusted vertically

and horizontally for purposes of alignment to ion beam .

-114-

-115-

FIGURE 10
Sketch of UHV system pictured in Fig. 9.

COLD TRAP

COV-500
PUMP

MS-075
ISOLATION
VALVE
I I

PRECISION
LINEARMOTION
FEEDTHRU

1 1

1 1

SPUTTERING
CHAMBER

PRECISION
MANIPULATOR

SP-11 SORPTION
PUMP

BVV-152
ISOLATION
VALVE

FARADAY
CUP a
VIEWPORT

GAUGE

0'1

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

-117FIGURE 11
The ion source and beam line.
are sketched inFig. 12.
of the photo at right.

(see p.15)

Essential components

The ultra-high vacuum system is just out

-11 8 -

-119FIGURE 12
Plan view of complete beam line pictured inFigs . 9 & 11.
to scale.

Not drawn

COLD
TRAP

STEERING
MAGNETS

QUADRUPOLE
FOCUSING MAGNET

STEERING
MAGNETS

DUOPLASMATRON
ION SOURCE
ACCELERATING
COLUMN

31° ANALYZING
MAGNET

SPUTTERING
CHAMBER

FARADAY CUP

_.

r·"

-121FIGURE 13
Fission-fragment tracks in mica after 15 minutes etch in 48% HF
(see p.26).

The photo was made in transmitted light at a magnification

of about 450X and enlarged by a factor of two.

The fission star is

fairly typical of the many observed in this experiment but is somewhat
smaller than average (see Section IIID).

-1 22 -

,, ----

\I, ,;,

--

/~

I _,.,

.,......

-123FIGURE 14

A larger-than-average fission star seen at lOOOX.
Section IIID).

(see p.26 and

-1 24-

-125FIGURE 15
Scanning electron micrograph of uranium foil target at 240X.
(see p.28)

-1 26-

-127FIGURE 16
SEM of uranium target at 2400X.

(see p.28)

This area of target

has not been sputtered, but was cleaned in HN0 3 .

-128-

-129-

FIGURE 17
Same area as Fig. 16 at 8400X.

-130-

-131FIGURE 18
SEM at 8400X of uranium target which has been sputtered by 3 x 1017
80 kev Ar+ ions per cm 2 . (see p.28)

-132-

-133-

FIGURE 19
The same area as Fig. 18 at 2400X.

-134-

~I

-135-

FIGURE 20
An area similar to the one in Fig, 19 at 2400X.

-136-

-137FIGURE 21

Another un-sputtered area of the target at 2400X.

(see p.29)

-138-

-139-

FIGURE 22
Scanning electron micrograph at 800X of uncleaned uranium target .
(see p. 29)

-140-

-141-

FIGURE 23
Angular distribution of 235 u atoms sputtered by 40 kev H~
at perpendicular incidence to target.

ions

(see p.30)

Nt= actual number of tracks counted per 130~ x 148~

field

of view i n the microscope .
N~= number of tracks per field of view

extrapolated to e= 0° .

e= sputtering angle rel ative to incident beam direction.
Error bars indicate uncertainty due to counting statistics only .
Angular distributions were found to be symmetric about the incident
beam direction.

-11!2-

00

Q)

en

<.D

c;;j

+I

Q)

c;;j

C\.J

<.D

r0

-143-

FIGURE 24
Angular distribution of 235 u atoms sputtered by 50 kev H~
(see p. 30)

Other information as in Fig. 23.

-144-

Q)

(/)

00

<.D

+I

Q,)

.::s:.

l{)

z-

1'0

l{)

-145FIGURE 25
Angular distribution of 235 u atoms sputtered by 40 keV H;. (see p.30)
Other information as in Fig.23.

-146-

CX)

(.!)

+N

><1>

CX)

(.!)

-147FIGURE 26
Angular distribution of 235u atoms sputtered by 120 keV He+.
(see p. 30)

Other information as in Fig. 23.

-14G-

00

Ct>

(/)

c.o

+Q)

Q)

..:%:.

C\J

C\J

c.o

r0

l{)

-149FIGURE 27
Sputtering yields S (in units of l0- 2 ) of 235 u under bombardment by protons with energies from 13 to 120 keV. (see Section IIIC and Tables 3&4).
Yields in upper two sets of points were from targets that had been
sputter-cleaned with an argon beam. (see p.l7)
Errors are discussed in Section IIIE.
The curve is S(E) calculated from the Sigmund theory (see Sections IVA
and IVB), but multiplied by 10-2 .

2~

20

41 t\ I
~ 31

5~

"'"

40

60
E (keV)

15 I

Ar pre-sputter 8 10

17

80

120

100

ps

ps

17

o No pre-sputter 8 10

t:t.

o Ar pre-sputter 8 .... 5x 10 p s

Hydrogen Sputtering

__.

(J'1

-151FIGURE 28
Sputtering yields S of 235 u atoms emitted per incident proton

bombardment by H1 , H2 and H3

(see

under

Section IIIC). Energies are per

proton.Targets had been sputter-cleaned in all cases except for the 60 keV

H~ data point, which was obtained by scaling up 60 keV point from
middle data set to upper set in Fig. 27.

-152-

...--.

Q)

-w

.::s:.

<..0

l()

+ - + C\J + !"()

r<>

(\j

~QI X ( W0~\1/SW0~\1) S

-153FIGURE 29
Sputtering yields S of 235 u under bombardment by 20 - 120 keV He+.
(see p. 33)
Curve is S(E) as calculated from
and !VB), but multiplied by 10-l .

Sigmund theory (see Sections IVA

-155FIGURE 30
Size distribution of 612 chunks emitted during proton sputtering of
pre-sputtered uranium targets.
tering runs with H~

Data was accumulated during five sput-

ion energies of 40-120 keV.

(see Table 5 and p.34)

The bars at the right show the number of chunks contain i ng 108 -2x10
235u and > 2 x 108 235 u atoms.
The two curves are visual fits to the histogram;
b= 10 3 cm- .

a= 2. 5 x

10 4 em -2 &

n is the actual number of tracks produced in mica by a given chunk
due to neutron irradiation.

-156-

------o

~----~~----~N

--------------~o

(j)(.C)

ooox
.::s:.

..Ole:

::J

1'-0

Q)

oa.
tO

C/)

o...-...'-----4 L{) <(
"+--

Q)

1.#-----t c;;:t ...0

::J

r-""'ll,..._._____--t oz
t0 ..

L{)

-------~---~--~----~--~o

CX)

tO

c;;:t

S}iUn4J JO .JaqwnN

-157FIGURE 31
The same data as Fig. 30 on an expanded scale,
fit to the data;

N = 180 and

A= 0.05.

The curve is a visual

(see Sections IIIE and IVC)

-158 -

1'-

----~----~~----~----~~----~----~0

00

<.0

s~un4:>

:J.O JaqwnN

C\J

-159FIGURE 32

Size distribution of 359 chunks emitted during 80 keV Ar+ sputtering
of a pre-sputtered uranium target.

(see Table 5 and p. 35)

The bars at the right show the number of chunks containing 108 - 2 x 108
235u atoms and > 2 x 108 235u atoms.
The curve is a visual fit to the histogram;

N0 = 60

and

A= 0,03.

(See Section IVC)
n is the actual number of tracks produced in mica by a given chunk
due to neutron irradiation.

-160-

"\.oI

oox

::)

r--6

'-

(])

(.!)

a.
(/)

oLO<(
'+--

~_a

::)

oz
f'() ••

LO

LO

r0

S>tUnliJ JO JaqwnN