Surface Modification by Design (or Accident?)
Aston University
Wednesday 10th January 1996
The Group was welcomed by Prof. Mike Cardwell, the Head of the Department.
John Sullivan (University of Aston) opened the morning session
with a talk describing work done in his group on low energy, up to ~5keV, ion beam damage
in semi-conductors and insulators. The collision can either result in reflection of the
incident ion, giving rise to the possibility of analysis by ISS, or in penetration of the
ion which causes sputtering or internal damage. For low energy incident ions, the
sputtered atoms are emitted from within 1 - 2 atom layers of the surface whereas the
implanted ion can go as deep as 10's nm . The damage produced in the form of dislocations
can, however, extend for several microns into the solid.
At low beam current densities the disruptive effects of any single
incident ion are complete before the next ion impact. This means that Sigmund's linear
cascade theory should be applicable to the process and this model works for pure elements.
However, for alloys or compounds the impact energy is shared differently between the atoms
in the solid. Therefore, preferential sputtering is inevitable and the theory cannot be
used in a useful predictive manner. The relative elemental sputtering yields of atoms in
an alloy are not a good predictor of the likely amount of preferential sputtering produced
in the alloy.
John has initiated a program to investigate the parameters that affect
surface composition changes of metal oxides under ion bombardment in order to investigate
a more appropriate theory due to Malherbe et Al. The amount of reduction suffered by the
oxides is found to be independent of incident ion energy but is related to the beam
current density used. Charge is also a significant factor since ions cause about twice the
reduction produced by a similar dose of atoms. These factors lead to the conclusion that
preferential sputtering is a combination of cascade sputtering and Gibbsian equilibrium
segregation. In this case a term related to the relative elemental surface energies
(surface tension modified by surface stress/strain effects) is found to be a better
predictor of preferential sputtering composition. The mechanism of Gibbsian segregation
requires self diffusion. However, at room temperature this is too slow to accomplish
equilibrium so it is clear that ion bombardment increases the vacancy and dislocation
population in the near surface region and thereby facilitates greatly enhanced surface
diffusion. Since the mechanism proceeds to thermodynamic equilibrium, only those compounds
that are thermodynamically stable are produced during sputtering, for example Al2+ is
never produced during reduction of Al3+ to Al0.
Graham Beamson (RUSTI, Daresbury) discussed the mechanism and
measurement of damage in polymers such as PTFE and PET under X-ray irradiation. Spreading
thin polymer films onto high secondary electron emitting substrates greatly increases the
degradation rate so it is concluded that the damage process is mediated by photo- and
Auger secondary electrons rather than the incident X-ray photons directly. Bremstrahlung
exacerbates degradation by generating both higher energy electrons and more low energy
secondaries. Thermal radiation from the X-ray source is not a serious problem, nor are the
very low energy, <10eV, flood gun electrons used for charge neutralisation with
monochromatic sources.
The breaking of a C-H bond in a polymer produces a radical chain
reaction which can only be terminated by mutual annihilation of two free radicals in
forming a C-C bond. This leads to cross-linking in the polymer and the formation of higher
molecular weight material. Alternatively, the breaking of C-C bonds can lead to production
of molecules of lower molecular weight which may be sufficiently small to be volatile in
vacuum, causing a pressure rise in the spectrometer. For example, PTFE emits CF4 and C2F6
fragments under X-ray irradiation.
Since these are chemical reactions, the rate of degradation can be
slowed by cooling. Alternatively, the degradation rate can be increased when there are
additional elements in the polymer which are comparatively strong X-ray absorbers, e.g.
fluorine. This leads to increased electron generation in the near surface region,
particularly if the element has an absorption edge just below the X-ray energy. However,
this propensity for increased degradation has to be balanced against the stabilising
effect of the element in quenching free-radical reactions. For example, a highly
fluorinated polymer should degrade faster than a less fluorinated variety due to the
absorption effect. This is not always the case experimentally because there is less
opportunity for radical reactions to take place. It is found, however, that -CF3 end
groups are formed from broken C-C bonds and the average molecular chain length is reduced.
Polymers containing aromatic ring structures are most stable, as might
be expected. Desorption of incorporated elements such as halogens depends on how long a
side chain the element is attached to. If it is an end group, one bond has to be broken to
release the halogen. If it is on a short chain, say C4, then breaking any one of the bonds
in the chain will probably result in loss of the halogen, albeit with some very short
molecular chain attached.
Ian Gilmore (NPL, Teddington) described a novel modification to
quadrupole SIMS which enabled a detailed study of the process of degradation of polymers
under ion bombardment. The value of XPS is limited in this application because information
on chain length and cross-linking cannot be extracted from the C1s or valence band peaks.
The modification involves placing an additional neutralisation plate in front of the
quadrupole. This is bombarded by a 1keV electron beam to produce secondary electrons which
compensate for charging of the sample. The essential feature is that a bias oscillating at
6.5kHz is applied to this plate and the sample so that the energy of the emitted ions is
swept. This has the effect of making the quadrupole insensitive to the different energy
distributions of the emitted ion and molecular fragments. As a consequence, accurately
reproducible spectra are obtained from any polymer and the variation as a function of ion
dose can be monitored quantitatively.
Ian showed curves of the fragment intensity variations with time of
bombardment, i.e. ion dose. Some fragments were unusually sensitive to degradation -
undergoing a 10% change in intensity after an ion dose equivalent to 1 incident ion per
3000 surface atoms. The damage in PET was found to occur at ~3 times faster than PTFE even
though this is opposite to the case in XPS.
In order to understand the formation of the fragments detected and
their temporal variation, Ian has devised a model based on the probability of multiple
bond breaking. As an example, to generate a C3 fragment from a polymer side chain may only
require one bond to be broken and sufficient momentum transfer in order to eject it from
the surface. In contrast, to produce this fragment from the backbone of a polymer requires
two bonds to be broken simultaneously from next nearest neighbour atoms without the bond
between them being broken. Sufficient momentum has also to be transferred to the fragment
at the same time. The probability of these cases can be calculated and the effect of
fragments produced by one impact being ejected subsequently by another could also be taken
into account. When this was done, the fits to the fragment intensity variation curves was
found to be excellent. The most easily produced fragments decay fastest and fragments
requiring prior damage grow in intensity from time zero through a maximum.
It was also found that there were limits to the size of fragments that
could be produced by this mechanism and this limit was related to the range over which a
single impacting ion can transfer its energy.
Roger Webb (University of Surrey) showed video recordings of
Molecular Dynamics Simulations of ion impacts on single crystals. The technique involves
solving Newton's Law of motion for a many body system. With past generation computers,
simple pair-wise potentials were used and systems of up to a few hundred atoms could be
solved in a few tens of hours. The latest machines use many body potentials applied to ten
of millions of atoms such that the deformation of a small chunk of material impacting on a
solid can be simulated.
By far the greatest majority of ions simply follow channels into the
solid, undergoing low impact events which produce no sputtered species and cause hardly
any disruption to the lattice. Occasionally, if the ion impacts close to a surface atom,
some sputtering occurs and considerable disruption of the solid results. Very rarely, a
high impact event occurs where the ion hits very close to an atom in the solid. This can
result in a large number (30-40) of sputtered atoms for one incident ion and produces vast
amounts of lattice disruption. Events which produce sputtered atoms are very rare but they
dominate our interest because it is sputtering that we are most interested in during, for
example, depth profiling.
The impact of a single ion on graphite was shown. As the ion penetrated
the crystal an acoustic wave was produced which radiated out from the impact site.
However, the weak bonding between the planes of graphite allowed the incident ion to
scatter in the gap between the planes and come to rest leaving a bump on the outer
surface. These bumps have been detected using STM but were not explained until this
simulation showed what was happening.
As a finale, Roger showed the impact of a large copper cluster onto a
surface which caused tremendous disruption. He also showed a carbon "Buckyball"
impact which was seen to bounce off a graphite lattice but stick to a silicon crystal.
Chris McConville (University of Warwick) described how low
energy electron loss spectroscopy, HREELS, could be used to characterise damage in III-V
semiconductors by studying the variation in plasmon energy as a function of depth. In
these materials the plasmon energy is related to the compound present or the effective
dopant concentration in the locality where it is generated.
From the commercial viewpoint it is important to produce clean, ordered
substrates upon which to grow specialised semi-conductor materials for device manufacture.
For {110} surfaces, cleaving in vacuum is sufficient. Clean {100} surfaces can be grown by
MBE techniques but some materials require ion bombardment cleaning and thermal anneal
cycles at close to their melting point to produce the desired surfaces. In these cases,
the dopant concentration is synonymous with lattice damage.
In HREELS the electron interacts with the solid through long range
dipole fields and, even though the electron penetration is very low, it is possible to
detect the consequences of plasmon generation in the material through peaks with extremely
small energy loss, ~100meV. In addition, the depth at which the plasmon is produced is
dependent upon the incident electron energy so the sampling depth can be varied over
~200nm for an electron energy up to 200eV.
Chris showed the small but clearly significant differences in the
optical phonon and conduction band electron energy losses on InSb{100} resulting from
oxidation. He had also "depth profiled" the oxidation level by sweeping the
incident electron beam energy.
Two experiments are possible to investigate the effects of damage.
Initially, the electron interaction depth can be fixed and the annealing temperature
increased after ion bombardment. Alternatively, the electron energy can be ramped. Using
both approaches gives information about the rate and mode of recrystallisation and changes
in the carrier concentrations as a function of depth. Chris showed that, on InSb, ion beam
damage always produces n-type material close to the surface and, since the first region to
anneal is the surface, a brief anneal can lead to a p-n-p type structure extending over
100nm depth.
The afternoon session was devoted to workshops and an introduction to
the UK Surface Analysis Forum World Wide Web page on the Internet. Anyone with access to the
Internet will find the site at
http://www.surrey.ac.uk/MSE/ESCA/ESCA/home.html
(the capital letters must be entered as shown).
This site has been set up on the initiative of Dr Simon Morton (Webed.
Fame at last ;-) (E_mail address S.Morton@uksaf.org)
who responded to a suggestion at one of the recent workshops that a web site be set up as
an information resource in surface science under the auspices of the UK Surface Analysis Forum.
Simon explained how the site was organised and demonstrated the features already
available. After only a very short time Simon has produced an extremely professional,
informative and technically useful facility which all members are urged to inspect in
their own time. Meeting reports and future meeting information and registration forms will
be available on the page to download directly.
Workshop Reports
Leader and Rapporteur Dave Sykes (ISST, Loughborough)
"Non-destructive depth profiling" was the title for the
workshop, "a discussion of how to get reliable depth information by XPS and ion
profiling" said the programme. I was really looking forward to this workshop which I
imagined would go along the lines........
Leader: Non-destructive depth profiling
Audience: You can't do that!
Leader: Oh yes you can!
Audience: Oh no you can't!
Then the leader would explain to a rapturous audience how it could be
done..... that was until I was asked to lead the workshop!
We got off to a good start by agreeing that we really needed to
re-define the title of the workshop, we were of one mind in thinking that non-destructive
ion profiling was a contradiction in terms. We had heard, in the morning sessions, how ion
beams damaged surfaces, both chemically and physically and that even exposing the sample
to the X-ray excitation necessary for XPS analysis caused sample damage in some systems.
Against that background our topic was a non-starter. We discussed
whether non-perturbing depth analysis was possible and, having considered angle resolved
XPS, variations in photon energy and ion scattering (RBS, MEIS), we came to the conclusion
that in some cases it may be possible.
Could we get reliable information by XPS and ion profiling?
Again we were not certain about this but felt that sometimes we could
get useful information. In general, there was a feeling that depth profiling in XPS was a
last resort, something one would not choose to do, but, if the customer insisted and was
prepared to pay, then..... and sometimes there was a useful result at the end. A pragmatic
view, from a user with a range of techniques available, was that given a sample from which
depth information was needed, choose SIMS first, Auger second but XPS only as a last
resort, a view I have some sympathy with.
Another faction of the audience, however, chose XPS every time. The
question of depth resolution and depth range over which one might profile in XPS was
raised and it seemed that our requirements were based on an acceptance of what instruments
were capable of delivering rather than demanding more of them. For example, the depth one
might sputter to was limited by the time taken to etch a crater large enough for XPS
analysis rather than the need to profile a given distance; the depth resolution that was
acceptable was defined by crater geometry rather than the fundamental limit of atomic
mixing.
So, at the end of the day, can we get reliable depth information by XPS
and ion profiling? No one was sufficiently confident to say "YES, YOU CAN!",
equally no one was shouting "OH NO, YOU CAN'T!". It was more a case of not being
able to show that the information was not reliable.
| Leader |
Len Hazell (CSMA Ltd) |
| Rapporteur |
Kathy England (University of Manchester) |
The workshop started with a brief discussion of current ideas on the
cause of damage under X-ray and electron bombardment. This was an extension of the results
presented in the talks in the morning session and included a description of Auger induced
ion emission by electrostatic field effects in ionic solids. All damage mechanisms were
discussed, including mechanical and direct vacuum desorption.
The worst case situations were where high irradiation dose was required
because long acquisition times were necessary, e.g. looking for low levels of elements in
small areas and/or take-off-angle studies, particularly on very sensitive materials such
as thin polymer films on metals.
Methods of acquiring spectra with the minimum possible X-ray dose were
discussed. Use of largest possible area at the poorest acceptable resolution with the
minimum settings of instrument "settle times" were recommended. Reducing the
sample temperature was useful provided adsorption from the vacuum was not a problem.
Acquiring and keeping each individual sweep and only adding together those that showed
acceptable damage had been tried. Datasystems do not allow this routinely but it is a way
of inspecting the data before quantification.
Use of monochromators was universally accepted as the most advantageous
way of damage limitation provided the flood gun energy is a low as possible and definitely
<10eV.
On large area samples that are "known" or expected to be
uniform over the surface it is possible to use small spot X- ray sources and move the
sample between acquisition of each spectral region or sweep. This "fresh
sampling" strategy would give the minimum possible amount of degradation but there
was some scepticism as to whether the quantified data would be acceptable. It was felt
that the initial assumption of surface uniformity may have to be previously justified,
although conventional large area sources effectively average any spatial variation out in
a similar way.
There was similar scepticism as to whether post acquisition correction
could be carried out in any scientifically valid way after the rate of degradation had
been established on a similar material. It was pointed out that the rate would have to
determined on the same material and establishing what this was required damage free
analysis in the first instance.
Finally, data processing methods such as deconvolution were discussed
which would enable poorer data quality to be tolerated.
In summary, it was agreed that samples should always be checked for the
likelihood of rapid degradation and a strategy such as "fresh sampling" adopted
to minimise this in the first place. As always, the minimum amount of data manipulation is
advisable and it was generally accepted that comparison of good and bad samples was the
safest way to proceed in commercial analysis.
Leader & Rapporteur: Bob Wild (IAC, Bristol)
This workshop was organised to help with the production of two proposed
standards for Instrument Performance Description in AES and XPS. The scene was set by
describing the progress during the last year. All the major instrument manufacturers had
been approached with a request that they provide a description of their current methods of
specifying instruments. Specifications had been received from all the manufacturers and
the information used to produce tables showing where there was and where there was not a
common approach. These tables had been used during discussions at the last ISO meeting to
determine where a specification could be accepted. Following that meeting two draft work
items have been produced for both AES and XPS. These were considered by those present at
the workshop.
Items discussed included the anode power to be used for maximum counts.
It was felt that a power should be used at which the anode would be guaranteed for a given
period, but how was this to be specified. Performance drift was also discussed with some
wanting data over short periods of say ten minutes while others were more concerned with
long periods of, say, 12 hours. A number of useful comments were made during the workshop
and the draft work items will be modified prior to sending out to member countries. |
