untitled - MACprl06.pdf

Evidence of Simultaneous Double-Electron Promotion in

F

Collisions with Surfaces

J. Mace, M. J. Gordon, and K. P. Giapis

Division of Chemistry and Chemical Engineering, California Institute of Technology, Pasadena, California 91125, USA

(Received 15 August 2006; published 22 December 2006)

A high-flux beam of mass-filtered

at low energy (100 –1300 eV) was scattered off Al and Si surfaces

to study core-level excitations of

and

. Elastic scattering behavior for

was observed at energies

300

(500) eV off Al (Si) for a 90

lab angle. However, above this energy threshold, orbital mixing in the

hard collision step results in electronic excitation of F via molecular orbital promotion along the

(

), significantly reducing the observed ion exit energy. In addition, despite the electronegativity of F,

scattering at energies

450

(700) eV off Al (Si) produces

— behavior which is remarkably similar to

off the same surfaces. Inelasticities measured for single collision events agree well with the energy

deficits required to form (doubly excited)

and

states from

and

, respectively; these excited

species most likely decay to inelastic

and

via autoionization.

DOI:

10.1103/PhysRevLett.97.257603

PACS numbers: 79.20.Rf, 34.50.Dy, 34.50.Fa, 34.70.+e

Scattering of electronegative gas atoms and ions (i.e., O,

F, Cl) off surfaces is generally accompanied by highly

efficient electron capture processes; as such, the scattered

particle flux is largely composed of anions and neutrals [

–

]. Negative ion formation via resonant capture is quite

probable because image charge effects near a metal surface

tend to shift the projectile anion level below the Fermi level

of the surface [

]. As such, charge exchange in these

systems is largely governed by nonlocal processes as the

projectile approaches or rebounds from the surface.

However, local interactions involving double-electron pro-

motion have been shown to be important in some systems,

resulting in significant positive ion production. For in-

stance,

scattering off light targets (Mg, Al, Si, and

P) at low energy results in very high positive ion yields

(

and even

)[

]. Since the electron energy level

ordering of F and Ne is identical with respect to these

targets, electron promotions in F should also occur. In

fact, such effects should be seen at collision energies

even lower than those for

In this Letter, we show that double promotion along the

molecular orbital (MO) does occur for both

and

when

is scattered off Al and Si surfaces. Inelastic

and

exit channels were observed at low collision

energies (

450

–

700 eV

lab angle); energy losses

for single-scatter events show that

and

, followed by autoionization, are responsible for the

and

exits, respectively. Furthermore, the threshold

min

calculated for

MO promotion in the F-Si quasi-

molecule agree quite well with the experimental onset of

production (

min

–

Experiments were performed in a custombuilt ion beam-

line scattering apparatus described in detail elsewhere [

Ions were produced in a 13.56 MHz rf-driven inductively

coupled plasma discharge with a feed of

1:1 CF

(5 sccm), operating at 2 mTorr and 600 W. Ion energy

was adjusted by floating the plasma above ground with dc.

After ion extraction, magnetic mass-filtering, and fast neu-

tral removal, a

–

beam of

(3 mm diameter) was

delivered at 45

onto a grounded target (

-type Si wafer or

polycrystalline Al, freshly sputtered and annealed).

Scattered products were monitored at a 90

lab angle

with a triply differentially pumped 90

hemispherical-

sector energy analyzer in series with a quadrupole mass

spectrometer.

Figure

shows the mean exit energies of

and

resulting from single-scatter (SS) collisions of

projec-

tiles (

100

–

1300 eV

) with Si, along with the elastic

prediction for

lab

(kinematic factor

191

[

]). Two distinct regions with respect to incident energy

are readily apparent where the collision kinematics are

quite different. For

500 eV

, scattering is essentially

elastic with minor losses attributable (at first glance) to

FIG. 1. Exit energies of

and

resulting from single

binary collisions of

projectiles with Si at a 90

lab angle

along with elastic prediction (solid line). Error bars are partially

shown to avoid clutter. (Inset) Exit energy distributions for

when

1290 eV

PRL

97,

257603 (2006)

PHYSICAL REVIEW LETTERS

week ending

22 DECEMBER 2006

0031-9007

97(25)

257603(4)

257603-1

©

2006 The American Physical Society

electron straggling (electron drag) [

] on the incoming

and outgoing trajectory paths as the projectile encounters

the target bands. However, careful inspection of the energy

offset in region 1 reveals that straggling losses (a few eV in

our case [

]) cannot be solely responsible for the deviation.

Identical experiments with

projectiles on bare and

fluorinated Si (

) show that the surface

layer

(which commonly forms in F etching environments [

])

significantly increases straggling: i.e.,

off bare Si is

perfectly elastic (within 3–5 eV [

]), while

and

off

show much larger deviations. This observation can be

rationalized on the grounds that chemisorbed fluorine will

draw a considerable amount of electron density toward the

surface which could likely increase the total electron drag

on the projectile. In any case, we see that region 1 is

effectively elastic in nature overall because the collision

apsis (distance of closest approach

min

) is relatively large

at low

(i.e.,

min

A@400 eV

, Thomas-

Fermi-Molie

re potential with Firsov screening length

[

]); as such, significant overlap of the core-shell atomic

orbitals (AO’s) of the collision partners (

) has not yet

occurred.

However, when the collision energy is raised above a

critical value (

700 eV

), the overall kinematics of the

system change entirely —

becomes quite inelastic and

suddenly appears with an even greater energy offset.

The measured average exit energy of

is 35– 40 eV

lower than that predicted by the binary collision approxi-

mation (BCA) for a 90

elastic deflection [

]. The simul-

taneous appearance of these two phenomena above a

critical collision energy is a clear indicator of electronic

transitions taking place in the hard collision step as the

AO’s of the colliding partners intermix at small

min

.For

instance, AO overlap at a close distance (

) gives

rise to hybrid MO’s during the close encounter portion of

the scattering trajectory when a short-lived (

)

quasimolecule is formed. In certain cases, particular

MO’s of the quasimolecule may be highly promoted in

energy (due to degeneracy) so as to cross other MO’s

which will ultimately decay into Rydberg states as the

atoms separate (Barat-Fano-Lichten theory [

]). At these

crossings, electron exchange can occur, leaving one of the

exiting species in an excited state after the collision with a

kinetic energy deficit. Indeed, very similar behavior has

been seen for

off Mg, Al, Si, and P targets [

]. In this

situation, inelastic

below a critical apsis is

caused by double promotion of electrons from the Ne

(

MO) to the Ne

in the hard collision step

(i.e.,

and

). These excited

states (

and

) then decay far from the surface

via autoionization to give

and

with large kinetic

energy deficits that directly correlate with the energy re-

quirements for excitation [

(41– 45 eV) and

(69 –72 eV)]. The fact that both the

and

energy losses remain relatively constant with

,so

long as the threshold

min

(700 eV at 90

) has

been reached, further suggests that well-defined electronic

excitations are at play in the

system.

Additional evidence for a fundamental excitation

mechanism comes from energy loss measurements using

an Al target. Since the energy ordering of AO’s of Al and Si

is identical with respect to the F projectile (i.e., the MO

correlation diagram is the same [

]), collision kinematics

for SS events off Al and Si should be quite similar. Indeed,

this is the case (Fig.

), except that the

inelasticity and

production begin at a lower collision energy (larger

min

–

). This latter effect is simply due to the

shell of Si being closer to the nucleus than that for Al.

Small differences also occur at low

;

scattering off Al

is almost perfectly elastic. It is likely that the fluorinated

layer (

) on the Al surface is significantly thinner than

that for Si because F spontaneously reacts with Si [

] but

not with Al. Smaller straggling losses would be expected

due to lower electron density in the surface region because

of less F on the surface. We also see that the scattered

intensity [

] makes a large jump when inelastic losses

become significant (Si is analogous). In general, the scat-

tered ion yield increases exponentially with inverse pro-

jectile velocity because of contact time arguments [

];

i.e., a slower projectile has a larger probability of being

neutralized by Auger or resonant charge transfer with the

surface as a whole. However, the rapid increase in

at the

threshold indicates the sudden turn-on of ‘‘additional’’

generation. Finally, it should be noted that the total

production continually increases with collision energy and

that it is smaller (

–

less) than

Previous experimental studies of F collision kinematics

at low energy are sparse [

] and

off Si has been

reported only once [

]. Hird

et al.

detected

off Si

using

projectiles when

min

(

8 keV

lab

); as a comparison, the same study found that

was formed when

min

573

(

800 eV

lab

). This large apsis discrepancy between F and

FIG. 2.

Exit energies and intensities of

and

resulting

from single binary collisions of

with Al at a 90

lab angle.

PRL

97,

257603 (2006)

PHYSICAL REVIEW LETTERS

week ending

22 DECEMBER 2006

257603-2

Ne is contrary to intuition given that the energy ordering of

AO’s for F and Ne with respect to Si is identical. In our

experiments, the onset of inelastic

and

generation

occurs at much larger apsides — on par with those expected

in light of

. To look more carefully at possible

mechanisms for

, raw energy losses measured in

the lab reference frame must be converted to the center-of-

mass system and adjusted for straggling. When

lab

the binary collision inelasticity (

bin

), i.e., the total energy

that is directly available for electronic excitation in the

hard collision, can be evaluated quite simply [

exit

bin

;

(1)

where

is the projectile energy,

the kinematic factor,

is the target-to-projectile

mass ratio, and

;

represents the straggling losses [

]on

the incoming (1) and outgoing (3) parts of the collision

trajectory. Inelasticity values determined in this fashion are

given in Fig.

along with the energy requirements for

several

excitation channels [

] using the

system for inspiration. Transition energies have

been referenced to the slowly increasing baseline because

of the aforementioned augmented straggling affects due to

surface fluorination. On this graph, one immediately sees

that the

bin

values for

and

are large, relatively

constant, and fall within well-defined ranges when

min

. Incidentally, this apsis value is exactly that pre-

dicted for

shell overlap of

(crosshatched re-

gion) [

], suggesting that AO mixing is ultimately

responsible for the opening of new excitation channels.

As indicated, the energy loss data can be conveniently

assigned to transitions initiating from ground state

(subject to the sloping baseline):

neut

bin

35 eV

;

bin

62 eV

In the first case,

is neutralized to

on approach to the

surface through nonlocal processes (Auger and/or resonant

charge transfer [

]), followed by double excitation to

via molecular orbital promotion (analogous to

) in the hard collision step. Since the lifetime of

is likely to be longer than the collision time (i.e., the

lifetime is

[

], so

should be similar),

can

leave the surface intact and subsequently autoionize (AI) to

far from the surface. This mechanism is probable

because most projectile ions approaching the surface will

be efficiently neutralized before the hard collision occurs

[

]. For

, an analogous process occurs where the

precursor state entering the hard collision is instead

which has survived neutralization on the incoming trajec-

tory path. This hypothesis is strongly supported by the

much smaller

signal intensity compared to

(inset

in Fig.

). Although double collisions are frequently sug-

gested as the cause of a

channel [

] (i.e.,

goes to

in the first collision, followed by

by direct ion-

ization in the second), we see that the magnitude of the

energy loss is too great for such an explanation [62 – 98 eV

vs.

collision no.

no.

4eV

]. In fact,

these two mechanisms are identical to those proposed for

inelasticity and

formation off Si; indeed, one

FIG. 3.

Binary inelasticities (

bin

)of

and

off Si

determined with Eq. (

). Transition energies for several projec-

tile excitations are shown. Grey areas represent the range of

bin

values possible due to the presence of several different doubly

excited final states. The crosshatched region denotes the colli-

sion apsides where the maximum electron density overlap of the

) and Si (

) orbitals would occur.

FIG. 4.

Orbital energies for the F-Si molecule as a function of

internuclear distance. The calculated

orbitals (

orbitals) are

indicated by solid (dashed) lines; the atomic character of each

MO is shown to the right. Strong promotion (heavy dashed line)

along the

MO occurs near

min

, allowing projectile

excitation to the

(

) and

(

) states.

PRL

97,

257603 (2006)

PHYSICAL REVIEW LETTERS

week ending

22 DECEMBER 2006

257603-3

would surmise that such processes should be rather generic

when orbital energy orderings and promotion are similar.

To further support this claim, molecular orbital calcula-

tions [

] were carried out to determine the range of

min

needed for

MO promotion (Fig.

). These results

demonstrate that excitation of the F

along the

MO should occur for

min

–

— identical to the

experiment.

Given the above mechanism, it is now possible to ex-

plain the

intensity variations with collision energy

(Fig.

). The sudden increase in scattered

can be

associated with opening of the

channel.

For

, the continual increase with

results from less

efficient neutralization of the projectile

as the energy is

raised. The large difference overall between the

and

intensities is due to the

projectile being neutralized

most of the time — and

cannot produce

in a

single collision event. Finally, the tapering off of the

signal at high energies is similar to the case of

off Si,

which has been explained as due to collision-induced

neutralization in the hard collision step [

], i.e., resonant

transfer from the target bands to the

MO (filling the F

)— which can occur only if the

is highly promoted

at small

min

. Although the present study has emphasized

only ions and their associated energy losses, identical

processes surely occur for neutral projectiles with respect

to an inelastic

exit, i.e., formation of

(which auto-

ionizes to

) that descends from an

precursor. There-

fore, energetic neutrals formed by ion neutralization will

show similar interaction dynamics and inelastic losses.

In conclusion, scattering of F ions off Si and Al surfaces

at low energies follows the BCA closely until a threshold is

reached where electron promotions result in significant

inelasticity for the exiting projectile. Core-level electronic

excitations of

and

via double promotion were seen to

be responsible for the inelastic

and

exit channels,

respectively, when

min

(

700 eV

) and

0.60 A

(

450 eV

) for the F-Si and F-Al collision pairs,

respectively. Despite its electronegativity,

scattering off

Al and Si is qualitatively identical to

from the per-

spective of a well-defined

turn-on, continually increas-

ing

yield with impact energy, and inelasticity values

which point to the same

double promotion mecha-

nism. Thus, in the range of energies where electron promo-

tion is observed, it appears that chemical interactions are

irrelevant. Finally, for incident energies larger than thresh-

old, scattered F atoms and ions will possess kinetic ener-

gies lower than those predicted by BCA. Since scattered

ions are important in patterning of semiconductor devices

using plasma etching, energy losses suffered in sidewall

collisions will influence the etch rate and pattern profile

[

This work was based on research funded by the National

Science Foundation (No. CTS-0317397).

Corresponding author.

Electronic address: giapis@cheme.caltech.edu

[1] M. Maazouz, S. Ustaze, L. Guillemot, and V. A. Esaulov,

Surf. Sci.

409

, 189 (1998).

[2] C. Auth, A. G. Borisov, and H. Winter, Phys. Rev. Lett.

2292 (1995).

[3] P. Roncin, A. G. Borisov, H. Khemliche, A. Momeni,

A. Mertens, and H. Winter, Phys. Rev. Lett.

, 043201

(2002).

[4] A. G. Borisov, V. Sidis, P. Roncin, A. Momeni,

H. Khemliche, A. Mertens, and H. Winter, Phys. Rev. B

, 115403 (2003).

[5] H. Winter, A. Mertens, C. Auth, and A. G. Borisov, Phys.

Rev. A

, 2486 (1996).

[6] S. Ustaze, R. Verucchi, S. Lacombe, L. Guillemot, and

V. A. Esaulov, Phys. Rev. Lett.

, 3526 (1997).

[7] L. Guillemot, S. Lacombe, V. A. Esaulov, and I. F.

Urazgildin, Surf. Sci.

334

, 224 (1995).

[8] M. J. Gordon, J. Mace, and K. P. Giapis, Phys. Rev. A

012904 (2005), and references therein.

[9] M. J. Gordon and K. P. Giapis, Rev. Sci. Instrum.

083302 (2005).

[10] J. W. Rabalais,

Principles and Applications of Ion

Scattering Spectrometry: Surface Chemical and

Structural Analysis

(Wiley, New York, 2003).

[11] O. Oen and M. Robinsen, Nucl. Instrum. Methods

132

647 (1976).

[12] F. R. McFeely, J. F. Morar, N. D. Shinn, G. Landgren, and

F. J. Himpsel, Phys. Rev. B

, 764 (1984).

[13] M. Barat and W. Lichten, Phys. Rev. A

, 211 (1972).

[14] Scattered intensities were corrected by the cross section

and normalized for beam current. See Ref. [

[15] J. P. Grouard, V. A. Esaulov, R. I. Hall, J. L. Montmagnon,

and V. N. Tuan, J. Phys. B

, 1483 (1986).

[16] B. Hird, R. A. Armstrong, and P. Gauthier, Phys. Rev. A

, 1107 (1994).

[17] Since the

parameters for the Oen and Robinsen strag-

gling loss formalism are unknown for

, the correspond-

ing values for

[

] were used instead.

[18] NIST Atomic Spectra Database, http://physics.nist.gov/

cgi-bin/AtData/levels_form.

[19] Distances are based on the location of maximum radial

electron density. Orbital data are taken from J. Slater,

Quantum Theory of Atomic Structure

(McGraw-Hill,

New York, 1960), Vol. 1.

[20] R. Morgensten, A. Niehaus, and G. Zimmermann, J. Phys.

, 4811 (1980).

[21] R. Souda, K. Yamamoto, W. Hayami, T. Aizawa, and

Y. Ishizawa, Phys. Rev. Lett.

, 3552 (1995).

[22] Unrestricted Hartree-Fock calculation at the STO-3G*

level using the

GAMESS

package; W. H. Hehre, R. F.

Stewart, and J. A. Pople, J. Chem. Phys.

, 2657

(1969); M. W. Schmidt

et al.

, J. Comput. Chem.

1347 (1993).

[23] For smaller deflection angles, corresponding to ions in-

volved in more glancing collisions, it can be shown that

even larger deviations from BCA and, thus, lower exit

energies are possible.

PRL

97,

257603 (2006)

PHYSICAL REVIEW LETTERS

week ending

22 DECEMBER 2006

257603-4