1136
Langmuir 1987, 3, 1136-1140
Chemical Aspects of Reactive Metal and Energetic Ion Interactions on Polyimide Ralph G. NUZZO,* Y.-H. Wong, and G. P. Schwartz AT&T Bell Laboratories, Murray Hill, New Jersey 07974 Received May 28, 1987 X-ray photoemission and in situ oxygen uptake studies have been performed on polyimide surfaces expeaed to controlled ion etching, reactive metallization, or the two acting in concert. Although the ion etching process utilized in the current study was highly destructive in terms of bond scission and was found to exhibit enhanced selectivity for the removal of carbonyl oxygen atoms, the resulting surface was not observed to be reactive to a subsequent in situ oxygen exposure. This observation suggests that the bond cleavage induced by energetic ions does not produce a surface containing significant densities of stable organic radicals but, rather, results in the extensive cross-linking of the surface with the concomitant formation of highly unsaturated or graphitelike residues. Exposure of an unsputtered polyimide surface to A1 deposited under UHV conditions results in the reduction of surface carbonyl functional groups. We conclude, based on subsequentoxygen titration data, that there is little evidence for significant meta-carbon bond formation. Aluminum deposited on a sputtered polyimide surface shows little evidence for charge transfer between the metal and the carbonaceous, surface residue. The design and fabrication of very large scale integrated circuits (VLSI) presents challenges of unprecedented complexity. Many of these issues, involving both fundamental questions of chemistry and physics as well as applied engineering, relate directly to the extremely small sizes (of the order of lpm) and high number densities ( S O 6 transistors) of active elements in such devices. There is, therefore, a strong interest in examining material technologies that might simplify and/or improve the processes by which these devices are manufactured as well as in enhancing the reliability with which they operate. As a result, organic polymers, whose importance in such areas as lithography1y2and packaging2 is well-appreciated, are currently being considered with regard to their utility in replacing both active and passive elements in semiconductor devices. Much of this attention has focused on the development of improved insulators for reasons that are readily understood. First, many polymers have excellent dielectric properties. Second, they are easily patterned and/or processed. Third, polymers provide some promise with respect to being able to planarize the complex topographies of VLSI devices. All of the above-mentioned features make organic polymers extremely attractive candidates for the dielectric layers used in multilevel interconnection s y s t e m ~ . ~ The demands made on a thin film insulator in these systems, both in performance and in processing, are extreme. A candidate material must possess, among other things, high strength and thermal stability. It is also equally important that the material be chosen so as to minimize the differential thermal expansion coefficient between itself and other elements of the device. On the basis of these criteria, polyimide films are receiving considerable attention for such application^.^ In practice, however, several significant problems remain extant. Among these, the durability and strength of the adherent interface between polyimide films and various metals represent a key issue in technical applications. In this paper, we present the results of a study that seeks to address several points relevant to the interfacial chemistry associated with aluminum-polyimide junction^.^ (1)For general references see: Materials for Microlithography; Thompson, L. F., Willson, G. E., FrEchet, J. M. J., E&.; ACS Symposium Series 266;American Chemical Society: Washington, D.C., 1984. (2) For general overviews see: Sze, S. M., Ed. VLSI Technology; McGraw-Hill: New York, 1983 and references cited therein. (3) Wilson, A. M. Thin Solid Films 1981, 83, 145-163.
Since current process technologies typically utilize an ion sputtering processing step prior to depositing the aluminum in order to promote metal continuit9 and adherence! we have examined various aspects of the reactive chemistry of polyimide thin films with both energetic ions and aluminum. Using X-ray photoelectron spectroscopy as well as oxygen-uptake studies conducted in UHV, we have been able to characterize some of the extremely complex surface reactions that occur on this material. Of particular significance is our finding that, in gross chemical terms, Ar ion sputtering and interfacial reaction with aluminum metal seem to have qualitatively similar outcomes for a polyimide substrate surface. Both treatments yield a thin and largely carbonaceous surface, presumably of strong aromatic character, which is selectively depleted in specific surface functional groups. The data also suggest that the contributions made by reactive surface functionalities, such as radicals, anions (or u metal-carbon bonds), or cations, to the total chemical composition of the surface are minimal. Beyond these similarities, however, the two types of surface reactions produce materials with distinctly different physical and morphological characteristics.
Experimental Section Core level spectra were acquired with a Kratos Analytical Instruments XSAM-800 electron spectrometer. The data were obtained in the fixed analyzer transmission mode using Mg Ka (1253.6 eV) excitation with an instrumental resolution of -1.1 eV. The base operating pressure of the analysis chamber is 1 5 x lo-" Torr. This system was equipped wiii an effusive molecular beam source for gas dosing, an ion gun for sample cleaning and sputtering, and a resistively heated furnace for metal deposition. The aluminum metal used was of >99.99% purity. The depositions were conducted at a rate of 5-50 A/min, and the background pressure was maintained below 1 X Torr during metal deposition. The research purity argon and oxygen gases used were obtained from Matheson. (4) We have chosen aluminum as the metal in this study for two reasons. First, because of ita low resistivity and excellent resistance to corrosion, this metal is and will continue to be used extensively in semiconductor devices. Second, aluminum is a good representative of the class of highly reactive metals that one might expect to form stable metal-carbon bonds. One could, of course, make other choices, limited examples of which are available in the literature. See, for example: Sanda, P. N.; Bartha, J. W.; Clabes, J. G.; Jordan, J. L.; Feger, C.; Solverman, B. D.; Ho, P. s. J. Vac. Sci. Technol. A 1986, 4 , 1035-1038. ( 5 ) Vossen, J. L., Kern, W., Eds. Thin Film Processes; Academic: New York, 1978 and references cited therein. (6) See, for example: Chapman, B. Glow Discharge Processes; Wiley: New York, 1980; pp 279-283 and references cited therein.
0743-746318712403-1136$01.50/0 0 1987 American Chemical Society
Langmuir, Vol. 3, No. 6, 1987 1137
Metal and Energetic Ion Interactions on Polyimide 0
0
i
01s
I
PMDA- PDA POLY M I D E
n
P M D A - O D A POLYM I D E
H FDA - ODA POLY M I DE
Figure 1. Polymeric repeat units for various forms of polyimide.
The polyimide thin film samples used in this study were a proprietary formulation of Hitachi Chemical LM. This material (described by the trade names PIQXlOO and L110) is characterized by an extremely low coefficient of thermal expansion and is believed to have a composition primarily derived from the monomers pyromellitic dianhydride and p-phenylenediamine (PMDA-PDA).' The samples were prepared by spin coating a 1rM film of the "amic acid" precursor on a 4-in. silicon wafer. The samples were then fully cured at 400 OC. Prior to any of the analyses reported below, the samples were again heated in the UHV chamber at temperatures ranging up to 400 "C. The desorption and evolution of water from these samples were monitored with a quadrupole mass spectrometer positioned so as to directly "image" the sample surface.
-
Results and Discussion Our interest here is to examine the chemical basis associated with ion etching polyimide in order to promote adhesion to a reactive metal such as Al. Because the reaction products are potentially quite complex, we have attempted to isolate in a systematic fashion the individual effects of (1)energetic ions and (2) reactive Al atoms prior to examining surfaces that have been exposed to a combination of the two. The determination of absolute binding energies is complicated by sample charging artifacts associated with the insulating nature of the cured polyimide. In order to compare our results with recent photoemission studies of pyromellitic dianhydride-o~ydianiline~?~ (PMDA-ODA) and hexduorodianhydride-oxydianilinelo (HFDA-ODA) derived polyimides, we have adopted the following ad hoc convention for referencing our experimental binding energies on pristine samples. Examination of the three polyimide structures shown in Figure l indicates that the carbonyl site is essentially equivalent among these materials. Consequently, we adopt and assign the value of 288.9 eV for the C 1s "carbonyl" peak as reported in ref 10. Relative to this value then, our N 1s peak lies at 400.7 eV and the dominant C 1s signal, composed of unresolved contributions from both PMDA and PDA rings, peaks at (7)Pryde, C., personal communication. (8)Silverman, B. D.;Bartha, J. W.; Clabes, J. G.; Ho, P. S.; Rossi, A. R. J. Polym. Sci. Polym. Chem. Ed. 1986,24,3325-3333.Silverman, B. D.;Sanda, P. N.; Ho,P. S.; Rossi, A. R. J. Polym. Sci., Polym. Chem. Ed. 1985,23,2857-2863. (9) Sanda, P. N.; Bartha, J. W.; Clabes, J. G.; Jordan, J. L.; Feger, C.; Silverman, B. D.; Ho, P. S. J. VQC.Sci. Technol. A 1986,4,1035-1038. (10)Buchwalter, L.P.; Silverman, B. D.; Witt, L.; Rossi, A. R. J. VQC. Sci. Technol., A 1987,5 , 226-230.
, I
400
200
-
0
BINDING ENERGY ( e V )
Figure 2. Low-resolutionXPS data for three polyimide surfaces:
(a) a cured sample which had been annealed at 350 "C in UHV for 1h; (b) a sample annealed as in a but then sputtered at room temperature with 1000 eV Ar+ at 2 X Torr for 5 min; (c) a sample prepared and then sputtered as in b except for a duration of 1h; (d) a sample prepared as in c and then exposed to >1OOOOO L of Op The spectrum in a contains a poorly resolved shoulder at higher binding energy attributable to carbonyl groups.
285.5 eV. On the basis of these values, which are in agreement with the data and ab initio calculations of ref 8-10, we observe a uniform charging shift of 1.4 eV. On samples exposed to large ion fluences or heavy A1 coverages, we no longer observe sample charging. In fact the C 1s signal on heavily sputtered samples coincides both in binding energy (284.4 eV) and in line shape (asymmetric on the high binding energy side) with that previously reported by Hamrin et al.ll for graphite. For low ion fluences or small metal overlayer coverages, however, we have observed evidence for inhomogeneous charging. Since we do not have a flood gun in this system for neutralizing such surfaces, we have limited our discussion to qualitative observations in these cases. (A) The Influence of Argon Ion Bombardment. Argon ion sputtering dramatically alters the nature of a polyimide substrate surface. The qualitative characteristics of these effects can be seen clearly by inspection of the data shown in Figure 2, which presents survey spectra for four representative surfaces. The upper trace in Figure 2a corresponds to that of a cured polyimide sample that has been annealed at 350 "C for 1 h in ultrahigh vacuum. As revealed by mass spectroscopic analysis, this procedure results in the liberation of copious quantities of water from samples that have been air-exposed after their curing bake. (11)Hamrin, K.;Johansson, G.; Gelius, U.; Nordling, C.; Siegbahn, K. Phys. Scr. 1970,I , 277-280.
1138 Langmuir, Vol. 3, No. 6, 1987
Nuzzo et al.
It is interesting to note that, despite this considerable gas evolution, this spectrum is indistinguishable in almost every regard from that obtained for a cured sample that had not been annealed further in vacuo. The water evolution we observe presumably involving both physically entrained material as well as that derived from dehydration of residual amic acid moieties then must derive largely from the sample bulk. Spectrum b in Figure 2 shows data obtained for a sample that had received a “light“ dose of argon ions. In this instance, we used a 1-kV accelerating voltage and a gas pressure of -2 X 10” Torr. The sample was sputtered for -5 min.12 Under these operating conditions and given the placement of the ion gun, we estimate the beam current density at the sample to be 1 pA/cm2. The most significant feature to note in this particular spectrum is the apparent selectivity for the removal of the oxygen heteroatom on this surface. In this instance, the integrated intensity of the 0 1s core level relative to that of the C 1s peak has decreased by -60% while that of the N 1s core level has declined by -30%. The observation of a preferred removal of surface oxygen demonstrates that the ion milling process is not merely an indiscriminate ablation of the polymer. The consequences of continued sputtering of the substrate are revealed by the spectrum shown in Figure 2c. Using conditions comparable to those described above, only continuing the sputtering for a longer length of time (in this case 1h), we find that the surface is extensively degraded. There is evident an almost complete depletion of the surface in oxygen (originally present in carbonyl groups), and the imide-derived nitrogen atom, while still present, has been greatly diminished. The visible alteration of the surface stoichiometry is an immediate demonstration that the incident ions are effective agents for chemical bond scission. An extremely energetic bond-breaking process such as this potentially can generate radicals (and less likely anions and cations) in the degraded surface layer of the polyimide. Conventional wisdom would suggest that enhanced metal adhesion on such a surface is directly related to the enhanced reactivity and bond-forming characteristics of such defects. The initial generation of charged and radical defects does not address the more important issue, however, of whether such moieties are stable and, if so, over what time scale. Since organic radicals and most organic anions react efficiently with 0 2 , 1 3 one can test the reactivity of the sputtered surface by performing an oxygen exposure and XPS 0 1s intensity measurement, in essence a titration of reactive surface functional groups. The data in Figure 2d address this point. This spectrum corresponds to a polyimide surface that was sputtered for 1 h and then exposed to >100000 L (1L = 1langmuir = lo* Torrs) of OF As is evident, this spectrum is essentially identical with that in Figure 2c, indicating that very little O2 is incorporated via reaction with sputter-induced radicals or anions. The XPS data thus indicate that the residual density of such defects must be low (
