function of (ElQ-)rt3 versus spectral energy E. Apparently the .... 300. ,i yes yes no no yes yes yes yes yes no. 1.0. 0.75. 0.54. 0.24. 0.75. 0.7 t220. 1000. 4O :43 : ...
I
@ Gordon and Breach
Radiation Effects
t'''"ï.ti:lìlä.:iläJ'?J,:
1978, Vol. 36, pp. 189-196
86B
THE SPUTTERING PROCESSES DURING 6 keV Xe ION BEAM BOMBARDMENT OF HALIDES M. SZYMOÑSrI1, H. OVEREIJNDER and A. E. DE FOM-Instituut uoor Atoom-
en
VRIES
Molecuulfysica, Kruislaan 407, AmsterdamlWgm, The Netherlands (Receiued SePtember 23, 1977)
Mass selected energy distributions of sputtered atoms
a
measured using a time of flight method. The analysis of ex which can be explained by a simultaneous appearance of thermal evaporation of decomposed target material. The an sputtering mechanisms are in very good agreement with the (mean t tributions: E, (surface binding energy), are obtained änd their physical meaning ts discussed. investigations is presented.
I",
1
reasons mentioned above one may express the total sputtering yield as a sum of three
For the
INTRODUCTION
The collisional mechanism of sputtering of
different parts:
ele-
mentáry solids has been well established by a large number of highly accurate experiments (see recent
S:
S"ot.
*
S,¡.r0. + S,n..""'.,
ference of the target components'6
In several experiments, however, the sputtering cannot be explained satisfactorily by the binary collision cascade model'?-12 In order to describe an
beam.
2 beams.T
Energy distributions of sputtered atoms and mole-
Kelly and coworkers have studied the stoichio-
cules have been measured with the apparatuS sche-
of several oxides and metry óhanges - of the surfaces ¡"¡i¿"r.r-tt 1'¡. observed loss of oxygen and halogen from the surfaces has been interpreted
matically shown in Figure l. The ion source, described in detail by Politiek et al,ta produced a 6 keV Xe ion beam with a current density of 0'5
as
being due to the bombardment induced decom-
posiiion and subsequent evaporation of oxygen and halogen,
t Present address: Instytut Fizyki, Uniwersytet Jagielloñski' 30-059 Kraków, ReYmonta 4, Poland.
EXPERIMENTAL
189
'
M SZYMT)NSKI, II.
OVEREIJNDER AND
NEUTRAL
OUTPUT SIGNAL
¡
BEAM
I
I.l( iLJl{l: rr¡rr:rl
.á¿.y'¿ TRAP
o -
I.
MULTIPLIER OUADRUPOLE
x
l0-7 torr during a run.
r5
All targets were polycrystalline, obtained from chemically pure powders by pressing them at 5000 atm. during 5 min.
3 ON THE ENERGY DISTRIBUTIONS
OF
SPUTTERED PARTICLES
According to Politiek and Kistemakerr6 the flux (Þ(¿) of particles leaving the surface with energy E is connected to the flux Õ(E') inside the solid by the following transformation formula
:
orn'l:
fi
TARGE T
MONITOR
II
and
III)
is
¡h,rrl + E6),
(r)
tarv target'
I
otE,) _ _.
(2)
has the lorm:18
o'or.('E)-
I
¿,."*-r"
,
(4)
where n,
= n j- = 2 refer to the powers in the atomic distributions IEq, (2)1. This so-called sratistical model predicts a much steeper decrease olthe molecular flux with the spectral energy than formula (3) does for atomic spectra, Il a primary ion dissipates its energy to the lattice at a high rate almost all particles within a small volume of the collision cascade can be in motion and a so-called "spike" event takes place.le There is a tendency in such a system to reach local spectrum
of
particles forming the spike.?'te The
binding energy Eo restricts the number of sputtered particles from the spike but does not influence the Maxwellian distribution [see formula (l)] and one can write:
oth.,e.(E)-Eexp
t #,)
(5)
u'here the temperature fro. should correspond to a mean energy of the particles within the spike volume. There is some doubt about the name "thermal" for the spike because of the collisional origin of this phenomenon: the mean energy ol the particles
forming the spike can even be in the eV range.re
E'¿
(l)
and (2) lead to an energy distriatoms sputtered b1' momentum transfer
Thc expressions
Il simultaneously sputtered atoms X, Y have a relative energy smaller than the dissociation energy of the XY molecule. cluster formation is possible and an asymptotic expression for the high energy region
equilibrium and for this reason aMaxwell-Boltzmann distribution has been assumed to describe the energy
;
where .8, is a surface binding energy of the particle as a consequence of the binding force directed perpendicularly to the surface. Let us consider atoms sputtered as a result of a collision cascade. It is known from theory6 as well as from experimentrT that in a binary target with a mass ratio between the components less than 2 we can use the same approximation as for an elemen-
bution of
IONISER
tn.
ln order to obtain energy distributions a time of flight method, the correlation technique, has been used. This method, the mass selected detection of neutrals and all technical details are described
o(E)
{
at
potcnt¡al ol 6 kV on the target. The base_pressure in thc collision chamber was below I x l0 7 torr,
elsewhere.
BEAM \ION BEAM
Schcmatic vicw of thc apparatus. The effective flight path for sputtered panicles (chambers I,
to 1.4ì
increasing to 8
VRIES
.:; .b
f
A. E. DE
fronr thc bínarv collision cascade:
Horvever. as lar as the Maxwellian distribution can
be used teq. (5)l there is always a possibility to define a temperature 7..o. and in this paper we prefer
to retain the term "thermal" and to
stress the
thermal properties olthe spike. (3)
Up to now very little is known about so-called "ox!'gen sputtering"e-rr or. more general. about ion
SPUTTERING OF HALIDES
beam-induced decomposition and evaporation of target material. According to the work of Chadderton and Torrens2o we think that an ion beam produces a large number of displaced atoms near the suface. These interstitials are highly mobile and can easily diffuse to the surface, then reach a thermal equilibrium with the surface atoms and evaporate if the vapour pressure of such material is sufficiently high, Another possibility is that the decomposition is induced by inelastic energy loss of the Xe ion (like
exciton and hole formation during electron irradiationrs'21). Nevertheless the expected energy distribution for thermally evaporating surface particles is the same as we have measured during sputtering of alkali halides with an electron beam.15'21 The spectra of evaporated alkali atoms as well as thermal halogen atoms and molecules were accurately described by the expression:
o,n..u.
- 1/8"*p (-
(6)
-1r)
The temperature l in expression (6) should now correspond to the macroscopic temperature of the surface, equal to or slightly above room temperature because of heating by the incident beam.
4 THE ANALYSIS OF ENERGY
SPECTRA
As will be shown in the next section we have observed a large contribution of molecules to the sputtering. Induced dissociation of these molecules may give extra atomic ions in the ionizer of our detection system, which could influence the measurement of the distributions of sputtered atoms with the same mass. Fortunately we have used the time of flight method and the atomic ions produced by dissociative ionization of molecules would contribute mostly to the low velocity parts of the spectra. However, we have not observed particles at all in the low velocity range of the investigated spectra while in the same velocity range of the corresponding molecular distributions quite large signals have been
observed. This fact indicates that the influence of induced dissociation in the ionizer on the measured energy distributions was negligible. Three exceptions for Br, F and CdI are considered in Section 5. The time of flight distribution of I atoms sputtered
from a CdI, target is shown in Figure 2. The structure of the spectrum suggests that the sputter-
ênergy (.V)
-
r8 a I
56 .ct
I
È4 ln zl¡J tZq
timc (ms)
- of I atoms sputtered FIGURE 2 The time of flight distribution from cdl2 with a 6 keV Xe+ beam. ing is caused by more than one mechanism. Let us assume that the collision cascade mechanism is responsible for the sharp peak at the high energy part ofthe plot. Thus formula (3) is valid. Now this expression can be rewritten as follows: (E/O.oll.)1/3
(3',)
-E+Eo.
In Figure 3 our experimental data are plotted as a function of (ElQ-)rt3 versus spectral energy E. Apparently the experimental points form a straight line in the energy range above 0.7 eV. According to Eq. (3') the corresponding binding energy Eocanbe calculated from the plot. With the known parameter in distribution (3) we subtract the collisional part from the total experimental energy distribution'
I 0
'e
t
¡i
g -r^
g
€. e
x
-t¡.¡
encrgy (eV)
3 I
atoms sputtered from CdI, with a 6 keV Xe+ FIGURE beam. The experimental energy distribution O(E) is plotted as a function of (E/@)t/3 versus the energy E.
-
M. SZYMONSKI, H. OVEREIJNDER AND A. E. DE VRIES
94
5.2
Molecular Spectra
As can be seen in Table I we have observed a large contribt¡tion of molecules to the total sputtering yield lor all investigated targets. In several cases the
cnergy distributions of those particles cannot be clescribed by the statistical model mentioned in
Section 3 llormula (4)1. Figures 8 and 9 show the energy distributions of Br, and I, molecules sputtered from AgBr and CdI, targets respectively. The spectra have quite different shapes and do not correspond with the asymptotic behaviour E-a.5 given by
Eq. (4). Considering other possibilities for the formation of molecules during ion bombardment we have used the same analytical procedure, as for atoms (see Section 4). We have found that the Br, flux is purely thermal and fits very well distribution (6) with a temperature T equal to 345 K. However the I, flux has a more complex structure. The high energy part can be expressed by a collisional distribution tEq. (3)l with E, : 0.24 eV. The rest of the spectrum again is thermal tEq. (6)l with a temperature 1:275 K. The same shape of the distribution has been observed for I, molecules sputtered from a PbI, sample. The observed flux of CdI also has a complex
structure. Low energy particles arc very well described by a Maxwell-Boltzmann function [Eq. (5)l which indicates the spike origin of these molecules, but the temperature 7.0. : 614 K is significantly lower than the mean temperature obtained from atomic fluxes. This suggests that in our ionizer a dissociation of CdI, may occur similarly to Br,
\ ì
¡r
l"*J1 I I
0r1t0(eV) -.----* energy
FIGURE9 Xe+ beam. as a sum ol broken lines Asymptotic collision cascade theory are marked respectively.
CdI2witha6keV lid [ñe) is presented .,. [Eqs. (3) and (6), êxperimental points, mechanism and the
by -E-2 and -E-.'5
dissociation. Taking into account this possibility the temperature of the CdI2 distribution is found to be T,o. 940 K which is very close to that from the
:
atomic spectra. The high energy part of the CdI spectrum and the flux of AgBr can be attributed to a statistical mechanism but both distributions decrease slower with energy than is predicted by the theory. Unfortunately very little can be said about the Ag, and Cd, distributions because of a very weak signal for such molecules. The flux of Cd, reaches a maximum at about 0.1 eV and decreases quite
rapidly at higher energies. A Maxwellian distribution gives a spike temperature which does not correspond with temperatures obtained for other particles and together with the steep slope of the higher energy paft of the spectrum this suggests a statistical mechanism for formation of metallic dimers. We have not measured spectra of heavier molecules like CdI, Pb2, PbI and PbI, because the mass range of our mass spectrometer goes up to about 250 atomic mass units.
6 oo0l1
energY (eV)
FIGURE 8 Br, molecules sputtered -from AgBr with a 6 keV Xe+ beam. The theoretical distribution (6) (thermal evaporation) is plotted as a solid line together with the experimental points.
DISCUSSION
The present results clearly show that the sputtering of halogen compounds has a complex character. The collision cascade model describes in our case only part of the sputtered particles, which for CdI, covers less than half of the total emission. However, energy
SPUTTERING OF HALIDES
193
TABLE I Results of the analyses of energy distributions. The relative signals are not corrected for the efficiency of the detection system. The means that the information is not available from the measurements because of too low signal-to-noise ratio. sign
"-"
Thermal
Observed Relative Collisional
spike
Target particles signals distribution Er(eV) distribution
AgBr
Ag
0.5
Br
1.0
AgBr AE, Br,
0.22 0.03 0.28
Ag
1.0
yes
F
0.r3
AgF
0.5
ï
Cd
0.28
I
r.0
yes yes
cdl
0. l1
Cdt r2
0.008 0.01
Pb
0.21
I
1.0
l2
0.03
AgF
cdl,
PbI2
a u
Thermal evapor.
7'.e.ß) distribution T(K) S"or. :S,n.,0.
1.0
no
no
t.2
no
no no
no
no
no
yes
1.0
no no
no b_
0.75
0.54
yes yes
no
b
no
no
no
yes
0.24
no no
yes yes yes
0.75 0.7
yes yes
yes yes
Statistical :
S,n,.f.
mechanism
b
t220 1000
1050 I 140
no
yes
yes yes
no
345
400 400
:43 : l7 51:29:14
4O
yes yes
yes
275
95: 0:
5
yes yes yes
300
,i
76:21:
J
no
70:17:13
Relative numbers with the assumption Srot"r: S"orr. * Str,.sp. + S,n..,. : 100. "yes" but ions are from dissociative ionization in the ionizer (see text, sections 4,5.L,5.2'),
system which can differ substantially for each mass and are given only for qualitative information.
5.1
Atomic Spectra
The energy distributions of sputtered atoms form two different groups. In the case of AgBr the spectrum of Ag atoms is presented in Figure 7. This distribution is collisional and can be fitted with formula (3) Golid line in Figure ?) and a surface binding energy of I eV. Also bromine atoms sputtered from the same target show only the collisional energy distribution with a slightly higher binding energy. The last spectrum however has a small fraction of very low energy particles which we have attributed to the dissociation of Br, molecules in the ionizer of our mass spectrometer because of the same time of flight of these particles and Br2
parts [Eqs. (3), (5) and (6)] originating from collision cascades, thermal spikes and thermal evaporation of decomposed target materials. From the fits to the experimental data we have obtained the surface binding energies, the mean temperatures of the spikes and the macroscopic temperature of CdI, and PbI, surfaces. The results are given in Tablé I.
similar behaviour was observed for AgF but here^the low energy contribution was even larger. molecules (see Section 5.2).
Completely different distributions have been for atoms sputtered from CdIr and PbI2. A typical example of such a distribution is presented in Figure 6 and has already been analysed in Section 4. All spectra (Cd, Pb and I for both targets) are very well described by the sum of three different
1
measured
t0....-
lo0
energY (eV)
FIGURE
?
Ag atoms sputtered from AgBr with a 6 keV Xe+
beam. The theoretical
"collisional" energy distribution Iformula
(3)ì is plotte¿ as a solid line together with the experimental points. ihe asymptotic behaviour is marked by -¿-2.
M, SZYMOÑSKI, H. OVEREIJNDER AND A. E. DE VRIES
192
¡
xe._
I I
1 -¿. f
I tt !-
cdlz
c
)
€d
rd
rt
I
0 b
ro5
l¡J
Þt
-x
ro'
10'
l¡J
Íx UJ
energy (eV)
FrcuRE
4 I atoms,rJ:ä::11,,;, :
O(E) beam. The distribution O'(¿) function of O'/-E versus the energy -E'.
-
kev Xe+
O"o',.(E) is plotted as a
leading to the difference spectrum O'(E)
:
O(E)
- -ElkT"Þ..
(5')
I I
.9
E
''oo
.ö
x
3
10
calcu-
. to be Points
to the
sputtering at very low energies' We now subtract the tùermal spike distribution to get @"(E): O'(E) (Þ(Ð,n..o. and use the logarithmic form of Eq. (6):
-
- -Elkr.
(6')
As can be seen in Figure 5 the spectrum (Þ"(E) fits very well the linear dependence [Eq. (6')] down to 0.0i eV which is the lowest energy detectable' The
incident beam.
-
r 1000
experimental points exhibit a linear dependence on the spectral energy, in this case from 0.5 eV down to 0.15 eV, indicating the spike origin of this part of the
temperat equal to perature,
-
1
The difference spectrum @'(E) is now plotted according to Eq. (5') in Figure 4' Once more the
los(@.^.",.1\Þ)
-
-
G
spectrum lated the equal to bélow 0.
:
beam. The distribution
o"or.(E). -îire Maxwellian distribution (5) can be written in a logarithmic form: log (@,n..'./E)
5 I
atoms sputtered from CdI, with a 6 keV Xe+ O"@) O(.8) O."r.(E) O,¡.,'.(E) is plotted as a function $" / \r/ E versus the energy E'
FIGURE
f this line is room tem-
001 FIGURE
6 I
0l
I
10
energY (eV)
atoms sputtered from-CdI, with a 6 keV Xe+
beam. The theoretical energy distribution (solid line) is presented (D,n."". as a sum of three anal¡icai distributions O"orr., Ot¡.sp. and (broken lines) and plotted together with the experimental points'
We have applied this method for the analysis of all energy distributions for atoms as well as for molecules. Of course, in several cases the spectra consist
only of two or even one contribution'
g due to the
5
RESULTS
The observed signals and results of the analyses of energy distributions of the investigated targets are summarized in Table I. The relative numbers describing the intensity of the observed signals are not corrècted for the efficiency of our detection
SPUTTERING OF HALIDES
distributions Õ.o,,. are in good agreement with theoretical formula (3). This fact indicates a linear energy transfer in the collision cascade, even for compounds with a wide mass difference between the components (mrlm, at least up to about 2). Such behaviour has been predicted theoretically by Andersen and Sigmund.6 The surface binding energy, which determines the efficiency of sputtering
by momentum transfer, is for all investigated targets significantly smaller than for pure metals (see Table I: Er-< I eV). Experimentally we have found a large contribution of "thermal" particles to sputtering which can
be ascribed to thermal spike and
thèrmal
evaporation mechanisms. There are two theoretical treatments of the spike event using a modern damage,theoly.l3'le'22 According to Sigmundle the appearance of spikes and the sputtering yield ,S,n..'. are determined by two quantities: the mean density of the energy deposited into
atomic collisions by an incident ion and the time constant r describing the dissipation of this energy. The maximum energy density in the central core of the spike áo and the value of ¡ can be estimated using Sigmund's contour plots from P.ef . 22. Generally for fixed energy and type of the projectile, do is proportional to ¡/t (N is the density of the sample) and decreases with decreasing mean mass of the target (M). The behaviour of r is just opposite. For com-
parison we also evaluated the experimental energy density, that is the average energy per particle 9.r0. : 312 kT,Þ..The numbers, together with important properties of the investigated targets, are presented in Table IL We have included our experimental data from sputtering ofalkali halides23 and a silver target. A pronounced spike effect may occur when do is larger than the minimum energy needed for sputter-
195
ing and the time constant r is larger or of the order of the slowing-down time ro of the collision cascade.re The minimum energy for sputtering should be approximately equal to the binding energy Er, while ro is of the order of l0-r3 s. As we can see in Table II these criteria are fulfilled for all investigated systems and indeed in all cases the spike effect has been observed except for AgBr. However the experimental energy density 4,o. of the spike particles is not described correctly by'the quantity do itself: d.*0. and 0o do not even show a clear corre-
lation.
As was mentioned before we have not observed spike sputtering in the case of AgBr. We think that the reason lies in stoichiometry changes on the AgBr
surface. As we will discuss below, during an ion bombardment the bromine vaporizes quickly from the surface and silver is left. This effect can cause a porous structure ofthe surface layers, decreasing the effective density i/ and then lowering the energy density 90. Therefore the effective spike temperature cannot be high. Also the vapour pressure of a heavily silver enriched surface is probably too low to get a spike.
The last contribution to the sputtering S,n..u. has an energy distribution given by Eq. (6). The experimental spectra, for all investigated targets, indicate that a vaporization of decomposed target particles occurs if the vapour pressure of the components is sufficiently large.
The pre-exponential factor in the energy distribution of vaporizing particles tEq. (6)l is different from that for a three-dimensional Maxwellian distribution tEq. (5)1. This observation, together with the same results for sputtering of alkali halides with an electron beam,rs gives a strong argument for the specific character of vaporization from halogen
TABLE II Comparison between experimental and theoretical parameters of spikes. For detailed inlormation text, Section 6. The sign
Target
u b
"-"
(M) (amu)
means that the information is not available.
Nx
10'z
A v0
(Ä-')
(ev)
0 - (\n,
T
(lQ-'r
s)
(e\''¡
Rbcl
6l
2.8
2.1
6.1
0.20"
RbBr AgBr Ag
83
2.5
2.6
4.7
0.23u
94
8.8
t.4
108
4.2 5.9
t22
2.8
PbI2
154
2.4
29.0 7.2 6.7
0.5
cdI,
t.'l 1.6
Experimental results from sputtering of alkali halides with a 6 keV Xe+ beam. Ref. 23' Experimental result lrom sputtering of Ag with a 6 keV Xe+ beam. Ref. 24.
3.00b 0. r4 0. t4
see
I É
7
M. SZYMOÑSKI, H. OVEREIJNDER AND A. E. DE
r96
compound surfaces. Only
if
VRIES
we assume that the surface is restricted, the energy distribution of these particlès
voor Zuiver Wetenschappelijk Onderzoek. One of us (M.Sz.) is indebted for the hospitality and the support by FOM.
leaving the surface can be described by the observed distribution of Eq. (6). Large differences in the vapour pressure between the target components can induce stoichiometry changes on the surface with an enrichment of the less volatile element. This type of selective sputtering can be caused by spikes as well as by thermal
REFERENCES
motion
of the
evaporation
of
particles
on the
l. 2.
the Physics oflonized Gases, Dubrownik, 1976), p.461.
3. M. W. Thompson, Påil. Mag tt,377 (1968).
4. P. Sigmund, Phys. Reu. lt4, 383 (1969). 5. P. Sigmund, Reu. Roum. Phys. 17, t01.9 (1972). 6. N. Andersen and P. Sigmund, Man Fys. Medd. Dan.
decomposed particles. Since we
observed a large silver enrichment on AgBr where a thermal spike does not occur, the stoichiometry change apparently is caused mainly by the evdporation process. So it seems reasonable to ascribe the
selective sputtering
7. M. W. Thompson and R. S, Nelson, PåiL Mag. 7,2015 (r962).
ll,
of halides and oxydes to the
same mechanism. Finally a few remarks are needed about sputtering of molecules, In Table I we see that besides a statistical mechanismrt both thermal mechanisms, spikes
and evaporation of molecules, are necessary to explain the experimental results. We would like to stress that sputtering of molecules from spikes is a favour'able process for chemical compounds because the molar heat of vaporization is comparable or even
lower than the heat of vaporization for atoms.ro From Figure 9 is is clear that direct sputtering of molecules by momentum transfer from the collision cascade must be taken into account for I, molecules. ACKNOWLEDGEMENTS We wish to thank Dr. P. Sigmund for very valuable discussions abou Kistemaker, ProL J. Los and Dr. interest in this work, helpful discussions and a careful reading of the manuscript. The work is part of the research programme of the Stichting voor Fundamenteel Onderzoek der Materie and was made possible by financial support from the Nederlandse Organisatie
Vid.
Selsk. 39, no. 3 (1974).
8. R. S. Nelson, P/¡í1. Mag. 291 (1965). 9. R. Kelly and N. Q. Lam, Rad. Etfects t9,39 (19?3).
H. M. Naguib and R. Kelly, Rad. Ellects 25, | (1975). ad. EJfects 25,19 (lg1-i). t2. H. L. Bay, H. H. Andersen, W, O, Holer and O, Nielsen, Nucl. Instr, Meth. 132,301 (1976). 13. R. Kelly, Ãad. Elfects 32,91 (1977). t4. J. Politiek, P. K. Rol, J. Los and P. G. Ikelaar, .Reu. Sc. 10.
I
l. H. M. Naguib and R. Kelly,.R
Instr.39, I 147 (1968). 15.
H. Overeijnder, M.
Szymoúski,
A. Haring and A. E.
de
Yries, Rad. Etfects 36, 63. 16. J. Politiek and J. Kistemaker, Rad. Ellects 2, 129 (1969). 17. M. Szymoriski, R. S. Bhattacharya, H. Overeijner and A, E. de Vries, submitted to J. Phys. D. 18. G. P. Können, A. Tip and A, E. de Vries,.Rad. Elfects 26, 23 (te7s).
19. P. Sigmund, to be published in Inelastic Ion-Surface Collisions, ed. by N. H. Tolk, J. C. Tully, W. Heiland and C. W. White, Academic Press Inc., New York. 20. L. T. Chadderton and L Mc C. Torrens, Proc. Roy, Soc, 294A,80, e3 (1e66). 21. H. Overeijnder, M. Szymoñski, A. Haring and A. E. de Yries, Phys. Stat. Sol, Btl, K11 (1977). 22. P. Sigmund, Appl. Phys. Lett. 25, 169 (1914). 23. H. Overeijnder and A. E. de Vries, in Physics of Ionized Gases 1976, ed. by B. Navinðek (Proc. VIII Int. Symp. and Summer School on the Physics of Ionized Gases, Dubrownik, 1976), p.461. 24. M. Szymoñski and A. E. de Vries, Plys. Leît. 63A,359 (r911't.
