Low energy (⩽ 100 eV) sputtering of thin molybdenum nitride films

Surface Science 130 (1983) 36 l-372 North-Holland Publishing Company

361

LOW ENERGY (c 100 eV) SPUTTERING NITRIDE FILMS David A. BALDWIN, J. Wayne RABALAIS Department Received

of Chemistry,

1 November

Noah SHAMIR

MOLYBDENUM

*, Petr HOCHMANN

University of Houston, Houston

1982; accepted

OF THIN

for publication

** and

, Texas 77004, USA

15 April

1983

Molybdenum nitride films formed by 100 eV NC bombardment to saturation of polycrystalline MO have been sputtered to high fluence by normally incident 100 eV ArC and He+ and 15 eV Ne+ while the surface nitrogen concentration was monitored by Auger electron spectroscopy (AES). The penetration distance of nitrogen atoms during film formation and subsequent sputtering is assumed to be small enough that AES will detect, to some degree, all of the nitrogen in the film. The nitrogen AES signal decays exponentially to unsputterable levels for the 15 eV Ne+ and 100 eV He+ cases and decays bi-exponentially to near the bulk contamination level for the 100 eV Ar+ case. The results are interpreted according to existing concepts, but the kinetics of nitrogen loss is modeled using a two-layer adaptation of the standard model for adsorbate monolayer sputtering kinetics. Fitting our proposed model to the data yields apparent cross-sections which are interpreted as composites of cross-sections for experimentally indistinguishable elementary processes; these elementary cross-sections for each process are geometrical averages on the polycrystalline surface. Processes considered, in addition to nitrogen sputtering, include bombardment-induced transport of nitrogen within the film and sputtering of the MO lattice itself.

1. Introduction Conventional concentration versus depth profiles [ 1,2] result from studies in which both the sputter ion range and surface analytic sampling depth (e.g., lo-15 A for AES) are small relative to the characteristic length of concentration gradients. On the other hand, ion-induced desorption cross-sections result from studies of sputtering of single monolayer adsorbates [3,4]. Our low energy reactive ion bombardment experiments [5-S] have resulted in the formation of films which are a few atomic layers thick, i.e. intermediate between the two extremes. In order to interpret these experiments it is necessary to determine the sputtering contributions from the low energy active ions. This is studied in

* Present address: Nuclear Research Center-Negev, Beer-Sheva, Israel. ** Present address: College of Sciences and Mathematics, The University Antonio, San Antonio, Texas 78285, USA.

0039-6028/83/0000-0000/$03.00

0 1983 North-Holland

of Texas

at San

362

D.A. Baldwin et al. / Low merg), ( G 100 e V) sputterrng

the present work by using noble gas ions in the same energy range to sputter a dilute tracer layer of nitrogen implanted at 100 eV into the outer atomic layers of MO. While experiments involving sputtering of thin tracer layers have been performed [9-121 at energies 2 1 keV. there appears to be a lack of data at lower energies. Primary results of the higher energy studies are that bombardment induced atomic mixing is prominent and that tracer atomic concentrations decrease exponentially with ion dose. We expected to observe similar effects below - 500 eV: however, if the sputtering ion energy is so low that the average penetration depth is less than that of the nitrogen atoms used to form the film, one might find that some of the subsurface-distributed nitrogen population is unsputterable, at least with the sputter ion fluences used here, because of shielding by the MO overlayers. However. when sputtering of the MO lattice is efficient, even subsurface nitrogen should eventually be sputtered away. Based on mass-matching energy transfer considerations, i.e., T = YE,,. where E, is the energy of the primary ion of mass m,, T is the maximum energy transferred to a collision partner of mass m, and y = 4m,m,/( m, + ion/energy combinations to demonstrate m2)2, we selected three bombarding the above effects: (1) 100 eV Ar+ should sputter (transfer energy to) both nitrogen and MO effectively; (2) 100 eV He+ should transfer energy to, and hence sputter, nitrogen preferentially; (3) 15 eV Ne+ should hardly disturb the MO lattice but may remove accessible nitrogen. The study of multicomponent sputtering is already a large field [ 131 which is becoming well understood. Central to most theories of multicomponent sputtering is the quantity N,(x) versus ion fluence, i.e., the number density (atoms cm-‘) of atomic component j at depth X. x = 0.0 being the surface. Because AES samples to comparable or greater depth than the nitride layer thickness in the present experiment, depth-averaged AES signals are obtained and N,(X) cannot be uniquely extracted. Since this situation is rather common, we propose a simplified kinetics model which is adapted to this measurement limitation. This model incorporates sputter-induced migration of an initially surface layer into the substrate lattice, as we observe in this work.

2. Experimental methods and results 2.1. Surface nitride preparation

and measurement

The UHV spectrometer system has been described elsewhere [8] and a complete description of the ion beam system is forthcoming [ 141. The polycrystalline MO surface was cleaned by 3 keV Arf bombardment at 15” off grazing and annealing (this procedure left - 7 X 10” nitrogen atoms cm _ ’ as an impurity [8]) after which it was nitrided by a fluence of 2.7 x 10” atoms cm-’ of 100 eV N-j+ at normal incidence. Both the cleaned surface and the

D.A. Baldwin et al. / Low energy ( < 100 eV) sputtering

363

- 25% of a monolayer combined of C and 0 nitrided surface exhibited contaminants, and these prevented the adsorption of residual nitrogen gas, as has been previously observed [ 15,161 and verified [8]. This layer, passivated with respect to nitrogen readsorption, was then bombarded with 15 eV Ne+, 100 eV He’, or 100 eV Ar’ at normal incidence and AES measurements were made as a function of ion dose. The energy spread of the ion beam is l-2 eV and the current densities, Cp, employed were I-10 ,F~A crnm2 = (6-60) X lOI ions cm-’ s-’ on a - 2 mm diameter spot size. Tests for various artifacts and complications were performed. The AES primary electron beam caused no desorption of nitrogen. Some C- and O-bearing gases accompanying the ion beam from the non-UHV ion source were found to adsorb in small quantities (< 10% of a monolayer), but the attenuation of the N and MO AES signals that this caused was corrected for by making blank runs, i.e., electrically deflecting the ion beam while letting gases through. The surface concentrations approach a steady-state due to the competition of adsorption and sputtering, as we verified by changing the ion flux density and observing changes in the steady-state C and 0 signal intensities. A final point concerns the uniformity of the ion beam current density at the target. We have measured the nitrided-spot profile by monitoring the N(KLL) Auger signal as the sample is translated through the AES primary electron beam. In all cases, a smooth gaussian-like profile was obtained. Since the electron beam is - 0.3 mm in diameter (- 90% of the intensity within a cirde of this size), there are no irregularities on this scale. We do not expect irregularities on a smaller scale either, because the target is placed (using a Faraday cup [8]) at the “waist” of the pinch-focus formed by the decelerator lens [ 17,181; at the tightest focus, the beam should be the most uniform in flux density over the central area of the spot.

2.2; AES sa~pi~~g

The Auger signal host (MO) lattice is

i=

collected

from a guest (nitrogen)

atom population

in a

I

where i indexes successively deeper atomic layers of the lattice such that N,(t) is the nitrogen concentration (atoms cmv2) in the outermost atomic layer, N2(t) is the concentration in the second atomic layer, etc. The /3, are unitless attenuation coefficients (0.0 < & < 1.O) for each layer which are determined by mean electron escape depths and geometrical considerations. The layer concept is in a geometrically averaged sense on this polycrystalline surface. If the nitrogen depth distribution N;(t) goes to zero with increasing i more rapidly

than the fl, distribution goes to zero. then setting all j3, = 1.O starts to become a reasonabIe approximation* This allows sampling of the entire nitrogen poputation (atoms cm--“> as

N(t) = f

q(t).

i= 1 This approximation is employed in our treatment. AES will also sample the bulk or volume contamination cm-‘). as

(2)

level, N, (atoms

where X is the mean escape depth of nitrogen KLL electrons. Thus, a constant (N,iD;,), where Da is the average poIy~rystalline Layer spacing, should be subtracted from each N,(t) in order to achieve convergence of eq. (2). We take N, to be characterized by the AES nitrogen signal on the cleaned, annealed MO surface. Using the same cleaning procedure for all runs presented herein as in previous studies [S-S], we have accumulated many AES measurements of the cleaned MO surface which yield an averaged N/MO signal ratio of 0.020 with a standard deviation of TV= 0.003, as corrected for AES sensitivity factors [l9]: this characterizes what we label as our “standard initial surface”. Assuming that the N impurity and the MO are uniformly mixed in the standard initial surface, we approximate ~s~,~~(N)/~~~~~(Mo) - 0.02. Since this is a small level of conta~nation, we calculate ~s~“,~(~o) = 3.23 x 10” atoms cm ’ by using the pure MO values N,,(Mo) = 6.40 x 1012 atoms cm-j and X(Mo) = 5.0 A (MNN line at 186 eV>. Using these values and the ratio I\is,,,rN)/i\il,,,,fMo), we find iNsRa,p(N) = 6.61 x LO” atoms crnm2 on the standard initial surface, which implies N,(N)= 7.8 X 10zO atoms cm-’ (given eq. (3) and A(N)= 8.5 A). The nitrogen concentration after bombardment to saturation with 100 eV N: was calculated to be 7.6 x lOI atoms cm -‘2 by assuming that the N AES signal is proportional to F_,,(N). Due to changing AES sensitivity factors during nitriding or bombardment (changing electronic structure, matrix effects. etc.), this linear relation between AES intensity and area1 number density is only approximate. although satisfactory for present purposes.

Sample characterization by XPS and UPS indicates nitrogen binding energies consistent with nitride formation, while the Auger lineshape we obtain matches that given for ammonia nitridation of MO at 450°C [20]. Based on this information, we assume that the nitrogen atoms are bound to the MO lattice by - 5-6 eV as estimated from the bulk heat of formation of Mo,N 121). The

D.A. Baldwin et al. / Low energy ( < 100 eV) sputtering

365

Fig. 1. Nitride surface concentration versus noble gas ion dose for 15 eV Ne+, 100 eV He+ and 100 eV Arf sputtering of molybdenum nitride layers formed by 100 eV NC bombardment of MO.

results of 15 eV Ne+, 100 eV He+ and 100 eV Ar+ sputtering of nitride films are shown in fig. 1 as nitride concentration N(t) versus primary ion dose (dose = +t). The principal qualitative results are that (1) some nitrogen remains unsputterable in the 15 eV Ne’ and 100 eV He+ cases up to the largest ion doses used and (2) there is still a substantial nitrogen surface concentration (37%) remaining after 100 eV Ar+ bombardment to a dose of - 1.7 x 10” ions cm -*. Adsorption of nitrogen gas was dismissed as the reason for either of these two observations. In the latter case, the sputtering yield of the MO lattice by 100 eV Ar+ is = 0.13 atoms ions- ’ [22-241 and there are - 1.6 x lOI MO atoms cm -2 in an atomic layer of polycrystalline MO, so we expect that > 8 monolayers of MO will be removed after a dose of 1 X 10” ions cme2 has been applied. Thus, our result implies that either (1) nitrogen from the preparatory 100 eV NC bombardment penetrated > 8 atomic layers or (2) some transport mechanism was operative during either the NC or rare gas ion bombardment which facilitated movement of nitrogen into the bulk from a shallow surface layer where it was originally trapped. 3. Discussion 3. I. General Primary processes occurring during nitride formation posure [5] include collisional dissociation of neutralized

by 100 eV NC exNC into N atoms

366

D.A. Baldwin et ul. / Low energ,’ ( i 100 eV)

sputtmng

followed by penetration and capture or backscatter of the atoms. Secondary processes (which may also occur during noble gas ion bombardment) include self-sputtering of nitride by impingent NC, diffusion [25] into the bulk, and. based on evidence presented herein, a set of bombardment-induced transport mechanisms for moving nitrogen into the bulk. The mechanism of nitrogen sputtering is expected to be typical of the case where both the ion and sputtered atom are lighter than the host lattice atoms [3.26]. i.e.. (I) the ion backscatters from a near surface MO atom and sputters N on its way out of the lattice and (2) the ion strikes N which recoils onto the lattice atoms from which it backscatters. For the cases of more vigorous bombardments (100 eV Ar’ and 100 eV N:), the MO lattice is being sputtered and, at least for Ar+. N (and impurity C and 0) is being driven ahead of the receding surface. Mechanisms for such inward transport of a guest atom can be classified as direct (collisional knock-on or recoil implantation) and indirect (impact or vacancy/defect enhanced diffusion). Such bombardment induced migration or transport processes are well known [27] in general; we note several features specific to this system: (1) With these low ion energies, well developed cascades and multiple vacancy/interstitial pair production are improbable. (2) The nitrogen is bound intersititially [21] so that the barrier to migration between interstitial sites will be lower than the lattice binding energy of nitrogen for some diffusion paths and the lattice does not need to expand much to accommodate interstitial transport. (3) Thermochemical driving forces favor the formation of bulk Mo,N, both at thermal energies [28,29] and under N2f bombardment [30,3 11. 3.2. Kinetics of nitrogen sputtering The models [32] for ion mixing, multicomponent sputtering, and high fluence implantation/active ion bombardment are typically for ion energies > 1.0 keV and involve the depth concentration distribution N,(X). Since AES samples deeper than our nitrogen ion penetration ranges, we have depth-averaged AES signals from which N,(x) cannot be uniquely extracted. We note that in the theory of Sigmund, et al. [32], integration over depth x is employed and the resulting expressions have the same functional form as those used herein. Using the model [3,4] for sputtering of adsorbate layers, we postulate the most rudimentary concept of depth, i.e., there are two nitrogen populations important in the sputtering process, (1) outer-surface sites N,(t) = N,(t) and (2) sub-surface sites N,(t) = Cp”=,N,( t). Under the action of a sputter ion flux density $ (ions cm ~’ s ’ ), atoms from the two guest populations are sputtered according to sputter cross-sections us and (I, (cm2) at the rate d N,/d t = -u$N,( t) for IV, and a corresponding rate for Nu. Both a, and u, are averages

Baldwin

AA.

et al. / Low

energy ( 4

100eV) sputtering

367

because of the several binding sites available on polycrystalline surfaces and because a, decreases with depth. We assign average cross-sections uk and ug to N, -+ N, and N, -+ N, transport, respectively. The uk includes knock-on and radiation enhanced diffusion while the ud includes similar processes as well as lattice sputtering. Thus dN&)/dt

= -a,+

N,(f),

so that dN,(t),‘dt=

+ek+ N,(t).

Recession of the surface plane is the dominant contribution to o, and will uncover atoms of the N, population so that they become part of the N, population. Since we are dealing with dilute nitrogen concentrations ( 5 8 x 102’ atoms cme3 compared to - 6.4 x 1O22 atoms cmP3 for pure MO), we assume that the lattice sputtering cross-section to be that of pure MO, i.e., up= Y/N,,,r, where Y (atoms/ion) is the MO sputter yield and Nsurf (atoms cmm2) is the average number of host atoms in the outermost layer. The macroscopic speed (cm s-l) of recession of the surface plane is dx,,,r/dt = a&Da. For the depth distribution of the N, population, we use a uniform volume density (atoms cm -3) down to a certain depth after which the concentration goes to zero. Recession of the surface plane into the N, population reduces it at the rate dN,,(t)/dt=(-dx

,,,Jdt)

K’

N,(t)

= -ea#N&),

with the corresponding rate d N,( t) = a,rpN,( t). The parameter Da cancels out. Collecting the sputtering and transport terms affecting the two populations, we find the overall rates dNs(t)/dt

= -@iv,(t)

dN&)/dt

= -e&N,(t)

-e,&%(f)

+ @N,(t),

(4)

-e&N&).

(5)

+eo,+%(t)

In the general case a, * u,,, the solution N(t) = N,(t) + N,(t), (5) have been integrated and summed, is N(t)=AI

exp(a,t)

+A,

exp(cY,t),

(6)

where a,.2

=

-

(

$ u,+u,+u”+up+~

after eqs. (4) and

I (u~+cJk-e”-uap)Z+4u,ap

I/2 1 >$3 (7)

and where A 1.2 =

N,o + N”0 2 f

Q&O + ealv,,

- (0, + ok - eu - +)(N,o

[( 0, i Uk - UU- up)* + 40,cJ~ The zeroes denote

initial

concentrations.

- X,)/2 r/2

1

(8)

368

D.A.

Baldwtn

3.3. Fitting and interpretation

et ul. /

Low merg,

( < I00 rV) sputterrng

of the kinitics data

Semilog plots of In N(t) versus dose do not yield straight lines, indicating that the single monolayer model [3,4] does not apply. The I5 eV Ne+ and 100 eV He+ data do give straight line semilog plots, shown in fig. 2. when fitted to an expression that is the sum of an exponential plus a constant. The equation is written in the form [N(t)-A]=Bexp(-C$,t) for plotting, and the constants, A, B and C are listed in table 1. The value of A was found by using a computer program that varies A to obtain the best straight line least-squares fit to the semilog plots. We lack the experimental information to extract the individual cross-sections in our model. Since the sputter yield of MO by 100 eV He+ is - 5 X lo-’ atoms ion’ [22-241 and by 15 eV Ne+ is even less, it is reasonable to approximate up= uU = 0, allowing the formation or maintainance of an unsputterable N, underlayer. Thus, for 15 eV Ne+ and 100 eV He+, the general model, eq. (6), reduces to N(t)=N,,+N,+(N,,-N,)

exp[-(~,+a,)@].

(9)

where (10)

Nr=N,,,[Q(~>+~,Jl. The fitting

of ln[ N(t) -A]

DOSE

versus +t yields numerical

values for A = N,,,, + N,.

(atoms/cm2)

Fig. 2. Resolution of the 100 eV Ar+ nitride sputtering curve into two exponentials single exponential fitting of the 100 eV He+ and 15 eV Net curves after unsputterable surface concentration and the bulk contamination level.

(left panel) and subtracting the

Fraction

initial sputter

of molybdenum

2.5

x lo-’

7.7 x 10’3

10.1

3.2 x IQ-” kO.6 x IO-”

6.8 x lOI 6.5 x lOI

15 eV Ne+, eq. (9) a)

Noble gas ions

of noble gas sputtering

film

4.1

x10-s

2.8 x lOI

36.1

1.5x10-” kO.1 x lo-”

4.8 x lOI 2.7 x 1Oi4

100 eV He+, eq. (9) a)

nitride

100 eV Ar+, eq. (6) ‘)

3.6 x 1O-2

3.6 x lOI

47.0

of determination

9.3 x 1om4

3.4 x lOI

45.4

3.8 x 10’4 2.7 x IO-” *0.9 x lo- ‘s

Slow exponential

fit. Coefficients

3.3 x lOI 1.0 x l.0-h kO.9 x lo- I6

Fast exponential

‘) Model equation used. b, 95% confidence limits were approximated using the “standard error of estimate” approach in the least-squares all > 0.9994; the large error bars for 100 eV Ar+ are due to the small number of data points

yield (atoms/ion)

Nitrogen

sputtered

sputtered

of nitride

(atoms/cm*)

Population

population

(W)

(intercept)

C (cm*) (slope) b,

A (atoms/cm*) B (atoms/cm*)

Table 1 Result of model analysis

are

370

D.A. Baldwn

et ul. / Low mergv ( < 100 ev) sputlerin~

B = N,, - Nf and C = us + uk, and the initial AES measurement gives N,, + I%‘(,,,. but this is not enough information to solve for cr, and uk. If N,,, = 0, such as for an ideal chemisorbed layer, then all constants can be uniquely determined in the present type of experiment. The 100 eV ArC data do not fit this expression for any value of A. Instead, the Ar+ data can be fitted by the sum of two exponentials. This is shown in fig. 2 and the optimum constants are listed in table 1. Resolution into two exponentials was based on the fact that for a sum of two exponentials, N(t) = N,(t) + N,( t ) where C, > C,, N( t ) = N2( I) at large t values. Hence, N*(t) was obtained by a least-squares straight line fit to the three highest dose data points, N,,,,,(t) results from N,.exp = N(r) - Nz,r,t(‘), and N,,r,,(t) was obtained from a least-squares fit to the first three data points curve. Again, even if we assume uU = 0 in accordance with the of the N,,,,,(t) findings of Harrison [33], we cannot resolve the constants in eqs. (7) and (8). The constants listed in table 1 can be interpreted in terms of qualitative sputtering concepts. Our data show substantial differences between 15 eV Ne’ and 100 eV He+ sputtering. The apparent cross-section (C in table 1) for 15 eV Ne’ sputtering is approximately twice that for 100 eV He+ sputtering, but the population sputtered by 1.5 eV Ne+ is small so that the initial sputter yields, Y = u,N,, are reversed, i.e., Y( 15 eV Ne+) < Y( 100 eV He+). The depth probed by the Ne+ is expected to be small compared to that probed by He’ due to the lower impact energy, the larger fraction of energy it loses in collisions, and its larger size. Thus the 100 eV He+ sputters a larger (deeper) guest atom population than does 15 eV Ne+, but it does so less efficiently than 15 eV Nei sputters the small population available to it. In the 100 eV Ar+ case an even larger population is susceptible to being sputtered in the initial transient than in either the 15 eV Net or 100 eV He+ decays, and the apparent cross-section is the largest as well. One expects that, in addition to the two sputtering mechanisms given above, the 100 eV Art can also disrupt the host lattice, sputtering both N and MO. We interpret the slow decay of the N concentration at long times to be due primarily to effective knock-on or bombardment-enhanced interstitial diffusion of nitrogen as the MO lattice is eroding. Any readsorption of nitrogen during this stage is minimal or absent.

4. Conclusions The kinetics of multi-atomic-layer tracer sputtering is experimentally characterized in terms of apparent cross-sections and population sizes much as is the kinetics of monolayer sputtering. The apparent cross-sections are averages of several microscopic cross-sections and may even be composites characterizing the probabilities of two different processes, e.g., knock-on and bombardment-enhanced-diffusion. We have attempted to enumerate and account for all the dominant processes within a simplified model by assigning a cross-section

to each group of distinguishable processes, and the resulting expressions can be readily fitted to the experimental data. The data and their interpretation within the proposed model are reconciled with existing knowledge of low energy sputtering processes.

Acknowledgements We thank J.A. Schultz for helpful discussions and two referees for some guiding comments. This material is based upon work supported by the National Science Foundation under Grant No. DMR-8007036 and by the Robert A. Welch Foundation under Grant No. E-656.

References [I] [2] (31 [4] [5] [6] [7] [8] [9] [lo] [1 l] [l2] [ 131

[14] [IS] [l6] [ 171 [IS] [19]

[20] [Zl] [22] [23] [24]

J.W. Coburn, J. Vacuum Sci. Technol. 13 (1976) 1037. H.H. Andersen, Appl. Phys. 18 (1979) 131. H.F. Winters and P. Sigmund, J. Appl. Phys. 4.5 (1974) 4760. E. Taglauer, W. Heiland and J. Onsgaard, Nucl. Ins&. Methods 168 (1980) 571; A. Sagara and K. Kamada, J. Nucl. Mater. 11l/l 12 (1982) 812. D.A. Baldwin. N. Shamir and J.W. Rabalais, Appl, Surface Sci. 1l/l2 (1982) 229. N. Shamir, D.A. Baldwin and J.W. Rabalais, Appl. Surface Sci. 1l/12 (1982) 222. D.A. Baldwin, P.T. Murray and J.W. Rabalais, Chem. Phys. Letters 77 (1981) 403. N. Shamir, D.A. Baldwin, T. Darko, J.W. Rabalais and P. Hochmann, J. Chem. Phys. 76 (1982) 6417. K. Wittmaack, J. Appl. Phys. 53 (1982) 4817. K. Wittmaack, Nucl. Instr. Methods 209/210 (1983) 191. Z.L. Liau, B.Y. Tsauer and J.W. Mayer, J. Vacuum Sci. Technol. 16 (1979) 121. R.R. Hart, H.L. Dunlap and O.J. Marsh, J. Appl. Phys. 46 (1975) 1947. For recent reviews, see: H.H. Andersen, in: Physics of Ionized Gases, Ed. M. Matic (Boris Kidric Inst. Nucl. Sci., Belgrade, Yugoslavia, 1980) p. 43 1; H.H. Andersen, in: Advances in ion Implantation, Eds. J;M. Poate and J.S. Williams (Academic Press, New York, 1982). D.A. Baldwin, T.R. Schuler, P.T. Murray and J.W. Rabalais, Rev. Sci. Instr., to be published. K. Haytek, H.E. Farnsworth and R.L. Park, Surface Sci. 10 (1968) 429. E.I. Ko and R.J. Madix, Surface Sci. 100 (1980) 1449. J.H. Freeman, W. Temple, D.G. Beanland and G.A. Gard, Nucl. Instr. Methods 135 (1976) 1. G.D. Alton, J.B. Roberto, C.W. White and R.A. Zuhr, Nucl. Instr. Methods 177 (1980) 273. LE. Davis, N.C. MacDonald, P.W. Palmberg, G.E. Riach and R.E. Weber, Handbook of Auger Electron Spectroscopy (Physical Electronics Div., Perkin-Elmer Corp., 6509 Flying Cloud Dr., Eden Prairie, MN 55343, USA). T. Kawai, K. Kunimori, T. Kondow, T. Onishi and K. Tamaru, Phys. Rev. Letters 33 (1974) 533. W.L. Jolly, The Inorganic Chemistry of Nitrogen (Benjamin, New York, 1953) p. 250. J. Bohdansky, J. Nucl. Mater. 93/94 (1980) 44. N. Laegreid and G.L. Wehner, J, Appl. Phys. 32 (1961) 365. J. Bohdansky, J. Roth and H.L. Bay, J, Appl. Phys. 51 (1980) 2861.

[25] N. Shamir. D.A. Baldwin and J.W. Rabalais. J. Electron Spectrosc. Related Phenomena 28( 1982) 219. /26] R. Behrisch. G. IMaderlecher, B.M.U. Scherzer and M.T. Robinson. Appl. Phys, 18 (1979) 39t. 1271 For examples, see: P. Sigmund and A. Gras-Marti. Nucl. Instr. Methods 182/1X3 (1981) 25: S.M. Myers, Nucl. Instr. Methods 168 (1980) 265: U. Littmark and W.0 Hofer, Nucl. Instr. Methods 168 (1980) 329. 1281 P. Schwarzkopf and R. Kieffer, Refractory Hard Metals (MacMillan, New York. 1953) p. 250. 1291 J.H. Evans and B.L. Eyre. Acta Met. 17 (1969) 1109. [30] E. Wirz, H.R. Oswald and S. Vepfek, in: Proc. 4th Intern. Symp. on Plasma Chemistry. Zurich. 1979, Eds. S. Vepiek and J. Hertz, p. 492. [3l] C. Braganza. S. Vepiek, E. Wirz, H. Stussi and M. Textor, in: Proc. 4th Intern. Symp. on Plasma Chemistry, Zurich, 1979. Eds. S. Vepiek and J. Hertz. p. 100. [32] For examples, see: P. Sigmund, A. Oliva and G. Falcone, Nucl. lnstr. Methods I94 (1982) 541; N.J. Chou and M.W. Shafer, Surface Sci. 92 (1980) 601; R. Collins and J.J. Jimenez-Rodriquez, Radiation Effects Letters 68 (1982) 19. 1331 D.E. Harrison. Jr. and R.P. Webb. J. Appl. Phys. 53 (1982) 4193.