Reactive magnetron sputtering of copper, silver, and gold

Reactive magnetron sputtering of copper, silver, and gold

Thin Solid Films 478 (2005) 196 – 205 www.elsevier.com/locate/tsf Reactive magnetron sputtering of copper, silver, and gold J.F. Piersona,*, D. Wiede...

901KB Sizes 5 Downloads 35 Views

Thin Solid Films 478 (2005) 196 – 205 www.elsevier.com/locate/tsf

Reactive magnetron sputtering of copper, silver, and gold J.F. Piersona,*, D. Wiederkehra, A. Billardb a

De´partement CREST, Institut FEMTO-ST (UMR CNRS 6174), Universite´ de Franche-Comte´, Poˆle Universitaire, BP 71427, Montbe´liard Cedex 25211, France b Laboratoire de Science et Ge´nie des Surfaces (UMR CNRS 7570), Ecole des Mines, Parc de Saurupt, Nancy Cedex 54042, France Received 19 September 2003; revised 19 October 2004; accepted in revised form 31 October 2004 Available online 8 December 2004

Abstract Copper, silver, and gold targets were sputtered in various reactive gas mixtures (Ar–N2, Ar–O2, and Ar–CH4) to compare the reactivity of noble metal atoms during the sputtering process. The evolution of the film’s growth rate and the variation of the reactive gas partial pressure vs. the reactive gas flow rate were investigated for each kind of metal. The structure of the deposited films was characterised by X-ray diffraction. Electrical resistivity of the coatings was determined at room temperature. The optical band gap of oxides and nitrides films were deduced from UV–visible transmission measurements. The reactive sputtering of copper target leads to the synthesis of Cu3N, Cu2O, Cu4O3, or CuO films. No copper carbide films were deposited. Depending on the methane flow rate, Cu/C films were either nanocomposite coatings (nc-Cu/a-C:H) or amorphous. Silver oxide (Ag2O) films were formed by reactive sputtering of a silver target in Ar–O2 mixtures. On the other hand, the reactive sputtering method did not allow the synthesis of silver nitride nor gold oxide films. D 2004 Elsevier B.V. All rights reserved. Keywords: Sputtering; Noble metals; Structural properties; Electrical and optical properties

1. Introduction Reactive magnetron sputtering is a well-known process used to synthesise various metallic oxides, nitrides, or carbides films using a metallic target and a reactive gas mixture (i.e., Ar–O2, Ar–N2, or Ar–CH4) [1,2]. Thus, this method has been widely used for mechanical, optical, electronic, or magnetic applications. Some studies have already been devoted to the deposition by reactive sputtering and the characterisation of copper oxide [3–6], copper nitride [7–9], or silver oxide films [10,11], and to the reactive sputtering of gold in oxygen [12,13]. However, these studies are only related to one kind of noble metalbased coatings (copper, silver, or gold). To the best of our knowledge, there is no published paper devoted to the comparison of reactivity of noble metals during reactive * Corresponding author. Tel.: +33 3 81 99 46 72; fax: +33 3 81 99 46 73. E-mail address: [email protected] (J.F. Pierson). 0040-6090/$ - see front matter D 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.tsf.2004.10.043

sputtering and to the characterisation of the deposited films. The aim of this paper is to characterise noble metal-based coatings reactively sputter-deposited in the same reactor (except for the gold ones) to compare their reactivity during the sputtering process. A copper target has then been sputtered in Ar–N2, Ar–O2, and Ar–CH4 reactive mixtures; a silver target has been reactively sputtered in Ar–N2 and Ar–O2 mixtures; and, finally, a gold target has been sputtered in Ar–O2 mixtures. For each kind of experiment, details on the sputtering process are provided and relevant properties of each of the deposited films (structure, electrical resistivity, and optical band gap) are investigated.

2. Experimental details Copper- and silver-based coatings were deposited on glass and silicon (100) substrates by RF (13.56 MHz) reactive magnetron sputtering of a 200-mm-diameter metallic target, using an Alcatel SCM 650 sputtering system. Details on the deposition reactor, substrate cleaning, and

J.F. Pierson et al. / Thin Solid Films 478 (2005) 196–205

197

Table 1 List of the different deposition parameters Target diameter (mm) Argon flow rate (sccm) Reactive gas Reactive gas flow rate (sccm) Total pressure (Pa) RF power (W) Target current (A) Substrate–target distance (mm) Deposition time (min)

Cu/N

Cu/O

Cu/C

Ag/N

Ag/O

Au/O

200 25 N2 0–10 0.9–1.2 600 – 60 20

200 25 O2 0–12 0.9–1.25 600 – 60 30

200 10 CH4 0–15 0.4–1.3 500 – 60 15

200 25 N2 0–15 0.9–1.4 600 – 60 20

200 25 O2 0–15 0.9–1.1 600 – 60 20

50 20 or 0 O2 0 or 20 0.4 – 0.5 80 10

deposition procedure can be found elsewhere [14]. The deposition conditions for the different coatings are listed in Table 1. Since gold is very expensive compared to copper or silver, gold-based coatings have been deposited from a 50mm-diameter target in a 30-l sputtering device described elsewhere [15]. The total pressure in the deposition chamber has been measured using an absolute gauge and the reactive gas partial pressure has been determined by subtracting the argon partial pressure to the total pressure. The thickness of the deposited films is measured using a tactile profilometer by allowing calculation of the growth rate for a given deposition time. The structure of the films is studied by X-ray diffraction (XRD) using Cu Ka or Co Ka radiation in h/2h mode or in grazing incidence (48), respectively. The grain size has been estimated from the full width at half maximum of the diffraction line using the Scherrer law and the Warren correction. The film’s morphology has been studied by scanning electron microscopy on the sample surface and on the film’s cross-sections. The 200-nm-thick copper nitride and Cu/C films deposited on both sides polished silicon substrates have been characterised by Fourier transform infrared spectroscopy (FTIR) in the 400–4000 cm 1 range with a 2-cm 1 resolution. The electrical resistivity at room temperature is deduced from sheet resistance measurements using the four-point probe method. Optical band gap has been determined from UV–visible transmission measurements in the 250–1100 nm range.

6 to 7 Am h 1 of the deposition rate between 0 and 4 sccm should result mainly from the deposition of less and less dense coatings preceding the variation of the film’s nitrogen concentration. The following smooth decrease of the deposition rate to 5 Am h 1 for 10 sccm should be related either to the poisoning of the target by a growing fraction of reaction product, the sputtering yield of which is close to that of copper, or to the increasing contribution of sputtering by nitrogen, which is less efficient than argon. In the tested deposition conditions, as soon as nitrogen is injected into the reactor, copper nitride is formed. Indeed, by FTIR, a weak absorption band, which is characteristic of Cu3N, is observed near 667 cm 1 [16]. However, when Q(N2)=1 sccm, XRD analysis shows the occurrence of weakly crystallised copper: the (111) and (200) Cu3N diffraction peaks exhibit a shoulder at higher angle position that can be assigned to copper (Fig. 2). Thus, the films deposited with Q(N2)=1 sccm are biphased coatings. They are composed of a mixture of a nanocrystalline Cu3N phase and a weakly crystallised copper one. For higher nitrogen flow rates, the copper phase is no more detected and only Cu3N is visible by XRD. The structure of the copper nitride films is strongly dependent on the nitrogen flow rate. When Q(N2)=4 sccm, the (111) diffraction peak is greatly predominant, indicating that the Cu3N grains are oriented in the [111] direction. On the other hand, for higher values of Q(N2), the texture changes and only the (100) and (200)

3. Reactive sputtering of copper in Ar–N2, Ar–O2, and Ar–CH4 mixtures 3.1. The Cu–N system 3.1.1. Influence of the nitrogen flow rate The deposition rate and the nitrogen partial pressure plotted vs. nitrogen flow rate (Fig. 1) are representative of a rather low reactivity of nitrogen with copper: the proportionality between nitrogen flow rate and nitrogen partial pressure as well as the rather high slope show that no significant getter occurs on the chamber walls. Moreover, the quite constant deposition rate indicates no catastrophic poisoning of the target. Nonetheless, the slight increase from

Fig. 1. Evolution of the copper nitride film’s growth rate (full squares) and the nitrogen partial pressure without plasma (open squares) and with plasma (open circles) vs. the nitrogen flow rate.

198

J.F. Pierson et al. / Thin Solid Films 478 (2005) 196–205

Fig. 2. Typical X-ray diffractograms of copper nitride films.

diffraction peaks are observed. This texture is the most frequently encountered for reactively sputtered copper nitride films [16–20]. In addition to the texture modification, the diffraction peaks shift progressively to lower angle position when the nitrogen flow rate increases. Thus, the lattice constant changes with Q(N2). The Cu3N lattice constant estimated from the position of the (100) diffraction peak is close to its theoretical value (i.e., 0.3815 nm) for Q(N2)=4 sccm. Then, the films are substoichiometric at low nitrogen flow rate and they are overstoichiometric in terms of nitrogen content for high values of Q(N2). This value of 4 sccm for the nitrogen flow rate is also a critical value for the films morphology. At low Q(N2), the films show a highly facetted morphology, whereas they show a nodular-like one at high Q(N2). However, whatever the nitrogen flow rate, the copper nitride films are columnar. The properties (electrical resistivity and optical band gap) of copper nitride films are also dependent on the nitrogen flow rate [21]. The electrical resistivity increases from 98 to 34,700 AV cm when Q(N2) varies from 1 to 10 sccm. The optical band gap increases also with increasing Q(N2) from 0.25 to 0.83 eV. The low values of optical band gap can be explained by the low electrical resistivity of the Cu3N films. This point is in agreement with Maruyama and Morishita [17], who observed that the optical band gap of Cu3N films decreased with their electrical resistivity. 3.1.2. Effect of the RF bias voltage Wang et al. [22] report that copper nitride loses nitrogen when it is submitted to argon ion bombardment (e.g., to remove the surface contamination of the films before an Auger or XPS analysis). To study the effect of the bias voltage (V b) on the elaboration of copper nitride films, four values of nitrogen flow rate have been chosen: Q(N2)=2, 4, 7, and 10 sccm. The influence of the bias voltage on the Xray diffractograms of films deposited with Q(N2)=2 sccm is shown in Fig. 3. When V b= 50 V, the (111) diffraction peak of copper is observed with Cu3N peaks. For a higher

negative bias voltage, only the copper phase is detected by XRD. The occurrence of the copper phase results from a preferential resputtering phenomenon of nitrogen atoms during the film’s growth when a bias voltage is applied to the substrates. For higher nitrogen flow rates and whatever bias voltage, copper and copper nitride are always detected together by XRD. In these deposition conditions, the films are biphased: Cu+Cu3N. Thus, to totally remove nitrogen from films deposited with Q(N2)N2 sccm, it is necessary to use more negative bias voltage than 160 V. Unbiased films deposited with Q(N2)=7 or 10 sccm exhibit a strong preferred orientation in the [100] direction (Section 3.1.1). When a bias voltage is applied, the film’s texture changes [e.g., the (111) diffraction peak of Cu3N is observed for Q(N2)=7 sccm and V b= 50 V] [23]. Three hypotheses may be considered to explain this result. The first one involves the nitrogen concentration modification due to the resputtering phenomenon. Indeed, it has been mentioned in Section 3.1.1 that the film texture is dependent on the nitrogen flow rate and then on the film stoichiometry. The second hypothesis corresponds to a slight increase of the substrate temperature during the deposition due to the ion bombardment. This temperature increase may enhance the adatom’s mobility and then modify the film’s texture as reported in Ref. [24]. However, it must be noted that the substrate temperature is always lower than 150 8C in the tested deposition conditions. Since the temperature variation when a bias voltage is applied seems to be too low to explain the film’s texture change, this hypothesis cannot be further considered. The third hypothesis is related to the increase of the adatom’s mobility during ion bombardment due to energy transfer from the impinging ions. The effect of the bias voltage on the film’s resistivity is depicted in Fig. 4. When Q(N2)=2 sccm, the electrical resistivity decreases slightly and becomes nearly constant at 4 AV cm when V b= 90, 120, or 140 V. This value is close to that obtained for a nonreactively sputtered copper film

Fig. 3. Influence of the bias voltage on the X-ray diffractograms of films deposited with Q(N2)=2 sccm.

J.F. Pierson et al. / Thin Solid Films 478 (2005) 196–205

Fig. 4. Comparison of electrical resistivity vs. the bias voltage for films formed in different Ar–N2 reactive mixtures.

(about 2 AV cm) [21]. These results are well consistent with XRD analyses where only the copper phase is detected in this bias voltage range (Fig. 3). For higher nitrogen flow rates, the electrical resistivity decreases continuously with the negative bias voltage increase due to both the decrease of the nitrogen concentration and the occurrence of the copper phase. However, in the tested deposition conditions, the electrical resistivity of copper is never reached in Cu+Cu3N films. 3.1.3. Thermal decomposition of Cu3N films Copper nitride is known to thermally decompose into copper and nitrogen. Depending on the authors, the thermal decomposition of Cu3N ranges between 100 and 470 8C [16,25]. The thermal decomposition of Cu3N has been successfully used to realise copper dots and copper lines [26], or to demonstrate the suitability of this material for write-once optical data storage [27]. In the present study, 2.5-Am-thick Cu3N films deposited on glass substrates with Q(N2)=8 sccm have been vacuum-annealed at 200 8C for 8 h. After this thermal treatment, the samples show a copperlike aspect at the film–substrate interface, whereas the film’s surface keeps its metallic silver colour. The annealed samples seem to be biphased. XRD analyses confirm this hypothesis: the (200) Cu diffraction peak is detected on the Cu3N-annealed film (Fig. 5). It must be noted that the vacuum annealing and the partial decomposition of the sample do not change the texture of the remaining Cu3N grains. Since the film–substrate interface has a copper-like aspect, it can be concluded that the copper phase is located near this interface and that the remaining copper nitride phase is located in the upper part of the film. Thus, the thermal decomposition of Cu3N occurs at the film–substrate interface and the copper phase grows perpendicularly to this interface. This kind of growth has been reported, for example, in the crystallisation of hydrogenated amorphous silicon films [28–30]. When the annealing duration is higher than 8 h, it is possible to totally decompose the Cu3N films. However, the copper films formed do not adhere any more

199

Fig. 5. X-ray diffractogram of a copper nitride film before and after a vacuum annealing.

to the substrate due to the high amount of tensile stresses induced by the nitrogen emission. Thus, to form copper by thermal decomposition of copper nitride, it is necessary to employ thin Cu3N films. 3.2. The Cu–O system The deposition rate and the oxygen partial pressure plotted in Fig. 6 as a function of the oxygen flow rate are typical of a rather reactive system with the existence of two distinct sputtering modes [31]. Up to about 5 sccm O2, oxygen partial pressure remains low due to the getter effect of Cu. The average sputtering yield of the target, which remains in the elemental mode, is close to that of pure copper and leads to a high deposition rate in the range 6–8 Am h 1. As for nitrogen enrichment, the slight increase of deposition rate is attributed to the deposition of a little amount of a phase less dense than copper (i.e., a copper oxide). Over 6 sccm O2, the target operates in the reactive sputtering mode with a high level of oxygen partial pressure and a low deposition rate around 1 Am h 1. One can observe

Fig. 6. Evolution of the copper oxide film’s growth rate (full squares) and the oxygen partial pressure (open circles) vs. the oxygen flow rate.

200

J.F. Pierson et al. / Thin Solid Films 478 (2005) 196–205

the abrupt transition between both modes, which is assumed to be associated with a well-known hysteresis, the width of which is lower than 1 sccm. As for the Cu–N system, the solubility of oxygen in copper is very low. As soon as oxygen is introduced into the deposition chamber, Cu2O (cuprite) diffraction peaks are observed by XRD analyses. Up to Q(O2)=3 sccm, the films are biphased: Cu+Cu2O (Fig. 7). Single cuprite phase can be formed only when Q(O2)=4 sccm. Indeed, when Q(O2)=5 sccm, another copper oxide is synthesised: Cu4O3 (paramelaconite). This compound crystallises in the tetragonal system (space group: I41/amd with a=0.5837 nm and c=0.9932 nm; JCPDS file 83-1665). Although paramelaconite has been discovered as a natural mineral during the late 1870s, this compound has never been synthesised in bulk form. Since, in Cu4O3, copper is present in two valence states (Cu(I) and Cu(II)), conventional processes cannot simultaneously stabilise both. Thus, only sputtering technique was proved to allow the deposition of Cu4O3 films [32,33]. Note that paramelaconite films can be formed using either RF or DC reactive magnetron sputtering [34]. As for Cu2O films, the oxygen flow rate range for the deposition of paramelaconite films is narrow. Indeed, when Q(O2) is higher or equal to 6 sccm, a third copper oxide is deposited: CuO (tenorite) (Fig. 7). The increase of the oxygen flow rate does not change the nature of the deposited phase. Furthermore, the preferred orientation of the CuO grains and their mean crystal size are unaffected by the variation of Q(O2). Thus, the control of the oxygen flow rate during the reactive sputtering of a copper target allows the synthesis of all the different phases of this binary system. Similar results have been already published concerning the Cr–N or Ti–O systems [35,36]. The properties of the different copper oxides films, presented in a previous paper [37], can be summarised as follows: Cu2O, Cu4O3, and CuO films are highly resistive with an electrical resistivity at room temperature close to

Fig. 8. The plots of (ahm)2 vs. hm of an as-deposited paramelaconite film and paramelaconite film annealed at 300 8C.

3108, 6108 and 109 AV cm, respectively. The optical band gap has been determined from UV–visible transmission measurements. Cuprite and tenorite films have a direct optical band gap close to 2.45 and 2.11 eV, respectively. Since the band structure of paramelaconite is not reported in the literature, there is no information about the nature of the gap in this phase: direct or indirect. Thus, to determine the optical band gap (E g) of Cu4O3 films, these two cases have to be considered: E g=1.34 or 2.47 eV for an indirect or a direct gap, respectively. The oxidation resistance of copper oxide films in air has been also investigated in the 250–350 8C range. It has been shown that the minimal temperature required for the partial conversion of cuprite into tenorite is influenced by the film’s texture. Indeed, Cu2O films with a preferential orientation in the [100] direction are partially oxidised into CuO at 300 8C, whereas untextured films are partially oxidised as soon as 250 8C. It must be noted that CuO films formed after air oxidation of Cu2O show a lower optical band gap (1.70 eV) than as-deposited CuO films (2.11 eV). Since copper atoms are present in two valence states in paramelaconite, this phase is also sensitive to air oxidation. For a 0.55-Am-thick Cu4O3 film, the conversion of paramelaconite into tenorite begins at 300 8C (Fig. 8). Indeed, the direct optical band gap decreases from 2.47 to 1.98 eV after this thermal treatment. XRD analyses of the annealed film show that the Cu4O3 phase is totally converted into CuO at 300 8C. Contrary to Cu2O or Cu4O3, a thermal treatment in air of CuO films does not modify the X-ray diffractogram of the films. They present the same texture, the same peak position, and the same full width at half maximum. Only a slight decrease of the optical band gap is noticed. However, the optical gap of annealed CuO films is always higher than that obtained after Cu2O conversion into CuO, which should indicate that the conversion is not complete. 3.3. The Cu–C system

Fig. 7. Effect on the oxygen flow rate on the X-ray diffractograms of copper oxide films.

The sputtering of copper in the presence of methane is consistent with a very low reactive system: deposition rate

J.F. Pierson et al. / Thin Solid Films 478 (2005) 196–205

remains constant with the methane flow rate and methane partial pressure is proportional to methane flow rate. In the literature, the Cu–C system has been rarely studied by sputtering, although this system leads to the synthesis of copper–carbon composites [38]. In our deposition conditions, the copper phase is detected by XRD when the methane flow rate ( Q(CH4)) is lower or equal to 8 sccm. Above this critical flow rate, no diffraction peak is observed and the films are amorphous. The effect of Q(CH4) on the mean crystal size of the copper phase, estimated from the full width at half maximum of the (111) diffraction peak of Cu, is displayed in Fig. 9. It decreases when Q(CH4) increases until the films become amorphous. The lowest grain size is obtained for Q(CH4)=8 sccm: 6 nm. It must be noted that although the films are amorphous, copper is still detected by X-ray energy-dispersive analysis. These results suggest that increasing the methane flow rate leads to the incorporation of carbon in solid solution into the copper lattice, which induces a refinement of the copper grains, as reported by Chu et al. [39], until a complete amorphisation of the films. These results are in agreement with FTIR analyses (Fig. 10). Indeed, when Q(CH4)=2, 4, or 6 sccm, no absorption band is observed on the FTIR spectrum. On the other hand, when Q(CH4)=8 sccm, different absorption bands are evidenced in the FTIR spectrum. They have been indexed using the results of Clin et al. [40]. The doublet at 1370 and 1450 cm 1 confirms the presence of sp3 C–H3 groups. However, the 1450 cm 1 band also contains a contribution from sp3 C–H2 groups. The band at 1580 cm 1 indicates that CMC bonds are also present in the deposited films. The bands located approximately at 2870 and 2820 cm 1 have been assigned to C–H vibrational modes. When the methane flow rate is fixed at 15 sccm, the intensity of the previous bands decreases and a new asymmetrical band is observed at 960 cm 1. It corresponds to the C–H wagging mode [41]. The FTIR results suggest that an amorphous hydrogenated carbon phase is formed when the methane flow rate exceeds 6 sccm. When Q(CH4)=8 sccm, the

201

Fig. 10. FTIR spectra of reactively sputtered Cu/C films (the carbon dioxide pattern is a doublet in the 2300–2400 cm 1 region).

deposited films are nanocomposite coatings: nc-Cu/a-C:H. For higher methane flow rates, the films show an amorphous-like structure. These structural changes vs. the methane flow rate are in agreement with the evolution of the electrical resistivity of the films vs. Q(CH4). Due to the grain refinement and to the possible incorporation of carbon atoms into the copper lattice, the electrical resistivity increases from 2 AV cm in pure argon to 9.0 AV cm when Q(CH4)=4 sccm. A value of nearly 86 AV cm is measured for films deposited with Q(CH4)=6 sccm. On the other hand, as soon as the amorphous hydrogenated carbon phase is detected by FTIR, the film’s electrical resistivity cannot be measured using our four-point probe apparatus (i.e., the electrical resistivity is higher than approximately 1012 AV cm). This result indicates that in the nc-Cu/a-C:H films, the amorphous phase may percolate.

4. Reactive sputtering of silver in Ar–N2 and Ar–O2 mixtures 4.1. The Ag–N system

Fig. 9. Variation of the copper mean crystal size vs. the methane flow rate.

As with copper, the sputtering of silver in the presence of nitrogen is significant of a very low reactivity (Fig. 11). Despite a deposition rate of silver in argon (11 Am h 1) about twice that of copper (6 Am h 1), one can observe a lower consumption of nitrogen: the nitrogen partial pressure when 10 sccm N2 is introduced is close to 0.3 Pa when silver is sputtered, whereas it is close to 0.23 Pa when copper is sputtered. Moreover, this value is very close to that obtained without discharge, which should indicate no or very few nitrogen incorporation into silver, whatever the nitrogen flow rate. Note that the slight decrease of deposition rate vs. nitrogen flow rate is ascribed to the increasing fraction of sputtering with nitrogen ions, which is less efficient than argon.

202

J.F. Pierson et al. / Thin Solid Films 478 (2005) 196–205

free energy of formation. Furthermore, Ag3N is known to be very unstable and to be easily decomposed into silver and nitrogen [44]. Note that from target potential measurements, Rizk et al. [45] have suggested the hypothetical formation of a silver nitride during the sputtering of a silver target in Ar– N2 reactive mixtures. However, no structural analysis was provided in their paper to support this hypothesis. 4.2. The Ag–O system

Fig. 11. Influence of the nitrogen flow rate on the growth rate (full squares) and the nitrogen partial pressure (open circles) during the sputtering of a silver target in Ar–N2 reactive mixtures.

As expected, XRD analyses reveal only the occurrence of the fcc silver phase in the deposited films whatever the nitrogen flow rate (Fig. 12). Furthermore, the lattice constant of the silver phase (estimated from the position of the (111) and (200) diffraction peaks) is very close to the theoretical value of bulk silver (0.40862 nm), indicating that the amount of nitrogen atoms incorporated in the silver network is very low. This hypothesis is consistent with the mean crystal size of the silver grains, which ranges between 50 and 60 nm for the different nitrogen flow rates tested in this study. Both nitrogen partial pressure measurements and XRD analyses show that the reactivity of sputtered silver atoms with nitrogen is either very low or null, as reported by Depla and De Gryse [42]. Indeed, Shanley and Ennis [43] have reported a standard free energy of formation for Ag3N of 314.4F2.5 kJ mol 1. Although the reactive sputtering method is known to be a nonequilibrium process, in the deposition conditions tested in this study, it did not allow the synthesis of a compound with a so large positive standard

Fig. 12. Evolution of X-ray diffractograms of films deposited by reactive sputtering of silver vs. the nitrogen flow rate. Q(Ar)=0 sccm and Q(N2)=25 sccm for the pure nitrogen condition.

The deposition rate and oxygen partial pressure vs. oxygen flow rate of the reactively sputtered silver target are representative of an intermediate reactive system (Fig. 13). A noticeable getter is observed and the transition between the elemental (up to about 6 sccm O2) and the reactive (over about 14 sccm) modes is continuous. As with copper, the increase of deposition rate with oxygen flow rate up to 6 sccm O2 is attributed to the deposition of a less dense coating than Ag, whereas its continuous decrease over 6 sccm is expected to provide from the target poisoning by a compound, with the sputtering yield of silver from this compound being lower than that of silver. One can, moreover, observe that the ratio of deposition rate in pure argon to that in reactive mode in the presence of oxygen is significantly higher for silver (about 1.5) than for copper (about 4), which can clearly be related to a stronger reactivity of copper than silver with oxygen. As for Cu–O, the solubility of oxygen into silver is very low. Indeed, a weak diffraction peak of Ag2O (111) is detected as soon as oxygen is introduced into the deposition chamber (Fig. 14a). Note that the detection of Ag2O (200) is difficult because its position is very close to that of Ag (111). The increase of the oxygen flow rate induces a progressive increase of the Ag2O diffraction peak intensity while those of silver decrease. When Q(O2)=7 sccm, the Ag (200) diffraction peak is still unambiguously detected, indicating that in this range of oxygen flow rate, the films are biphased: Ag+Ag2O. When the oxygen flow rate increases up to 8 sccm, the silver diffraction peaks disappear and the films become transparent, indicating that they are

Fig. 13. Variation of the silver oxide film’s growth rate (full squares) and the oxygen partial pressure (open circles) as a function of the oxygen flow rate.

J.F. Pierson et al. / Thin Solid Films 478 (2005) 196–205

Fig. 14. Effect of the oxygen flow rate on the X-ray diffractograms of (a) biphased films Ag+Ag2O and (b) Ag2O films. m-(hkl) and o-(hkl) denote silver and Ag2O diffraction planes, respectively.

composed of single silver oxide (Fig. 14b). This behaviour has been also reported for the Cu/O system (Section 3.2). However, in copper-based films, the oxygen flow rate required for the transition from biphased films to oxides ones is lower than for the Ag/O system (4 and 8 sccm, respectively). This phenomenon is ascribed to both the difference in deposition rate and in reactivity for copperbased than for silver-based films (Figs. 6 and 13). Ag2O films deposited with Q(O2)=8 sccm show a strong preferred orientation in the [100] direction. The crystal size of Ag2O grains has been estimated to nearly 14 nm (i.e., slightly higher than for Cu2O films: 10 nm) [37]. The strong texture of Ag2O films disappears gradually for higher oxygen flow rates (Fig. 14b). However, it must be noted that even with Q(O2)=12 sccm, only the cubic Ag2O phase is detected by XRD analysis. This result is very different from that obtained in the Cu/O system where Cu2O, Cu4O3, and CuO films are successfully deposited by adjusting the oxygen flow rate (Section 3.2). Indeed, in the deposition conditions tested in this study, the reactive sputtering

203

method cannot be used to synthesise pure silver oxide such as: AgO, Ag3O4, or Ag2O3. When the silver target is sputtered in pure oxygen, the XRD analysis shows the occurrence of Ag2O diffraction peaks (111), (200), and (311) (Fig. 15). Furthermore, a weak diffraction peak is detected close to 2h=34.18. It has been assigned to the (002) plane of the monoclinic AgO phase (JCPDS card 43-1038). Since silver atoms in AgO are present in two valence states: Ag(I) and Ag(III) [46]; the occurrence of Ag2O and AgO diffraction peaks shows that the reactivity of sputtered silver atoms with oxygen is not high enough to obtain fully oxidised silver-based films. As abovementioned, only Ag2O film is transparent. The optical transmittance of a 4.3-Am-thick Ag2O film is presented in Fig. 16a. For a wavelength lower than nearly 600 nm, the transmittance is lower than 0.01. It increases strongly between 600 and 800 to reach approximately 0.6 at over 800 nm wavelength. This optical absorption has been used to estimate the optical band gap of the Ag2O films, assuming that it is a direct semiconductor (i.e., as for Cu2O). The plot (ahm)2 vs. hm is presented in Fig. 16b. The intercept of the abscissa axis with the full line of the (ahm)2 vs. hm plot allows the determination of the optical band gap: E g=2.23 eV. In our deposition conditions, the optical band gap of Ag2O films is hardly dependent on the oxygen flow rate. The mean value of E g is 2.27F0.07 eV. This value is in fair agreement with that published by Varkey and Fort [47] for Ag2O films produced by chemical bath deposition (2.25 eV). Note that although Cu2O and Ag2O crystallise in the same cubic structure (spatial group Pn3m), the optical band gap of Ag2O is about 0.2 eV lower than that measured for our cuprite films. The variation of the film’s electrical resistivity vs. the oxygen flow rate is depicted in Fig. 17. Nonreactively sputtered silver films exhibit an electrical resistivity at room temperature close to 1.86 AV cm. This value is consistent

Fig. 15. X-ray diffractogram of a silver oxide film deposited in pure oxygen [ Q(Ar)=0 sccm and Q(O2)=25 sccm]. Vertical lines correspond to the theoretical position of Ag2O diffraction peaks.

204

J.F. Pierson et al. / Thin Solid Films 478 (2005) 196–205

120 AV cm until Q(O2)=6 sccm. The low resistivity is due to percolation through the Ag grains embedded in Ag2O because the resistivity of single-phase Ag2O is very high. For higher oxygen flow rates, an abrupt increase of the electrical resistivity up to approximately 1010–1011 AV cm is observed. Since films deposited with Q(O2)=7 sccm are biphased (Ag+Ag2O), this drop in electrical resistivity suggests that the electrical conduction takes place in the Ag2O phase, which is expected to percolate. Thus, when Q(O2)=7 sccm, Ag grains are embedded in an Ag2O matrix. For single Ag2O films, the electrical resistivity ranges between 1010 and 1011 AV cm. Note that these values are about two orders of magnitude higher than that measured for Cu2O films (Section 3.2).

5. Reactive sputtering of gold in Ar–O2 mixtures Deposition of gold was performed by DC sputtering in pure argon and pure oxygen atmospheres. Fig. 18, which presents X-ray diffractograms performed in both cases, shows that no significant modification occurs, neither in the full width at half maximum nor in the peak positions. Despite neither deposition rate nor resistivity measurements, this behaviour clearly shows the absence of reactivity of gold with oxygen.

6. Conclusion Fig. 16. Optical properties of a Ag2O film deposited with Q(O2)=9 sccm. (a) Optical transmittance of the film; (b) (ahm)2 vs. hm plot for the same film.

with that usually reported for bulk silver (approximately 1.59 AV cm), indicating a very low level of contamination of the deposited films. As soon as Ag2O is detected by XRD, the electrical resistivity of the films increases up to

Fig. 17. Evolution of the electrical resistivity at room temperature as a function of the oxygen flow rate during the reactive sputtering of a silver target.

In this article, copper, silver, and gold targets have been reactively sputtered in various mixtures (Ar–N2, Ar–O2, and Ar–CH4 for Cu, Ar–N2, and Ar–O2 for Ag and Ar–O2 for Au). Among the three noble metals, copper is the more reactive. Indeed, Cu3N and three copper oxides (Cu2O, Cu4O3, and CuO) can be synthesised using the reactive sputtering process. On the other hand, the reactive sputtering

Fig. 18. X-ray diffractograms in grazing incidence (48) using Co Ka radiation of gold films deposited in (a) pure argon and (b) pure oxygen.

J.F. Pierson et al. / Thin Solid Films 478 (2005) 196–205

of copper in Ar–CH4 mixtures cannot be used to form any copper carbide. Depending on the methane flow rate, the deposited films are either nanocomposite (nc-Cu/a-C:H) or amorphous. Since silver is less reactive than copper, silver nitride films cannot be deposited by sputtering of Ag in Ar– N2. Even in pure nitrogen, XRD analysis shows only the fcc silver phase. Furthermore, only one kind of silver oxide has been synthesised: Ag2O. XRD analysis of a silver oxide film deposited using pure oxygen shows Ag2O diffraction peaks and a weak diffraction peak of another silver oxide (AgO). However, pure AgO films have not been synthesised in this study. Finally, the reactive sputtering of gold even in pure oxygen leads to the deposition of pure gold. Some properties of the different defined compounds have been characterised. Copper nitride is a semiconducting material with a wide level of stoichiometry. Depending on the nitrogen flow rate, the Cu3N electrical resistivity ranges between 98 and 34,700 AV cm and the optical band gap varies from 0.25 to 0.83 eV. The application of a bias voltage during the Cu3N synthesis leads to the preferential resputtering of nitrogen as confirmed by XRD analyses and electrical resistivity measurements. The thermal decomposition of 2.5Am-thick Cu3N films has also been studied. It has been demonstrated that the thermal decomposition occurs at the film–substrate interface and that the copper phase grows perpendicularly to this interface. Cu2O, Cu4O3, and CuO films exhibit a high electrical resistivity (higher than 108 AV cm) and direct optical band gaps of 2.45, 2.47, and 2.11, eV respectively. The optical band gap of CuO can be lowered by full oxidation of a Cu2O film. Information about the stability of Cu2O and Cu4O3 during air annealing has also been presented. Contrary to Cu2O, a single Ag2O phase can be formed in a wide range of oxygen flow rate. Depending on Q(O2), Ag2O films exhibit a strong preferred orientation in the [100] direction. The mean crystal size has been estimated up to 14 nm, indicating that the coatings are nanocrystallised. The optical band gap of Ag2O films is close to 2.27 eV and their electrical resistivity ranges between 1010 and 1011 AV cm. Acknowledgements The authors are greatly indebted to Dr. J. Jolly (LPTPUMR CNRS 7648) for vacuum annealing of Cu3N films. The CTITS/LERMPS is acknowledged for its technical support. References [1] R.F. Bunshah, Handbook of Deposition Technologies for Films and Coatings, Noyes Publications, Park Ridge, NJ, 1994. [2] M. Ohring, The Materials Science of Thin Films, Academic Press, San Diego, CA, 1992. [3] A. Parretta, M.K. Jayaraj, A. Di Nocera, S. Loreti, L. Quercia, A. Agati, Phys. Status Solidi, A Appl. Res. 155 (1996) 399. [4] N. Nancheva, P. Docheva, M. Misheva, Mater. Lett. 39 (1999) 81. [5] S. Ishizuka, T. Maruyama, K. Akimoto, Jpn. J. Appl. Phys. 39 (2000) L786.

205

[6] S. Ghosh, D.K. Avasthi, P. Shah, V. Ganesan, A. Gupta, D. Sarangi, R. Bhattacharya, W. Assmann, Vacuum 57 (2000) 377. [7] S. Terada, H. Tanaka, K. Kubota, J. Cryst. Growth 94 (1989) 567. [8] T. Maruyama, T. Morishita, Appl. Phys. Lett. 69 (1996) 890. [9] K.J. Kim, J.H. Kim, J.H. Kang, J. Cryst. Growth 222 (2001) 767. [10] U.K. Barik, S. Srinivasan, C.L. Nagendra, A. Subrahmanyam, Thin Solid Films 429 (2003) 129. [11] R. Snyders, M. Wautelet, R. Gouttebaron, J.P. Dauchot, M. Hecq, Surf. Coat. Technol. 174/175 (2003) 1282. [12] D.M. Mattox, Thin Solid Films 18 (1973) 173. [13] C.R. Aita, N.C. Tran, J. Vac. Sci. Technol., A, Vac. Surf. Films 9 (1991) 1498. [14] J.F. Pierson, E. Tomasella, Ph. Bauer, Thin Solid Films 408 (2002) 26. [15] F. Sanchette, Tran Huu Loi, A. Billard, C. Frantz, Surf. Coat. Technol. 74/75 (1995) 903. [16] L. Maya, J. Vac. Sci. Technol., A, Vac. Surf. Films 11 (1993) 604. [17] T. Maruyama, T. Morishita, J. Appl. Phys. 78 (1995) 4104. [18] T. Nosaka, M. Yoshitake, A. Okamoto, S. Ogawa, Y. Nakayama, Thin Solid Films 348 (1999) 8. [19] L. Soukup, M. Sicha, F. Fendrych, L. Jastrabik, Z. Hubicka, D. Chvostova, H. Sichova, V. Valvoda, A. Tarasenko, V. Studnicka, T. Wagner, M. Novak, Surf. Coat. Technol. 116/119 (1999) 321. [20] S.O. Chwa, K.H. Kim, J. Mater. Sci. Lett. 17 (1998) 1835. [21] J.F. Pierson, Vacuum 66 (2002) 59. [22] D.Y. Wang, N. Nakamine, Y. Hayashi, J. Vac. Sci. Technol., A, Vac. Surf. Films 16 (1998) 2084. [23] J.F. Pierson, Surf. Eng. 19 (2003) 67. [24] S. Ghosh, F. Singh, D. Choudhary, D.K. Avasthi, V. Ganesan, P. Shah, A. Gupta, Surf. Coat. Technol. 142/144 (2001) 1034. [25] Z.Q. Liu, W.J. Wang, T.M. Wang, S. Chao, S.K. Zheng, Thin Solid Films 325 (1998) 55. [26] T. Nosaka, M. Yoshitake, A. Okamoto, S. Ogawa, Y. Nakayama, Appl. Surf. Sci. 169/170 (2001) 358. [27] R. Cremer, M. Witthaut, D. Neuschqtz, C. Trappe, M. Laurenzis, O. Winkler, H. Kurz, Mikrochim. Acta 133 (2000) 299. [28] R. Bisaro, J. Magarino, K. Zellama, S. Squelard, P. Germain, J.F. Morhange, Phys. Rev., B 31 (1985) 3568. [29] T. Mohammed-Brahim, D. Briand, K. Kis-Sion, D. Guillet, A.C. Salaqn, O. Bonnaud, Mater. Res. Soc. Symp. Proc. 398 (1996) 387. [30] J.F. Pierson, K.S. Kim, J. Jolly, D. Mencaraglia, J. Non-Cryst. Solids 270 (2000) 91. [31] A. Billard, C. Frantz, Me´m. Etud. Sci. Rev. Me´tall. 11 (1992) 725. [32] J. Li, G. Vizkelethy, R. Revesz, J.W. Mayer, K.N. Tu, J. Appl. Phys. 69 (1991) 1020. [33] A. von Richthofen, R. Domnick, R. Cremer, Fresenius’ J. Anal. Chem. 358 (1997) 312. [34] A. Thobor, J.F. Pierson, Mater. Lett. 57 (2003) 3676. [35] M.A. Djouadi, C. Nouveau, P. Beer, M. Lambertin, Surf. Coat. Technol. 133/134 (2000) 478. [36] F. Lapostolle, A. Billard, J. von Stebut, Surf. Coat. Technol. 135 (2000) 1. [37] J.F. Pierson, A. Thobor-Keck, A. Billard, Appl. Surf. Sci. 210 (2003) 359. [38] Y. Pauleau, F. Thie`ry, Mater. Lett. 56 (2002) 1053. [39] J.P. Chu, C.H. Chung, P.Y. Lee, J.M. Rigsbee, J.Y. Wang, Metall. Mater. Trans., A Phys. Metall. Mater. Sci. 29 (1998) 647. [40] M. Clin, M. Benlahsen, A. Zeinert, K. Zellama, C. Naud, Thin Solid Films 372 (2000) 60. [41] A. Kumbhar, S.B. Patil, S. Kumar, R. Lal, R.O. Dusane, Thin Solid Films 395 (2001) 244. [42] D. Depla, R. De Gryse, Vacuum 69 (2003) 529. [43] E.S. Shanley, J.L. Ennis, Ind. Eng. Chem. Res. 30 (1991) 2503. [44] L. Grocholl, J. Wang, E.G. Gillan, Mater. Res. Bull. 38 (2003) 213. [45] A. Rizk, L.N. Makar, N.S. Rizk, R. Shinoda, Vacuum 40 (1990) 285. [46] J. Curda, W. Klein, H. Liu, M. Jansen, J. Alloys Compd. 338 (2002) 99. [47] A.J. Varkey, A.F. Fort, Sol. Energy Mater. Sol. Cells 29 (1993) 253.