Studies on structural and electrical properties of silicon nitride films deposited by unbalanced magnetron sputter deposition

Materials Science and Engineering B90 (2002) 90 – 98

www.elsevier.com/locate/mseb

Studies on structural and electrical properties of silicon nitride films deposited by unbalanced magnetron sputter deposition S.K. Patra, G. Mohan Rao * Vacuum and Thin Films Laboratory, Department of Instrumentation, Indian Institute of Science, Bangalore 560012, India Received 30 April 2001; accepted 22 October 2001

Abstract Unbalanced magnetron sputtering has been used to deposit silicon nitride films. Films were deposited at substrate temperature of 500 °C, nitrogen partial pressure of 1 ×10 − 4 mbar and sputtering pressure of 1 × 10 − 3 mbar. Substrate bias has been varied in order to change the energy of the ions bombarding the substrate. The structural and compositional characterization has been done using FTIR spectroscopy, atomic force microscopy, and Auger electron spectroscopy. A detailed study of electrical properties has been carried out in metal–insulator– semiconductor and metal– insulator– metal configurations. From the C –V measurements dielectric constant and interface charge density were calculated, while the I –V characteristics of the films were used to determine resistivity, dielectric strength and critical field. The deposited films showed resistivity of the order of 1012 V cm, dielectric constant 7.25, with an interface state density of 8 × 1010 eV − 1 cm − 2. The observed variation in the properties of the films has been explained in terms of changes in composition and microstructure of the films. © 2002 Elsevier Science B.V. All rights reserved. Keywords: Unbalanced magnetron sputtering; Ion assisted growth; Silicon nitride films

1. Introduction Thin films of amorphous silicon nitride have been widely used as insulating layers in multilevel interconnects [1], capacitor dielectric in dynamic random access memory (DRAM), gate insulator in metal – insulator – semiconductor field-effect transistor [2 – 4] (MISFET) and so forth. Silicon nitride coatings for these purposes have been formed mainly by low pressure chemical vapor deposition (LPCVD) or by atmospheric pressure CVD (APCVD) at high substrate temperature above 700 °C. For many device applications it is not desirable to heat the substrate to such a temperature, as it will degrade the substrate. It has been found that the deposition temperature can be considerably lowered with plasma-enhanced CVD (PECVD). However, since the surface damage due to the energetic ions may pose a serious problem in high quality devices [5 –7], new * Corresponding author. Tel.: + 91-80-3092349; fax: +91-803345135. E-mail address: [email protected] (G.M. Rao).

techniques such as electron cyclotron resonance (ECR) PECVD [8–10] have been developed to lower the deposition temperatures and to reduce the radiation damage as well. But the main problem that arises in the deposition of silicon nitride film is that, these films contain bonded hydrogen which degrades the device performance occasionally [11,12]. Considering the above problems, reactive sputtering seems to be a promising technique. With sputtering, even room temperature deposition of stoichiometric silicon nitride is possible. It is well known that ion bombardment during the growth of the films causes improvement in the quality of the film [13]. The two important parameters that play a major role in controlling the film properties are ion energy and ion-todepositing atom ratio. There are two ways in which ion bombardment on the growing film can be accomplished. One way is to use a separate ion gun for assisting the growth of the film, which is conventionally termed as ion assisted deposition (IAD). There are different types of ion sources, which are being used for IAD; Kaufman ion source, End Hall source and ECR

0921-5107/02/$ - see front matter © 2002 Elsevier Science B.V. All rights reserved. PII: S 0 9 2 1 - 5 1 0 7 ( 0 1 ) 0 0 9 1 6 - 3

S.K. Patra, G.M. Rao / Materials Science and Engineering B90 (2002) 90–98

plasma source. Except for ECR source, the difficulty in using these sources in sputtering is that of focusing the ions on the substrate, as target-to-substrate distance is small. The other fact is that, in all these cases, a separate ion source has to be used, making sputtering system more and more complicated. The other method is bias sputtering, wherein the application of a negative voltage to the substrate extracts ions from the parent plasma and thus results in the ion bombardment of the growing film. Using unbalanced magnetron (UBM) sputtering system, this can be done in a most efficient way. In UBM sputtering configuration due to the magnetic field strength difference between the inner and outer magnets, magnetic flux lines extend towards the substrate, and thereby result in plasma near to the substrate. By changing the substrate bias one can vary the ion flux as well as ion energy. Earlier efforts to deposit good quality silicon nitride films by ion assisted growth proved to be unsuccessful. The high energy of the ions resulted in stress in the films and most of the time the films were highly non-stoichiometric [14]. However, the unbalanced reactive magnetron sputtering, due to low ion energy, may be a good alternative for the purpose. Various aspects on the UBM sputtering process have been well documented [15– 17]. In this paper we present details of the deposition of silicon nitride films by unbalanced reactive magnetron sputtering. The effect of substrate bias on the properties of the deposited films has been investigated.

91

crystal with a Ž100 orientation (p-type) and a resistivity of 1.7 V cm. Rotary and diffusion pump combination was used to get the desired vacuum. The base pressure of the system is less than 1× 10 − 5 mbar. After attaining the base pressure, nitrogen pressure is set using a needle valve. Later on argon was let in and sputtering pressure was maintained. In order to check the stability of the partial pressure, after each deposition argon flow was stopped and nitrogen partial pressure was checked. It was confirmed to be at the set value. Such practice is generally followed in reactive sputtering processes [18]. Prior to each deposition, presputtering has been carried out in pure argon discharge. All the studies were carried out at a total pressure of 1× 10 − 3 mbar. During the film deposition, the substrates were maintained at a temperature of 500 °C. Distance between the target and the substrate was kept at 10 cm. Films were deposited on silicon (p-type, z= 10 V cm, Ž100 orientation), platinum coated silicon. For bare silicon RCA [19] method was used to clean the substrate. For the platinum coated silicon, ultrasonic cleaning in alcohol was employed. Films were deposited at different substrate biases, i.e. 0 V, floating potential (around − 15 V), −25, −50, −75 and −100 V. Initially UBM plasma analysis was carried out, in the substrate plane over a radius of 6 cm by Langmuir probe method. It showed that the ion density is of the order of 2×109 cm − 3 on the substrate plane.

3. Results and discussions 2. Experimental

3.1. Current–6oltage characteristics of the target The UBM cathode is built in the laboratory based on electromagnetic coils. The design and performance details of the UBM have been discussed elsewhere [17]. Silicon nitride films were deposited by DC reactive sputtering of silicon target. The silicon target is a single

Fig. 1. Cathode voltage vs. nitrogen pressure for different target currents: (A) 175 mA; (B) 200 mA; (C) 225 mA.

To determine the nitrogen partial pressure at which stoichometric silicon nitride films can be deposited, the current–voltage (I–V) characteristics of the target were studied at different nitrogen partial pressures while the working pressure remained at 1× 10 − 3 mbar as shown in Fig. 1. It is observed that as the nitrogen pressure increases, the voltage required for maintaining a fixed target current decreases and, after a certain nitrogen pressure, it almost remains constant. The ionization potential for nitrogen and argon is 14.53 and 15.76 eV, respectively. So there is not much difference in the ionization potential. But, due to its heavier mass, argon has more sputtering yield than nitrogen. As the nitrogen pressure increases, target poisoning starts taking place and also sputtering rate will be less and so the rate of formation of silicon nitride on the target increases. The secondary electron emission coefficient of silicon nitride is higher than bare silicon [20]. Due to this fact, the number of electrons generated in the discharge will be more and hence the voltage required for maintaining a fixed current will decrease. At certain nitrogen pressure for a fixed target current, the rate of

S.K. Patra, G.M. Rao / Materials Science and Engineering B90 (2002) 90–98

92

Fig. 2. FTIR spectra of the films deposited at substrate bias: (a) 0 V; (b) −25 V; (c) − 50 V; (d) − 75 V; (e) − 100 V.

sputtering and the rate of formation of silicon nitride on the target will be equal. This will be the steady state condition. Upto that pressure, the voltage required for maintaining a fixed current keeps decreasing. But, above that pressure, due to the formation of excess silicon nitride, there will not be any appreciable change in the voltage. But instead there will be a slightly higher voltage required due to the change in the resistance of the target. This behavior is similar to the glow discharge characteristics of reactive sputtering of metals in oxygen and argon environment [21]. It is observed that at a nitrogen pressure of 5×10 − 4 mbar and above, the cathode potential is almost constant for a target current of 200 mA. This condition also represents the formation of stoichiometric silicon nitride film on the substrate as confirmed by the FTIR studies discussed later.

3.2. FTIR spectroscopy FTIR spectroscopy was used to estimate the chemical configuration in the deposited SiNx films. The IR transmission spectra were recorded in the range 300–4000 cm − 1 and corrected for the substrate absorption. Fig. 2 shows IR spectra for five films deposited at different bias voltages. All the spectra show a strong absorption band centered around 830– 870 cm − 1 that can be identified with the SiN stretching vibration mode [22]. The large value of full width at half maxima (FWHM) of this band is consistent with the previously reported data [22,23] and can be accounted for amorphous nature of SiNx films. The additional small absorption band ob-

served at 490 cm − 1 corresponds to SiN breathing mode [23]. In addition to these two bands, there are small absorption peaks at 3340, 2350 and 2190 cm − 1. The band at 3340 cm − 1 is believed to originate from water molecule absorption on the film after the deposition. Earlier, Yoon et al. [24] detected the broad 3340 cm − 1 feature in reactively evaporated silicon nitride films and interpreted this as due to the formation of OH bonds resulting from the reaction of water molecules with the nitride films to form hydrogenbonded silanol (SiOH) groups. Liao et al. [25] also reported a hydrogen-bonding broadened SiOH stretching band ranging from 2800–3700 cm − 1 in PECVD silicon nitride films and confirmed that it is due to absorption of H2O molecules (moisture) from the air. The band at 2190 cm − 1 due to SiN2 stretching band (w(NN) [26] in defective silicon nitride is close to w(SiH). Knollel and Osenbach [27] observed that SiH stretched at about 2030–2140 cm − 1, depending upon Si/N ratio in the film studied. The position of w(NN) in SiN2, on the other hand, occurs at 2195 cm − 1, just above SiH range. The other band at 2350 cm − 1 indicates the presence of trapped nitrogen gas [28]. All the deposited films showed this absorption band of trapped nitrogen.

3.3. Morphology Atomic force microscopy (AFM) was used to estimate surface roughness and morphology. Ion bombardment during the growth of the film results in the

S.K. Patra, G.M. Rao / Materials Science and Engineering B90 (2002) 90–98

evolution of smooth surfaces. Fig. 3a– c shows three-dimensional AFM images of the films deposited without substrate bias, and with − 50 and − 100 V substrate bias. It is clear from the images that the irregularities become smaller with an increase in bias voltage. The surface roughness of the samples deposited without any substrate bias is about 2.06 nm. The surface roughness of the samples deposited with −50 and − 100 V substrate bias is 1.47 and 0.56 nm, respectively. As has been observed in the previous studies [29] these surface features are directly related with columnar structure surrounded by void regions. The films deposited without substrate bias, as shown in Fig. 3a exhibit columnar microstructure defined by voided open boundaries. This structure has the appearance of classic zone 1-type films in structure zone model (SZM) of Movchan and Demchishin [30]. The voided growth structure is a fundamental consequence of atomic self-shadowing in concert with the low atomic mobility and clearly undesirable in most of the applications [31]. At − 50 V substrate bias, as seen in Fig. 3b, the atomic network is more interconnected and the voids (low-density regions) surrounded by columns are decreased. For − 100 V substrate bias, shown in Fig. 3c, the voided regions almost disappear and the film is extremely dense and featureless, indicating zone T-type microstructure [30,31]. From these results it can be clearly acknowledged that a progressive densification of the film microstructure occurs as the bias voltage is increased. It is generally recognized that energetic particle bombardment on the growing film, induces the mobility of the surface atoms, thereby reducing the atomic self-shadowing [32] conditions. Such bombardment process minimizes or removes the inter-columnar network of voids or low-density regions which are evident at low or zero substrate bias. The observed transition of film microstructure from open columnar structure to a typical zone T structure with an increase in bias voltage is in good agreement with molecular dynamics (MD) simulation developed by Muller [33] for microstructure of sputter deposited thin films under energetic particle bombardment. Kim and Chung [34] have also reported a similar behavior for silicon nitride films prepared by reactive bias magnetron sputtering. Table 1 Composition of the films deposited at different substrate bias Substrate bias (V)

N/Si ratio

0 −25 −50 −75 −100

1.04 1.07 1.16 1.23 1.19

93

3.4. Composition Auger electron spectroscopy (AES) analysis was performed on the samples to determine Si, and N content in the films. Table 1 shows the composition of the films deposited at different substrate bias. Nitrogen content of the films increases as the substrate bias increase. This is attributed to the higher reactivity between nitrogen and silicon. It can be seen that at substrate bias higher than −75 V, the silicon content in the films increases, probably due to re-sputtering of nitrogen from the deposited film. The properties of the films are in conformity with those compositional changes as explained later.

3.5. Electrical properties To study the electrical properties, films were prepared on silicon substrate (p-type Ž100, 10 V cm) and platinum coated silicon to carry out the measurements in metal–insulator–semiconductor (MIS) and metal–insulator–metal (MIM) configuration. Aluminum electrodes (2.8× 10 − 3 cm2) were deposited on silicon nitride using a shadowing mask. Aluminum was deposited on the back of silicon substrate to provide ohmic contact for MIS configuration. The C–V measurements were carried out using a Keithley C–V analyzer (model 590) at a frequency of 1 MHz. To carry out the I–V characteristics a Keithley electrometer (model 614) with separate voltage source was used. The I–V characteristics were measured only in the MIM configuration.

3.5.1. Capacitance–6oltage characteristics The dielectric behavior of the silicon nitride films in MIM configuration was studied in terms of dielectric constant measurement at a frequency of 1 MHz at room temperature. For this purpose capacitance– voltage (C–V) characteristics of the films deposited at different bias have been studied. The C–V characteristics are given in Fig. 4, and from this data dielectric constant has been calculated. A noticeable feature of C–V behavior is that the capacitance does not show any voltage dependence. In common ferroelectric films like PZT and PLT, the capacitance depends on the magnitude of the polarity of the applied voltage [35]. In silicon nitride films, due to the absence of domain contributions, such dependency was not observed. Another feature is that the dielectric constant decreases as the substrate bias voltage increases as shown in Fig. 5. This is attributed to the fact that due to higher ion bombardment at higher substrate bias, the films were more nitrogen-rich compared to unbiased samples. So the dielectric constant decreases. But at − 100 V substrate bias, an increase in dielectric constant was observed. It is due to the fact that at high-energy ion

94

S.K. Patra, G.M. Rao / Materials Science and Engineering B90 (2002) 90–98

Fig. 3. AFM pictures of the films deposited at different substrate bias (sin1 =0 V, sin3 = −50 V, sin5 = −100 V substrate bias.

S.K. Patra, G.M. Rao / Materials Science and Engineering B90 (2002) 90–98

Fig. 4. Gate voltage vs. capacitance of the films deposited at substrate bias: (A) 0 V; (B) −15 V; (C) − 25 V; (D) −50 V; (E) − 75 V; (F) −100 V.

95

interface state density (Dit), decreases as the substrate bias voltage increases as shown in Table 2. This is attributed to the increase in nitrogen content in the film due to the greater number of nitrogen ions bombarding the growing film, as the bias voltage increases. The interface state density Dit is related to structural defects at insulator/semiconductor surface, the value of Dit is expected to be higher for silicon rich films [41]. So the films having more nitrogen content will show less interface state density. As substrate bias increases, due to increase in nitrogen ion bombardment, the nitrogen component will be higher. The interface charge density for the sample deposited at − 75 V substrate bias is 8.63× 1010 eV − 1 cm − 2 compared to 1.114×1012 eV − 1 cm − 2 for the films deposited without any substrate bias. But, at − 100 V substrate bias, the interface charge density increases (Dit = 5.18× 1011 eV − 1 cm − 2) as high-energy ion bombardment causes structural disorder and breaks the SiN bonds [42].

3.5.2. I–V characteristics To study the insulating properties of the films, the leakage current measurements of the samples were carried out. The measurements were done in MIM configuration at room temperature. Measurements were also carried out by changing the polarities of the electrodes, but both results were identical. It clearly indicates that the current conduction mechanism does not depend on electrode but is controlled by the bulk of the film. It has generally been agreed that Poole–Frenkel [43] emission process leads to current density of the form: Fig. 5. Dielectric constant vs. substrate bias.

bombardment, bonds are broken and re-sputtering of nitrogen occurs. So the films tend to be silicon-rich compared with those deposited at lower bias voltages. The C–V characteristics of the samples in MIS configuration are given in Fig. 6a– e. The figures show the accumulation, depletion and inversion regions of MIS capacitors. The C – V characteristics of the samples exhibit considerable change in terms of flat band voltage. The dielectric constant of the films has been calculated in the accumulation region which is consistent with the value obtained from MIM configuration. The forward and reverse trace of the C – V curve showed anti-clockwise hysterisis, which indicates the hole injection into the silicon nitride film [36,37] which can also be associated with silicon dangling bonds (°SiN3) [38]. The interface state density has been calculated using Terman’s analysis [39] from the high frequency C–V measurements. The flat band offset voltage Vfb shows a shift towards the negative value which indicates that there are positive fixed-charges present [40] at the SiNx /Si interface. The other noticeable feature is that the flat band voltage, and hence the

J= CE exp[− q(ƒb − (qe/pm0md)/kT] where J is current density, C is a constant determined by trap density in the film, ƒb is Poole–Frenkel barrier height and m0 and md the free space and dynamic dielectric constant, respectively. The dynamic dielectric constants obtained from the slope of ln(J)–E 1/2 in the high field region of the curve shown in Fig. 7 is 5.27. The dielectric constant obtained from UV–visible transmission spectra is 3.96. For the Poole–Frenkel conduction mechanism, the dynamic dielectric constant should be between 3.96 and 6.2. These results confirm that the electrical conduction mechanism in silicon nitride film is Poole –Frenkel only. The resistivity and critical fields are obtained from Fig. 8. The critical field is defined as the field at which the current density is 1 mA cm − 2 and also the resistivity has been calculated at that current. The resistivity as well as the critical field, as shown in Fig. 9, is the lowest for the sample deposited without any substrate bias and both show an increase as the substrate bias increases. This is due to the fact that the samples deposited with lower substrate biases have a columnar microstructure surrounded by void regions as shown in AFM. These columnar structures have an

96

S.K. Patra, G.M. Rao / Materials Science and Engineering B90 (2002) 90–98

unacceptable level of leakage and so the resistivity as well as the critical field decreases. For higher substrate bias, the films have zone-T like microstructure, without any voids. So the critical field and resistivity increases. But, at − 100 V substrate bias, the films show a decrease in resistivity and critical field. This can be attributed to the fact that high-energy ion bombardment leads to the rupture of short range ordering in SiN network [41] which creates high defect density of charges. All the films show low resistivity as well as low critical field compared to the CVD silicon nitride films. This can be attributed to the fact that, due to high-energy ion bombardment, considerable amount of structural damage occurs in sputtering which makes the films inferior compared to the CVD films in terms of electrical properties [14].

4. Conclusions The variation in glow discharge characteristics of silicon target during UBM sputtering in nitrogen and argon environment has been explained in terms of difference in secondary electron emission coefficient of silicon and silicon nitride. From the I–V behavior of the target at different nitrogen partial pressure optimum nitrogen partial pressure was set at 5×10 − 4 mbar to deposit stoichiometric silicon nitride films. FTIR spectra show pronounced absorption peak around 830–870 cm − 1 and confirms the formation of silicon nitride films. AFM has been used to estimate the roughness of the films. It shows that the films deposited without any substrate bias have a roughness of 2.06 nm while films deposited at substrate bias of −50 and − 100 V, shows roughness of 1.47 and 0.56 nm, respec-

Fig. 6. Capacitance –voltage characteristics of the films deposited at substrate bias: (a) 0 V; (b) −15 V; (c) − 25 V; (d) − 50 V; (e) −75 V; (f) −100 V.

S.K. Patra, G.M. Rao / Materials Science and Engineering B90 (2002) 90–98

97

Fig. 6. (Continued)

tively. The smoothness of the films has been explained to be due to the bombardment of higher energy ions at high bias voltage. Electrical characterization has been carried out using both MIM and MIS configurations. Dielectric constant, interface charge density has been estimated by C–V measurements at 1 MHz frequency. The variations in the dielectric constant and interface charge density with substrate bias voltage has been explained in terms of nitrogen content in the film. The optimum dielectric constant and interface charge density are 7.25 and 8.63× 1010 eV − 1 cm − 2, respectively. The I–V characteristics have been carried out to determine the conduction mechanism, resistivity and critical field. The I –V characteristics show the conduction mechanism to be Poole– Frenkel emission. The critical field and resistivity were found to be 1.6 MV cm − 1 and

1012 V cm, respectively. The variation in resistivity and critical field with substrate bias has been explained to be based on microstructure of the film.

Table 2 Flat band shift and interface state density with substrate bias Substrate bias voltage (V)

Flat band shift (V)

Dit at flat band eV−1 cm−2

0 −15 (float) −25 −50 −75 −100

−7.46 −8.29 −4.44 −3.84 −1.00 −4.64

1.14×1012 1.17×1012 6.60×1011 6.15×1011 8.63×1010 5.18×1011

98

S.K. Patra, G.M. Rao / Materials Science and Engineering B90 (2002) 90–98

Fig. 7. I – V characteristics of the film deposited at substrate bias − 75 V.

Fig. 8. I – V characteristics of the films deposited at substrate bias: (a) 0 V; (b) − 15 V; (c) −25 V; (d) −50 V; (e) − 75 V; (f) −100 V.

Fig. 9. Resistivity and critical field vs. substrate bias voltage.

References [1] T.Y. Huang, J.L. Paterson, J. Electrochem. Soc. 132 (1985) 1406. [2] M.M. Moslehi, C.Y. Fu, T.W. Sigmon, K.C. Saraswat, J. Appl. Phys. 58 (1985) 2416.

[3] E. Poloura, K. Nauka, J. Lagowsky, H.C. Gatos, Appl. Phys. Lett. 49 (1986) 97. [4] H. Matsumara, J. Appl. Phys. 66 (1989) 3612. [5] B.-C. Ahn, K. Shimizu, T. Satosh, H. Kanoh, O. Sugiura, M. Matzumura, Jpn. J. Appl. Phys. 30 (1991) 3695. [6] N. Nickel, W. Fuhs, H. Mell, J. Non-Cryst. Solids 137 and 138 (1991) 1221. [7] M. Boehm, S. Salomon, Z. Kiss, Mater. Res. Soc. Symp. Proc. 118 (1988) 243. [8] T. Hirao, K. Setsune, M. Kitagawa, Y. Manabe, K. Wasa, S. Kohiki, Jpn. J. Appl. Phys. 26 (1987) L544. [9] T. Hirao, K. Setsune, M. Kitagawa, T. Kanada, T. Ohmura, K. Wasa, Y. Izumi, Jpn. J. Appl. Phys. 27 (1988) L21. [10] Y. Manabe, T. Mitsuyu, J. Appl. Phys. 66 (1989) 2475. [11] G.N. Parsons, J.H. Souk, J. Batey, J. Appl. Phys. 70 (1991) 1553. [12] S.K. Roy, S. Das, C.K. Maiti, S.K. Lahiri, N.B. Chakraborti, J. Appl. Phys. 75 (1994) 8145. [13] A.W. Stephens, J.L. Vossen, W. Kern, J. Electrochem. Soc. 123 (1976) 303. [14] M.F. Lambrinos, R. Valizadeh, J.S. Collign, J. Vac. Sci. Technol. B 16 (1998) 2951. [15] B. Window, N. Savvides, J. Vac. Sci. Technol. A 4 (1986) 196. [16] S.L. Rohde, Physics of Thin Films, vol. 18, Academic Press, New York, 1994. [17] M. Komath, G. Mohan Rao, S. Mohan, Vacuum 52 (1999) 196. [18] M.H. Suhail, G. Mohan Rao, S. Mohan, J. Appl. Phys. 71 (1992) 1421. [19] W. Kern, D.A. Puotien, RCA Rev. 6 (1970) 187. [20] W.D. Westwood, Physics of Thin Films, vol. 14, Academic Press, New York, 1989. [21] G. Mohan Rao, S. Mohan, J. Appl. Phys. 69 (1991) 6652. [22] G.N. Parsons, J.H. Souk, J. Batey, J. Appl. Phys. 70 (1991) 1553. [23] T. Serikawa, A. Okamoto, J. Electrochem. Soc. 131 (1984) 2928. [24] J.S. Yoon, C.V. Deshpandey, R.F. Bunshah, Thin Solid Films 220 (1992) 80. [25] W.S. Liao, C.H. Lin, S.C. Lee, Appl. Phys. Lett. 65 (1994) 2229. [26] L. Huang, K.W. Hipps, J.T. Dickison, U. Mazur, X.D. Wang, Thin Solid Films 229 (1997) 104. [27] W. Knollel, J.W. Osenbach, J. Appl. Phys. 58 (1985) 1248. [28] J.W. Lee, R. Ryoo, M.S. Jhon, K.I. Cho, J. Phys. Chem. Solids 56 (1994) 293. [29] R. Messier, R.C. Ross, J. Appl. Phys. 53 (1982) 6220. [30] B.A. Movchan, A.V. Demchishin, Phys. Met. Metallogr. 28 (1969) 83. [31] J.A. Thornton, J. Vac. Sci. Technol. A 4 (1969) 3059. [32] R. Messier, A.P. Giri, R.A. Roy, J. Vac. Sci. Technol. A 2 (1984) 500. [33] K.H. Muller, J. Appl. Phys. 62 (1987) 1796. [34] J.H. Kim, K.W. Chung, J. Appl. Phys. 83 (1998) 5831. [35] H. Hu, S.B. Krupanidhi, J. Appl. Phys. 71 (1992) 376. [36] D. Landheer, K. Rajesh, D. Masson, J.E. Huise, G.I. Sproule, T. Quance, J. Vac. Sci. Technol. A 16 (1998) 2931. [37] S. Sitbon, M.C. Hugon, B. Agius, F. Abel, J.L. Courant, J. Vac. Sci. Technol. A 13 (1995) 2900. [38] W.L. Warren, J. Kanicki, F.C. Rong, E.H. Poindexter, J. Electrochem. Soc. 139 (1992) 880. [39] L.M. Terman, Solid State Electron. 5 (1962) 285. [40] J. Robertson, M.J. Powell, Appl. Phys. Lett. 44 (1984) 425. [41] D.K. Basa, M. Bose, D.N. Bose, J. Appl. Phys. 87 (2000) 4324. [42] W.A.P. Claassen, W.G.J.N. Valkenburg, M.F.C. Willemsen, W.M.v.d. Wijgert, J. Electrochem. Soc. 132 (1985) 893. [43] S.M. Zee, J. Appl. Phys. 38 (1967) 2951.