Surface Science 67 (1977) 373-392 0 North-Holland Publishing Company
QUANTITATIVE AUGER ANALYSIS OF GOLD-COPPER-OXYGEN AND GOLD-NICKEL-OXYGEN SURFACES USING RELATIVE SENSITIVITY FACTORS
P.M. HALL and J.M. MORABITO Bell Telephone Laboratories, Incorporated, Allentown, Pennsylvania 18103. USA Received 21 March 1977; manuscript received in final form 17 June 1977
Quantitative Auger analysis is becoming routine for unoxidized binary aUowys. Many specimens of technological importance, however, contain large amounts of oxygen, which can complicate quantitative analysis by changing (a) peak shapes (e.g. line widths, 6), (b) energy position of the peaks, (c) Auger currents, and (d) the sputtering correction factor (R). This paper describes quantitative Auger analysis of the gold-copper-oxygen and gold-nickeloxygen systems, based on relative sensitivity factors before (P,,I) and after (qe,) sputtering. The relative sensitivity factors, eel were found to be independent off the gold and oxygen concentrations for both systems. In the gold-copper-oxygen system this is attributed to the sharpness of the 920 eV (LMM) copper transition compared to the energy resolution (0.6%) of the CMA analyzer used. For the gold-nickel-oxygen system it is ascribed to a fortuitous cancellation of two effects (1) increased line width, 6, hence a decreased 848 eV (LMM) peak height upon oxidation, and (2) increased Auger emission of the 848 eV peak due to oxidation. A single crystal standard of Cu20 was used to confirm the 2 : 1 ratio of Cu to 0 in the ternary samples. The sputtering correction factors were found to be R(0, Cu)*= 1 .O for the composition of Cu20 and R(0, Ni) = 0.72 for the composition NiO. The variation in R with Au concentration is negligible. The major limitation on the quantitative Auger arualysis of these ternary systems has been found to be uncertainty in the sputtering correction factor R. Nomographs for calculating concentrations for constant Co/CC, and CQ/CNt using eel are also presented.
1. Introduction Quantitative Auger electron spectroscopy is most commonly [l-lo] performed using relative sensitivity factors and peak heights in the derivative (d(EN)/dE) spectrum. There are at least two (Prer, p”,1) different relative sensitivity factors defined for such analysis. Pier is generally defined [l] as Prer(A, B) = VA/CA W'L&'B> ,
where PA and PB are the peak heights of prominent peaks associated with the elements A and B, and CA and cn are the actual concentrations (in atomic fraction) of A and B. These concentrations are assumed uniform within the volume (escape depth times beam area) being analyzed. It should be noted that P/C ratios can 373
defend on concentration [I 1-l 33, so P&A, 3) is not necessarily equal to aeon, as is commonly assumed [14,15]. (The superscript “p” refers to a pure element.) Three par~ete~ which are known to affect the relation between P&A, I?) and ~~~~ are (1) the ratio of escape depths ~~~~~~ [3,16] I (2) the ratio of the backs~atte~ng correction factors (1 + r&l + tn) fl I ], and (3) the total number of atoms per unit volume (nA + nn) [13,1? 1. If any one of these three is a function of conce~tm~on, then Pret is not equal to ~~/~~ In ad~tion, if either ~~~~~~ or (1 + r&/fI + rn) depend on co~~entmtio~, ~~~~~A, B) will also depend on co~ce~tration. Ion beam sputtering is routinely used in Auger analysis, and this can change the surface concentration of an initi~~y ~iform specimen. Let Ci and Cb be the concent~t~ons at the surface (escape depth region) after sputter~g, and Pi arid P”uthe peak heists after sputte~~g. Then the same value ofPrer can be used to relate the concentrations to the peak heights: G&A, I9 = ~~~/C~~~~~~lC~,
(if the concentration varies significantly within the escape depth region, then Cfh and C& represent the average concentra~ons.~ It is very often of interest to relate the peak heists after sputtering to what the ~on~e~~atio~s were before the sputtering. This can be done by ~trod~~ing a new factor, P&r, which is defined by
&+orneqs. (2) and (3) one can relate the two relative sensitize %~A, B) = J’RICA,31 R(A, B),
factors by (4)
where R(A, I3) = ~C~/C*)~~C~C B) is called the sputter correction factor . Because of this relation, P& is called the sputter~o~e~ted relative sensiti~ty factor. it is clear that P&i incorporates any va~a~ons in the sputter correc~on factor if R lapins to depend on concentration. The factor R has been ide~tl~ed with a ratio of sputter efficiencies ]4,S] t but this is not necessary for the present work. Since stutter ef~~iencies can depend on concentration, no attempt was made here to interpret R in terms of sputter yield for pure elements, In a binary alloy, since CA f Cn = 1, it follows by de~nition from eq. (1) that one can find the concentrations from the peak height ratios using the expre~o~ fV S] CA = [I + (PAPAS ~~~I(A,@] -‘. Si~arly,
if peak heights are measured after sputte~g,
CA= fl f ~~~~~~~~,~A, B)]-“.
one fmds from eq. (3) that (6)
Eq. (6) is only valid, of course, if CA does not change appreciably in the distance req~red for the preferential sputte~~g effect to reach steady state. cadence is rapidly a~c~mulat~g [I ,6--S, 12-14,17,I9,2~] to show that both
P.M. Hall, J.M. Morabito /Auger analysisof Au-Cu-0
Prer and P:er are independent of concentration for many (but not all) binary metal alloy systems, and this knowledge is being used to simplify quantitative Auger analysis. There are many times in surface analysis, however, when other elements are present (most commonly oxygen) and the experimenter is faced with (at least) a ternary analysis. This is complicated by the possibility that the N(E) peak shapes may be different when the element is in an oxidized state. Since oxidation can affect relative sensitivity factors , it would be desirable to develop a method to obtain “oxide” relative sensitivity factors, analogous to those found in the literature [ 151 for pure metals. The present study describes such a method. In a ternary case, even if one component is oxygen, the relative sensitivity factor definitions (eqs. (1) and (3)) can be used unchanged without making assumptions as to matrix effects which might cause the sensitivity factor to vary with composition. Eqs. (5) and (6) must be generalized, using C,, + Cu t Co = 1. The result is CA = a.% P;[email protected]
Cc = aps: PL(B, C),
where a = [Ps, + PiP$,(B,
A) + P;P$,(B,
The quantity a is simply a normalization constant which, at least to first order,
remains independent of composition if experimental conditions are held constant. These equations, found in slightly different form elsewhere [9,15,18], provide the basis for ternary analysis. Note that only two sensitivity factors are required, since by definition (eq. (3)) PLt(A, C) = P&r(A, B) X PS,I(B, C). Also note that any effect which increases all peaks proportionally has no effect on the calculated value of CA, Cu or Co. Also note that these equations guarantee CA t Cu t Cc = 1. The two relative sensitivity factors (and their dependence on composition) must now be evaluated for the ternary case. The present work describes such an evaluation for the cases of Au-Ni-0 and Au-Cu-0.
2. Experimental The analysis was performed in a commercial scanning Auger microprobe (Physical Electronics Industries, Model 545). The sputter-ion beam at 2 kV and 5 mA had a 2 mm diameter and impinged on the sample at about 30” incidence (from the normal). The primary electron beam at 5 kV and about 1 PA had a 10 pm diameter and 60” angle of incidence. The energy resolution of the CMA analyzer was 0.6% and the modulation amplitude was 5 V peak-to-peak. Samples were prepared from multi-layer e-gun evaporated films of Ti-Ni-Au and Ti-Cu-Au. The original thicknesses and deposition rates are given in table 1. The substrate temperature during deposition is estimated to be within 20°C of room temperature. These films were then heated in air, with the result that the Ni and Cu came through the Au via gram boundary diffusion  and covered the Au
Table 1 Cbara~teristi~
Ti Ni Au Composite
of films as deposited Thickness w
Deposition rate (A/set)
200 2350 3400 5950
16 6 26 -
Ti CU AU Composite
-_ Deposition rate (Afsec)
100 3860 2180 6140
66 22 37 -
surface via surface effusion. As the ~deriying metal reaches the surface, however, it becomes oxidized. Thus, the cornFosit~o~ profile staring from the surface ~cl~des: first a region of Ni (or Cu) oxide, then a tr~sit~on resow where the Au signal becomes me~urabie and increases as the oxide context decreases until a region of (~most~ pure Au is observed, fo~owed by a gradual transition to pure Ni (or Cu>. This type of sample makes both the binary and ternary systems ava~abie in one sample, accessible without refocussi~g or readj~ting the Auger CMA analyzer in any way. Each successive layer is merely exposed by ~on~p~tter removal of the layers above it.
3. Results 3.1. Au-Ni-O Fig. la is a plot of the unprocessed peak he~~t data of the de~~~ve spectrum for a Ti-Ni-Au film aged in air at 35O*C for ten hours, The peaks monitored were the Au -69 eV, 0 -511 e’?, and Ni -848 eV. Auger data from unsputtered surfaces are not generally reliable because of surface ~o~t~na~on. In our case there were 1 cm peaks of carbon and sulphur before sputtering. taken after these peaks ~sap~ared are shown in this paper. Rata from fig. la for the first three minutes of sputtering are re-piotted in fig. 1b. The Ni intercept was obt~ed by a least squares fit with a ~o~el~tion coefficient of 0.989. The curve for 0 was obtained by a me~od explained later, These intercepts should be viewed with some caution, as transient effects might have occurred before the data were obt~ed. Once can safely say, however, that the sputter rate of the ternary system is less than that of Au, which was measured by the total sputter time. From this we conclude that the first points shown (at 0.2 and 0.4 min) correspond to a removal of less than 4 and 8 W [email protected]
The escape depth of these Auger electrons is of the order of 10 irl. Thus these intercepts are probably a reasonable appro~mation to what the unshuttered surface would have produced had it not been contami-
P.M. Hail, J.M. Mcmbito /Auger
analysis of Au-CL-0
100 120 TIME (MINUTES)
Fig. 1. (a) Compositionprofile of T&M-Au from unproce~d
Augerdata. (b) expanded view
of first portion of (a).
nated. The ratio of these intercepts is 1.59 whereas the contaminated 0 : Ni peak height ratio was 1.62. In fig. la it appears that the Oxy peak height (P&) near the surface is proportional to the Ni peak height (P&i). (Fig. 2 is a plot of the 0 peak height versus the Ni
P.M. Nult, J.M.
analysis of Au-i%4
Fig. 2. Oxygen peak height versus nickel peak height white sputtering a Ti-Ni-Au
peak height, showing rather good correlation with the exception of the very surface (P& > 12). This is interpreted in the following way. The proportions region corresponds to a constant Ni-0 ratio. The steady state surface concentration, however, is not established untiI a few monolayers have been removed. Thus, the top part of the data deviate from proportionality. This region might be due to an adsorbed layer cont~~ng oxygen or to an approach to sputter “equilibrium” (Le. steady state). In preferential sputtering, one normally expects one component to decrease exponentially to a constant value while the other increases exponenti~y to some other value. In our case, both Ni and 0 decrease because the Au increases. The ratio Pb/P&i might still decrease exponentially, however. The solid curved line in fig. 2 is a representation of the equation P~~P~i = 1.14 t 0.45 exp(-t/1.5) where t is the sputter time in minutes, and P&i has been taken as the least squares straight line fit on fig. 1b. This equation was also used to generate the curve for P& on fig. 1b. The decay constant of 1.5 min would correspond to 30 A if the sputtered rate of the ternary composition were comparable to that of Au. In fact, it is probably closer to 15 a, which is reasonable for a sputter decay const~t . One would expect adsorbed layers to be gone before that, however. Thus, although an adsorbed layer cannot be ruled out, we interpret this effect in terms of preferential sputtering of the oxygen. Now there is only one form of Ni oxide [23,24] generally accepted (NiO), so it is assumed that the region of constant stoichiometry represents the monoxide (before sputte~g). Thus, the slope of fig. 2 (w~ch is 1.14) is equal to P&r (0, Ni),
P.M. Hall, J.M. Morabito /Auger analysisof Au-Cu-0
the relative sensitivity factor in the presence of sputtering (eq. (3)). The point marked “t” in fig. 2 is based on the extrapolations of fig. 1b and represents what the surface would have given without contamination or sputtering. It gives a value of P&O, Ni) = 19.6/l 2.3, or 1S9. Taking the ratio of P&/P,r gives f?(O, Ni) = 0.72 indicating that the oxygen atoms are preferentially sputtered. This value is not directly comparable to the ratio of sputter yields for pure elements because sputter efficiencies can depend on composition, and our value of R is strictly valid for only one composition (Cr&!, = 1). Near the origin in fig. 2 there is a slight deviation from the straight line. This is interpreted as an oxygen deficient region, being partially protected from the atmosphere by the layers above it. Fig. 3 is a plot of the peak height of Au versus the peak height of Ni measured while sputtering through the specimen. The upper curve (labeled oxygen-free) is for the Au-Ni interfacial region. This interface is so deep within the specimen that no 0 is measurable, and the straight line that results for this binary case confirms that P,,r(Ni, Au) is independent of composition, as described previously . The lower curve on fig. 3 represents the region closer to the surface, where the Au signal goes from 0 to 21 (cm) as the Ni and 0 steadily decrease. Here a region of linearity is observed representing the same data as the region of proportionality in fig. 2. This line has the same Au intercept as the oxygen-free curve, but its (extrapolated) Ni intercept is 14.0. The ratio of the Ni intercept for pure Ni to that of Ni
Fig. 3. Gold peak height versus nickel peak height while sputtering a Ti-Ni-Au
P.M. Hull, J.M. Morabito /Auger analp& of Au-G-0
and Au-N&-O surfaces
in NiO, based on the extrapolated line, is 23,0/14.0, or 1.64. The extrapolated line, however, represents NiO as modified by sputtering. A more fundamental number is the ratio of the Ni intercept in pure Ni to the actual intercept (‘dashed line, fig. 3) for NiO, which should represent unsputtered NiO. This ratio is 23.0112.0, or 1.92. It might be compared to the ratio of the number of Ni atoms per unit volume in Ni metal to the number of Ni atoms per unit volume in NiO, which, based on bulk densities, is 1.70. The discrepancy is only I3%, but this agreement is mostly fortuitous because it is based on the derivative peak heights, PA, not the Auger currents, IA, and closer inspection shows that the peak width changes during oxidation, which changes the proportionality between 1, and PA. This will be discussed in more detail later. In order to characterize the ternary system using eq. (7), two of the three P&I values must be known. We have already established a vaiue for P&(0, Ni) of 1.14. It has been demonstrated previously [ZS] that, of the sputtering correction factorR is independent of concentration, one can set P&r(Ni, Au) = P&/Pxu with very little change in apparent concentration gradient. This is the same assumption made by Hammer et al. [ 141 in their analysis of pseudo-binary systems. This ratio is the red* procal of the slope of the upper curve on fig. 3, which is I .08. Thus, as an interim approbation we set P&(Ni, Au) = 1.08. Then P&(0, Au) = P&(0, Ni) X P&(Ni, Au) = 1.23. Then one can calculate the atomic! concentrations CAu, Co, and CNr from the equations
a = [P& f P&J’~~(O, AU) + P~iP&~(O, Ni)] -r. Thus, with only two calibration constants any alloy in the ternary diagram carr be analyzed within the accuracy of the assumptions mentioned above. These equations are convenient for computation with a programmable pocket calculator. Once the calibration constants are stored, one need only enter the three peak heights in order to calculate the three concentrations and the concentration ratio (co/cNi)* Fig. 4 is a replot of the profiles shown in fig. 1, where the data are now expressed in atomic percent. To generate the curves on fig. 4, it was assumed that P&(Ni, Au) was equal to PP Ni/PPAu, @Oring the Sp~tte~g correction factor [l2] and any dependence of the backscatter ratio and escape depth ratio on concentration. It is very unlikely that these combined effects represent more than a factor of five. Therefore, as a “worst case” example, a reca~cu~a~on of these profdes was made using P&(Ni, Au) = 5 X P&/P& The results of these calculations are shown as dashed lines on fig. 4. Depending on the use that is made of such profiles, the difference may be important or insignificant. For example, in the plateau region the profile with the higher P& gives 2% Ni, whereas the lower P& gives 0.4%. Secondly, if the amount of Ni which has arrived at the surface is estimated as the area under the Ni curve (for
P.M. Hall, J.M. Morabito /Auger analysisof Au-Cu-0
Fig. 4. Composition profile using relative sensitivity factors to obtain concentrations in atomic percent for a specimen of Ti-Ni-Au.
sputter time from 0 to 20 min), then the higher P,S,rprovides an estimate of the amount of Ni as only 47% of that from the lower Pkr . Thirdly, at the Ni-Au interface, the major effect of using the higher P&J is to shift the apparent position of the interface from 109 to 128 mln (sputter time). If the absolute depth at which the interface exists is critical, then this approximation is not useful. In that case, however, the use of sputter time for a depth scale is not warranted either, especially in the region of non-dilute alloys where sputter rates are largely unknown. In diffusion measurements, the important parameter is the gradient at the interface. This gradient is only different by 14% between the two curves. Since diffusion measurements usually involve measuring a gradient before and after the diffusion , most of this error is cancelled out in the correction for the undiffused profile. In principle, absolute determinations of the profile require complicated deconvolutions of the data by Fourier transformations or iterative numerical procedures . Most of the problems can be.adequately corrected for by a very simple deconvolution , however, and given the accuracy of currently available profiles, the time required to further refine the data reduction is normally not justfied. The value of B which results from the simple deconvolution is quite insensitive to the value used for PAr. Thus, depending on what use is made of the profile, the unknown factor may or may not be significant. The oxygen data have been omitted from fig. 4, since they follow the Ni so closely. A display of this fact is presented in tIg. 5, which is a plot of C&‘r,u as a function of Au content. Again the initial region represents the approach to sputter equilibrium, the plateau at 1 .O represents the region of NiO, and the last region
P.M. Hall, J.M. M~ra~i~o[Auger arrafysis of Au-Cu-0
and Au-&i-O surfaces
CAU Fig. 5. Atomic ratio of oxygen-t~~ickel
versus gold content for a Ti-N&-Au specimen.
represents the oxygen deficient oxide. The scatter in the data about the plateau gives an indication of the precision of the measurements. Note from eq.  that the calculation of the Ni : 0 ratio at any point is independent of the value used for P&(Ni , Au).
It is import~t to realize that the peak height PA is measured from the derivative (d(~~~~ spectrum, whereas the Auger current associated with a given transition is IA = I(N - Nb)d.E where A$ is the background, and the interval of integration is over the extent of the peak (a few eV). Of course the Auger current 1, is a more fundament~ measurement [Z8]. If the peak in the N(E) spectrum maintains its shape as a function of composition, however, PA is proportional to 1~. But if the shape (e.g. width or skewness) changes, then PA is no longer proportions to IA, and thus PA is not a good measure of 82,. In the N(E) spectrum, these peaks are moreor-less ga~si~ (bell shaped), and PA is measured between the two erection points. Thus a change in peak shape would be expected to show up first as a change in peak width as measured by the energy separation (6) between the positive and negative excursion in the derivative spectrum. If a peak retains its general shape, but simply becomes broader upon a change in composition, then one expects 1~ to be
P.M. Hall, J.M. hiorabito /Auger analysisof Au-Cu-0
proportional to PAS1 [ 1,29,30] ,. This is simply because PA is in effect a second derivative of I,. Besides changing shape, a peak can shift in energy especially upon oxidation. If a specimen consists of a metal partially oxidized, it is even possible to resolve two peaks, one for the oxidized metal, one for the unoxidized. An effect of this type will clearly change the shape of the peaks, especially for a high resolution instrument. Changes of this type have been shown upon oxidation of such metals as Ti , Si [9,15], Al , Mg , and Ta . The general appearance of the Ni spectrum is similar before and after oxidation , but some differences were observed in the present work in focussing on the region from 828 eV to 857 eV, encompassing the 848 eV (the largest) Ni peak. The separation in energy between the positive and negative excursion changed from 5.93 to 9.75 eV upon oxidation, as shown in fig. 6. It seems not to be a function of Au content, as shown by its constancy over the Au-Ni interface region. As mentioned above, one expects the Auger current to be approximately proportional to the peak height in the derivative spectrum times the square of the separation between the positive and negative peaks. Thus, upon oxidation the Auger current (INi) associated with the 848 eV peak actually increased by a factor of (INi)oxide __ -_._ (INi>metal
(fiia&)oxide ______ @%is%netal
_ (14*o) (9.75)’ _ ________
This is far from what one expects based on the atomic ratio calculated earlier
Ni SIGNAL TOO SMALL FOR ACCURATE READINGS
too 120 TIME (MINUTES)
Fig. 6. Peak width of Ni 848 eV peak showing increase due to oxygen.
P.M. Hall, J.M. Morabito /Auger anaiysis of Au-Cu-O
and Au-N&-O surfaces
(l/l .70, or 0.59). A similar increase in the case of oxidation of Nb was ascribed to a larger escape depth in the oxide [lo], The other peaks (0 5 11 eV and Au 69 eV) did not noticeably (within 1 eV) change width or position as the sputtering proceeded. Also, the Ni 61 eV peak did not change its position upon oxidation. Throughout this study there was no evidence of oxygen desorption due to the primary beam of electrons.
Diffusion in the case of &-Au  is quite ~~ogous to that of Ni-Au , except that B is much higher in the Cu-Au case. The Auger data for a Cu-Au film aged in air at 200°C for 25 h are shown in fig. 7. Analysis controls are the same as for the Ni-Au case, except that the sputter ion beam emission current was 8 mA and the Cu 920 eV peak was monitored along with the Au 69 eV and 0 5 11 eV peaks. It appears that Psj is roughly proportional to P&, in a manner similar to that of the Ni case. Fig. 8 is a plot of P& versus P& where data from two other profiles have been combined with the 25 h data to provide more data points. There is more scatter in this plot than in fig. 2, but it too shows reasonable proportionality and an oxygen deficient region near the origin. In the case of the copper oxide, however, there is no clear evidence of a region of approach to sputter eq~b~um. Perhaps in the case of copper oxide the sputter efficiency of oxygen is not much greater than that of copper, and R(Cu, 0) = 1.O for C&O.
SPUTTER TIME (MINUTES)
Fig. 7. Compositon profile of Ti-&-Au from unprocessedAugerdata.
P.M. Hull, J.M. Moral&o /Auger analysisof Au-0-O
Fig. 9 is a plot of Pi, versus P& for the data shown in fig. 7. As in fig. 3, two straight lines are generated, although here we do not have a region of completely pure Au. Therefore, the Au intercept is not quite as well determined. Also, the intercepts on the Cu axis are not confused in this case by an approach to sputter equilibrium. The ratio of the intercepts on the Cu axis (i.e., P&(in Cu)/P&(in CusO)) is 10.9/6.9, or 1.58. If the oxide is assumed to be CuaO, then the ratio of the number of atoms per unit volume in Cu metal to the number of Cu atoms per unit volume in CuaO, based on bulk densities, is 1.67. This is only a 6% discrepancy, even less than in the case of Ni. But in this case the line shape of the 920 eV copper peak does not seem to change upon oxidation. The separation between the positive and negative peaks (8.0 eV) remains the same within 5%, and the ratio of the positive-to-negative peak height (which is a measure of skewness) is also unchanged within 5%. The lack of change for Cu is directly related to the energy resolution of the spectrometer used, which is 0.6%, corresponding to 5.5 eV at 920 eV. The peak is further broadened by the 5V peak-to-peak modulation used in this study. Schon [36,37] using a higher resolution spectrometer, indicates the L$Q,sM4,s (920 eV) Auger Cu peak to be less than 3 eV wide, whether oxidized or not. Since this is less than the energy resolution of the spectrometer used in the present study, the measured peak width of 8 eV is resolution limited. In the caSe of Cu there are at least two oxides known to exist, CuO and CuzO. Some evidence [38,391 has been shown for CuOo_e7, but it is suspected to be a defect state of CuaO. Early work  suggested that Cu at temperatures below 375°C oxidizes in air to form CuO, but it is now generally agreed f41] that CuaO is normally formed. It has been found 1421, however, that the presence of Au in the Cu can shift the energy balance toward the formation of CuO so that, at lOOO*Cwith more than 77 at% Au, the stable oxide form is CuO. This result is not directly applicable to the present work because of the temperature difference. Still, it creates some question as to the stoichiometry of the oxide in our samples. Because of this uncertainty, a single crystal of CuaO was analyzed in the same Auger system. The crystal had been stored in air for many months, so surface cont~nation precluded obtaining a reliable value for PRr(O, Cu). After sputter cleaning (2 kV, 8 mA) for several minutes, however, the Cu 920 eV and 0 503 eV peaks came to steady state at a ratio of P&P& of 0.73. This agrees well with data from a spectrum of the same crystal taken some time ago , which gave a peak height ratio of 0.74. Thus, the value of P&(0, Cu) is 1.46 for this system. Now the slope of fig. 8 is P&/P&, = 0.82. Putting these values into eq. (3), the original concentration ratio in the film sample is calculated to be C&& = 0.8211.46, or 0.56, which corresponds to Cui.aO, reasonably close to CuaO. It also agrees well with the ratio found  for the oxide on bare copper aged in air at 150 and 2OO’C: P&/P&, = 0.80. This value also corresponds to Cur,aO. In addition, evidence  from weight gain during oxidation of bare copper indicates that most (if not all) of the oxide is CuaO. Also note that compositions down to Curs0 have been postulated [38,39] to be a defect structure of CuaO, Thus, it is concluded that the oxide on
P.M. Hali, J.M. Morabito [Auger
anulysis of Au-Cu-O
OOOO*C-sti A 200*C-12.6H
6 6 pcu
Fig. 8. Oxygen peak height versuscopper peak height while sputteringinto a Ti-C&Au specimen.
top of the Au is also CuaO, and from eq. (3) we calculate P&(0, Cu) of 2(0.82), or 1.64. This conclusion (along with the ratio of Cu intercepts of fig. 9) implies that the Auger current associated with the 920 eV Cu peak is almost proportianal to the atomic concentration of Cu, at least for oxygen concentrations up to 33%. An approximate value for &!&Au, Cu) could be obtained from fig. 9 in a marmer analogous to that used for P&(Au, Ni). The result would be &(Au, Cu) = lO.O/ 10.9, or 0.92. As explained above, however, this assumes that R(Au, Cu) = 1. In the case of Au-Cu, this assumption was not necessary because a series of bulk alloys of known composition was available for an independent measurement of P&(Au, Cu). There are three major peaks in the usable Au spectrum, 69,239, and 2024 eV. The 69 eV peak is usually the strongest, so that was used in the present work. A spectral interference exists, however, between the 61 eV Cu peak and the 69 eV Au peak for &rich alloys of Cu-Au. Thus, in previous work [ 121 on the same set of alloys, this peak was not used. The 2024 eV peak was preferred in that study over the 239 eV peak because its greater escape depth makes it less sensitive to sputter corrrection factor variations. In that study, the value of &(Au 2024, Cu) varied from 0.11 for Au-rich alloys to 0.25 for Cu-rich alloys. This variation was attributed to a variation in the sputter correction factor (R), which should affect the 69 eV cabbration in a similar manner, if not more so. The results for the 69 eV calibration are shown on fig. 10. For the Au-rich alloys (CA, greater than 25%) P&r(Au 69, Cu) = 0.55; but for Cu-rich alloys, the apparent value is less (0.30), contrary to the results from the 2024 eV peak. Probably the main reason for this discrepancy is the underestimation of Piu 69 eV when measured so close to the Cu 61 eV peak. Also, the
P.M. Hall, J.M. Morabito /Auger analysisof Au-Cu-0
h -i 6
Fig. 9. Gold peak height versus copper peak height while sputtering into a Ti-Cu- .Au spe cimen.
Pi”60 /Pc”u Fig. 10. Calibration of the Au 69 eV/Cu 920 eV ratio for a series of bulk alloys.
P&f_ HnN, J.M_ Morabito /Auger nnuiysisof Au-&-O
Fig. 11. Composition trations in at%.
profile of Ti-Cu-Au
using relative sensitivity factors to obtain concen-
alloys in this range are so dilute that accuracy suffers. At C..&C~U of 0.04, the signal-to-noise ratio is only about two. in view of this, one might be tempted to use the 2024 eV results to increase J’&r by a factor of two for the Cu-rich alloys. But the same ~terference problem exists for the ~~0~ profiles as for the standards. On the other hand, it seems too drastic to decrease J$& by a factor of two (as in fig. 10) for this region. Perhaps a better compromise is to use the same value of I’&r for the Cu-rich as for the Au-rich alfoys. This approach has been followed here, using ~~(Au, Cu) = 0.55 and P&(0, Cu) of 1.64 to construct fig. 11, following eq. (7). 3,4. Nomographs for calculating concentrations For both the Au-Cu-0 and Au-N-0 systems studied here, there were substantial regions where the ratio of G-&To, (or CQ&;IN~) was relatively constant. This fact can be used to simplify eq. (7) by setting C&rr = m, giving
P.M. Hall, J.M. Morabito /Auger analysisof Au-Cu-0
PEAK HEIGHT RATIO
Fig. 12. Nomograph for obtaining concentration
CONCENTRATION Of B (ATOMIC W
from Auger peak height ratios (binary case).
1000 600 400
10 6 4
.A Fig. 13. Nomograph for obtaining concentration
from Auger peak height ratios (termUY case).
P.M. Ha& J&f. ~~~abir~ /Auger a~I~sis ofAu-Cu-0
The quantity f is thus seen as a modi~ed sensiti~ty factor and operates just like the binary se~siti~ty factor in solving for CA. A nomograph has been constructed using this equation. It is based on the nomograph shown in fig. 12, which is for the binary case eq. (6). If the peak height ratio (P&P&) is, for example, 3.0 and the relative sensitivity factor (P:“el(A,B)) is 0.5, then Cu is 14 at% (solid line). If PvPb is less than one, the designations A and B may be interchanged, causing the peak height ratio to be on scale. If this is done, one must use the reciprocal of the relative sensitivity factor in the middle scale and the right hand scale reads CA rather than C,, If in the above example the peak height ratio (PgPh) had been 0.1 and &&(A, B) were still OS, the dashed line shows that CA would be 17%. Fig. 13 is a similar nomograph for the ternary case, where CJe, is a constant (m). In the case of [email protected]
, we set CQ,/C~ = m = 2. Now, as we have seen, P&(Au, Cu) = 0.55, so ~~~Cu, Au) = 1.8. Connec~ng these two points (dashed line) gives a value for f on the center line of 1.2. Then, if P&/P$,, = 3, the concentration of Au is given as 30 at%, The remai~g 70% is l/3 0 (23%) and 213 Cu (47%). 4. Conclusions In this paper it has been shown that q~n~tative ternary Au-Mets-~ analysis can be performed using Auger electron spectroscopy, In both the Au-Cu-0 and Au-Ni-0 cases,the relative sensiti~ty factor, P=i was found to be independent of CA, and G-J in the concentration ranges studied. For the case of Cu, this was due to the sharpness of the peak compared to the experimental resolution of the Auger analyzer. For the case of the Ni, it was ascribed to the fort~to~ cancellation of two effects: (1) increased line width and (2) increased emission of Auger electrons upon oxidation. No predictions are made for the case of other systems, where it may be nece~a~ to use a ~~erent Prel due to oxidation, i.e. an “oxide” sensitivity factor. In most practical problems, P&r (the sputter-corrected relative sensitivity factor) is more useful than Per, but P&r involves sputter effects as well as Auger effects. Thus our knowledge of P& for ternary analysis is limited by ~cert~nties in R, the sputtering correction factor. R(O, Ni) has been found to be 0.75 for NiO, and R(O, Cu) is about 1.O for C&O. These factors seem to be relatively in~nsitive to CAM, but it is not known how they depend on &,/Co or &&J. Since the sputtering process in an alloy (especially one ~ont~g oxygen) is quite different from that in a pure element, no attempt has been made to compare R with sputter yields for pure elements. It should be noted that the ternary systems studied here were by no means homogeneous, No doubt the 0, Cu and Ni were heavily concentrated near the grain boundaries. Also, oxidation is generally a non-uniform process on a larger scale, so conclusions should be made with care. Nevertheless, the techniques developed here should be useful in those areas where it is desirable to obtain as much info~ation as possible form the AES analysis of gold-metal-oxygen systems.
P.M. Hall, J.M. Morabito /Auger analysisof Au-Cu-0
Acknowledgments The authors gratefully acknowledge R.H. Mills for his careful operation Auger system used in this analysis, and C.A. Haque for his comments.
References [l] J.M. Morabito and P.M. Hall, Scanning Electron Microscopy/l976 (Part 1) (IIT Research Institute, Chicago, IL, 1976) pp. 221-230.  A.P. Janssen, C.J. Harland and J.A. Venables, Surface Sci. 62 (1977) 277.  M.P. Seah, Surface Sci. 32 (1972) 703.  H. ShimIzu, M. Ono and K. Nakayama, Surface Sci. 36 (1973) 817. [S] C.C. Chang, Surface Sci. 48 (1975) 9.  L.A. West, J. Vacuum Sci. Technol. 13 (1976) 198.  H.J. Mathieu and D. Landolt, Surface Sci. 53 (1975) p. 228.  T. Narusawa, T. Satake and S. Komiya, J. Vacuum Sci. Technol. 13 (1976) 514.  C.C. Chang, in: Characterization of Solid Surfaces, Eds. P.F. Kane and G.B. Karrabee (Plenum, New York, 1974) p. 509. [lo] A. Joshi, L.E. Davis and P.W. Palmberg, in: Methods of Surface Analysis, Ed. A.W. Czandema (Elsevier, Amsterdam, 1975) p. 159. [ 111 D.M. Smith and T.E. Gallon, J. Phys. D7 (1974) 151. [ 121 P.M. Hall, J.M. Morabito and D.K. Conley, Surface Sci. 62 (1977) 1. [ 131 P. Holloway, to be published. [ 141 R. Hammer, N.J. Chou and J.M. Eldridge, J. Electron. Mater. 5 (1976) 557.  L.E. Davis, N.C. MacDonald, P.W. Palmberg, G.E. Riach and R.W. Weber, Handbook of Auger Electron Spectroscopy, 2nd ed. (Physical Electronics Industries, 1976) pp. 39,41, 43,47,49 and 53. [ 161 D.R. Penn, J. Electron Spectr. Related Phenomena 9 (1976) 29. [ 171 P.W. Palmberg, Anal. Chem. 45 (1973) 549A.  P.S. Ho, LE. Lewis, H.S. Wildman and J.K. Howard, Surface Sci. 57 (1976) 393. [ 191 R. Bouwman, L.H. Toneman and A.A. Holscher, Vacuum 23 (1972) 163.  K. Christmann and G. Ertl, Surface Sci. 33 (1972) 254.  P.W. Palmberg, J. Vacuum Sci. Technol. 13 (1976) 214.  P.M. Hall and J.M. Morabito, Surface Sci. 59 (1976) 624. [ 231 M. Hansen, Constitution of Binary Alloys (McGraw-Hill, New York, 1958) p. 1024.  R.P. Elliott, Constitution of Binary Alloys, First Supplement (McGraw-Hill, New York, 1965) p. 661.  P.M. Hall, J.M. Morabito and J.M. Poate, Thin Solid Films 33 (1976) 107.  P.M. Hall and J.M. Morabito, Surface Sci. 54 (1976) 79.  P.S. Ho and J.E. Lewis, Surface Sci. 55 (1976) 335.  J.T. Grant, T.W. Haas and J.E. Houston, Surface Sci. 42 (1974) 1.  M.P. Seah, Surface Sci. 40 (1973) 595.  N.J. Taylor, Rev. Sci. Instr. 40 (1969) 792.  P.J. Bassett and T.E. Gallon, J. Electron Spectr. Related Phenomena 2 (1973) 101.  T.W. Haas and J.T. Grant, Phys. Letters 30A (1969) p. 272.  G.E. Riach and W.M. Riggs, Ind. Res. 18 (No. 9) (1976) 74.  P.M. Hall, J.M. Morabito and N.T. Panousis, Thin Solid Films 41 (1977) 341. 1351 P.M. Hail, J.M. Morabito and R.J. Nika, to be published.
P.M. Hall, J.M. Morabito /Auger
analysis of Au-Cu-0
 G. Schon, J. Electron Spectr. Related Phenomena
1 (1972/1973) 377. G. Schon, Surface Sci. 35 (1973) 96. E.G. Clarke and A.W. Czanderna, Surface Sci. 49 (1975) 529. H.Wieder and A.W. Czanderna, J. Phys. Chem. 66 (1962) 816. M. Hansen, Constitution of Binary Alloys (McGraw-Hill, New York, 1958) p. 606. F.A. Shunk, Constitution of Binary Alloys, Second Supplement (McGraw-Hill, New York, 1969) p. 292.  N.G. Schmahl, Z. Anorg. Allg. Chem. 266 (1951) 1.  J.M. Morabito, Surface Sci. 49 (1975) 318.  R.J. Nika and P.M. Ha& to be published.
 (381