Professional Documents
Culture Documents
KFK4221
KFK4221
Mai 1987
Kernforschungszentrum Karlsruhe
KERNFORSCHUNGSZENTRUM KARLSRUHE
KfK 4221
ISSN 0303-4003
Contents
page
I. lntroduction 1
II. Experimental 2
IV. Conclusions 17
v. References 18
Abstract
Auger electron spectroscopy (AES) is often used for the determination of the
quantitative composition and the speciation of the materials under investigation.
The present work Blustrates the quantitative analysis of titanium oxides and its
application to the characterization of a partially oxidized titanium overlayer an
top of a copper substrate. The quantitification procedure described provides an
average composition of the surface area investigated. Care has to be taken to
convert this composition into an information about the chemical species present at
the sample surface because the observed fine structures of the AES spectra cannot
be correlated unambiguously with the chemical species present. The AES spectra
are not suited to distinguish conclusively between a homogeneaus composition of
the analyzed volume and the presence of a mixture of several titanium species.
Additional information is required for a final distinction between the two
possibilities, which can be achieved, for example, by X-ray photoelectron
spectroscopy (XPS).
Zusammenfassung
One of the important examples for the application of AES in materials science is
the investigation of titanium meta!, of its alloys, and of its reaction products with
various reactants. These investigations are of interest in context with bonding and
adhesion properties of titanium containing materials (1), with the corrosion
resistance of titanium alloys (2), and with the reaction (e.g. gettering) behavior of
titanium in generat (3) and for the production of wear resistive cornpounds (1 a,4).
More basic investigations using AES dealt with problems of the speciation (3 b,5)
and quantification (3 c,6) of titanium meta! and its oxides and with the explanation
of specific spectral features (7). Additional work was performed for understanding
the influences an the AES spectra of electron beams used for the excitation of AES
spectra and of ion beams used for the removal of surface near material by
sputtering (6 c,B).
major aim of this publieation is to show that this approaeh is not straight-forward
as soon as the lateral sample eomposition is heterogeneous. In this ease the
interpretation may be erroneous unless independent methods e.g. X-ray
photoelectron speetroseopy (XPS) permit additional eharaeterization.
II. Experimental
Titanium coated copper plates with a thiekness of the titanium overlayer of about
one mierometer are used as support for the X-ray sensitive polymer (polymethyl-
methaerylate) in deep-etch synchrotron radiation lithography (11). Ta get a
suffieient adhesion of the polymeric mierostructures, whieh are produced by
synehrotron irradiation and subsequent development, it is neeessary to pretreat the
support by chemieal etching with alkaline hydrogen peroxide solution. It has been
observed that the adhesion properties of these microstructures depend strongly on
eteh time.
To investigate the reason for the etch time dependent adhesion properties the etch
time has been varied systematieally and the effeets of etching on structure and
chemistry of the titanium overlayer have been followed by AES depth profiles (12).
The observed Variations of the depth profiles made it worthwhile to examine
whether or not the AES spectra were suited for a quantifieatinn and speeiation of
the overlayer after etehing.
in doubt whether this treatment can eliminate line shape broadening effects with
certainty. In cantrast to the powder sample, the titanium dioxide overlayer formed
by reactive sputtering showed no eherging even without any pretreatment. ln
addition, an XPS analysis as weil as X-ray diffraction indicated that substantial
differences in composition between the powder and the thin layer sample were
absent (9).
Titanium metal samples were made from 0.5 mm thin sheets. Any surface
contaminations (adsorbed hydrocarbons, oxidic overlayer) were removed by ion
bombardment.
The AES and XPS measurements were performed in a Vacuum Generators (VG)
ESCALAB 5 electron spectrometer described in detail elsewhere (10). Sampies of
appropriate size were sticked to meta! probes by adhesive and conducting silver
paint, and were measured after introduction into the ultra-high vacuum system as
soon as a base pressure of 1o-9 mbar was achieved.
Auger spectra were recorded in the derivative mode at a constant retard ratio of
ten, a modulation valtage of 2.0 eV and a time constant of the lock-in amplifier of
300 ms. Spectra acquisition and storage was controlled by a POP 11/03 computer.
The measured spectra resulted from the summation of several repetitive scans with
a step width of 0.2 eV per channel and a scan speed of 17 channels per second.
Auger electrons were excited by electrons with a primary energy of 5 keV, a total
current of 2.0 11A and a current density of 33 I-lA mm-2. The electron gun used was
a VG LEG 100.
XPS spectra were recorded by excitation with Al K CY. radiation (100 W X-ray
power) in the constant analyzer pass energy mode. The pass energy was choosen to
be 20 eV for the acquisition of the spectra of the major photoelectron Iines. The
spectra were digitized with a step width of 0.1 eV and a scan speed of about 25
channels per second. Normally, the spectra represent the sum of several repetitive
scans.
This chapter gives an example of a depth profile acquired after etching a titanium
coated copper plate for about five minutes. After a short description of the in-
depth distributions as compared to those of an unetched sample, the following two
chapters illustrate the approach choosen for quantification and speciation and its
short-comings with regard to the sample described in this chapter.
Fig. 1 a shows the depth profile of a Ti/Cu sample before etching. The in-depth
distributions of the main constituents of the sample are represented by the peak-
to-peak intensities of the corresponding differentiated Auger lines. The profile
directly reflects the layered structure of titanium on top of copper. 8oth elements
exhibit nearfy constant signal intensities over the depth ranges of the overlayer and
the substrate, respectively. Very close to the surface one observes an increase of
the titanium signal intensities accompanied by a steep decrease of the O(I<LL)
intensity. This results from the fact that each titanium metal surface after contact
with air is covered by a thin titanium dioxide layer. This layer is quickly removed
by sputtering which Ieads to the observed variations of the signal intensities of
titanium and oxygen.
It is remarkable that the oxygen signal intensity and thus the oxygen concentration
does not vanish in the bulk of the titanium metal overlayer, but remains at an
easily detectable Ievel. This could, in principle, mean that during the formation of
the overlayer by sputter deposition oxygen has been incorporated.
However, the sputter conditions used for the overlayer formation are believed to
lead to a crystalline product which exhibits open grain boundaries (11 a,l3).
Therefore, not only the visible surface of the overlayer is oxidized during contact
with air but also those grain boundaries. within the overlayer, which are accessible
- 5 -
to the oxidizing agent. This will lead to the detection of certain amounts of oxygen
also within the layer. In this latter case, however, oxygen is not distributed
homogeneously, but concentrated at the oxidized open grain boundaries.
a)
____ ----------------...
..............
,..--, 1
"-
Vl
.....
c
::J
::>.. 3:0(KLL)
....
0
....
..... 0
.0
.... interface ,.,.,..... ...
+I . . ,.. . . --~-----""- - - - - 1
0 b)
L....J
r !i
.....::>.. I
I
Vl I
c -------overlayer------141----substrate---
-<1.1
c
1 2 3
I T T
I I
..---- ------
I
0 ~--------~--------~~--~~--------~----~
0 100 200 300 400
sputtering time [min]
Fig. 1 Depth profile of Ti/Cu samples (a) before and (b) after oxidation with
peroxide solution. One minute of sputtering time corresponds to the
removal of about 2.8 nm of surface material.
- 6 -
For comparison, Fig. 1 b shows the depth profile of another Ti/Cu sample after
treatment with peroxide solution for several minutes. Two major effects of the
treatment can be evaluated directly from a comparison of the two profiles. The
primary effect is that the overlayer has been reduced in thickness as indicated by
the shorter sputtering time needed to reach the interface. In addition, cornparison
of the signal intensities of the O(KLL) transitions in both figures shows a
significant increase of the oxygen content by the etching process, i.e., the
oxidative attack Ieads to a partial dissolution of the overlayer and to a residue of
undissolved titanium oxide.
The oxygen signal intensity passes a maximum close to the sample surface. It then
drops to a value comparable to that of the titanium Auger lines remaining constant
in the major part of the overlayer. The occurrance of a maximum indicates that
the visible surface of the overlayer after peroxide treatment is covered with an
oxidic layer thicker than the natural dioxide formed during air contact.
XPS spectra taken directly after introduction of the samples into the spectrometer
showed that the surfaces of the meta! and of the suboxides had been oxidized to
titanium dioxide. Only in the case of the meta! sheet a small additional signal was
observed, which could be attributed to titanium meta! originating from meta!
atoms underneath the dioxide overlayer. Therefore, three of the four samples could
not be used "as received" for referencing purposes. It was necessary to lay bare the
underl yi ng materials, which is favourably done by ion induced sputtering in the
ultra-high vacuum system of the electron spectrometer. However, titanium oxides
are known to undergo preferential sputtering effects, which Iead to the formation
- 7 -
of lower oxides in the surface near region. Titanium dioxide is partly reduced to
the threevalent and the divalent oxidic stete but not to the meta! (8),
Consequently, titanium sesquioxide should, in part, be converted to titanium
monoxide, while the oxidation stete of the latter is not affected by sputtering.
Fig. 2 shows AES spectra of the four reference materials. The spectra were
recorded under continuous ion bombardment and after sputter equilibrium had been
reached. It is clear that the spectra da not correspond to those of the pure oxides
Ti02, Ti203, and TiO, respectively, but represent only the result of a convolution
of several effects. However, the spectra are characteristic of the chemical (and
eventually the physical) state of the starting materials and of the treatment before
measurement of the AES spectra. In addition, effects of the electron beam (6 c) as
well as of instrumental parameters of the electron energy analyzer have to be
regarded as further parameters, which may influence the shapes of the Auger
spectra. It is, therefore, necessary to state that the experimental conditions have
been kept constant during the acquisition of all spectra.
For the quantitative evaluation it has been assumed that the intensity ratio
(termed Q) of the O(KLL) and the Ti (L3 M23 M23) transitions is proportional to the
stoichiometric ratio in each reference material. The intensity of each Auger peak
has been taken to be the peak-to-peak height without consideration of the course
of the background and of any fine structure of the transitions. The Ti (1_3 M23 V)
transition has been deliberately not used since the excitation probability of this
transition varies significantly with the chemical environment (5,6 c) rlue to the
participation of the valence band.
- 8 -
Ti Oz
II
.....U') Ti 2 03
c
::J
0
.....
1-
..c
1-
0
L,_J
Ti 0
.....
~
U')
c
.....Cl)
c
Tio
rfv'--~-
Fig. 3 shows the intensity ratios determined for the various oxides. The square data
points correspond to the surface states of the oxides before sputtering, which have
been shown by XPS to be titanium dioxide in all three cases. Those data points
have, therefore, been entered at a stoichiometric ratio of two. The dashed line
reflects the expected slope in case of a linear dependence of Q on the
stoichiometric ratio.
3.0 ..........
0 ... / II
-,...
.....
.2 ...........
a ......
-
I""') II- ......
......
N ......
......
......
I""') U) ......
- 2.0
N /
c
/~i
~ ClJ
I""')
....I c
~
1--
._ aCl)
c.. .....
......
. . . . !/
/
--
....I I ......
....I
~
0
1.0 .....
...... "'
I ......
......
~
0 {j ......
......
Cl) ......
/
c.. ......
/
......
......
......
......
0
0 0.5 1.0 1.5 2.0
0 I Ti stoichiometric ratio ( bulk composition)
The fuil circles give the Q values of the oxides in sputter equilibrium, and the full
line corresponds to the least squares fit of the data points. It is remarkable that
the variation of Q can be well represented by a straight line with nearly no
deviations. It is further remarkable that the value of Q of titanium monoxide
almost coincides with the value expected from an interpolation between zero and
the values of unsputtered titanium dioxide but that the deviation increases with
increasing stoichiometric ratio of the bulk composition. This can easily be
explained by preferential sputtering of oxygen because it can be expected that for
a higher oxygen content in the unsputtered sample the percentage of sputter-
reduced lower oxides is also higher.
Other authors have used the same approach for a quantification of titanium oxides
(3 c,6 c). The results of these investigations are summarized in Fig. 4 and, tagether
with our own measurements, in Table 1. oth groups observe a linear relation
between the intensity ratio Q and the oxidation stage of the reference materials.
This confirms that the empirical approach is a reasonable one. The slopes of the
three fitted lines (Fig. 3,4) are slightly different. In our data and in the results of
Mathieu et al. (6 c) the straight line does not pass through zero, while in the case
of Gandon's results (3 c) it does very weil. This may be an indication for the
absence of preferential sputtering effects in Gandon's investigation. However, we
believe that the observed differences lie within the error Iimits of the three
investigations. It is rather astanishing that, remembering the various influencing
factors on the surface composition of the samples (electron and ion beam effects,
sample preparation etc.), the results of the investigations agree so weil.
0 3.0
-
0
0
I-
~
("")
N
("")
-
::>..
I I)
c:
- 2.0
N
(!)
("") c:
...J
~
1- 0
(!)
.........
a.
I
--
...J
...J 0
~
1.0
I
~
0 0
(!)
a.
0
0 0.5 1.0 1.5 2.0
0 I Ti stoichio metric ra tio ( bulk co mpo si tion)
Fig. 4 Intensity ratio Q as a function of the bulk composition; values taken from
literature (3 c, 6 c). Different results of one author for a specific oxide
represent different sample preparation methods (cf. Tab. 1).
- 12 -
Fig. 1 b shows, as already described in Chap. III.1, the depth profile of the oxidized
Ti/Cu sample. The composition in the profile region with constant oxygen content
can be determined using the peak-to-peak intensities for calculation of the value
of Q. Fig. 3 then permits the calculation of the averege stoichiometry. The
relevant curve in Fig. 3 is the one, which is derived from the oxide reference
samples in sputter equilibrium since the value of Q being characteristic for the
composition of the Ti/Cu sample has also been derived during depth profile
measurement. The evaluation is illustrated in Fig. 5. The result indicates a
stoichiometry close to that of titanium monoxide.
3.0 .....
_..111
0 /
/ 111
/
0
...... / """
d
.... /
('V')
/
N /
>.
...... /
~ /
ooy
. / "'
('V') II)
.....
2:
!
Cu(L 3M"VI /.
~
N c 2.0 II I //
<U
...... 0
_.
10~111
('V')
c too 200 Joo
spultering time [ min) " ,.
~ II
1- 0 11
/v
/
"'/
I
......... <U
_. c. ----~--------;~I
_. 0
I
"" I
~ ...... 1. 0 .,.."' I
I ""/ I
~ / I
0
d // I
(!)
c. // I
-" I
// t
-" I
0 / I
a)
TiO
Fig. 6 .L\uger spectra of titanium monoxide (in sputter equilibrium) and of the
titanium overlayer an copper (after removal of about half of the overlayer
by sputtering, compare Fig. 1 b). The lower spectrum results from the
flUmmation of the Auger spectra of titanium meta! and titanium dioxide in
sputter equilibrium. The intensities of the two spectra were about equal
with regard to the intensity of the Ti (L3M23M23) transition.
- 15 -
When comparing the AES spectra from the plateau region of the profile with those
of titanium monoxide in sputter equilibrium (Fig. 6 a and 6 b) one observes no
significant differences even in those parts of the spectrum which vary
characteristically with oxidation stage (compare Fig. 2). A simple straight-forward
interpretation based on the experimentally determined stoichiometric ratio and the
peak shapes could, therefore, Iead to the conclusion that TiO has been formed over
the major part of the overlayer and that the substrate is covered homogeneously
with this oxide.
However, as discussed previously, the presence of open grain boundaries and the
Variation of the peak intensities of oxygen and titanium close to the Ti/Cu
interface indicate the possibility of an oxidation at the grain boundaries leading to
a heterogeneously distributed oxidation product (supposedly titaniurn dioxide). It
can easily be shown that the Auger spectrurn in Fig. 6 b is not only very similar to
that of TiO but that it can be equally well reproduced by the surnrnation of Auger
spectra of titaniurn meta! and titaniurn dioxide (in sputter equilibriurn) of
appropriate relative intensities. Fig. 6 c shows the result of such an attempt. The
intensities of the contributing spectra are in this case approximately equal with
regard to the intensity of the Ti (L3 M23 Mz3) Auger transitions in titanium meta!
and titanium dioxide, respectively. Obviously the ratio Q as weil as the peak shapes
of the sampte spectrum are weil reproduced by this method.
Fig. 7 a shows the Ti 2p XPS spectra of the Ti/Cu sample aftersputtering until the
plateau rElgion of the profile (Fig. 1 b) had been reached. For comparison Fig. 7 b
shows the XPS spectrum of titanium meta!. The binding energies of the Ti 2p
photopeaks of the upper spectrum are very close to the values of the metBI and are
definitely different from those of titanium monoxide, which are shifted by about
1.5 eV to higher binding energies (2,8). Differences between the two spectra are
indicated by the hatched area in Fig. 7 a. These differences correspond to
contributlons of titanium oxides (for a detailed evaluation compare Ref. 12). These
observations show that, indeed, the mechanism, which in based on an oxidative
attack of the peroxide solution at the open grain boundaries, must be the relevant
- 16 -
one. On the other hand, the AES measurements are suited only for the
determination of the integral composition of the surface analyzed. It can be
estirnated from these results that titaniurn rnetal and titaniurn oxide cover about
one half of the analyzed surface each (12).
a)
c
~
.....
.....
....,
.....
L,_"J
....,
~ b)
V)
....,
GJ
Tio
IV. Conclusions
It has been shown that Auger spectra have to be evaluated very carefully if one is
interested not only in the averege composition of the sample but also in a chemical
speciation. In the case of partially oxidized titanium the Auger spectra cannot be
interpreted unambiguously, even if details of the spectral features are additionally
taken into consideration. This is of course only true if the characteristic
dimensions of the heterogeneity are much smaller than those of the surface area
analyzed by AES. However, a reduction of the probed area is only helpful if one can
get access to homogeneously composed sample volumes. In other words, the
dimensions of the heterogeneity have to be known and, in addition, one must be
able to localize differently composed parts of the sample, which is only possible
with an appropriate instrumentation. This means that in the case of "unknown"
samples, there remains some uncertainty about the interpretation of the results.
V. References
3. a) M.C. Burrell and N.R. Armstrong, J. Vac. Sei. Techno!. Al, 1831 (1983).
b) N.R. Armstrang and R.K. Quinn, Surf. Sei. 67, 451 (1977).
c) J. Gandon and J.C. Joud, J. Less-Common Metals 69, 277 (1980).
5. G.D. Davis, M. Natan and K.A. Anderson, Appl. Surf. Sei. 15, 321 (1983).
6. a) M.C. BurreH and N.R. Armstrong, Appl. Surf. Sei. 17, 53 (1983).
b) G. Sieking and A. Eisenhuth, Ber. Bunsenges. Phys. Chem. 89, 1311 (1985).
c) H.J. Mathieu, J.B. Mathieu, D.E. McClure and D. Landolt,
J. Vac. Sei. Techno!. 14, 1023 (1977).
7. a) J.S. Solomon and W.L. Baun, Surf. Sei. 51, 228 (1975).
b) Bui Minh Duc, C. Jardin, J.P. Gauthier, G. Thailet and P. Michel,
J. EI. Spectrosc. Relat. Phenom. 20, 213 (1980).
c) S. Nishigaki, Surf. Sei. 125, 762 (1983).
8. S. Hofmann and J.M. Sanz, J. Trace Microprobe Techn . .!_, 213 (1982-83).