Annealing of niobium coatings deposited on graphite

ARTICLE IN PRESS Vacuum 79 (2005) 171–177 www.elsevier.com/locate/vacuum Annealing of niobium coatings deposited on graphite S. Barzilaia,b, A. Raveha,b,, N. Fragea a Department of Material Engineering, Ben-Gurion University, P.O. Box 653, Beer-Sheva 84105, Israel b Division of Chemistry, NRC-Negev, P.O. Box 9001, Beer-Sheva 84190, Israel Received 14 December 2004; received in revised form 16 January 2005; accepted 28 March 2005 Abstract Niobium coatings 8–12 mm thick were deposited by magnetron sputtering on ATJ graphite substrate. The kinetic growth of carbide layers into niobium coatings and the properties of the Nb2C and NbC phases obtained after annealing in the temperature range of 1073–1773 K were studied. It was found that the carbide layer growth displayed parabolic behavior patterns establishing the growth rate constants (K) of Nb2C and NbC layers, as follows:     190 kJ m2 164 kJ m2 K Nb2 C ¼ 3:7  108 exp  ; K NbC ¼ 4:5  109 exp  . RT RT s s r 2005 Elsevier Ltd. All rights reserved. Keywords: Sputtering; Coating; Annealing; Niobium carbide 1. Introduction The surface of commercial graphite can be improved by the formation of carbide coatings due to their chemical inertness, hardness and wear resistance. Niobium carbide has been well documented as an important technological material because of its unusual combination of physical and mechanical properties [1]. Therefore, NbC is of Corresponding author. Division of Chemistry, NRC-Negev, P.O. Box 9001, Beer-Sheva 84190, Israel. Tel.: +972 544 757 915; fax: +97 286 568 686. E-mail address: aviraveh@hotmail.com (A. Raveh). great importance in a wide range of applications, in corrosive, erosive and wear environments. In this study, we fabricated niobium-carbide coatings on graphite substrates by the deposition of niobium metal followed by thermal annealing in order to improve the coating to graphite adhesion and surface properties. In a previous investigation [2], we showed that the annealing of niobium coatings produces niobium carbide phases, which improve the hardness and reduce the porosity of the coating on graphite. The density, morphology, and structure of the niobium layer formed by magnetron radio-frequency (rf) sputtering and of the carbide derived by thermal annealing are 0042-207X/$ - see front matter r 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.vacuum.2005.03.005 ARTICLE IN PRESS S. Barzilai et al. / Vacuum 79 (2005) 171–177 172 strongly affected by a negative bias voltage (Vb). It was observed that three distinct categories of Nb and NbC coatings were formed depending on the negative bias voltage, Vb: (a) at V b o50 VDC, a singularly nucleated columnar structure of the Nb coating was formed, which was transformed into a highly porous NbC coating by heat treatment; (b) at 50oV b o80 VDC, a singularly nucleated columnar structure composed of a continuously nucleated sub-columnar structure was formed, which was transformed into a dense NbC coating with the highest micro-hardness, of 13 GPa; and (c) at V b X80 VDC, an imperfect structure of the Nb coating was formed, which was transformed into a NbC coating with the highest density and an intermediate range of micro-hardness. Several authors [3–7] studied the interaction between Nb and graphite in the temperature range of 1673–2573 K and they determined the carbon inter-diffusion coefficients as Nb2C and NbC layers. Woodford and Chang [3] observed that the intrinsic diffusivity of carbon into NbC1x is reduced by the carbide stoichiometry, and that the diffusion coefficient of niobium is negligible compared to that of carbon. They also found that the intrinsic diffusivity of carbon into NbC1x was sensitive to the phase composition rather than the activation energy. Brizes et al. [4] demonstrated that niobium atom diffusion into the carbides is negligible compared to carbon atom diffusion into the NbC phase. The data summarized in Table 1 display the niobium-carbide layer growth by the diffusion of carbon into a niobium metal bulk in the temperature range of 1673–2573 K. It summarizes the kinetic parameters of the interaction between bulk niobium and carbon, i.e., the constant growth rate represented by an Arrhenius-type equation, as shown by other authors [3–7]. Miyake et al. [8] reported the effect of the graphite substrate temperature on the deposition of Nb coatings produced by chemical vapor deposition (CVD). They observed that the coating deposited at 1473 K was composed of two layers, identified as Nb and Nb2C phases, while that deposited at 1523 K was composed of Nb2C and NbC layers. However, only deposition above 1563 K produced a single NbC layer. They concluded that the rate of carbon diffusion and carbide layer formation, originating from a niobium coating on graphite, is significantly higher than that observed by other authors [3–6] for niobium metal bulk. Results similar to Miyake et al. [8] were obtained by Isobe et al. [9] while investigating the interaction between molybdenum coatings on a graphite substrate. Isobe et al. [9] studied carbon diffusivity in molybdenum carbide by measuring the carbide layer growth in both CVD molybdenum coating on a graphite substrate and in bulk molybdenum that was coupled directly to graphite. They concluded that the diffusion coefficient for coatings were higher than those obtained for the bulk specimen. Moreover, the activation energy was lower than those obtained for the bulk specimen. This study was motivated by the lack of data regarding the growth of carbides fabricated from niobium coatings. We therefore investigated the interaction between niobium and graphite by Table 1 The growth rate constants and carbon diffusivity into niobium carbide bulk as determined by the layer growth Carbide phase K0 (m2/s) QK (kJ/mol) T (K) Reference Nb2C NbC1x Nb2C NbC Nb2C NbC Nb2C NbC 1.5770.35  105 2.6570.25  105 3  105 1.76  103 2.2  106 4.5  106 5.970.5  107 1.170.1  104 302.5725.8 312.5711.6 337.4 402.2 287.7 305.3 260.1714.1 344.979.1 1673–973 1673–1973 1973–2573 1973–2573 1700–2090 1700–2090 2173–2573 2173–2573 [3] [3] [4] [4] [5] [5] [6] [6] ARTICLE IN PRESS S. Barzilai et al. / Vacuum 79 (2005) 171–177 measuring the carbide layers by means of metallographic and depth profile analyses. The metallographic cross-sections enabled us to distinguish between sub-layers and the depth profiles verified the sub-layer composition. These were used to evaluate the growth rate of the carbide layers into the niobium coatings formed by thermal annealing. 173 diluted solutions containing nitric acid and hydrofluoric acid. The thickness of the NbC and Nb2C layers, obtained after heat treatment of the niobium coating, were measured by SEM. Measurement of the thicknesses of the sub-layers was enabled by defining the location of interfaces after the chemical etching. These measurements were also verified by depth profile analyses using wavelength dispersive spectroscopy (WDS), and micro-probe analyzer. 2. Experimental 2.1. Layer deposition and thermal annealing 3. Results and discussion The niobium coatings 8–12 mm thick were deposited in a custom-designed rf magnetron system. The ATJ graphite substrates (10  40  1.5 mm) were polished with a 600 mesh SiC paper, then ultrasonically degreased, cleaned and mounted on a substrate holder. The sputter Nb target (80 mm in diameter and 6 mm thick, 99.9%) was mounted on an rf magnetron source at a distance of 5 cm from the substrate holder. The base pressure was below 6.67  104 Pa and the sputtering process was performed in argon (99.999%) at a constant pressure of 0.67 Pa and an rf input power of 400 W. The substrate holder was subjected to V b ¼ 80 V with reference to ground. The maximum substrate temperature resulting from the deposition process was about 473 K. After deposition, the Nb layers were annealed in a vacuum of 6.67  104 Pa and in the temperature range of 1073–1773 K for 12–480 min. The heating rate was 3001/min and the cooling rate was 2001/min. 3.1. Nb coatings on graphite 2.2. Layer characterization Phase analysis was carried out using an X-ray diffractometer (XRD) with Cu–Ka radiation ðl ¼ 0:154 nmÞ and a graphite monochromator. Scanning electron microscopy (SEM) micro-graphs of the cross-sections after fracturing the sample by bending tests and also that of metallographic sections, which were prepared by standard polishing of the samples at several stages of 600, 1000 and 2400 mesh followed by polishing with a synthetic cloth and 1 mm diamond paste were taken. The metallographic samples were etched in The structure and properties of the Nb coatings were studied as a function of gas pressure, rf power, negative bias voltage and deposition time. It was found that the bias voltage and gas pressure were the two parameters that mainly affect the micro-structure and the deposition rate [2]. The change in the micro-structure was seen from singularly nucleated columnar structure to continuous nucleated structure. In addition, the bias voltage also causes the nucleation of sub-columnar structure, while coatings deposited at V b X80 V show broken columnar structure with a smoother surface. The effect of the pressure and bias voltage were seen to be in qualitative agreement with the structure zone T of Thoronton diagram [10,11]. The coatings deposited at various Vb and at 5 mTorr contained micro-pores [2]. Three ranges of micro-porosity were observed: (a) at Vb ¼ 0–50 V, the coatings show 10 vol%; (b) at Vb ¼ 50–80 V, the coating show 5–8 vol%; while coatings deposited at range (c) V b 480 V showed 2–3 vol% of micro-porosity. It is probable that the micro-structure and the micro-porosity concentration affect the carbide growth rate and the final structure after annealing. Fig. 1 depicts the fracture of the Nb layers deposited at various bias voltages. The micrographs represent the three categories of Nb coatings which were formed by the various negative bias voltages, Vb: (a) at V b ¼ 0 V, singularly nucleated columnar (SNC) structures with relatively smooth domed tops were formed (Fig. 1a); (b) at Vb of ARTICLE IN PRESS 174 S. Barzilai et al. / Vacuum 79 (2005) 171–177 became smoother (Fig. 1c). The differences in the columnar structures are attributed to the increased intensity of ion bombardment with the increase in negative bias voltage. 3.2. Nb– C layer growth rate XRD peak patterns of the deposited 8–12 mm thick coatings were examined after annealing for 3 h at various temperatures (Fig. 2). In thin 2 mm coatings treated at low T ¼ 1173 K, X-ray depth sensing could only identify a Nb2C phase, but could not identify it in thick coatings (curve b). At T ¼ 1373 K, a Nb2C phase was formed (curve c), but only after treatment at T ¼ 1773 K did the treated Nb layer transform to a single NbC phase (curve d). It is probable that Nb2C and NbC phases were formed in the interface between the coating and graphite at 1173 and at 1373 K, respectively. However, these phases could not be detected in thick coatings, X5 mm, due to the XRD method of depth sensing. Metallographic cross-section micro-graphs enabled us to distinguish between the phases by displaying clear boundaries, as can be seen in Nb 700k Nb2C Intensity [arbitrary units] 600k Fig. 1. Effect of the negative bias voltage on the fracture morphology of Nb coatings deposited on graphite: (a) V b ¼ 0; (b) V b ¼ 50 V; and (c) V b ¼ 80 V. 500k NbC (d) 1773 K (c) 1373 K (b) 1173 K (a) as deposited 400k 300k 200k 100k 0k 50 V, secondary nucleation occurred, causing the columnar structure to transform into a continuously nucleated sub-columnar (CNSC) structure with a rough surface (Fig. 1b); (c) at V b ¼ 80 VDC, the columnar structure broke down and the surface 30 35 40 45 2θ [degrees] Fig. 2. Typical XRD patterns of Nb coatings: (a) after deposition; (b) after 3 h heat treatment at temperatures of 1173 K; (c) 1373 K; and (d) 1773 K. ARTICLE IN PRESS S. Barzilai et al. / Vacuum 79 (2005) 171–177 0 67 133 t [min] 200 267 333 400 467 533 1073 K 1173 K 1273 K 1373 K 1473 K 2.0 W2Nb2C*1011 [m2] Fig. 3. The metallographic section after chemical etching revealed the boundaries caused by a preferred chemical attack between the phases. In addition, fractured SEM micro-graphs (not shown) also demonstrated distinct Nb, Nb2C and NbC phase structures. Fig. 4 depicts the square thickness (W2) of the NbC and Nb2C phases (Fig. 4) versus duration time of treatment t at various temperatures T. A linear relationship was observed using the least square method which enabled us to define the growth rate constant (K) at various T. By plotting the ln K vs 1/T for the NbC and Nb2C phases (Fig. 5), Arrhenius-type relationships were obtained. It was observed that the K values are expressed by Eqs. (1) and (2), as follows:   190 kJ m2 8 K Nb2 C ¼ 3:7  10 exp  , (1) RT s 175 1.5 1.0 0.5 0.0 0 4 8 12 (a) 0 83 167 16 20 t*10-3 [s] 24 t [min] 250 333 28 417 32 500 1.4 1073 K 1173 K 1273 K 1373 K 1473 K 1573 K 1673 K 1773 K W2NbC*1010 [m2] 1.2 1.0 0.8 0.6 0.4 0.2 0.0 0 (b) 5 10 15 20 25 30 t*10-3 [s] Fig. 4. The square thicknesses (W2) vs heat treatment duration: (a) for Nb2C layer at 1073–1273 K; (b) for NbC layer at 1073–1773 K.   164 kJ m2 K NbC ¼ 4:5  109 exp  , RT s Fig. 3. Metallographic and fractured cross-section of niobium coating: (a) after treatment at 1273 K for 12 min; (b) after treatment at 1273 K for 100 min. (2) where K Nb2 C and KNbC are the growth rate constants of Nb2C at T ¼ 1073–1473 K and NbC at T ¼ 1073–1773 K. The calculated values of K Nb2 C and KNbC are given in Table 2. For comparison, K values for carbon in bulk niobium obtained by other authors [3–6] are plotted in Fig. 5. It was observed that the Q and K values obtained in our study for niobium thin ARTICLE IN PRESS S. Barzilai et al. / Vacuum 79 (2005) 171–177 176 1E-11 Resnick [6] 287.7 kJ/mole KNb2C [m2/s] 1E-12 Woodford [3] 302 kJ/mole 1E-13 1E-14 Brizes [4] 337.5 kJ/mole 1E-15 Bornstein [5] 259.7 kJ/mole this work 190 kJ/mole 1E-16 1E-17 4 5 6 8 9 10 Resnick [6] 344.6 kJ/mole 1E-11 Woodford [3] 312.4 kJ/mole 1E-12 KNbC [m2/s] 7 104/T [K-1] (a) 1E-13 1E-14 Brizes [4] 402.3 kJ/mole 1E-15 this work 164.2 kJ/mole Bornstein [5] 305.3 kJ/mole 1E-16 4 (b) 5 6 7 8 9 10 104/T [K-1] Fig. 5. The growth rate constant (K) for: (a) Nb2C phase; and (b) NbC phase. Table 2 The growth rate constants of NbC and Nb2C coatings Temperature (K) KNb2C (m2/s) KNbC (m2/s) 1073 1173 1273 1373 1473 1573 1673 1773 3.4  1017 2.5  1016 7.8  1016 2.8  1015 7.3  1015 1.7  1014 7.8  1014 3.1  1014 2.6  1017 7.7  1017 4.6  1016 2.8  1015 6.1  1015 — — — —, not examined. coatings differ significantly from those obtained for bulk niobium, i.e., Q is 60% and K is 3–7 times greater than those values obtained for bulk niobium. Similar differences in the Q values obtained for thin coatings versus bulk molybdenum were found by Isobe et al. [9]. They found that the Q value for the formation of a carbide phase from a thin molybdenum layer is in the range of 138–180 kJ/mol depending on the deposition process, in comparison to values reported for bulk material which varied between 319 and 380 kJ/mol [12–14]. It is worth mentioning that the data presented in Fig. 5 were taken from different temperature ranges. According to Matzke [15,16], the activation energies can vary with varying temperatures due to the presence of different diffusion mechanisms, as will be discussed below. In fact, the difference between the kinetic carbide growth rate in a thin metal layer compared with bulk material is well documented [9,13,14,17]. However, these differences are not fully explained and apparently not well understood. We believe that it may be caused by the micro-porosity found in the thin layer [2]. It is possible that the diffusion rate onto the micropore surface in the coating is higher than the one which occurs in the already-formed carbide phase. In addition, it is also possible that the activation energy through the layer surface is lower compared to that through bulk. The growth rate constants of NbC and Nb2C are dependent on carbon diffusion into niobium carbide. Possible reasons why the thin layer diffusivities differed from those of bulk can be related to high defect concentration such as that caused by micro-pores, dislocations induced by the energetic bombardment, and high quenching rates inherent in the deposition technique. Microporosity causes surface diffusion on the interface between the pores and can generate an enhanced diffusion path for carbon. This may cause the increase of the carbide layer growth rate predominantly at low temperature interdiffusion ranges, i.e., 0.35 of the melting temperature. In this temperature region, short-circuiting processes, such as surface and dislocation diffusion, are dominant mechanisms contributing to carbide growth rate. Carbon atom activation energy (representing the migration enthalpy) is necessary for carbon atom movement from one sub-lattice site to ARTICLE IN PRESS S. Barzilai et al. / Vacuum 79 (2005) 171–177 another [18,19]. Migration enthalpy is highly dependent on atom proximity. While carbon atoms in the lattice are entirely surrounded by other atoms, those on the surface or in the dislocation are not. This facilitates atom migration in the presence of any lattice defects. For example, in FCC structures it is well known that the surface activation energy is about 2.6 times lower than that in bulk diffusion. In conclusion, thin layers diffusivities differed from those of bulk because of micro-porosity and/or high dislocation density inherent in the deposition technique, enabling atom carbon migration in surface, dislocation and bulk diffusion. This raises the diffusion coefficients (mainly at low temperature) and lowers the activation energy of thin layers in contrast to bulk diffusion processes. 4. Summary and conclusions Niobium-carbide layers were formed after niobium coatings were deposited on graphite by magnetron sputtering followed by heat treatment at various temperatures and time durations. Our study arrived at the following conclusions: (a) Structure and micro-porosity are important factors for determining carbide growth rate and can explain the difference in the values found in thin layers versus bulk materials. (b) The growth of the carbide phases displays parabolic behavior enabling us to determine the growth rate constants of the NbC and Nb2C sub-layers. (c) The activation energies of the growth of the NbC and Nb2C sub-layers were 190 and 164.2 kJ/mol, respectively. 177 Acknowledgments The authors wish to thank Mr. Avi Ben-Shabat for his expert technical assistance, and Mr. E. Boublil for SEM micro-graphs. This work was supported by a grant from the Israeli Council of High Education and the Israeli Atomic Energy Commission. References [1] Pierson HO, editor. In: Handbook of carbide and nitride. NJ: 1996, Noyes Publication, (Chapter 5). [2] Barzilai S, Weiss M, Frage N, Raveh A. Surf Coat Technol, in press, available on-line. [3] Woodford J, Chang YA. Metall Mater Trans A 1998; 29A:2717. [4] Briezes WMF, Cadoff LH, Tobin JM. J Nucl Mater 1966; 20:57. [5] Bornstein NS, Hirakis EC, Friedrich LA. Part & Witney (USA): TIM-927, 1965. [6] Resnick R, Steinitz R, Siegle L. Trans AIME 1965;222: 1918. [7] Gel’d PV, Liubimov VD. Isv Akad Nauk SSSR Otd Tekhn Met I Toplivo 1961;6:19 as cited in Ref. [3]. [8] Miyake M, Hirooka Imoto YR, Sano T. Thin Solid Films 1979;63:303. [9] Isobe Y, Son P, Miyake MJ. Less-Common Metals 1989;147:261. [10] Thornton JA. Ann Rev Mater Sci 1977;7:239. [11] Messier R, Giri AP, Roy RA. J Vac Sci Technol A 1984;2:500. [12] Kidson GV. J Nucl Mater 1961;3:21. [13] Rosa CJ. Metall Trans A 1983;14A:199. [14] Eremeev VS, Panov AS. Izv Akad Nauk SSSR Neorgan Mater 1968;4:1507 as cited in Ref. [8]. [15] Matzke H. Solid State Ionics 1984;12:25. [16] Matzke H. Defect Diffusion Forum 1992;83:111. [17] Andrievki RA, Eremeev VS, Zagryazkin VN, Panov AS. Akad Nauk SSSR Neorgan Mater 1967;3:2158 as cited in Ref. [8]. [18] Gupta PSH. Diffusion phenomena in thin films and microelectronics materials. NJ: Noyes Publication; 1987 (Chapter 1). [19] Ohring M. The materials science of thin films. San Diego: Academic Press Inc.; 1991 (Chapter 8).