GORrsi05.pdf

Low-energy ion beamline scattering apparatus for surface science

investigations

M. J. Gordon and K. P. Giapis

Division of Chemistry and Chemical Engineering, California Institute of Technology,

Pasadena, California 91125

Received 9 May 2005; accepted 14 June 2005; published online 28 July 2005

We report on the design, construction, and performance of a high current

monolayers/s

mass-filtered ion beamline system for surface scattering studies using inert and reactive species at

collision energies below 1500 eV. The system combines a high-density inductively coupled plasma

ion source, high-voltage floating beam transport line with magnet mass-filter and neutral stripping,

decelerator, and broad based detection capabilities

ions and neutrals in both mass and energy

for

products leaving the target surface. The entire system was designed from the ground up to be a

robust platform to study ion-surface interactions from a more global perspective, i.e., high fluxes

100

A/cm

of a single ion species at low, tunable energy

50–1400±5 eV full width half

maximum

can be delivered to a grounded target under ultrahigh vacuum conditions. The high

current at low energy problem is solved using an accel-decel transport scheme where ions are

created at the desired collision energy in the plasma source, extracted and accelerated to high

transport energy

20 keV to fight space charge repulsion

, and then decelerated back down to their

original creation potential right before impacting the grounded target. Scattered species and those

originating from the surface are directly analyzed in energy and mass using a triply pumped, hybrid

detector composed of an electron impact ionizer, hemispherical electrostatic sector, and rf/dc

quadrupole in series. With such a system, the collision kinematics, charge exchange, and chemistry

occurring on the target surface can be separated by fully analyzing the scattered product flux. Key

design aspects of the plasma source, beamline, and detection system are emphasized here to

highlight how to work around physical limitations associated with high beam flux at low energy,

pumping requirements, beam focusing, and scattered product analysis. Operational details of the

beamline are discussed from the perspective of available beam current, mass resolution, projectile

energy spread, and energy tunability. As well, performance of the overall system is demonstrated

through three proof-of-concept examples:

elastic binary collisions at low energy,

core-level

charge exchange reactions involving

with Mg/Al/Si/P targets, and

reactive scattering of

/CF

off Si. These studies clearly demonstrate why low, tunable incident energy, as well as mass

and energy filtering of products leaving the target surface is advantageous and often essential for

studies of inelastic energy losses, hard-collision charge exchange, and chemical reactions that occur

during ion-surface scattering.

DOI: 10.1063/1.1994987

I. INTRODUCTION

Ion-surface interactions are important in a variety of

fields such as plasma physics, surface analysis, materials

growth

plasma-enhanced chemical vapor deposition and

ion-beam assisted deposition

, and semiconductor dry etch-

ing. In many of these application areas, collision energies are

specifically kept below 1 keV to have preferential energy

deposition at the target surface only, i.e., short collision cas-

cades to provide high selectivity to the topmost atomic layers

without damaging the underlying material. This regime of

energies

1 keV

represents an important window in the

energy space of ion-surface interactions because many

threshold processes occur at low energy

sputtering, core

shell electronic excitation, momentum-assisted desorption,

secondary electron release, etc.

1,2

In addition, the kinetic

energy of the incoming ion can even “activate” surface-

specific processes, allowing epitaxy at lower temperature,

stimulation of surface reactions/adatom diffusion, synthesis

of metastable material phases,

and control of film density,

texturing, and stress

for a review, see Ref. 1

. However, the

research literature on low energy ion-surface collisions is

relatively sparse because of the experimental challenges as-

sociated with providing sufficient ion beam current at low

impact energy for scattering, material growth, and etching

studies. We have set out to probe this little-studied energy

range by developing an ion beamline scattering system to

investigate a wide variety of ion-surface interaction phenom-

ena below 1 keV. This article describes the design and con-

struction of our system and its application to three proof-of-

Author to whom correspondence should be addressed; electronic mail:

giapis@cheme.caltech.edu

REVIEW OF SCIENTIFIC INSTRUMENTS

, 083302

2005

0034-6748/2005/76

/083302/15/$22.50

, 083302-1

Downloaded 14 Dec 2005 to 131.215.225.9. Redistribution subject to AIP license or copyright, see http://rsi.aip.org/rsi/copyright.jsp

concept examples: elastic binary collisions at low energy,

core-shell charge exchange collisions involving Ne

, and re-

active scattering of CF

/CF

off Si.

Our design philosophy has been to take an inductively

coupled plasma

ICP

source and couple it to a high-voltage

ion beam transport line with magnetic mass filtering to pro-

vide a clean ion beam surface probe with high current

100

A/cm

and tunable energy

50 eV–1 keV

Space charge repulsion between the ions, which usually pre-

cludes high current at low energy, is circumvented using the

accel-decel scheme for transport. In this arrangement, ions

are created at the desired collision energy in the plasma

source, extracted and accelerated to high transport energy

fight space charge repulsion

, and then decelerated back

down to their original creation potential right before impact-

ing the grounded target. In this way, the beam current is high,

the collision energy is easily tunable

by floating the plasma

source above ground

, and the target is always kept

grounded. The ICP-based beamline is a generic and robust

system because any ion created in the plasma can be indi-

vidually singled out and delivered to the target as a clean

surface probe composed of only one species and one charge

state. Species scattered by or originating from the target sur-

face are analyzed with a hybrid detector which allows simul-

taneous mass and energy filtering—so the kinematics, charge

exchange, and chemistry which occurs upon scattering can

be investigated.

II. DESCRIPTION OF THE INSTRUMENT

The low-energy, ion beamline scattering system is

shown schematically in Fig. 1, with technical specifications

given in Table I. The system is composed of four main sec-

tions:

ICP ion source with extraction optics;

high

voltage beamline with magnetic mass filter, focusing ele-

ments, and deceleration optics;

scattering chamber; and

scattered product detector. Specific construction details of

each section of the instrument can be found elsewhere.

The

design philosophy of the beamline was to provide high fluxes

100

A/cm

of a single ion species at low impact energy

50–1400±5 eV full width half maximum

FWHM

onto a

grounded target to facilitate ultrahigh vacuum

UHV

scat-

tering studies using both inert and reactive beams. Space

charge repulsion is controlled by transporting the ions at high

voltage, with mass filtering, beam steer, and quality adjust-

ments all occurring on the accelerated beam.

The particle flux from the target is analyzed with a

triple-differentially pumped hybrid detector with both energy

and mass dispersion. This system combines an electron im-

pact ionizer, hemispherical electrostatic sector, and quadru-

pole mass filter in series with single ion counting on the rear

end so that small signals of both ions and neutrals can be

analyzed. With such a scheme, the collision kinematics,

charge exchange, and chemistry occurring on the target sur-

face can be separated by fully analyzing the scattered prod-

uct flux. In the next few sections, key design aspects of the

plasma source, beamline, and scattering system are empha-

sized to highlight why low-energy ion beam experiments are

so demanding and how to work around physical limitations

associated with high beam flux at low energy, pumping re-

quirements, beam focusing, and scattered product detection.

A. Design philosophy

As inferred in the introduction, a high-current ion beam

cannot be exposed to decelerating fields during transport,

otherwise it will diverge. Unfortunately, space charge repul-

sion between the ions can only be “controlled” using

high transport energy

Langmuir–Child law

continual

refocusing,

or by artificially neutralizing the beam by add-

ing electrons externally

see Lawson

. The latter two op-

tions are usually reserved for very high current beams

used in particle accelerators. On the laboratory scale, con-

tinual refocusing is cumbersome and usually limited to

highly symmetric systems

without a mass filter

; as well,

artificial neutralization is not an option because a neutralized

beam cannot be mass filtered. The solution is therefore

simple—just transport and mass filter the beam at high ac-

FIG. 1. Schematic of the low-energy ion beamline scattering apparatus showing various sections:

ICP plasma source and extraction optics,

high-voltage

floating region including both steering magnets and decelerator,

scattering chamber, and

scattered product detector.

083302-2 M. J. Gordon and K. P. Giapis

Rev. Sci. Instrum.

, 083302

2005

Downloaded 14 Dec 2005 to 131.215.225.9. Redistribution subject to AIP license or copyright, see http://rsi.aip.org/rsi/copyright.jsp

celeration voltage.

8–11

What one desires is a symmetric

usu-

ally cylindrical

, pumpable, high-voltage “shield” around the

ion beam column throughout the beam transport line

includ-

ing the mass filter

while working with a grounded target.

This situation is experimentally demanding because it re-

quires a pumpable high-voltage envelope around the propa-

gating ion beam to shield it from the grounded beamline

chambers which maintain the vacuum. Such an approach has

been used successfully for a 10 keV beamline by construct-

ing an electrostatic mesh chamber inside a grounded vacuum

shell.

We have taken a different approach whereby the entire

vacuum system itself

chambers, valves, feedthrus, etc.

floating below ground at the beam transport energy

−20 keV

. This philosophy avoids all the problems associ-

ated with high-voltage isolation inside vacuum where space

is tight and pumping conductance must be maximized. All

the isolation problems are brought outside of vacuum and we

can avoid making a chamber within a chamber.

The high-voltage floating section of our beamline sys-

tem, shown in Fig. 1, extends from the end of the premagnet

Einzel triplet up to and including the decelerator where it

joins to the grounded scattering chamber. The floating sec-

tion includes the 60° sector magnet, mass slit, quadrupole

doublet, 10° steering magnet, and decelerator. Electrical

breaks are used at both ends to isolate the floating section

from the rest of the system with the premagnet Einzel triplet

lens number 3

and decelerator

lens numbers 2 and 3

func-

tioning as transfer lenses to shield the beam between sec-

tions. The vacuum chambers themselves are supported using

a poly-vinyl chloride scaffold system to stand off the floating

potential and the vacuum pumps

CT7 cryopump after the

60° sector, 200 L/s turbo on the quadrupole doublet

are also

isolated with electrical breaks.

B. ICP plasma source and ion beam extraction

One of the most important choices in any beamline de-

sign is the ion source itself. The source should have high

conversion efficiency, low operating pressure

to avoid gas

loading the beamline

, well-defined ion creation potential,

narrow ion energy spread, and compatibility with both inert

and reactive working gases. Traditionally, Freeman-type,

low-voltage dc arcs have been the ion source of choice be-

cause of their high current and low pressure operation.

3,12,13

However, arc sources require frequent filament changes

few tens of hours for high beam currents

and contaminate

the beam with useless ions resulting from reaction of the

working gas with the hot filament used to sustain the dis-

charge. We have solved these problems by designing an ICP

source and extraction system for our beamline which can

satisfy all the above criteria along with clean, sustained

operation—even with corrosive working gases. The induc-

tive coupling scheme is ideal because it can provide high ion

density

–10

/cm

with low plasma potential

20 eV

and narrow ion energy spread

2–5 eV

, all at low

working pressure

0.1–5 mTorr

for a summary of ICPs, see

Refs. 5 and 14

A schematic of the ICP plasma reactor and extraction

electrode system is shown in Fig. 2. The plasma reactor is

constructed from a 4 in

.o.d.

4 in. alumina tube

a Pyrex

version is also used

with alumina end caps sealed by o-rings

to an 8 in. Conflat flange fitted to the front of the beamline.

The plasma is excited by a two-turn, rf-solenoid antenna ex-

cited through a

network at 13.56 MHz. A grounded, cop-

per Faraday shield

similar to Ref. 15

is situated between

the antenna and reactor to eliminate all capacitive coupling

between the plasma and rf antenna. It is imperative that ca-

TABLE I. Design and operational specifications for the low-energy ion beamline scattering system.

Ion source

Inductively coupled plasma, 2-turn solenoid, 1 kW @ 13.56 MHz

Source pressure

0.5–10 mTorr

Beamline acceleration voltage

10–20 keV

Magnetic mass filter

60°, 300 mm radius of curvature, 50 mm air gap

0.8 T max field, water cooled, 300 V–16 A supply

Field stability: Hall feedback, ±3 G

Ion deflector magnet

fast neutral removal

10°, 1300 mm radius of curvature, 50 mm gap

1800 G max field, 600 V–2 A supply

Extractor

four elements

1 floating extraction electrode+3 biased separately

Beam shaping

four element asymmetric Einzel

before 60° magnet

-axis divergence corrector

after 60° magnet

dc quadrupole doublet

eight elements — after mass defining slit

Decelerator

Asymmetric Einzel triplet+steering quadrupole+guard electrode

Beam energy

20–1500 eV±5 eV FWHM

maximum

measured @ target with electrostatic sector

Beam size

1–3 mm diameter, tunable

Beam current

250

A/cm

@50eV,150

A/cm

@50eV

Scattering chamber pressure

−8

Torr

beam on @ 500

A/cm

, plasma source @ 5 mTorr

Scattered product detector

Electron impact ionizer—10 mA emission

90° hemispherical sector filter with Herzog plates, 58 mm radius

rf/dc quadrupole: 19 mm diam poles @ 2.22 MHz, 300 W

Ion counter

Daly-type conversion dynode @ -15–30 keV

Fast organic scintillator+PMT tube

in vacuum

Other facilities

Sputter cleaning, anneal to 900 °C, Ar

SIMS

scattering chamber

XPS

in-vacuum transfer

083302-3 Low-energy ion beam scattering apparatus

Rev. Sci. Instrum.

, 083302

2005

Downloaded 14 Dec 2005 to 131.215.225.9. Redistribution subject to AIP license or copyright, see http://rsi.aip.org/rsi/copyright.jsp

pacitive coupling be minimized to keep the plasma potential

low

20 V

and ion energy distribution narrow.

The en-

tire plasma volume can be floated above ground by an inter-

nal molybdenum bias electrode

Ti and Ta were also used,

depending on the plasma chemistry

connected to a dc sup-

ply which sets the final beam energy. Operating pressures are

typically 1–5 mTorr for input powers ranging from 300 to

800 W. In addition, highly electronegative gases

were frequently blended with small amounts of Ar to provide

sufficient electrons for good inductive coupling and power

transfer from antenna to plasma. Mixing gases in the source

is not a great concern because the beam is mass filtered

downstream.

The ion beam

plasma beam

is extracted from the ICP

source through a floating extraction aperture

5 mm sharp-

edged hole, 120° bevel

and triode accel-decel lens system

for example, see Refs. 16 and 17

. Smooth field variation in

the vicinity of the “virtual” sheath of the plasma is necessary

to form a uniform cross section ion beam column.

In close

proximity are three cylindrically symmetric electrostatic

lenses in series

puller, buncher, and front beamline accel

electrodes

with tantalum cover plates to prevent excessive

sputtering. The final design for the extractor electrode system

was the result of both space charge simulations

SIMION

well as extensive operational testing. The front beamline

electrode is the main acceleration step and is joined to the

premagnet Einzel triplet

see next section

through stainless

steel mesh tubing. Since the gas load to the beamline is of

paramount importance, the extractor was design to be as

“open” as possible to maximize pumping, so that background

gas collisions are minimized. A 1000 L/s drag turbo directly

below the extractor functions as the main differential pump-

ing step, keeping the extraction chamber at 5

−6

Torr

with the plasma source running at 5 mTorr.

C. High voltage beamline

After the ion beam is extracted from the plasma and

accelerated to

20 keV by the front beamline electrode, the

beam waist is adjusted by an Einzel triplet to coincide with

the source point of the magnetic sector field. In this way,

angular divergence of the beam at the source point is auto-

matically removed by the sector field when the pass mass is

refocused after the magnet at the field image point

the mass

slit

17,19

The 60° sector magnet has 8 in. diam poles, hex-

agonal shoes, 50 mm air gap to allow clearance for the beam-

line flight tube

1.5 in. o.d., floating at −20 keV, insulated by

mica sheets

, and adjustable mass slit

0–13 mm wide

18 mm high

downstream. Since the air gap was not small

compared to the shoe height

and pole size

, a significant

fringe field extended beyond the physical pole boundaries.

As such, it was necessary to account for the “virtual” pole

boundaries using an extended fringe field method

20,21

to cal-

culate the effective radius of curvature

11 in. instead of 8

in., as given by the pole diameter

. In addition, it was abso-

lutely necessary to actively control the magnet at

10 Hz

with Hall probe feedback to keep the ion beam from drifting

on the target. With this system, the field could be held within

±3 G with the magnet at maximum strength

0.8 T

One should note that a magnetic sector with parallel pole

shoes does not produce a stigmatic image of the source

point.

Such an effect manifests as an ion beam which enters

the magnet having circular cross section, but leaves the field

elongated in the nondispersive

direction as an ellipse

with less on-axis brightness.

Stigmatic focusing with a sec-

tor usually requires tilting the pole shoes to create a slightly

inhomogeneous field in the radial direction to offset the ra-

dius of curvature for different flight paths. The lack of a

stigmatic image was not deemed so critical in our design;

however, the loss of current due to divergence in the

di-

rection was considered important. Therefore, a weak electro-

static “bunching” field in the

direction

+500–1000 V off

the beamline floating potential

after the magnet was added

to recover the otherwise lost beam current rather than tilting

the pole shoes.

When the ion beam passes through the mass slit, it ex-

hibits a crossover point in the dispersive direction

, but an

elliptic cross section in the nondispersive direction

. Un-

fortunately, the sector removes the inherent cylindrical sym-

metry of the ion beam as extracted from the plasma because

of the field “handedness.” Since it is desirable to decelerate a

symmetric, parallel beam having circular cross section and

small waist, we felt that an additional focusing step was nec-

essary after the mass slit. A dc quadrupole doublet was used

in the drift space between the mass slit and decelerator to

affect the symmetry correction. A quadrupole field was cho-

sen because its planar symmetry seemed better suited to fix

unequal divergences of the beam in the

and

axes. In

addition, a quadrupole field provides stronger focusing ac-

tion and is shorter overall than an axially symmetric lens

system.

The doublet is operated with

correction first, fol-

lowed by

first quad - ±

plates variable, second quad - ±

plates variable

with the plate voltages derived from power

supplies floating on the beamline acceleration potential.

It is well known that collisions of energetic ions with

background gas atoms can generate fast neutrals through

charge exchange processes.

22,23

In fact, fast neutral beams

are produced using this method by sending a fast ion beam

through a charge exchange cell containing background gas.

The process of fast neutral generation, although much less

significant in our system, could possibly influence scattering

FIG. 2. Design schematic of the ICP plasma source and ion beam extraction

electrodes. Both the reactor and electrode set have cylindrical symmetry.

The extraction aperture where the virtual plasma sheath forms is electrically

floating with the plasma itself. Also, two versions of the plasma reactor

pyrex and alumina

were used.

083302-4 M. J. Gordon and K. P. Giapis

Rev. Sci. Instrum.

, 083302

2005

Downloaded 14 Dec 2005 to 131.215.225.9. Redistribution subject to AIP license or copyright, see http://rsi.aip.org/rsi/copyright.jsp

results. Since it is extremely difficult to measure fast neutrals

quantitatively, we have therefore taken the approach to rid

the beam of fast neutrals on purpose, even if they may not

exist to a significant degree in our system. This is simply

done by deflecting the ion beam 10° with a small magnetic

sector field right before the decelerator entrance so that any

fast neutrals in the upstream beam are not within line of sight

of the target. The short flight distance through the decelerator

to the sample is unimportant because this region is held at

high vacuum conditions

−8

Torr

. The small sector

has square pole shoes, 2 in. air gap, and gives a 10° deflec-

tion to the ion beam with

1300 mm turning radius. Since

the deflection is small and the curvature radius extremely

large, the beam symmetry is not disturbed to any great

extent.

D. Deceleration

Deceleration lens schemes for high-energy ion beams

have traditionally been approached from two very different

points of view. Experiments using ion implanters

50 keV

transport energy, mA currents

have demonstrated that ul-

trashort, extremely strong slowing fields

50 kV in 10–20

using two or three thin electrostatic lenses provide the

best spots.

13,24,25

On the other hand,

A beam currents at

lower transport energies are more successfully dealt with

using longer, more complex schemes

3–6 tube-like

electrodes

26–29

Our decel system is rather simple

Fig. 3

with an asymmetric Einzel triplet

short-long-short

followed

by a large inner diameter, short quadrupole with a grounded

end cap to shield the target from any high-voltage fields. The

Number 2 and 3 lenses of the triplet are run more negative

than lens Number 1 to pinch the beam after a first stage of

slowing occurs between the main beamline and decel Num-

ber 1. It is thought that the potential drop of a few kV in this

region

between decel Numbers 1 and 2

aids in neutralizing

the beam space charge for the rest of the deceleration be-

cause slow electrons are trapped in the beam channel.

The

end of decel Number 3 and the entire quad extend into the

grounded scattering chamber. The quadrupole exit setup for

final beam steer is short and stubby so the field asymmetries

near its plate electrodes are so far away from the bunched

beam that they are irrelevant. Operation of the quad is very

weak with only 200–400 V of asymmetric steer capability on

a centerline floating potential of −6 to −10 kV. A floating

circular shield encloses the quad to screen the beam from the

grounded walls of the scattering chamber. The quad shield,

quad centerline float, and the four steering plates are all in-

dependently adjustable. Finally, a fully grounded end cap

over the quad exit shields the target from any high-voltage

fields.

E. Beamline bias scheme

Due to the “floating beamline” design principle, it was

deemed easier to adopt a floating power supply bias scheme

for the focusing elements on the HV section. In this design,

smaller, more accurate supplies float on the main beamline

bias voltage to provide fine beam steering control. A sche-

matic of this system is shown in Fig. 4. In addition, all beam-

line power supplies

except the plasma bias

float on top of a

separate dc supply that is ramped up identically with the

plasma bias voltage. In this way, exactly the same extraction

and focus conditions with respect to earth occur for all beam

energies. It is necessary to separate these two voltages be-

cause the 13.56 MHz noise from the plasma electron oscil-

lation causes havoc in the regulator circuits for the rest of the

beamline power supplies. Incidentally, this oscillation is

too high in frequency to affect the inherent ion beam

self-neutralization.

Also, the vertical bunching plates and

quadrupole doublet float on the beamline high voltage

−15–20 kV

while the steering quad supplies and quad

shield float atop the steering quad centerline potential

−6–10 kV

F. Scattering chamber

The UHV scattering chamber was custom built in-house

to have two pumping stages mounted inside the overall

chamber to differentially pump the scattered product beam

leaving the target. Each stage ha

sa2mm

diam skimmer and

its own dedicated turbopump. Also, a capacitor deflector is

situated between the two stages so charged species can be

removed from the scattered product signal if desired. The

target sample rests on a five-axis goniometer with heating to

900 °C and scattering experiments are usually carried out at

a 90° lab angle in specular reflection. The scattering chamber

FIG. 3. Design schematic of the ion beam decelerator, high-voltage beam-

line electrical break, and target region.

FIG. 4. Simplified schematic of the power supply bias scheme for the beam-

line. All acceleration, focusing, and steering lens supplies float atop the main

beamline bias

20–1500 eV

which controls the final impact energy. A sepa-

rate plasma bias supply is ramped up in direct correspondence with the

beamline float bias.

083302-5 Low-energy ion beam scattering apparatus

Rev. Sci. Instrum.

, 083302

2005

Downloaded 14 Dec 2005 to 131.215.225.9. Redistribution subject to AIP license or copyright, see http://rsi.aip.org/rsi/copyright.jsp

itself is pumped by a CT8 cryopump and 1300 L/s maglev

turbopump to maintain the pressure in the 10

−9

Torr range

during bombardment at maximum beam current.

For specular reflection and 90° lab angle, the target is

inclined 45° from the beam propagation axis. With low en-

ergy beams, it is well documented that an inclined target can

have significant impact on the actual versus “desired” scat-

tering angle as the incident beam energy is varied.

To avoid

this problem, a rotatable, tantalum beam flag with 2 mm

diam hole is place directly in front of the target as a shield,

perpendicular to the beam propagation axis. In this way, the

beam sees a totally symmetric decelerating field, rather than

the inclined target. Thus, the target angle does not modify the

overall lab scattering angle for different incident beam ener-

gies

see Sec. III for verification of this fact

. In addition, the

current on both the beam flag and target can be monitored

separately to tune the ion beam steering and focus.

Finally, the scattering chamber is outfit with a miniature

180° hemispherical sector energy analyzer that can be moved

into and out of the beam path directly at the target position to

measure the incident beam energy distribution for every in-

cident beam condition. The analyzer is

23 mm radius with

Herzog corrector plates, inlet Einzel triplet, and channeltron

detector.

G. Scattered product detector

The most idealized scattering experiment would involve

the detection of all species leaving the target surface, both

ions and neutrals, within the space of mass and energy. This

scheme would allow the kinematics, charge exchange, and

chemistry of scattering phenomena to be separated. Toward

this goal, our detection system combines an electron impact

ionizer, electrostatic sector energy filter, and high transmit-

tance rf/dc quadrupole mass filter with an extremely sensi-

tive ion counting system. Figure 5

shows a simplified

schematic of the detector system. We believe the electrostatic

sector with sequential quad provides several advantages over

traditional time-of-flight

TOF

techniques:

the sector can

be placed very close to the sample surface;

energy analy-

sis does not require beam chopping;

energy and mass

dispersion are separated inherently; and

it is easy to

implement an electron impact ionization scheme to detect

low energy neutrals

TOF requires

neutral

1–3 keV for di-

rect detection with a microchannel plate

The design of the scattered product detector was based

on the ionizer and pole set

19 mm diameter

from an Extrel

QPS system, with the electrostatic energy filter, transfer op-

tics, and ion counting system designed and fabricated in-

house. The ionizer has a Whetstone bridge filament structure

with axial entrance and ion extraction, along with five lens

elements after the grid to enable different extraction modes.

Specifically, the ionization volume size and ion creation po-

tential can be easily changed by tailoring the grid, extractor,

and focus plate voltage ratios. This is a very important point

because any variation in ion creation potential in the ionizing

volume

most ionizers have this problem

causes an artificial

linewidth in the energy spectrum. We made specific intent to

minimize this effect by fully characterizing our ionizer sys-

tem. Experiments with the electrostatic sector show that the

ionizer can be sufficiently “tuned” to provide ionization at a

uniform ion creation potential of narrow energy width

1eV

All lens elements in the ionizer are fully tunable

and emission currents up to 10 mA at 100 eV electron energy

are possible. The ionizer itself is positioned directly behind

the second skimming aperture

2mm

on the back of the first

differential pumping stage with the ionizing volume situated

only 80 mm from the sample surface. An Einzel triplet lens

system was added to the rear of the ionizer to provide better

extraction and imaging of ions on the inlet slit of the elec-

trostatic sector. The triplet lens system floats off the electro-

static sector retard voltage ramp to provide a nearly constant

focusing power over all ion energies.

SIMION

was used to

design the lens system and operating voltages to provide

uniform transmittance of all ion energies irrespective of re-

tardation level.

For the 90° electrostatic sector, spherical electrodes

gold-plated oxygen-free high-conductivity copper

with 50

and 65 mm radii were used, giving a mean sector radius of

57.5 mm

2.25 in.

. Herzog corrector plates were posi-

tioned at the inlet and exit of the sector to appropriately

terminate the deflection field, and a four-element Einzel-type

transfer lens assembly is used to transport and tightly focus

FIG. 5.

Simplified schematic of the scattered product detector showing

differential pumping stages, ionizer, energy and mass filter, and Daly-type

ion counting detector.

Example ion trajectory calculation for 50 eV Ar

at 15 eV pass energy showing various lenses in the scattered product detec-

tor

Rp=repeller, G=grid, Ex=ion extractor, L1.

.3=ionizer focus, E1. .3

=Einzel transfer, Hz=Herzog plates, Ho, Hi=outer and inner hemispheres,

T1. .4=quad transfer, and Q=quad centerline float

. All lenses float off the

sector retard voltage used to scan the kinetic energy pass band.

083302-6 M. J. Gordon and K. P. Giapis

Rev. Sci. Instrum.

, 083302

2005

Downloaded 14 Dec 2005 to 131.215.225.9. Redistribution subject to AIP license or copyright, see http://rsi.aip.org/rsi/copyright.jsp

exiting ions onto the entrance aperture of the rf/dc quad. The

energy filtering action of the sector is accomplished by slow-

ing down or speeding up ions which enter the sector to the

pass energy of the analyzer

constant acceptance energy

CAE

mode

by sweeping the retarding voltage from

−

pass

kinetic energy

KE=0

up to some maximum scanned

energy. A pass energy of 15 eV was used for most of our

scattering experiments because it provided good energy reso-

lution and adequate transport energy through the quadrupole.

Both the inlet triplet and four-lens exit system on the sector

analyzer were designed through extensive

SIMION

simulation.

An example ion trajectory calculation for a 50 eV Ar

beam

retarded to the 15 eV pass energy of the analyzer is shown in

Fig. 5

. Very specific voltage ratios between the inlet and

exit lens elements were required for proper focus over a wide

range of ion energies. As such, a custom, computer con-

trolled power supply system was built for the sector analyzer,

inlet, and exit Einzel lens systems.

For the quadrupole, the dc rod voltages were derived

directly from the rf drive to the poles by vacuum tube diodes.

This unique feature allowed the entire rod set and match box

circuitry to be floated above ground. The ability to artificially

move the quad centerline potential above ground was abso-

lutely necessary because the ion flight energy within the

quad must ramp with the energy analyzer retard voltage—

otherwise, the flight energy through the quad is not constant.

The centerline quad potential was attached directly to the

main retarding voltage ramp of the energy filter. In this way,

ions fly through the quad at exactly the pass energy of the

energy filter, irrespective of what kinetic energy is being

scanned. Finally, to further decrease the residual gas back-

ground and prevent ion forming collisions within the quad, a

special cryocooled shroud at 30 K was placed around the

quad pole set

not shown

The extremely small ion signals

−15

–10

−17

Torr effec-

tive pressure

that are generated by electron impact ioniza-

tion of secondary neutrals leaving the target surface require

an extremely sensitive, single-ion type counting system. The

problem is unavoidable because ionization by electron im-

pact is quite inefficient. The most well-designed magneti-

cally confined ionizers running in space charge limited mode

at 10 mA of emission current can only provide a conversion

efficiency of maybe one part in 10

This dictates that the

ion detector must be able to count every single ion that is

generated. For this purpose, we developed a hybrid ion

counting system based on the approach of Daly.

Ions exit-

ing the quad are “bunched” by a three element immersion

lens

+200 V,−3 kV,−6 kV

at the rear of the quad and they

are subsequently accelerated to −15 to −30 kV by a conver-

sion dynode. When the high energy ions strike the aluminum

dynode surface

45°, angle of maximum secondary elec-

tron emission

, a shower of secondary electrons

15–30 keV

with respect to ground

is created. These secondary elec-

trons, in turn, are accelerated away from the dynode and

converted to light

max

=408 nm

by a plastic scintillator

3–4 photons/

−

for Bicron BC408

. Finally, the resulting

photon pulses are registered by a special bi-alkali photocath-

ode photomultiplier tube

PMT

with maximum sensitivity at

400–420 nm. The scintillator was metallized on the dynode

side by sputter coating with 1500 Å of aluminum to provide

a grounded surface and light shield for the PMT. The PMT

tube itself was housed inside vacuum with a

-metal mag-

netic shield and pressed firmly against the rear of the scintil-

lator. The tube is run in pulse count mode with the photo-

cathode at positive high voltage to further reduce spurious

electron pulses.

III. BEAMLINE PERFORMANCE

The major goal of our endeavor was to construct a mass-

filtered ion beam system which could deliver reasonable flu-

ence

100s of

A/cm

at low beam energy

50–1000 eV

for surface scattering studies. The real test of such a system

indeed lies in the mass-filtering capability, available current,

and scattered product detector performance. As such, we

sought to answer several questions, which are addressed in

the following sections:

Can isotopically pure beams be generated from a com-

plex plasma gas mixture

i.e., one commonly used for

dry etching

What is the current delivery versus beam energy?

Is the ion beam energy distribution at the target suffi-

ciently narrow for scattering experiments?

Can the impact energy be tuned easily by floating the

entire ICP plasma source above ground?

A. Mass separation

The mass resolution of the ion beamline system was

evaluated by forming isotopically clean ion beams from a

complex plasma gas mixture that is typically used for dry

etching of SiO

. This situation represents an extreme case of

beam contamination and can be used to test the ability of an

ion beam system to give high mass resolution and produce

clean, pure beams. Figure 6 shows a typical mass sweep of

the beams that can be formed by extracting all the ions from

aCF

/Ar/O

plasma mixture running at 500 W plasma

power. The plot was formed by sweeping the field strength of

both the 60° sector and 10° deflector magnets simultaneously

via computer control while maintaining the beam energy at

100 eV and beamline transport voltage at −12 keV. The mass

exit slit on the 60° magnet was set t

oa3mm

width and the

beam current at the target location was measured with a Far-

aday cup

500

m diam inlet aperture

. It can be seen that all

the ions in the plasma source are easily resolved by the sec-

tor magnet and can be transported to the sample as isotopi-

cally pure beams for scattering experiments. Higher resolu-

tion is possible with smaller exit slit size, but better mass

separation is not needed for most of the beams shown. Mass

sweeps of this type demonstrate the power of the beamline

system for plasma diagnostics as well as a direct feedback

mechanism for tuning plasma operation to obtain the highest

yield of the ion species of interest.

B. Target current

One of the most important design criteria for the beam-

line was high beam current at low impact energy. This is

made possible by the accel-decel scheme where ions are cre-

083302-7 Low-energy ion beam scattering apparatus

Rev. Sci. Instrum.

, 083302

2005

Downloaded 14 Dec 2005 to 131.215.225.9. Redistribution subject to AIP license or copyright, see http://rsi.aip.org/rsi/copyright.jsp

ated in the plasma at the impact energy, accelerated to high

transport energy, and then decelerated near the target back to

their initial creation potential. The success of such a scheme

depends on having high ion density in the plasma, efficient

extraction, minimal current loss during transport, and proper

focus correction during deceleration. However, the entire

system cannot deliver high current at low energy unless

all of the beamline components function together in unison

through proper beam transport between each stage. Thus,

overall system performance can be evaluated directly

through the available beam current. Typical data for our sys-

tem are shown in Fig. 7. The ion arrival rate is represented in

terms of beam current density on the target delivered through

the 2 mm diam flag aperture

see discussion in Sec. II F

Current density at the target is more useful for direct com-

parison with an arrival rate of one monolayer per second

1ML/s

320

A/cm

for Si

. Figure 7

shows an Ar

beam extracted from a pure Ar plasma discharge running at

500 W and 5 mTorr. Atomic and molecular oxygen beams

extracted from an Ar/O

plasma are also given in panel

Argon was added to the oxygen discharge to increase induc-

tive power coupling with the rf antenna using a mixing ratio

of Ar:O

of 1.3:1 and 5 mTorr operating pressure. The beam

currents on these plots were measured for identical extraction

conditions

puller and buncher voltages held constant

over

all beam energies with the main beamline transport energy at

13 keV. As one might expect, the beam current increases

with increasing impact energy until a saturation condition is

reached. The saturation current depends strongly on the

plasma operating conditions as well as the field strength in

the beam extraction region. When the beam energy is low-

ered below 100–150 eV, the beam current begins to fall off

because of excessive space charge spreading during extrac-

tion from the plasma and during the beam deceleration step.

In general, higher currents are possible if the ICP plasma is

driven harder and stronger extraction conditions are used to

compensate the increased space charge density that occurs

when ions leave the virtual sheath at the floating extraction

aperture

see Sec. II B

. Beam currents below 100 eV can

also be increased

up to

1.5 times

if the decelerator is

specifically tuned for stronger focusing action and the steer-

ing quad plates directly in front of the target are run signifi-

cantly positive to pinch the beam inward. The ion fluxes

shown for the Ar

and O

beams are at least 1 ML/s for

impact

100 eV. This current level int

oa2mm

spot at such

low impact energy represents several orders of magnitude

higher arrival rate than most mass-filtered ion beam sources.

For example, electron impact sources typically used for ion

scattering spectroscopy

ISS

even without mass filtering

can barely approach 0.01–0.1 ML/s.

These sources typi-

cally provide up to 100 nA into a beam spot of 1 mm at 1

keV with the spot size becoming significantly larger and

beam current smaller as the impact energy is lowered.

FIG. 6. Beam currents measured via Faraday cup

0.5 mm inlet diameter

the target position for a CF

/Ar

48:14:38

plasma running at 3.5 mT

and 500 W. Beamline transport voltage was −12 keV and the mass slit was

set to 3 mm width. SiF

and SiF

species result from attack of the pyrex

plasma reactor used for this experiment. Also, O

was added to the plasma

mix to prevent CF

-like polymer deposition on the reactor walls at high CF

blending ratios.

FIG. 7. Current density at the target for:

and

and O

ion

beams for different final beam energies. In each case, the beamline transport

voltage was maintained at

13 keV, decelerator was tuned to optimize

beam current through the 2 mm diam beam flag aperture. A neat Ar plasma

and O

/Ar plasma at

500 W and 3–5 mTorr were used.

083302-8 M. J. Gordon and K. P. Giapis

Rev. Sci. Instrum.

, 083302

2005

Downloaded 14 Dec 2005 to 131.215.225.9. Redistribution subject to AIP license or copyright, see http://rsi.aip.org/rsi/copyright.jsp

C. Ion beam energy distributions and energy

tunability

The success of any ion beam scattering experiment relies

heavily on a well-defined scattering geometry and incident

beams with narrow energy spread if the energy distributions

of surface scattered products are to be meaningful. Further-

more, tuning the projectile energy provides a direct means of

varying the distance of closest approach

min

that occurs

during the close encounter between the two colliding nuclei.

These two features, tunable energy and narrow width, are

fundamental requirements for any ion source and beam sys-

tem used for scattering studies, especially at low impact

energy.

ICP plasma discharges have been shown to have a rela-

tively narrow ion energy distribution function

IEDF

by sev-

eral different measurement techniques.

33,34

Typical IEDFs

are 5–10 eV FWHM with low- or high-energy tails in some

cases. However, an ICP plasma has not been used as a beam-

line ion source for low-

studies before and most of the

beamline experience that exists in the literature

Freeman

sources

pertains to much higher impact energies

1–5 keV

, where an energy width of 20–50 eV is usually

inconsequential. Therefore, we felt that a concrete measure

of the beam energy distribution at the sample location was

necessary. Just because the ICP can be tuned to give a nar-

row IEDF does not mean that the beam hitting the target

sample will be narrow as well. Broadening of the distribution

due to collisions with background gas atoms in the extraction

region and during transport as well as ion creation along the

flight path through electrode collisions can potentially pol-

lute the beam hitting the sample.

Any capacitive coupling

from the rf drive antenna to the ICP will also cause inher-

ently bimodal beams

IEDF for a capacitive discharge is

bimodal

5,14

Finally, our technique for adjusting the impact

energy, by floating the entire plasma volume above ground

with a dc bias, must be tested.

The energy tunability of the ion beamline system was

evaluated by measuring the IEDF of the incoming ion beam

directly at the target location with the 180° energy sector

analyzer mentioned earlier. A typical energy distribution for

both a low and high energy

beam measured at the

target location is given in Fig. 8. The sector was run in CAE

mode

pass

=15 eV

while ion energy scans were performed

by ramping the retard voltage on the whole analyzer. Figure

8 shows that the beam energy can be varied easily by in-

creasing the dc voltage on the plasma bias plate, where the

offset between applied bias and measured energy represents

the Ne plasma potential

10–12 eV

. The IEDFs are con-

sistently narrow

10 eV FWHM

and peaks are unimodal,

indicating that capacitive coupling from the rf antenna is

efficiently stopped by the Faraday shield, i.e., power transfer

to the plasma electrons is purely inductive. Similar plasma

potentials in the 10–20 eV range have been measured by

other authors with Langmuir probe techniques for inert gas

ICP discharges in the 2–5 mTorr range.

Thus, the beamline

can simply transport the plasma IEDF all the way to the

sample without significant broadening. A narrow energy dis-

tribution for the incoming ion beam is very important be-

cause this inherent energy width can be comparable to the

particle exit energies when hyperthermal projectiles

50–500

are scattered from a surface. Also, inelastic energy loss

mechanisms typically account for

10% of the total loss

upon scattering.

Careful measurements of these losses can-

not be accomplished if the incident beam energy width is too

large. Discrete losses during the close encounter may occur

but they can be frequently overshadowed or “smeared out”

by an incident energy distribution that is too broad.

D. Scattered product detector performance

Energetic ion bombardment of a solid sample generates

a whole range of particle fluxes leaving the target surface

see Ref. 36

. The major species are sputtered neutrals and

secondary ions with kinetic energies of a few eV up to tens

of eV. These species are a mix of projectile and target atoms

that are generated as a result of multiple collisions and sput-

tering processes. At higher exit energies, surface atom re-

coils, and directly scattered projectiles that survive neutral-

ization

or those that are reionized on the exit path from the

target surface

are observed. The energy spectrum of these

directly scattered species contains information about single

collision processes and the energy losses that can occur dur-

ing the close encounter between the projectile and target nu-

clei. The inelasticity in these single collisions can manifest

itself as electron excitation of the projectile or target atoms,

photon generation, or energetic electron release.

A product

detection system capable of distinguishing the energy spec-

trum of many of the species leaving the surface could prove

very useful in understanding the fundamental processes oc-

curring on the surface during bombardment.

As a test case, the neutral detection capabilities of our

system

secondary neutral mass spectroscopy

SNMS

were

evaluated by looking for directly scattered hyperthermal Ar

from Ar

bombardment of an Ag surface at 110 eV impact

energy. Results of this experiment are shown in Fig. 9. En-

FIG. 8. Ion beam energy distributions for

measured at the target

position using a miniature hemispherical sector energy analyzer with chan-

neltron detector. For each case, the difference between the plasma float

potential and the average beam energy

11–12 V

is dictated by the Ne

plasma self-potential. Narrow FWHMs over a wide impact energy range can

be seen.

083302-9 Low-energy ion beam scattering apparatus

Rev. Sci. Instrum.

, 083302

2005

Downloaded 14 Dec 2005 to 131.215.225.9. Redistribution subject to AIP license or copyright, see http://rsi.aip.org/rsi/copyright.jsp

ergetic Ar neutrals are formed from single collision binary

events of Ar

projectiles with Ag target atoms that are neu-

tralized along the incoming or outgoing portion of the trajec-

tory in the near surface region. Single collisions of Ar

with

the surface result in a large fast scattered ion signal near

45–50 eV. This energy represents the single scattering events

that occur in one binary collision, where the energy transfer

can be approximated by the binary collision approximation.

The kinematic factor

energy transfer ratio

for Ar

on Ag for

a 90° scattering angle is

0.46, which would predict

50 eV exit energy for the 110 eV incident projectile. Also

shown are the neutral signals, with one large peak at low

energy along with a much smaller, broad Ar

peak occurring

near 65 eV. All neutral scans were conducted with the ca-

pacitor ion deflector in the first pumping stage set to +200 V

to prevent any charged species from ever reaching the detec-

tor. The ionizer was run at 2 mA electron emission current

with 70 eV electrons and the quadrupole was locked at 40

amu with

1 amu pass mass width. In addition, the scan

was taken in a single pass with 0.5 s dwell time for each

energy bin

=1 eV

. The large Ar

peak at low energy

represents the residual gas background and thermal species

leaving the target which are ionized near the 15 V grid po-

tential of the ionizer. The

3 eV shift to lower energy occurs

because of the space charge potential depression created in

the ionizing volume by the high emission current

2mA

The 65 eV neutral peak can be attributed to single-scattered

projectiles that are neutralized along the incoming or outgo-

ing scattering trajectory by the Ag surface. This peak is

shifted upward by 15–20 eV from the directly scattered ion

peak because the ionizing volume is run 15 V above ground.

When fast neutrals enter the ionizer, they are given a “kick”

in energy the moment they are converted to ions by electron

impact due to the 15 V accelerating field. A question may

arise why the Ar

exit kinetic energy is not exactly equal the

exit energy taking into account the 15 eV kick. Fast

neutrals seem to appear

3–5 eV faster than the directly

scattered ions. The offset likely occurs because the electron

straggling of the projectile on the incoming and outgoing

trajectories near the surface is charge state dependent.

has also been shown by Xu

et al.

that the continuous strag-

gling loss, as represented by the Oen and Robinsen

approach,

depends on the particle charge state.

The neutral experiments conducted with our scattered

product detector indicate that having an energy filter between

the ionizer and the rest of the detection system can be suc-

cessfully used to separate residual gas neutrals from those

originating from the target. In addition, the peak energy po-

sitions of both the ion and neutral signals can be identified

and make good sense when the operation of the detection

system is clearly understood. As well, the detection of neu-

trals generated at the target surface has been demonstrated,

suggesting that scattering experiments with reactive systems

where most of the reaction products are neutral

will be

possible.

IV. APPLICATIONS

A. Elastic scattering at low energy

Low energy He

ion scattering has been used for a long

time to determine the composition and structure of the top-

most atomic layers of a target surface

for example, see Ref.

. This technique is based on the binary collision approxi-

mation

BCA

model, where the projectile is deflected by a

sequence of pairwise interactions with individual target at-

oms. Each deflection event is assumed to be totally elastic,

so the energy transfer during the hard collision event can be

described by conservation of energy and momentum. As

such, the projectile energy after a single collision can be

simply calculated using the well-known “kinematic factor”

exit

cos

− sin

where

and

exit

are the incident and scattered projectile

energies,

is the target-to-projectile mass ratio

and

is the lab scattering angle. Application of the BCA

theory to He

ISS using keV impact energies is beyond

debate.

39,42

However, systematic studies of scattering phe-

nomena involving heavier inert ions at lower impact energies

have not been so widespread. To this end, it seems most

natural that we use our beamline system to investigate BCA

predictions at low collision energies. Indeed, any discussion

of inelastic energy losses that occur in hard collision events

i.e., electronic excitation of projectile or target

must first

include an assessment of how realistic the BCA model is at

low energy. In addition, it has already been mentioned that

equipment issues, such as scattering angle change with im-

pact energy, broad IEDFs for the incident beam, and high

surface neutralization rates frequently hamper low energy

scattering experiments. Evaluating the BCA over a wide

range of target-to-projectile mass ratio serves as a test plat-

form to assess if such effects are important for our beamline

scattering system.

FIG. 9. Energy scans of Ar

and Ar

exit channels from Ar

scattering off

an Ag target at 110 eV and 90°. Neutrals were detected with the ionizer

running at 2 mA emission current, 70 eV electron energy, and the capacitor

deflector in the first pumping stage set to +200 V. In the hyperthermal Ar

case, the ion creation potential in the ionizer is

15 eV above ground,

giving the fast neutrals leaving the surface a kinetic energy kick of 15 eV

when electron impact ionization occurs. Thus, the single scattered Ar

which is neutralized to Ar

by the surface should occur at

15 eV above the

SS position shown for the ion peak.

083302-10 M. J. Gordon and K. P. Giapis

Rev. Sci. Instrum.

, 083302

2005

Downloaded 14 Dec 2005 to 131.215.225.9. Redistribution subject to AIP license or copyright, see http://rsi.aip.org/rsi/copyright.jsp

Figure 10

shows the measured exit energy for single-

scattered

and Ar

off several target surfaces

Pb, Au,

Ti, Ge

111

at 90° lab angle for collision energies starting at

50 eV. Regression lines using a hard zero intercept are also

given, allowing an experimental kinematic factor to be deter-

mined directly from the slope. For these experiments, each

target was sputter cleaned

5 keV Ar

, 60° incidence

, an-

nealed, and amorphized with the low-

projectile beam at

300 eV for

15 min before taking data. In each case, the

exit energy dependence is seen quite linear with collision

energy, even down to the low energy range. The small upturn

in the Ge data

exit

exit,BCA

is likely caused by increased

multiple deflection events

one glancing, one near 90°

due

to the underlying

111

crystal structure being insufficiently

amorphized at

150 eV.

Experimental

factors for

50–1000 eV impact were

also measured for a wide variety of target materials using

and Ar

projectiles. These results are given in Fig.

as a function of target-to-projectile mass ratio, along

with the BCA predicted

value for single scattering at 90°

Eq.

. Agreement between the experimental data and

theory is excellent, despite the oversimplified view of the

collision process given by the BCA model. It should be noted

here that the

factor determination for Ne–Al and Ne–Si are

based on impact energies

400–500 eV only. In the next

section, we will see that inelastic energy losses become im-

portant for the Ne–Al and Ne–Si systems once a critical dis-

tance of closest approach has been reached in the binary hard

collision. We can also determine that the constant

factor

seen in all cases verifies that the lab scattering angle does not

change as the impact energy is varied for our system.

B. Charge exchange involving Ne

with light targets

We have shown in the previous section that the BCA

model is quite successful in predicting energy transfer for

projectiles with impact energies between

50 and 1000

eV on a variety of targets. However, inelastic loss processes

can sometimes occur where the translational KE of the in-

coming ion is converted into electronic excitation or ioniza-

tion of the atoms in the colliding pair.

The existence of

such channels is often seen as shifts in exit energies of the

projectile or target recoil from their elastic positions, changes

in the scattered particle charge state, or emission of charac-

teristic electrons and photons as excited states

created in the

hard collision

decay

see Ref. 37 for a review

. In particular,

inelastic losses, high ion yields

50% Ne

off Mg

43–45

and multiply-charged scattered projectiles

and Ne

off Mg, Al, and Si

40,46–50

have been seen for single binary

collisions involving Ne

projectiles at keV impact energies.

Such “richness” in the scattering behavior is intimately

linked to local charge exchange phenomena which occur as

the atomic orbital

states of the collision partners quan-

tum mechanically mix into hybrid molecular orbitals

MOs

during the hard collision step

the Fano-Lichten MO

mechanism

Although many studies have been conducted

with Ne

on Mg, Al, and Si at high collision energies, little

experimental data exist for the threshold region at low im-

pact energy where local charge exchange processes and in-

elastic losses begin to occur. We present a brief summary

here of our experiments on Ne

collisions with light targets

Mg, Si, Al, and P

for impact energies

1400 eV to empha-

size why tunable incident energy, as well as mass and energy

filtering of scattered products, is necessary. Details of these

scattering experiments are presented elsewhere.

Simulta-

neous mass and energy dispersion of scattered products is

highly advantageous because the energy distributions of

=20

and Ne

=10

can be measured

separately—eliminating signal overlap and confusion from

multiple charge states which can occur for electrostatic-only

and TOF-type analyzers

see Refs. 45 and 49 for overlap

issues

Exit energies of Ne

and Ne

resulting from single bi-

nary collisions of Ne

with an Al target from

100 to 1400

eV are shown in Fig. 11

, along with the BCA prediction

=0.149

for single scattering

at 90°. The total Ne

scattered intensity is also given, where the raw detector

counts in the lab frame have been converted to a term pro-

FIG. 10.

Experimental exit energies for single-scattered

and Ar

off several targets showing the constant kinematic energy transfer factor.

Summary of experimental kinematic factors for collision energies between

50 and 1000 eV, measured from the slope of the

exit

data. The

theoretical prediction from the BCA model for 90° scattering is given by the

line. For Al and Si targets, only the data below

=500 eV was used for the

-factor determination.

083302-11 Low-energy ion beam scattering apparatus

Rev. Sci. Instrum.

, 083302

2005

Downloaded 14 Dec 2005 to 131.215.225.9. Redistribution subject to AIP license or copyright, see http://rsi.aip.org/rsi/copyright.jsp

portional to the ion yield by normalizing by the incident

beam current and scattering cross section.

Two regions are

seen where the Ne

and Ne

exit energies are markedly

different. At low impact energy

region 1

, the SS Ne

exit

appears totally elastic and Ne

is not seen in the scattered

ion spectrum. Since the collision

min

distance of closest

approach

is relatively large

0.8 Å

at low-

impact,

straggling losses are small

a few eV maximum

and signifi-

cant overlap of the core-shell atomic orbitals of the collision

partners does not occur. As such, elastic scattering behavior

for Ne

is seen and no excitation channels exist to form Ne

due to MO promotion in the hard collision

. However, as the

collision energy is raised

region 2

, a transition occurs

where the Ne

exit becomes inelastic and Ne

is suddenly

generated at

470 eV with a large energy offset from the SS

elastic line. At this threshold

minimum

min

requirement

three phenomena occur simultaneously in the scattered ion

spectrum:

generation;

reversal of the Ne

yield

trend with impact energy; and

significant hard collision

losses in the Ne

and Ne

exit channels. The

min

value

where Ne

first occurs

470 eV=0.64 Å Thomas–Fermi–

Molière or 0.56Å Ziegler–Bierszack–Littmark interatomic

potentials

agrees well with the theoretical distance required

for

-shell orbital overlap for the Ne–Al collision pair

0.57

This observation suggests that a local charge exchange

process, driven by atomic orbital overlap of the 2

and 2

shells of Ne and Al, is likely responsible. Going one step

further, binary collision inelasticities can be evaluated by ad-

justing the lab frame energy loss to the center-of-mass frame

and accounting for electron straggling on the incoming and

outgoing trajectory paths

see Ref. 52 for details

. This in-

elasticity data

bin

for Ne–Al is given in Fig. 11

, where

saturation-like behavior for both the Ne

and Ne

exit chan-

nels can be seen. Furthermore, we can identify the saturation

energy loss values

40–45 eV for Ne

, 70–75 eV for Ne

as associated with double excitation events, which occur in

the hard collision step as the 4

MO from the Ne 2

highly promoted at small

min

+ 41–45 eV

→

+ 69–72 eV

→

+**

In the first case, Ne

is Auger neutralized to Ne

in the in-

coming path to the surface and is doubly excited by transfer-

ring two electrons out of the promoted 2

as it

crosses the 3

.Ne

can leave the hard collision intact

lifetime

1.5

−14

and autoionize in vacuum far from

the surface region. For the latter, a projectile ion which has

survived neutralization on the incoming path is doubly ex-

cited in the same manner but significantly more excitation

energy is required to promote two electrons in the ion. Auto-

ionization decay of Ne

+**

then results in the Ne

exit chan-

nel, which shows

70 eV hard collision inelastic loss.

The reversal in Ne

yield with impact energy is dis-

cussed in depth in Ref. 52, but it is clear that a transition

from nonlocal neutralization at low-

impact

Auger type

min

-dependent collision induced neutralization occurs as the

min

decreases. The critical

min

requirement for the yield

reversal is intimately tied to the 4

MO promotion as well.

If the incoming Ne

vacancy evolves as 3

, reso-

nant transfer from the target bands to the promoted 4

can

neutralize the projectile, causing the total Ne

yield to de-

crease. This mechanism, because it involves resonant trans-

fer, should only be operable if the 4

is promoted enough

so as to cross the target conduction band energy levels. As

such, a strong

min

dependence, or minimum impact energy

threshold, would be seen. This is indeed the case.

C. Reactive scattering of CF

/CF

off Si

Scattering studies using reactive projectiles on Si

100

were conducted as preliminary work towards a larger goal of

understanding the fundamental energetics and reaction

FIG. 11.

Exit energies of Ne

and Ne

resulting from single binary

collisions of

projectiles with polycrystalline Al for 90° lab scattering

angle in specular reflection. Elastic scattering behavior for a single collision

is indicated with

=0.149. Error bars

±5 eV

on the energy data are par-

tially shown to avoid clutter.

Binary collision inelasticities determined

from the data in

. Energy requirements for various electronic excitations

of Ne

and Ne

are shown, along with the theoretically predicted overlap

distance of the

=2 orbitals of the collision partners

grey area

. The colli-

sion

min

has been calculated using the Thomas–Fermi–Molière potential.

083302-12 M. J. Gordon and K. P. Giapis

Rev. Sci. Instrum.

, 083302

2005

Downloaded 14 Dec 2005 to 131.215.225.9. Redistribution subject to AIP license or copyright, see http://rsi.aip.org/rsi/copyright.jsp

mechanisms that occur during reactive ion bombardment of

Si—as applied to dry plasma etching processes. These ex-

periments with pure, mass-filtered beams of reactive projec-

tiles in the low energy range from 50 to 500 eV are, in fact,

one of the main driving forces behind our ion beamline sys-

tem. The 50–500 eV range is required in plasma etching to

provide momentum assist to surface chemical reactions and

stimulate etch product removal from the wafer surface with-

out causing damage to the substrate from heavy sputtering at

high impact energies.

The experiments presented here only

provide a taste of the complex processes occurring in CF

etching of Si, but they demonstrate why one needs a single

ion species at well defined energy along with broad based

detection capabilities of the particle flux leaving the target

surface. Fundamental understanding of the complex mecha-

nisms involved requires all the beamline components:

robust, high density ion source,

mass-filtered ion beam-

line,

low tunable energy,

high current, and

detec-

tion of mass and energy of ions and neutrals leaving the

substrate surface.

As a first step, we looked at the ion species leaving the

target surface for CF

and CF

bombardment of Si

100

as a

function of impact energy to give hints about the influence of

ion energy on removal of surface reaction products in

Si/SiO

etching environments. Although this experiment

seems rather simple, it has not been done before with a

single ion species and simultaneous product detection. CF

and CF

beams were formed from CF

/Ar/O

plasmas run-

ning at 500–700 W and

5 mTorr, giving target currents

200

A/cm

and

A/cm

. For this scat-

tering experiment, we sought to answer two questions:

what is the fate of the molecular ion as it hits the surface?

and

which ion species are preferentially released from the

reactive SiC

layer that forms on the Si surface during

bombardment? Figure 12 summarizes the results for CF

and

bombardment of Si at 90° lab angle

45° incidence

for

impact energies from

50 to 500 eV. As shown, the main

ion species leaving the surface are CF

,Si

, and

surprisingly—SiF

. At the lowest impact energy

70 eV

some CF

and CF

are seen, but they quickly disappear

above 100 eV impact energy. The presence of CF

at low

energy

Fig. 12

is likely molecular ion survival while the

may be caused by F-atom abstraction of the incoming

projectile

dissociative scattering

or physical sputtering of

-like species from the surface. An analogy is drawn here

to CF

scattering off fluorinated liquid surfaces,

where the

perfluoropolyether

PFPE

surface is terminated entirely by

–CF

and –F species with the ether oxygen atoms buried

beneath the surface.

57,58

Collision phenomena should be

similar because it is a generally held belief that CF

etching

of Si proceeds through a reactive SiC

layer with dangling

–CF

and –SiF

moieties on the surface.

For the PFPE case,

the predominant ion exit for CF

scattering at

250 eV im-

pact was CF

, with little or no CF

or CF

. Our preliminary

work echoes this behavior as well. What is interesting at

energies above 100 eV in our case is the rise of C

and SiF

exit channels. The large increase in SiF

from the surface

suggests that as the impact energy gets higher, more projec-

tile F atoms react to form dangling SiF

species which are

sputtered away as SiF

only

if they exit the surface in a

charged state

. We specifically looked for SiF

and SiF

leav-

ing the surface and did not detect any signal for all the im-

pact energies tested. It is curious why SiF

is the only

charged Si-containing species leaving the surface, but one

could explain such an observation as being due to an

F-deficient surface. Finally, the spectra show the onset of Si

for impact energies above

200 eV, which is a sure sign that

the increased momentum of the projectile is beginning to just

sputter the target surface.

This simple study gives a flavor for the physical scatter-

ing behavior as well as some of the chemical reactions that

are occurring on the surface during bombardment with just

one incident ion species. It is also clear from this one experi-

ment that complex scattering behavior, especially for a reac-

tive system, absolutely requires mass filtering of the particle

flux leaving the target surface to sort out all the exit chan-

FIG. 12. Charged exit channels leaving an Si

100

surface for:

and

projectiles. SiF

and SiF

ion species were not seen in the scattered

ion spectrum for any impact energy tested. The total scattered ion intensity

has been normalized by the incident beam current and was obtained by

integrating the detector counts of each species

quad at fixed mass

obtained

during an energy sweep of the hemispherical sector. CF

/Ar plasmas

running at 1–5 mTorr and 500–700 W were used.

083302-13 Low-energy ion beam scattering apparatus

Rev. Sci. Instrum.

, 083302

2005

Downloaded 14 Dec 2005 to 131.215.225.9. Redistribution subject to AIP license or copyright, see http://rsi.aip.org/rsi/copyright.jsp