CHEMICAL INTERACTIONS BETWEEN CARBON SUBSTRATES AND METAL ATOMS

Transcription

1 CHEMICAL INTERACTIONS BETWEEN CARBON SUBSTRATES AND METAL ATOMS H. Toyota, T. Ide and H.Yagi Department of Mechanical Engineering, Faculty of Engineering, Ehime University 3 Bunkyo-cho, Matsuyama, Ehime , Japan Introduction Chemical interactions between substrates and metals appear to be very important in many industries. For example, adhesion and ohmic contact of a film on a substrate are significant factors in producing electronic devices. Strong adhesion and adequate chemical reactivity between the matrix and the filler are also key technologies in producing strong composite materials. Moreover, the weak adhesion and the chemical inactive between two contact materials are important in friction, lubrication, and the material purification technology. All such matters concerned with the interfacial chemical reaction can be considered as problems of wettability and atomic diffusibility. If they can be theoretically confirmed and then controlled, we can control any phenomenon related to the properties of solid surfaces, including friction, wear, molecular adsorption, and dispersion of fine particles. In general, the carbon material is used to refine the high-purity metal and semiconductor because of its chemical inactivity. Therefore, it is important to research the chemical reactions between carbon and metal atoms on the surface of carbon materials. In this study, we investigate the chemical interactions between carbon substrates and metal atoms by the experiments in ultra-high vacuum and ab-initio molecular orbital calculations. As the substrates, two types of substrates, glass-like carbon (GLC) and highly oriented pyrolytic graphite (HOPG), are adopted to study experimentally chemical reactions of solid carbon surface. Chemical reactions are investigated using pure liquid metals (Au, Ag, Cu, Fe, Al) on the substrate, and ultra-high vacuum ion bombardment process is used to keep the clean conditions of specimens during experiments. In order to explain the experimental results, ab-initio molecular orbital calculations according to Hartree-Fock method are done. Theoretical Expression In the majority of current theories, the wetting phenomenon is classified into three types; adhesional wetting, spreading wetting, and immersional wetting which we, however, consider as an apparently macroscopic one. Present experiments and calculations are carried out considering that the wetting phenomenon fundamentally depends on three elements, as shown in Figure 1. We consider that contact angles measured in wetting experiments are dominated by the element (I), and atomic diffusibility is dominated by the element (III). The interaction energy γ 0, that is, the adhesive work per unit area, is theoretically analyzed by considering the equilibrium of forces acting on the front atom of the liquid metal, and is described by equation (1) below, which is the well-known Young and Duprè expression. γ 0 = γ l (1+cosθ) (1) where γ l is the surface energy of liquid metal and θ is the contact angle. We adopt the γ l value from the Reference [1]. We can obtain γ 0 by measuring contact angle experimentally. Experimental To obtain a contamination-free contact between liquid and solid, experiments must be carried out using pure specimens under clean conditions in ultrahigh vacuum (UHV), because the wetting phenomenon is based on chemical reactions between the first-layer atoms of the surfaces in contact. Figure 2 shows the schematic of the apparatus used in the wetting experiment. A well-annealed substrate (GLC or HOPG (001) plate) with a purity higher than 99.99% is first cleaned by the ultrasonic method and fastened into an UHV chamber. Au, Ag, Cu, Fe, or Al with a purity of 99.99% is set in the chamber as a liquid metal specimen. Then the substrate is flushed under UHV at 1500 C during the degassing process of the chamber, and then cleaned together with the liquid metal specimen by glow discharge of Ar gas ( % purity). This cleaning process is intended to completely remove even slight contaminations such as an oxide film from the substrate and liquid metal surfaces [2]. The liquid metal specimen is evaporated onto the substrate, and then rapidly melted. The liquid metal is cooled after achieving stable wetting. During the wetting process, the substrate and the liquid metal surface are

2 continuously bombarded by Ar ions with an energy of 0.5keV. The steps of the wetting process are recorded using a video camera and an optical microscope of 100 magnification. After removing the substrate with the solidified liquid metal specimen from the UHV chamber, the contact angle and atomic diffusion are observed carefully using a scanning electron microscope (SEM). Figure 3 shows an example of the SEM image of wetting sample. Liquid metal changes into small spheres when atomic diffusion does not occur. The contact angle must be measured using a magnification higher than to obtain the correct contact angle expressed by Young and Duprè expression. Figure 4 shows a SEM image of HOPG substrate suraface damaged by liquid Al because of atomic diffusion. It is thought that the measurement of the contact angle is meaningless in such a case. We must distinguish the case of "atomic diffusion", shown in Figure 4, from cases of "wetting", shown in Figure 3. Because liquid Fe has different diffusibility between on the surface of GLC and on the surface of HOPG, ultimate analysis is done by the auger electron spectroscopy. Experimental Results and Discussion It cannot be specified by the experimental technique in this study whether the atom which belongs to either the substrate or the liquid metal diffuses into the other. Therefore, it is a somewhat vague expression but the word "atomic diffusion" is used, not especially specifying the direction of diffusion. Figures 5~14 show the experimental results of the atomic diffusibility of various liquid metals on GLC and HOPG substrates. Liquid Al atomically diffuses with both HOPG and glassy carbon substrates. Liquid Fe does only with GLC substrate, not with HOPG substrate. Other liquid metals do not with both GLC and HOPG substrate. When liquid Fe is on the HOPG substrate, it behaves same as liquid Cu, and hardly wet on the substrate comparing with liquid noble metals. To confirm this experimental fact concerning the liquid Fe, the liquid Fe and the interacting substrate surface were analyzed in the element with the scanning Auger electron spectroscopy. Figures 15 and 16 show the results of ultimate analysis, after 90 minutes contacts of liquid Fe on GLC and HOPG substrates, respectively. In each right figure, white area means element of Fe. Each left figure is the SEM photograph of analysis regions. Because the surface layer of each sample is removed by the ion bombardment before the ultimate analysis, the left upper half of the spherical Fe droplet is removed. It is understood that the spherical Fe droplet and the substrate surface is removed by the ion bombardment by about 1µm. Figures 17 and 18 show the auger electron spectra measured at the positions where the sign 1~4 adheres respectively. Atomic concentrations at all positions obtained from Auger electron spectra are shown in Table 1. From Figures 15 and 16, it is clearly understood that a lot of Fe is seen on the surface of the spherical droplet on the HOPG substrate while it is distributed to the entire GLC substrate. When Fe melts on the HOPG substrate, it changes into the spherical liquid droplets according to cohesive force, but when it melts on the GLC substrate, it is distributed in the entire substrate surface. It is understood from Table 1 that the purity of the Fe droplet is higher on the HOPG substrate than on the GLC substrate. The surface positions of 1 and 4 on the Fe droplet adhered on the HOPG substrate have high purities with about 80%. However the surface position of 2 on the Fe droplet adhered on the GLC substrate has low purity with about 30%, although it is the highest purity on the GLC substrate. Therefore, it can be concluded that liquid Fe and the HOPG substrate do not atomically diffuse compared with the case of liquid Fe and the GLC substrate. This experimental result, that liquid Fe does not have chemical reactivity with HOPG (001) surface, is the most important result, because liquid Fe has been generally thought to react easily or to have large atomic diffusibility. And this is important news for carbon industries. We reveal the reason using the ab-initio molecular orbital calculation, and report in next section. Averages of contact angles sampled in SEM images and estimated interaction energies using equation (1) are shown in Table 2, when atomic diffusion does not occur. All interaction energies are far smaller than those between liquid metals and tungsten substrate. The expression atomic diffusion means that the contact angles cannot be measured because of atomic diffusion. Calculations and Discussion We have calculated binding energies and hybridized orbitals of a two-atom molecular model, and reached the conclusion that wettability can be quantitatively determined by the binding energy of the two atoms. The calculated result also reveals that atomic diffusion occurs in cases where unpaired p- or d-electrons exist in both liquid metal and the substrate [3]. In order to estimate wettability more quantitatively and to obtain further information on hybridization during atomic diffusion, we perform cluster model calculations using the ab-initio molecular orbital system for supercomputer (AMOSS). The restricted open-shell Hartree-Fock (ROHF) method is used to calculate energy levels. Christiansen's effective core potential [4] is used and the electron correlation correction to energy is calculated using the Møller-Plesset second-order perturbation (MP2) theory. Basis sets for expressing molecular orbitals consist of

3 atomic orbitals that are linear combinations of Gaussian functions. Only the Gaussian function of each orbital which has the lowest exponent is separated and dealt with as a different base for expressing accurately spreading wave functions far from the atomic core. The Mulliken population analysis is applied to estimate the binding strength between target atoms. Figure 19 shows the calculation models. When we use models constructed by many atoms, we waste a lot of CPU time and hardly analyze hybridization between target atoms. Then we adopt the simplest models which have fundamentally similar frontier orbitals to those of actual material surface. We check the frontier orbitals of these models by comparing with results calculated using large cluster models which give us results quantitatively agreeable with the physical values of original material surface. C 2 H 4 model corresponds to a reaction between liquid metal and HOPG (001) surface. Distorted (Dis.) C 2 H 4 model corresponds to a reaction between liquid metal and diamond structure which is contained in micro grain boundaries of glass-like carbon. Although glass-like carbon surface contains both C 2 H 4 and Dis.C 2 H 4 structure, chemical reaction is thought to begin at the Dis.C 2 H 4 structure. Total energies are calculated approaching a liquid metal atom stepwise to a bond which connects carbon atoms of these models. Calculated results are shown in Table 2 and Figure 20. As shown in Table 2, the binding energies (B.E.) do not seem in good proportion to the experimental interaction energies. But comparing with calculated and experimental values on tungsten substrate introduced in previous papers, [2][3] e.g. Au on W, calculated B.E.= Hartree, experimental γ 0 = 2.24 J/m 2, binding energies qualitatively seems to be in good agreement with interaction energies. Atomic diffusibility is explained in terms of the configurations of frontier orbital hybridization. As observed in Figure 21 (a)(c), the Al or Fe p-orbital and the C-C antibonding orbital hybridize, and form HOMO, thus the C-C bond is weakened. In the other cases, as seen in Figure 21 (b)(d), the C-C antibonding orbital does not hybridize with the Fe orbital and does not generate HOMO, thus the C-C bond is not weakened. These orbital hybridizations seem to dominate the experimental phenomena of atomic diffusion. antibonding orbital of ethylene and forms HOMO in the cases of experimental atomic diffusion. Thus, we conclude that atomic diffusibility is dominated by the hybridization between the metal p- or d-orbital and the antibonding orbital of substrate atoms and by their stability. References 1. B. J. Keene, Review of data for the surface tension of pure metals, Int. Mater. Rev. 1993; 38 (4): H. Toyota, T. Ide, H. Yagi, Y. Mori, and H. Hirose, A New Measurement and Evaluation of Wettability between Solid and Liquid Metals in Ultra-High Vacuum, Jpn. J. Soc. Prec. Eng. 1998; 64 (5): H. Toyota, T. Ide, H. Yagi, H. Goto, K. Endo, H. Hirose, and Y. Mori, A new measurement and ab-initio molecular orbital calculation of wettability, Transaction of MRSJ 1996; 20: L.A. Lajon, P.A. Christiansen, R.B. Ross, T. Atashroo, and W.C. Ermler, Ab initio relativistic effective potentials with spin-orbit operators.iv. Cs through Rn, J. Phys. Chem. 1987; 87: Acknowledgments We are grateful to Dr. Toshikazu Takata for helpful discussion regarding the calculations using AMOSS. We are also indebted to Dr. Tetsuro Tojo and Dr. Sinsuke Goda for kindly providing the pure glassy carbon specimens. Conclusions Interaction energies between liquid metals and substrates were obtained. It was conclusively found that liquid Fe does not diffuse into a HOPG(001) surface but does into a glass-like carbon surface. By calculating the electronic states of simple cluster models using the AMOSS program, it was found that the p-orbital of the metal adatom hybridizes with the C-C

4 Figure 1. Three elements of wetting. Figure 2. Schematic apparatus for wetting experiment. TMP: turbo molecular pump RP: rotary pump Figure 3. Configuration of liquid Cu wetting on glass-like carbon (GLC) substrate. Measuring method of contact angle is also shown. Figure 4. Configuration of atomic diffusion. Liquid Al damaged GLC substrate. (1200 C, 90 min) Figure 5. Configuration of liquid Au wetting on GLC substrate. (1064 C, 1 min) Figure 6. Configuration of liquid Au wetting on highly oriented pyrolytic graphite (HOPG) substrate. (1064 C, 1 min)

5 Figure 7. Configuration of liquid Ag wetting on GLC substrate. (962 C, 1 min) Figure 8. Configuration of liquid Ag wetting on HOPG substrate. (962 C, 1 min) Figure 9. Configuration of liquid Cu wetting on GLC substrate. (1085 C, 1 min) Figure 10. Configuration of liquid Cu wetting on HOPG substrate. (1085 C, 1 min) Figure 11. Configuration of liquid Fe wetting on GLC substrate. (1540 C, 1 min) Figure 12. Configuration of liquid Fe wetting on HOPG substrate. (1540 C, 1 min)

6 Figure 13. Configuration of liquid Al wetting on GLC substrate. (660 C, 1 min) Figure 14. Configuration of liquid Al wetting on HOPG substrate. (660 C, 1 min) Figure 15. Ultimate analysis of liquid Fe on GLC substrate after 90 min contact. (a) SEM photograph, (b) Mapping indication of ultimate analysis. Figure 16. Ultimate analysis of liquid Fe on HOPG substrate after 90 min contact. (a) SEM photograph, (b) Mapping indication of ultimate analysis.

7 Figure 17. Auger electron spectra of measured at the positions where the sign 1 3 adheres on Fig. 15. Figure 18. Auger electron spectra of measured at the positions where the sign 1 4 adheres on Fig. 16. Table 1 Atomic concentrations at all positions obtained from Auger electron spectra. Substrate Element Position 1 Position 2 Position 3 Position 4 GLC Fe % C % HOPG Fe % C % Table 2 Averages of measured contact angles and estimated interaction energies. Liquid metal GLC substrate HOPG substrate γ 0 J/m 2 γ 0 J/m 2 Au Ag Cu Fe Atomic diffusion Al Atomic diffusion Atomic diffusion

8 (a-1) Atomic configuration of diamond surface reaction. (a-2) Corresponding simple model of diamond surface reaction. (b-1) Atomic configuration of graphite surface reaction. (b-2) Corresponding simple model of graphite surface reaction. Figure 19. Atomic configurations of Hartree-Fock calculation models. Left figures (a-1) (b-1) show actual surface reactions on diamond and graphite surfaces. Right figures (a-2) (b-2) show the simplest models which can explain frontier orbitals. (a) Remarkable adiabatic potentials. (b) Mulliken population analysis of the lowest potential curves. Figure 20. Calculation results. The vertical line in each figure means the position at the potential bottom of each case.

9 Table 2. Calculated binding energies and corresponding experimental interaction energies. A.D. : atomic diffusion. Distorted C 2 H 4 C 2 H 4 Metal Binding GLC Binding HOPG adatom Energy γ 0exp Energy γ 0exp Hartree* J/m 2 Hartree* J/m 2 Au Ag Cu Fe A.D Al A.D A.D. *1 Hartree = 27.21eV = J (a) Dis.C 2 H 4 - Al (b) Dis.C 2 H 4 - Au (c) Dis.C 2 H 4 - Fe (d) C 2 H 4 - Fe Figure 21. Configurations of hybridized frontier orbitals. The difference in orbitals is pronounced in the highest occupied molecular orbital (HOMO) and the lowest unoccupied molecular orbital (LUMO).