Emission spectroscopic studies of sputtering in a low-power glow

Stand.) 1971, NSRDS-NBS 35, 124-143. (10) Nojiri, Y., National Institute for Environmental Studies, Tsukuba, Iba- raki, Japan, unpublished work, Nov-D...
0 downloads 0 Views 538KB Size
2024

Anal. Chem. 1904, 56,2024-2028

(6) Kostyniak, P. J. J. Anal. Toxicol. 1983, 7 , 20-23. (7) Aiiain, P.; Mauras. Y. Anal. Chem. 1979, 51, 2089-2091. (8) Reader, J., Coriiss, C., Eds. "CRC Handbook of Chemistry and Physics", 59th ed.; CRC Press: Boca Ratan, FL, 1978; Line Spectra of the Elements, pp E-216-E-348. (9) Moore, C. E. "Atomic Energy Levels" Natl. Stand. Ref. Data Ser. ( U . S . , NaN. Bur. Stand.) 1971, NSRDS-NBS 35, 124-143. (10) Nojiri, Y., National Institute for Environmental Studies, Tsukuba, Ibaraki, Japan, unpublished work, Nov-Dec 1982.

(11) Aider, J. F.; Bombeika, R. M.; Kirkbright, G. F. Spectrochim. Acta, Part B 1980, 358, 163-175. (12) Nojiri, Y.; Tanabe, K.; Uchida, H.; Haraguchi, H.; Fuwa, K.; Winefordner. J. D. Spectrochim. Acta, Part 8 1983, 388, 61-74.

for review December 2, l98.3. Resubmitted April 30, 1984. Accepted May 1, 1984.

Emission Spectroscopic Studies of Sputtering in a Low-Power Glow Discharge Kazuaki Wagatsuma and Kichinosuke Hirokawa*

The Research Institute for Iron, Steel, and Other Metals, Tohoku University, Sendai, Japan

The Ion sputtering In a glow discharge plasma Is lnvestlgated by an emlsslon spectroscopic method. Prlmary Ions (Ar' ions In our work), which colilde wlth the cathode In a discharge tube, emit their characterlstlc radlation. The intenslties of Ar I I emhlon lines, whlch are asslgned to the optical transltions related to varlous exclted levels, reflect the relatlve populations among dlfferent energy states and, therefore, the energy dlstrlbution among the projectlle Ions. I t is useful to study how the emlsslon lntensitles of different kinds of Ar I I lines depend on Input power In order to compare each plasma condltlon created for some cathode materlais (Ag, Cu, NI, Co, and Fe). By use of the Ar+ ion line analysis, ll was found that when a glow lamp was operated at the same power level, slmllar sputterlng condltlons occurred for these elements. The emission lntensitles of some llnes of sputtered elements were recorded as a function of Input power, and the sputterlng yleid ratios of one element to the others were estlmated from the intenslty ratios for two related emission Ilnes.

Sputtering phenomena have been widely investigated for various reasons, for example, the application for producing thin films, wall attacks by high energy particles in a nuclear reactor or discharge tube, and analytical method for surface analysis. Different light sources are employed in optical emission spectroscopy. While the sample introduction in the usual light source such as an arc or spark discharge results from melting or vaporization, this process is based upon cathode sputtering in a glow discharge source. The glow discharge lamp suggested by Grimm (I, 2) has recently been used for the study of surfaces (3-6). The emission lines are sharp (7) and self-absorption is small compared to the other light sources (8) due to the discharge under reduced pressures. Knowledge on glow discharge sputtering can be obtained from the various spectral lines emitted by the glow discharge plasma. In the case of a high power glow discharge, as employed in the earlier works (9, IO), the sputtering rate is rather high and enough intensity of the emission lines can be observed due to the high ion density of projectiles. However, the resolving power in depth is insufficient and, because much deposit is heaped around the anode, it is difficult to control constant plasma conditions for a long time. On the other hand, mild sputtering conditions, realized in a low-power discharge, are suitable for the study of surfaces.

We have reported the results on several alloy systems using this low power glow discharge spectrometry (11, 12). The variation of emission intensities, which probably depends on the sputtering yields of constituent elements in an alloy, has been observed especially in very low wattage regions. Furthermore, the emission intensities give quantitative information on ejected atoms in the plasma. Emission lines of Ar+ ions (Ar 11) provide available knowledge on the populated argon ion states. In this paper, the power dependence of the emission intensities was observed and the sputtering conditions in the glow discharge could be deduced from the analysis of Ar I1 lines. Emission intensities from target materials were also monitored as a function of input power. Sputtering yield ratios were estimated from intensity ratio of some spectral lines pairs for Ag, Cu, Ni, Co, and Fe and compared with the publised values in earlier works. EXPERIMENTAL SECTION The equipment for our measurements was described elsewhere (12). Our glow discharge lamp was made according to the original model reported by Grimm (1). The inner diameter of the hollow anode was 8.0 mm and the distance between the anode and cathode was adjusted to be 0.4-0.6 mm and kept constant for each measurement. The lamp was evacuated to 4.0-1.3 Pa and then argon gas was introduced (99.9995%,purity) until a pressure of 9.3 X lo2 Pa was reached. Plates of pure Ag, Cu, Ni, Co, and Fe, with the purity of 98-99.99%, and their alloys were used as the cathode (target material). The surfaces were mechanically polished with water-proof emery papers (no. 600 to no. 1500) and then finished to mirror faces with emery cloth. THEORETICAL SECTION As already discussed elsewhere (12), the intensity ratio ( I a / I b ) of resonance emission lines for two different elements a and b is approximately given by

R = constant

(1)

where N,/Nb is the atomic ratio for two elements in the plasma if the following major assumptions are satisfied. The first is that the energy level of one excited state should be nearly equal to that of the other state. If these two excited levels have the same energy levels, exp[(Eb - E,)/kT] (the Boltzmann distribution) becomes unity (where E is the energy level of an excited state and Tis the excitation temperature in the plasma); that is, the measured intensity ratio would be

0003-2700/84/0356-2024$01.50/00 1984 American Chemical Society

ANALYTICAL CHEMISTRY, VOL. 56,NO. 12, OCTOBER 1984 558.2

Table I Selected Emission Lines of Each Element and Their Assignments

410$ ,\

silver copper

nickel

cobalt

iron

wavelength, nm 338.2 324.7 327.4 341.5 352.5 361.9 345.5 345.523 345.351 350.2 344.1 344.099 344.061

upper level

lower level

(3.66 eV) 2 p ~ (3.82 p eV) 'pa/' (3.79 eV) 3F4(3.66 eV) 3P2(3.54 ev) lF3 (3.85 eV)

'SljZ(0.00 eV) 2Sl/z(0.00 eV) 'S1/' (0.00 eV) 3D3(0.03 eV) 3D3(0.03 eV) 'Dz(0.42 eV)

4D1/z (3.81 ev)

4F3/2(0.22 ev) 4FQ/z(0.43 ev) 4FQ/z(0.43 eV)

'p3/2

4D~/2 (4.02 ev)

4D,/z (3.96 ev)

'PZ (3.65 eV)

5Ps(3.60 eV)

\

\v:','2

assignment element

2025

19.97 19.81

' 7

nm

5D3(0.05 eV) 5D4(0.00 eV) Figure 1. Energy level diagram of Ar' ion (extracted).

independent of the excitation temperature. The second assumption is that the degree of self-absorption for the emission lines should be comparble and small. On the other hand, the number of sputtered atoms per unit sample area is given by the following equation (13):

where Jpis primary ion current density and S is sputtering yield (the number of sputtered atoms per incident particle). In general, the sputtering yield can be obtained as a function of primary ion kinetic energy (Ep),the mass of a target atom ( M J or a primary ion (Mp),and the surface binding energy of the target material (UJ (14). The relation between emission intensities and sputtering yields can be derived from eq 1 and 2 as follows:

(3) As described above, by selecting the proper pairs of emission lines which have almost the same energy levels, the intensity ratio is approximately independent of the variation of the excitation temperature. It is also important to get rid of spectral interference from gas emission lines as well as other analytical lines. The selected emission lines for each element are listed in Table I. In order to estimate the degree of self-absorption, it is necessary to examine the intensity ratio of two different emission lines of the same element. If the ratio of these intensities is independently constant of the supplied power to a glow lamp, the contribution of self-absorption is not significant. The dependence of these ratios on the input power has been investigated in several binary alloys. It is found that the effect of self-absorption could be neglected when the emission lines summarized in Table I were employed in low power glow discharge regions (below ca. 8 W for Ag-Cu and ca. 15 W for the others). The constant (R)in eq 1 depends on some factors, i.e., the wavelength, transition probability, partition function, statistical weight for each transition, and the sensitivity function of a spectrometer. Though the transition probability and statistical weight for many transitions have been reported (15), it is difficult to obtain reliable values of all these parameters and to calculate the R value in the strict sense. We have experimentally estimated the proportionality factors for the emission lines shown in Table I (11,12). It is convenient to employ these factors instead of the R value in order to correct the sensitivity difference between the selected emission lines. The sputtering parameters (the density, ion current, and kinetic energy of primary ions) as well as the plasma characteristics (electron temperature, electron density, etc.) can be indirectly controlled by the variation of input power, while

Table 11. Selected Emission Lines of Argon Gas and Their Assignments

assignment wavelength, nm

upper level

lower level

I1 465.8 I1 459.0 I1 457.9 I1 454.5 I1 410.4 I1 358.2 1415.8

'P1/,(19.80 eV)

'P3/' (17.14 eV) 'D3/2 (18.43 eV) 'PljZ(17.27 eV) 2P3jz(17.14 eV) 4D,j2 (19.49 eV) 4D3/2(19.61 eV)

2F5/2(21.13 eV) 'P3/2 (19.97 eV) 'P3/2 (19.87 eV) 4P5j2 (22.51 ev) 4F6jz(23.07 eV) 14.53 eV

11.55 eV

the pressure of the fill gas was kept constant (12). Of course, these factors are not always the same evea if obtained at the same power level. As will be shown in the next section, the emission lines from the fill gas (Ar) were investigated to estimate the dependence of the plasma conditions on the supplied power. RESULTS AND DISCUSSION Analysis of Gas Emission Lines. When a glow lamp is operated with Ar gas, a large number of emission lines are assigned to Ar+ ions and metastable Ar atoms in low power discharge regions. The intensities of Ar I1 emission lines reflect the relative populations among different energy states of excited Ar+ ions. As the optical tansitions are forbidden from the excited energy levels belonging to the quartet terms (the spin multiplicity = 4) to the ground state level (2P3jz), it is considered that the lifetime of the quartet term levels is much longer than that of the doublets, and, therefore, it is interesting to compare the intensities of Ar I1 emission lines derived from the optical transitions among the quartets with those among the doublets. If different target elements (cathode materials) are used, the available information on the sputtering conditions (the state of the plasma) for each element is obtained by examining the power dependence of emission intensities of Ar I1 lines for different target materials. Figure 1 shows the selected energy diagram of Ar+, and the assignments of the selected emission lines are arranged in Table I1 (16,17). Figure 2 shows plots of the intensity ratio of Ar I1 465.8 nm (as standard) to several Ar I1 emission lines with the variation of input power in the case of a pure copper target. The intensity ratios of the doublet state to quartet state lines of Ar I1 465.8 nm to Ar I1 358.2 nm (Figure 2b), and to 410.4 nm (Figure 2 4 , decrease with increase in input power, while the ratios of lines in the doublet state, Ar I1 457.9 nm (Figure 2a) and 454.5 nm (Figure 2d), show no power dependence. These results can be explained in terms of the difference that exists between the energy levels of the excited states. The excitation

2026

ANALYTICAL CHEMISTRY, VOL. 56, NO. 12, OCTOBER 1984

ocu 0 Fa

eco

1.2

2

i

1.5-

> k

cn

>

D

z w

t v)

k-

z

1.0-

0.8 -

k

r

t

1 5

10

Watt

Flgure 2. Plots of the intensity ratios of Ar I 1 465.8 nm to several different Ar I1 lines: (a) 0 , 457.9 nm; (b) Q), 358.2 nm; (c) 0, 410.4 nm, (d) 8, 454.5 nm emission line, measured with a copper cathode. 1

"

"

Ar 465 8

I

"

"

0.6 -

15

POWER

1.2

0 Ni

'

l

5

10

15 Watt

POWER Flgure 4. Power dependence of the intensity ratio of Ar I 1 465.8 nm to 410.4 nm for the same elements as in Figure 3.

'

/ 459 0 nrn

1

1.0-

0.8 -

0 cu

0 Ni

5

10

15 Watt

0

2

4

POWER Figure 3. Power dependence of the intensity ratio of Ar I 1 456.8 nm to 459.0 nm for four target elements: 0, Fe; 0,Cu; 8, Co; 0,Ni.

energy of Ar I1 457.9 nm (19.97 eV) and 454.5 nm (19.87 eV) is approximately equal to that of the standard line Ar I1 465.8 nm (19.80 eV), while those of Ar I1 358.2 nm (23.07 eV) and 410.4 nm (22.51 eV) are higher than the level of the standard emission line. It is significant that the relative intensities from higher excited Ar+ states related to the quartets prominently increase with input power. The influence of target materials was also investigated. The power dependence of some intensity ratios of Ar I1 lines was examined for Cu, Ni, Co, and Fe targets. The intensity ratio (1465,8/Z454,5) for two emission lines, which have almost the same excited energy levels, is independently constant of both the input power and the kind of target element. In the case of ratios of emission lines with different excitation voltages, 1&5,8/1459,0 (Figure 3) and 1465 8/1410,4 (Figure 4), the particular

6 Watt

POWER Figure 5. Relation between input power (1-6 W) and the emission intenslties normalized with Ar I1 465.8 nm for spectral lines of two different target materials: (a) 0, Ag I 338.2 nm; (b) 0,Cu I 324.7 nm.

power dependence, mentioned above, has been found. However, regardless of the kind of target material, these intensity ratios show similar changes with input power. From this work, it is concluded that plasma conditions with the same power level are comparable for the Ar+ ion bombardment against Cu, Ni, Co, and Fe targets. Though not illustrated in figures, the same conclusion can be deduced for Ag and Cu targets in the power range 1-8 W. It is roughly estimated that, when a glow discharge is obtained with the same power level for different cathode materials, the resulting energy distribution of Ar+ ions is similar. The relation between the intensity ratio (1&5,8/1410 4) and the supplied power was

ANALYTICAL CHEMISTRY, VOL. 56, NO. 12, OCTOBER 1984

2027

Table 111. Observed Sputtering Ratios Calculated from the Porportionality Factor and the Normalized Intensity for Some Pairs of Elements and Published Ratios in Earlier Works

'2

sputtering yield ratio this study reporteda

ratio of normalized intens

element

selected pairs of emission lines (proportionality factor)

A g P

Ag 1338.2/Cu 1324.7 nm (1.0)

1.3-1.4

Cu/Ni

Cu 1327.4/Ni 1341.5 nm (1.0) Cu I327.4/Ni 1352.5 nm (0.74)

Co/Ni

Co 1345.5/Ni 1361.9 nm (1.6) Co I350.2/Ni 1361.9 nm (0.95)

Ni/Fe

Ni 1341.5/Fe 1344.1 nm (1.9) Ni 1352.5/Fe 1344.1 nm (2.6)

1.8-2.0 1.4-1.5 (10-14 W) 1.3-1.4 0.8-0.9 (10-15 W) 3.2-3.6 4.1-4.5 (10-15 W)

Co/Fe

Co 1345.5/Fe 1344.1 nm (1.6)

1.4 (18),1.3 (19) 1.3 (20) 1.3 (18), 1.8 (19) 1.7 (20), 1.7 (21)

1.3-1.4

(4.0-6.0 W)

1.8-2.0 1.9-2.0

2.0-2.1

0.81-0.88 0.8-1.0

0.9 (19)

1.7-1.9 1.6-1.7

1.0 (18),1.2 (19) 1.5 (21)

1.3

1.1 (19)

(10-15 W)

keV Ar in ref 18, 600 eV in ref 19, 1 keV in ref 20, and 600 eV in ref 21.

M /At

*

c-

0 0

90

e

v)

x

w

88

I-

0

*

5

8

k 6

P 2.0 W

f N

a E

@

z

P

0

w

-

e

N A a x

1.0

4

0

l-

a

P

0:

u

2

2

0

6

4

Wan

POWER

5

10

POWER

15

wan

Figure 6. Relation between Input power (5-15 W) and the emission intensities normalized with Ar I1 465.8 nm for the following lines of four different target materials: (a) 0, Cu I 327.4 nm; (b) (D, Ni I 341.5 nm; (c) 8,Co I 345.5 nm; (d) 0, Fe I 344.1 nm.

Figure 7. Power dependence of the intensity ratio (Ag I 338.2 nm/Cu I 324.7 nm) normalized with Ar I1 emission lines of (0)465.8 nm, (e) 459.0 nm, and (0)454.5 nm, and with Ar I 415.8 nm (a),respectively. '

I

"

'

'

I

.

'

"

I

Cu/Ni

*

3.c

Cu 327,4/Ni

3 4 1 . 5 nm

k to L

also examined when some binary alloys (Cu-Ni, Ni-Co, and Fe-Ni) were employed as a cathode. These plots are in good agreement with the results on pure elements indicating similar plasma conditions as in pure elements. As a result, if an intensity ratio for two different cathode materials is measured with the same input power, the sputtering yield ratio for these elements can be obtained according to eq 3. Analysis of Metal Emission Lines. The emission intensities of target elements were normalized with respect to those of Ar I1 lines to correct for the fluctuations of experimental conditions. Normalized emission line intensities, which were obtained with Ar I1 465.8 nm as the standard, are plotted against input power in Figures 5 and 6. The normalized intensity ratios of Ag I 338.2 nm and Cu I 324.7 nm as a function of input power, employing four different argon lines for normalization, are shown in Figure 7. All the plots converge onto a single curve. When the neutral atomic line Ar 1415.8 nm is used instead of Ar I1 lines, the same curve was obtained. The ratios of normalized intensities obtained by

W I-

z n w 2.0

N J

4

I

a 0 L r 0

1.0

0

+ U a

POWER

Wait

Figure 8. Power dependence of the intensity ratio (Cu I 327.4 nm/Ni I 341.5 nm) normalized with Ar I1 and I emission lines denoted in Figure 7.

2028

Anal. Chem. 1984,56,2028-2033 1





1



Co 345.5 /Ni



[

Co/Fe could be the stability difference between Fe and Ni(Co) oxide films. In addition to the elements described above, Al and Cr were also studied as target materials. However, these emission lines were absent or very weak in a spectrum measured in low power glow discharges ( 15 W). Since several suitable emission lines, which should be observable even in low power operations, do exist in both A1 and Cr (15),it is assumed that the number of sputtered atoms was small. It is generally known that very stable and firm oxide films (A1,03 or Cr203)are formed on A1 or Cr surfaces. Thus, sputtering yields are being influenced by the properties of oxide films and these effects are more pronounced with an increase in the stability of oxide films on the target surfaces as were pointed out with X-ray photoelectron studies (23). Registry No. Ar, 7440-37-1;Ag, 7440-22-4;Cu, 7440-50-8;Ni, 7440-02-0;Co, 7440-48-4;Fe, 7439-89-6.

361.9 nm

N

0

m

e

e m

,y

I

5



1



I



1



L

10

15

POWER

Watt

Flgure 9. Power dependence of the intensity ratio (Co I 345.5 nmlNi I 361.9 nm) normalized with Ar I1 and I emission lines denoted in Figure 7.

the same method as above are arranged in Figure 8 for Cu/Ni and in Figure 9 for Co/Ni. According to eq 2 and 3, the observed ratio of sputtering yields can be deduced from the results obtained in Figures 7-9 on the assumption that the proportionality factors which were estimated from calibration curves in binary alloys can be employed to analyze the results on pure elements and that the sputtering conditions for several target materials vary with input power in a similar manner. The observed ratios of sputtering yields are summarized in Table 111. Sputtering yields for several elements have been reported by theoretical or experimental methods in the earlier works (18-22). In comparison, the sputtering yield ratios calculated from these published values are also shown in Table 111. The observed ratios in our work are roughly in agreement with the reported values. Since the glow lamp was operated with Ar gas containing impurities, e.g., water and oxygen, these would be expected to have an influence on the sputtering. Especially, stability and strength of oxide films cause a major problem. A possible reason for the slight disagreements of Ni/Fe or

LITERATURE CITED (1) Grimm, W. Naturwlssenshaffen 1987, 5 4 , 588. (2) Grirnm, W. Specfrochin?. Acta, Part 8 1968, 2 3 8 , 443. (3) Berneron, R.; Charbonnier, J. C . S I A , Surf. Interface Anal. 1981, 3 , 134. (4) Ohashi, Y.; Yamamoto, Y.; Tsunoyama, K.; Kishidaka. H. S I A , Surf. Interface Anal. 1979, 1 , 53. (5) Waitlevertch M. E.; Hurwitz, J. D. Appl. Spectrosc. 1976, 3 0 , 510. (6) Belle, C. J.; Johnson, J. D. Appl. Spectrosc. 1973, 2 7 , 118. (7) Hirokawa, K. Bunko Kenkyu 1972, 2 2 , 317. (8) West, C. D.; Human, H. G. Specfrochin?. Acta, Part B 1976, 318, 81. (9) Boumans, P. W. J. M. Anal. Chem. 1972, 44, 1219. (IO) Dogan, M.; Laqua, K.; Massman, H. Spectrochin?. Acta, Part 8 1971, 2 6 8 , 631. (11) Wagatsuma, K.; Hirokawa, K. Anal. Chern. 1984, 5 6 , 908. (12) Wagatsuma, K.; Hirokawa, K. Anal. Chem. 1984, 56, 412. (13) Toyokawa, F.; Furuya, K.; Klkuchi, T. Surf. Scl. 1981, 110, 329. (14) Betz, G. Surf. Sci. 1960, 9 2 , 283. (15) Corliss, C. H.; Bozman, W. R. “Experimental Transition Probabilities for Spectral Lines of Seventy Elements”; US. Government Printing Office: Washington, DC, 1962; NBS Monograph 53. (16) Zaidel, A. N.; Prokof‘ev, V. K.; Raiskii, S. M. “Spektraltabellen”; VEB Veriag Technik: Berlin, 1961. (17) Moore, C. E. “Atomic Energy Levels“; NBS Circular 467, 1949. (18) Andersen, H. H.; Bay, H. L. Radiat. H f . 1973, 19, 63. (19) Laegreid, N.; Wehner, G. K. J . Appl. Phys. 1961, 32, 385. (20) Oechsner, H. 2.Phys. 1973, 261, 37. (21) Weijsenfeld, C. H.; Hoogendoorn, A,; Koedam. M. Physica (Amsterdam) 1961, 27, 763. (22) Andersen, H. H.; Bay, H. L. “Sputtering by Particle Bombardment I”; Behrish, R., Ed.; Springer-Verlag: Berlin, 1981. (23) Kim, K. S.;Winograd, N. Surf. Sci. 1974, 43, 625.

RECEIVED for review March 19,1984. Accepted May 14,1984. We are grateful to Nissan Science Fundation for the financial support of our work.

Glow Discharge Sputtering of Chromium and Niobium Disk Cathodes in Argon Soo-Loong Tong’ and W. W. Harrison*

Department of Chemistry, University of Virginia, Charlottesville, Virginia 22901 A glow dlscharge was used as an atomlzation/excitation source to study the Sputtering of chromium and nioblum dlsk cathodes with argon as the discharge gas. Voltage-current characterlstlc curves were determlned for both cathodes. The effect of dlscharge pressure and voltage on chromium and nlobium sputtered atom density was studied. Argon emission patterns were compared wlth chromium and nioblum emlsslons. On leave from t h e D e p a r t m e n t of Chemistry, U n i v e r s i t y of M a laya, Kuala Lumpur, Malaysia.

Although the glow discharge is one of the oldest spectroscopic sources, relatively few applications have developed of importance to analytical chemists. The most prominent of these has been its role as a hollow-cathode line source for atomic absorption spectroscopy ( I ) . In addition, the hollow cathode is one of several forms of the glow discharge which has been studied as an alternative excitation source for emission spectrochemical analysis in place of the conventional arc and spark sources (2). Both fundamental investigations and more applied studies have shown certain advantages of this technique, mainly because of the inherent stability of the

0003-2700/84/0356-2028$01.50/00 1984 American Chemical Society