INTERDIFFUSION OF Au/Ni/Cr ON SILICON SUBSTRATE B. Yan, D. Lin, D. Mao To cite this version: B. Yan, D. Lin, D. Mao. INTERDIFFUSION OF Au/Ni/Cr ON SILICON SUBSTRATE. Journal de Physique Colloques, 1988, 49 (C5), pp.c5-631-c5-634. <10.1051/jphyscol:1988580>. <jpa- 00228077> HAL Id: jpa-00228077 https://hal.archives-ouvertes.fr/jpa-00228077 Submitted on 1 Jan 1988 HAL is a multi-disciplinary open access archive for the deposit and dissemination of scientific research documents, whether they are published or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers. L archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.
JOURNAL DE PHYSIQUE Colloque C5, supplbment au n010, Tome 49, octobre 1988 INTERDIFFUSION OF Au/Ni/Cr ON SILICON SUBSTRATE B. YAN, D. LIN and D. MA0 Shanghai Jiao Tong University, Department of Materials Science and Engineering, Shanghai 200030, China ABSTRACT By using surface resistance measurement and Auger Electron Spectroscopy, the interdiffusion behavior of Au/Ni/Cr metallization on silicon substrate was examined. Cr was proved to be an excellent diffusion barrier between Au and Si up to 450 C for a Cr layer of 500A thick which would avoid serious degradation of solderability and resistivity of the metallization. Ni was found much less effective as a diffusion barrier and therefore was mainly to improve the solderability. It was shown that silicon outdiffusion was dominant in the interdiffusion between gold overlayer and silicon substrate. It was attributed to the large grain boundary area and high defect density in the polycrystalline gold film. I. Introduction Au/Ni/Cr multilayer metallization is widely used in microelectronics for interconnections, such as bonding and soldering. Especially, it is used in silicon power devices in order to improve the thermal fatigue life. Interdiffusion of the Au/Ni/Cr metallization with silicon substrate is of great practical importance for it has both beneficial and harmful effects to the service properties of the metallization. Proper interdiffusion can improve the cohesivity among the thin films, whereas excessive interdiffusion will deteriorate the conductivity and solderability of the metallization. Cr is an ideal element to improve the adhesion to silicon and silicon dioxide, moreover, it has the thermal expansion coefficient very close to that of silicon, which is an important factor in thermal fatigue. Au has good solderability and is an extremely inert element to protect the metallization from being oxidized during storage. Ni is expected to suppress the interdiffusion between Au and Cr (1) and has good solderability as well. However the effectiveness of Ni as diffusion barrier has never been proved experimentally. There have been extensive work on the metallizations. In order to simplify the problem, most of the work were carried out on the multilayer thin films deposited on sapphire substrate or Si02 to avoid the interdiffusion between the metallization and the substrates, and the focus has mainly been on the interdiffusion within the metallizations. However, silicon outdiffusion through the metallization which is of great importance in the service properties is frequently ignored. The primary objective of this paper was to study the interdiffusion between the gold base metallization and the silicon suhstrate under various annealing temperatures by using the sheet resistivity measurement and Auger spectroscopy profiling technique. Therefore the effectiveness of Cr and Ni as diffusion barriers could be verified and the proper annealing temperature could be determined. Article published online by EDP Sciences and available at http://dx.doi.org/10.1051/jphyscol:1988580
JOURNAL DE PHYSIQUE 11. Experimental Thin films used in this study were deposited on polished silicon substrates. Cr and Ni were deposited in a three-hearth EB evaporator without breaking the vacuum. The vacuum before evaporation was loe6 torr. Because of its limited availability, gold was evaporated seperately in the resistance-heated evaporator. The exposure time to the air between the two evaporations was kept as short as possible. The thicknesses of the films are listed in Table 1 measured by the stylus profilometer. After evaporation, the specimens were annealed in N2 at 250 C, 380 C, 450 C and 530 C respectively for 1 hr. Table 1 Film thicknesses of the specimens Thickness (A) 1 350 c r I N i. I Au 2200 Sheet resistance measurement was carried out at the STZ-1 digital four-point probe. Detailed information of interdiffusion was obtained by using the Auger spectroscopy profiling technique on the PHI 610 Scanning Auger Spectroscope. 111. Results 1. Sheet resistances The measurements of the sheet resistance vs. annealing t.emperature for the three specimens are plotted in Fig.1. The sheet resistances of Au/Cr and Au/Ni in the as-deposited state are similar. If we assume that the gold layer has the main contribution to the electric conduction, the resistiviti-es calculated from the sheet resistances and film thicknesses of Au are 3.06 and 3.08 pn-cm respectively, which is slightly higher than the bulk resistivity (2.35 PO-cm) due to higher defect densities. The higher value of Au/Ni/Cr was probably the result of measurement scatter, it fell down to a lower value after annealing at 250 C for 1 hr. The sheet resistances increased with increasing temperature. At lower temperature, they increased rather slowly and then increased drasticall-y with the annealing temperature. For Au/Cr the turning point was at about 380 C, but for Au/Ni/Cr the drastic increase of the sheet resistance occurred atalower temperature, 250 C, and the resistances were much higher than that of the bilayer metallizations. Notably, Au/Ni showed lower resistances when annealed at 530 C for 1 hr. 2. Auger spectroscopy So The depth profiles were obtained by Auger spectroscopy. It was shown,,--.- *urn, that no significant interdiffusion was observed after annealing at 250 C for I,.. 1 hr which was in good agreement with the minor changes of the sheet resistances. :,,- --.- I I AUINI~CI,L I A~,',S, I. The composition distribution of as-deposited Au/Cr specimens and after #,,- annealing at 450 C and 530 C are given in Figure 2. Oxygen was found to exist in the Cr layer which was proved to be incorporated into t.he film during depo-.l- -----_-=-*.' sition process. On annealing at 450 C,a,,."a 300.no 500 for 1 hr, Cr and O diffused to the outer rrrnverature rcl surface of the specimen. The ratio of Cr to 0 at the surface is 0.71, very Fig.1 Sheet resistances close to that of Cr203 (0.66). The outvs. annealing temperature diffusion of Cr observed in this study is in agreement with the previous re- I' I ' i i i / / / / / p;,.9,g+"
ports (2-4). It should be noted that although the concentration of the Cr layer became rather low,~20at%, very little interdiffusion took place between Au and Si, showing excellent behavior as diffusion barrier. Once the chromium-rich layer disappeared, e,g. at 530 C, interdiffusion between silicon and gold took place significantly. Figure 3 shows the depth profiles of BulNi. Notice that no oxygen existed in the Ni layer which was attributed to its lower tendency to oxidation. After annealing at 380 C, Ni layer diffused to both sides of the film as shown in Fig.3b. Contrary to Au/Cr, significant interdiffusion between Au and Si took place in spite of the existance of the Ni-rich layer with Ni concentration of -30at%. When the annealing temperature was raised to 450 C, Ni peak no longer existed. The composition profiles of Au/Ni/Cr specimens are given in Fig.4. Again, it confirms the existance of oxygen in the Cr layer. The interdiffusion behavior is almost the combination of Au/Cr and Au/Ni. It is interesting to note that Cr outdiffusion occurred at 250 C which was lower than that in the Au/Cr specimen, moreover, Cr layer also disappeared at a lower temperature, 450 C, instead of 530 C as in the Au/Cr specimen. IV. Discussion The present results clearly show the excellent property of Cr in suppressing the interdiffusion between Au and Si. As can be seen from Fig.Zb, Si had much less concentration in the Cr-rich layer than Au. This fact is also confirmed in the Au/Ni/Cr specimen shown in Fig.4b. It implies that Si has much lower diffusivity in the Cr layer than Au. However, after the diffusion barrier disappeared and the interdiffusion between Au and Si took place, Siconcentration in the Au layer was much higher than that of Au in silicon substrate as shown in Fig.4~. Since silicon and gold have negligible solubility between each other, the higher concentration of Si in the gold layer might be due to the higher defect density and large grain boundary area of the polycrystalline gold film, i.e. Si can be segregated at the crystal defects and grain boundaries, whereas no grain boundary exists in the silicon substrate resulting in the very low Au concentration. By the same reason, the diffuspvity of Si in the Au layer is higher than that of Au in Si substrate. In other words, silicon outdiffusion is dominant in the interdiffusion between Si and Au. Ni showed rather poor barrier behavior to the interdiffusion between Si and Au. As a matter of fact, Au has higher grain boundary diffusivity in,cr ( E=0.68eV(5)) than in Ni (E=2.6eV, estimated from its bulk value(6)). Since Au diffusion is not the controlling process, the poor barrier behavior shows a higher diffusivity of Si in Ni. More quantitative work is underway and will be reported elsewhere. It is observed that interdiffusion between the Au and Si is the dominating factor in t.he increase of the sheet resistivity. Surprisingly, the sheet resistance of Au/Ni specimens after annealing at 530 C decreased although significant interdiffusion between Au and Si occurred as shown in Fig.3~. The depth profile of Fig.3~ shows that even after high temperature annealing, the Au concentration at the surface I.ayer is still relatively high (-80at%). Our SEM observations of the surface morpholog,y, which will be published seperately, showed that many nickel silicide crystals surrounded by Au lumps were observed on the surface of the Au/Ni specimens. As the annealing temperature increased, the silicide grew and the surrounging Au lumps were more likely to be connected together so as to decrease the sheet resistance when four probe measurement was made. Contrary, in the Au/Cr and ~u/ni/cr specimens, due to the outdiffusion of Cr and the formation of chromium oxide at the surface, gold concentration at the surface layer was low (-50-60at%). The existance of the rather thick layer of chromium oxide (-3001) and the high Si concentration in the underneath layer contri.bute to, the- higher sheet resistance. It was.observ&d that the Au/Ni/Cr specimens showed higher interdiffusion than in the bilayer specimens. The chromium outdiffusion seemed to be enhanced in the three layer metallization. There are two possibilities First, the thickness of Cr layer in the triple layer specimen (2501) was less than that in the bilayer specimen (350ij.
C5-634 JOURNAL DE PHYSIQUE Second, the outdiffusion of Cr through the Au layer might be enhanced by the existance of Ni as solute. The enhancing effect of Ni on the self diffusivity of Cr was reported by Askill(7). We think the solute enhancement in the present case is very likely. Further study is needed. V. Conclusions 1. Interdiffusion between Au and Si substrate causes significanz increase of sheet resistivi-ty and deteriorate the solderability. 2. Cr is not only.an adhesive layer to the silicon substrate, but also an ideal diffusion barrier for the interdiffusion between Si and Au. A Cr layer of 5001 thick can maintain its barrier property up to 450 C. 3. Silicon outdiff usion is dominant in the interdif fusion between gold overlayer and siiicon substrate. It is attributed to the large grain boundary area and high defect density in the polycrystall.in~ thin fiim. Therefore, a good diffusion barrier must have low Si diffusivity. 4. Ni can neither suppress the interdiffusion between Au and Si nor the outdiffusion of Cr as expected. References 1. J.R.Rairden, C.A.Neugebauer, and R.A.Sigsbee, Met. Trans. 2, 719, 1971 2. P.H.Holloway, Solid State Technology, Feb. 1980, p.109 3. A.Munitz and Y.Komen, Thin Solid Films, 2, 177, 1980 4. L.S.Weinman, T.W.Orent, and T.S.Lin, Thin Solid Films, 72, 143,1980 5. G.Elajni, G.Ottaviani and tl.prudenziati, Thin Solid Films, 38, 15, 1976 6. T.G.M.Van den Belt and J.H.W.de Wit, Thin Solid Films, 109, 1, 1983 7. J.Askj.11, Appl. Phys. Letters, 2, 82, 1966 Fig.2 Depth profiles of Fig.3. Depth profiles of Fj.g.4. Depth profiles of Au/Cr spec:imens Au/Ni specimens Au/Ni/Cr specimens