Chem. Mater. 1993,5, 1701-1709
1701
Metalorganic Chemical Vapor Deposition of Copper from Copper(I1) Dimethylamino Ethoxide V. L. Young,? D. F. COX,* and M. E. Davis*?$ Department of Chemical Engineering, Virginia Polytechnic Institute and State University, Blacksburg, Virginia 24061 Received March 9, 1993. Revised Manuscript Received October 7, 1 9 9 9 Copper metal films were grown on single-crystal strontium titanate (100) by the thermal decomposition of copper dimethylamino ethoxide in inert atmosphere at temperatures between 150 and 270 "C. Films grown a t 200 "C are copper metal, free from contaminants, while higher temperatures result in significant carbon and oxygen incorporation. Deposition products were identified by Fourier transform infrared spectroscopic analysis of the reactor gas phase in situ and by mass spectroscopic analysis of the reactor exit gas during deposition. A t 200 "C, deposition occurs by interdependent ,&hydride elimination and reductive elimination reactions which produce (dimethylamino)ethanal, (dimethylamino)ethanol, and copper metal. @-Hydrideand reductive elimination reactions are also dominant at 250 "C; however, the competition of ligand fragmentation reactions with the whole-ligand eliminating reactions leads to carbon and oxygen contamination of the copper metal film.
Introduction
temperatures above 250 0C.13g Further development of precursors requires a greater understanding of deposition Because of ita high conductivity and resistance to chemistry. electromigration,l copper metal has potential microelecRecognizing the importance of copper thin film growth, tronics applications. Additionally, copper oxide film we have investigated in situ the decompositionchemistry growth may be valuable for semiconductor/superconductor of a volatile copper MOCVD precursor. Only a handful hybrid technology. Film growth by metalorganic chemical of studies address the chemistry of copper deposition, and vapor deposition (MOCVD) requires precursors of high few involve in situ analysis. It is usual to collect volatility and low decompositiontemperature. Recently, several potential copper MOCVD precursors have been decomposition products exiting the deposition zone for reported. Copper tert-butoxide,2 copper &diketonate analysis after deposition is complete. Some products may triakylphosphines,3 copper cyclopentadienyltrialkylphosfurther react between collection and analysk6 Here we phines? copper alkoxytrialkylphosphines,5copper alkyne use a combination of Fourier transform infrared and mass fl-diketonates,G fluorinated copper D-ketoimine~,~ and spectroscopies to investigate the MOCVD of copper from copper o-diketonate cycloalkenes8 have been tested, in a copper alkoxide. addition to the commercially available copper &diketoA copper alkoxide was selected for our study for two nates (fluorinated and unfluorinated). The thermal reasons. (i) Alkoxides decompose at relatively low temproperties of several of these were recently s ~ r v e y e d . ~ peratures. (ii)Alkoxides contain copper bonded to oxygen Although ease of handling and commercial availability and may, therefore, be useful for depositing either metal make the 8-diketonates convenient, they deposit copper or oxide films. Control of the copper oxidation state has metal only in the presence of a hydrogen cofeed1° and at been demonstrated for deposition from copper tertbutoxide under ultrahigh vacuum.2J1 We have synthesized * To whom correspondence should be addressed. and tested copper tert-butoxide and found that it is not + Preeentaddreee: CentreforAtmoephericChemiatry,YorkUniversity, 4700 Keele Street, North York, Ontario M3J 1P3,Canada. sufficiently volatile for study in moderate vacuum.12 8 Present address: Department of Chemical Engineering, California Institute of Technology, Pasadena, CA 91125. Copper(I1) dimethylamino ethoxide, Cu(OCHzCH2NAbstract published in Advance ACS Abstracts, November 15,1993. (CH3)2)2,is a commercially available, sublimable, water(1) Pei, P. L.;Ting, C. H.; Chiang, C.;Wei, C. S.; Fraeer, D. B. Mater. Res. SOC.Symp. Proc. VLSI V 1990,359. sensitive solid. Each ligand forms a fivemembered chelate (2)Jeffries, P. M.; Girolami, G. S. Chem. Mater. 1989,1, 8. ring with the central copper,13binding to the copper via (3) Shin, H. K.; Chi, K. M.; Hempden-Smith, M. J.; Kodas,T. T.; the nitrogen (lone-pair association with the central metal Farr, J. D.; Paffett, M. Adv. Mater. 1991,3, 246248. (4)Beach, D. B.; LeGoues, F. K.;Hu, C. K. Chem. Mater. 1990,2, ion) and the oxygen (covalent bond with the central metal 216219. ion). Hydrolysis of the alkoxide yields (dimethylamino)(5) Hampden-Smith, M. J.; Kodas,T. T.; Paffett, M.; Farr, J. D.; Shin, H. K. Chem. Mater. 1990,2,636-639. ethanol, HOCH&H2N(CH&. This alcohol also has afive(6) (a) Chi, K. M.; Shin,H. K.;Hampden-Smith, M. J.; Kodas,T. T.; membered ring structure,as the alcoholicproton hydrogen Duesler, E. N. Inorg. Chem. 1991,30,4293-4294. (b) Jain, A.; Chi, K.
M.;Kodas,T.T.;Hampden-Smith,M. J.;Farr,J.D.;Paffett,M.F.Chem. Mater. 1991,3,995-997. (c) Baum, T. H.; Larson, C. E. Chem. Mater. 1992,4,365. (7)Fine, S. M.; Dyer, P. N.; Norman, J. A. T.; Muratore, B. A,; Iampietro, R. L. Mater. Res. SOC.Symp. R o c . 1990,204,415. (8) Kumar, R.;Fronczek, F. R.; Maverick, A. W.; Lai, W. G.; Griffin, G. L. Chem. Mater. 1992,4,577. (9) Gross, M.E. J. Electrochem. Soc. 1991,138,2422-2426. (10) Arita, Y. Mater. Res. SOC.Symp. Proc. VLSI V 1990, 335.
(11)Jeffries, P. M.; Duboie, L. H.; Girolami, G. S. Chem.Mater. 1992, 4 , 1169.
(12)VandigrifftYoung, V.L.TheChemiatryofMetalorganicChemical Vapor Deposition from a Copper Alkoxide Precursor. Ph.D. Dissertation, VirginiaPolytachnicInstituteandStateUniversity,Blackeburg,VA, 1992. (13)Goel,S.C.;Kramer,K.S.;Chiang,M.Y.;Buhro,W.E.Polyhedron 1990,9,611.
0897-475619312805-1701$04.00/0 0 1993 American Chemical Society
1702 Chem. Mater., Vol. 5, No. 12, 1993
' 'tovent
Figure 1. Schematic of deposition system.
bonds intramolecularly with the lone pair on the nitrogen.14 The decomposition temperature of copper dimethylaminoethoxide is reported to be either 18413or 120 OC.15 Thus, this amino alkoxide may be useful for MOCVD below 250 OC. It is reported that copper dimethylamino ethoxide decomposes to a mixture of copper metal, copper(1) oxide, and copper(I1) oxide when heated under nitrogen flow from 25 to 300 "C over 2 h.'3 This observation suggests that the as-deposited oxidation state of the film may be controlled by adjusting the deposition conditions. The objective of our work is to elucidate the decomposition chemistry of copper dimethylamino ethoxide using in situ analytical techniques. An increased understanding of MOCVD chemistry should lead to improved precursor design and to intelligent selection of deposition conditions.
Experimental Methods and Procedures Figure 1 is a schematic of our deposition system. The deposition reactor/infrared cell was a standard, ultrahigh vacuum compatible, stainless steel cube with 2.75-in. nominal diameter conflat flanges. All gaskets were copper. The side windows were calcium fluoride, which pass infrared radiation down to a frequency of lo00 cm-l. Calcium fluoride was chosen for its resistance to hydrolysis. These infrared-transparent viewports allowed analysis of the gas within the reactor by infrared spectroscopy. A flow of warm helium (2 sccm, 99.999% pure) was directed against the windowe to reduce condensationon them. The reactor/IR cell was wrapped and warmed with heating tape to reduce condensation of the precursor on the walls. The reactor outside wall temperature was 110 OC. The reactor/IR cell was placed in the optical bench of a Fourier transform infrared spectrometer (FTIR). The sample stage (MACORmachinableceramic,Dow Corning) was mounted on a rotary feedthrough. A feedthrough opposite the sample stage provided ceramic-insulated electrical connections for a thermocouple and resistance heater. The substrate was clamped in place by titanium shims. Platinum foil sandwiched between the substrate and the sample stage served as a resistance heater. The foil was connected to the electrical feedthrough by copper braid insulated with ceramic braid (Omega). A hole through the bottom of the sample stage passed a type K thermocouple (Omega, sheathed), which touched the (14) Nyquist, R. A. The Interpretation of Vapor Phase Infrared Spectra; Sadtler Research Labs: Philadelphia, PA, 19W,pp 116-118, 12815) Frank Wagner, Strem Chemicals, Newbury, MA, personal communication.
Young et al. bottom of the substrate. "Substrate temperature" refers to this thermocouple reading. The system was thoroughly tested to ensure that copper deposition did not result from the copper hardware or from precursor decomposition on the warm reactor walls. No deposition occurred, and no organic species was observed by FTIR, without the precursor in the reactor. With the precursor present, deposition products were observedby FTIR only when deposition occurred on the substrate, never when the substrate was unheatd. The precursor was held in a flowthrough glass tube with a built-in bypass line. To reduce hydrolysis of the water-sensitive precursor, the tube was treated with dichlorodimethylsilane (Aldrich Chemical), rinsed with distilled water, and baked overnight. The carrier gas was 99.999 % helium, flow rate 5 sccm. All flow lines and fittings were 316 stainless steel warmed to an outside wall temperature of 100OC. The carrier gas and window purge flows were controlled by Brooks 5850E mass flow controllers. The reactor outlet gas could be sampled via a metering valve for mass spectral analysis. The reactor outlet line was connected to a direct drive mechanical vacuum pump. Experiments were conducted at reduced pressure (estimated Torr). Before each experiment, gas flow rates and system temperatures (except for the precursor tube) were maintained for at least 48 h to minimize water in the system. During each experiment, deposition conditions were maintained for 48 h. Our reactor was designed for in situ IR analysis, not for optimum film growth. Our aim was to grow films thick enough for surface analysis to be meaningful (after ion bombardment to remove surface contamination from air exposure after deposition), yet to keep the length of each experiment manageable. Unfortunately, deposition rates could not be measured directly in our system. Our deposition time does not indicate the intrinsic deposition rate possible with this precursor. The precursor, copper(I1) dimethylamino ethoxide (Strem Chemicals),was used as received. Elemental analysis (Galbraith Laboratories, Inc.) yielded the expected c0pper:carbon:nitrogen ratio. The precursor was transferred to the precursor tube in a drybox. The tube's built-in bypass line isolated the precursor while the flow system was flushed before each experiment. The precursor could be held undecomposed in the precursor tube for at least 10 days. We observed that copper dimethylaminoethoxide sublimed between 90 and 98 OC at an estimated pressure of 10-2 Torr. The substrates were single crystal strontium titanate (0.25 X 0.25 X 0.02 in.) oriented in the (100) direction and polished on one side (Commercial Crystal Laboratories, Inc., Naples, FL). Deposition was performed on the polished side to clarify the boundary between film and Substrate. Strontium titanate was selected because its bandgap was 'appropriate for planned photoassisted deposition experiments with copper alkoxides,and thermal deposition results on the same substrate were required for comparison.12 Before each deposition, the strontium titanate was held at 500-600 OC for 2-4 h under reduced pressure and 2.0-4.0 sccm helium flow. After cooling, the crystal ranged from colorless, with some black color centers, to mustard yellow. Such colors indicate an oxygen loss of less than 0.1 atom % This reduction was carried out in the reactor/IR cell, and the substrate was isolated from laboratory air until after deposition. Crystals were easily reoxidizing by heating in air. In addition to reducing the bulk (which reduces charging during surface analysis), this heat treatment cleans the surface of the crystal. Organic surface contaminants combine with surface lattice oxygen at high temperature and are burned off. The heat treatment also reorders the surface, which has been damaged by polishing and chemical cleaning. The influence of heat treatment on surface ordering and cleanliness was confirmed by LEED and XPS,respectively, in a separate vacuum system. The Fourier transform infrared spectrometer was an IBM IR/ 32 with an IBM 9OOO computer for control and data analysis. Apodization was triangular and resolution was 4 cm-l. Infrared peak positions were observed to be consistent within 1cm-1. Mass (16) Michael Urbanik,CommercialCrystalLaboratories,Inc.,Naples, FL, personal communication.
CVD of Cu from Copper(Il) Dimethylamino Ethoxide
Figure 2. Vapor-phase spectra of (a) (dimethy1amino)ethanol and (b) copper dimethylamino ethoxide. spectral analysis was performed with a Hewlett-Packard 5970A series mass selective detector (MSD) with a Hewlett-Packard 9825B calculator for control and data analysis. Atomic composition of the films was determined by Auger and X-ray photoelectron spectroscopy. Auger analysis was performed with a Perkin-Elmer PHIS10 scanning Auger microprobe. X-ray photoelectron spectroscopy was performed on a Perkin-Elmer PHI-5400. Scanning electron micrographs were obtained on a Cambridge Instruments Stereoscan 200. Upon removal from the reactor, films were classified as conductive or nonconductive. Conductivity was tested by touching the leads of a Fluke 75 multimeter to opposite corners of the film (separation about 1 cm). A film was defined as conductive if it triggered the continuity signal of the multimeter, indicating the resistance between the leads was less than 150 a. Four-point probe resistivity measurements of some films were also made. Values were corrected for the finite dimensions of the sample." Samples were stored in a vacuum desiccator after their initial classification as conductive or nonconductive and were replaced in the vacuum desiccator after four-point probe measurements. Resistivity was measured before any other film analysis took place. Films stored in the desiccator showed less postdepositioncontamination from carbon and oxygen than those held for similar times in air.
Results Spectral Standards. Vapor phase infrared and mass spectra of copper dimethylamino ethoxide, Cu[OCH2CH2N(CH&]2, and (dimethylamino)ethanol, HOCH2CHzN(CH& (Aldrich Chemical) were collected (Figure 2). Collected spectra of (dimethy1amino)ethanol match published spectra,la but no published spectra of the alkoxide are available. The IR spectra of the alcohol and alkoxide may be distinguished by differences in peak intensity and position that result from the substitution of copper for hydrogen. The asymmetric stretch of CH2 groups in the five-membered ring appears at 2960 cm-l for the alcohol and a t 2971 cm-l for the alkoxide. For (17) Logan,M. A. Bell Syst. Technical J. 1961,40, 885. (18) (a) Nyquist, R. A. The Interpretation of Vapor Phase Infrared Spectra; Sadtler &search Labs: Philadelphia, PA, 1988. (b) Sadtler Book ofStandardInjrared Spectra;Sadtler ReeearchLabe: Philadelphia, PA, 1978.
Chem. Mater., Vol.5, No. 12, 1993 1703
comparison, the asymmetric CH2 stretch appears at 2970 cm-I for cyc10pentane.l~ The peak at 1278 cm-' for the alcohol (1266 cm-l for the alkoxide) is related to C-N stretching in a tertiary N,N-dimethylamine,lg as is the peak at 1084 cm-l for the alcohol (1090 cm-l for the alkoxide). The alcohol spectrum has peaks near 3500cm-l associated with the O-H groups.lhJ9 Frequencies below 1000 cm-I are not accessible with our experimental apparatus because calcium fluoride windows do not transmit in that region. Certain of the possible volatile decomposition products of the alcohol and alkoxide (formaldehyde, acetaldehyde, (dimethylethyllamine,acetone) can be purchased (Aldrich Chemical) and handled readily in the experimental system. Vapor-phase infrared and mass spectra of these compounds were collected to allow their identification in spectra collected during deposition. Collected spectra were consistent with published standards.'*J9 The alkoxide cannot be distinguished from the alcohol by mass spectral analysis in our experimental apparatus. No copper-containing fragments are detected. (Coppercontaining fragments are identifiable by the isotopic ratios of 63Cuand 6sCu. A fragment containing copper has peaks in a ratio of 1.0:0.45,the second 2 a m u heavier than the first.) Copper-containing fragments may occur primarily as neutrals or negatives, which are not detected. This has been observed in the mass spectral analysis of other metalorganics using electron impact ionization.20 Copper fragments were also not observed from other volatile copper compounds insensitive to hydrolysis (Le., copper(I1) acetylacetonate) which indicates that premature decomposition of the precursor does not prevent coppercontaining fragments from reaching the detector. Use of another ionizationtechnique to obtain a standard spectrum of the alkoxide was considered and discarded. Only the electron impact mass selective detector was available for integration into the experimental flow system, so it would still be impossible to identify the alkoxide by mass spectrometry under deposition conditions. Most of the expected organic decomposition products have similar fragmentation patterns. The most abundant peak in the mass spectrum should be 58amu for any species containing the group (CH&NCH2. Products must be identified by small differences in the relative abundances of fragments, which are unreliable unless the instrument is running near steady state. In our system, pulsed samples were required to avoid flooding the mass spectrometer. The mass selective detector proved most useful for identifying low molecular weight products. For example, C02 observed in the IR spectrum might be due to a change in the IR bench purge gas (dry air) or in the reactor atmosphere. In contrast, the mass selective detector samples only the reactor atmosphere, so C02 detected by the mass spectrometer is certainly a reaction product. Decompositionof Copper DimethylaminoEthoxide. The purpose of this experiment was t o obtain a preliminary (19) (a) Szy"ki,H.A.ASystematicApproach to thelnterpretatio ofInfrared Spectra; Hertillon: Buffalo,NY, 1967. (b) Szymaneki, H.A. Interpreted Infrared Spectra; Plenum: New York, 1964; Vol. 1. (c) Szymanski, H. A. Interpreted Infrared Spectra; Plenum: New York, 1966; Vol. 2. (d) Szymanski, H. A. Correlation of Infrared and Roman Spectra of Organic Compounds; Hertillon: Philadelphia, PA, 1969. (e) Bellamy, L. J. The Infrared Spectra of Complex Molecules, 3rd ed.; ChapmanandHall: London,1975. (0Bellamy,L.J. TheInfraredSpectra of Compler Molecules, 2nd ed.; Chapman and Halk London, 1980; Vol. 2. (g) Socratea, G. Infrared Characteristic Group Frequencies; John Wiley and Sons: Chichester, 1980. (20) Miller, J. M.; Wharf,I. Can. J . Spectrosc. 1987, 32 (l), 1.
Young et
1704 Chem. Mater., Vol. 5, No. 12, 1993
il
117501
15001
12501
1OCO
Wavenumbor, em''
Figure 3. Infrared spectra of volatiles from copper dimethylamino ethoxide at (a) 150 "C, (b) 180 "C, and (c) 210 "C.
idea of the changes in chemistry with temperature in the thermal decomposition of the precursor. Copper(I1) dimethylamino ethoxide was placed in the precursor tube, and the system prepared as for deposition. The temperature of the precursor tube was ramped from 20 to 230 "C under vacuum with 5 sccm flow of helium carrier gas. The precursor decomposed in the precursor tube. The volatile species leaving the tube were analyzed as they passed through the reactor. There was no deposition on the substrate, which was held at 90 "C. Below 150 "C, infrared analysis shows the presence of copper(I1) dimethylamino ethoxide. A t a precursor tube temperature of 150 "C, the spectrum is still typical of the alkoxide, but a small peak near 1750cm-l appears (Figure 3a). Absorption near 1750 cm-' is characteristic of an aldehyde or ketone carbonyl stretch. The same peak at 1750cm-l appears when (dimethy1amino)ethanol contacts strontium titanate at 150-355 "C. With the precursor at 180 "C, the carbonyl stretch at 1750cm-l is more obvious (Figure 3b). Other peaks begin to change in intensity and position, indicating a mixture of the alkoxide and alcohol. With the precursor tube at 210 "C (Figure 3c) alcohol peaks dominate the spectrum. The peak at 1750 cm-1 is strong, and a new peak is just visible above the background near 1640 cm-1. The only change in the infrared spectra as the tube temperature reaches 230 "C is a moderate increase in the 1640cm-1 peak intensity, The same product results from decomposition of (dimethy1amino)ethanol over strontium titanate at 350 "C. The peak near 1640 cm-l may indicate a C=N or C=C stretch. Both imines and carbon-carbon double bonds absorb in this region.19 There is no basis for a conclusive identification of this species. A small peak at 1725 cm-l appears in each spectrum in Figure 3. This peak often appears in the vapor phase spectrum of copper dimethylamino ethoxide in our deposition system. The peak is absent under identical conditions with (dimethy1amino)ethanol in the system
41.
Figure 4. Residual spectrumshowing aldehydic decomposition
product.
rather than the copper amino alkoxide. Comparison with standard spectra shows that the 1725-cm-l peak is not a carbonyl stretch of acetaldehyde, formaldehyde or acetone, (cleaning solvents and potential decomposition products). During deposition in the reactor, the peak intensity depends not on substrate temperature but only on precursor temperature. We conclude that the peakat 1725 cm-l indicates a minor contaminant in the precursor which does not participate in the decomposition chemistry. A peak at 1750 cm-l is characteristic of a vapor-phase aldehyde or ketone C-0 stretch. Apparently, oxidative dehydrogenation of the alkoxy group begins at 150 "C. Likely aldehyde products are (dimethy1amino)ethanal ((CH&NCH2CHO), formaldehyde (CH20), and acetaldehyde (CH3CHO). No skeletal rearrangements are necessary to form these aldehydes. A ketone product is unlikely because its formation involves either migration of the oxygen on the ligand, or formation of a carboncarbon bond (i) to couple ligand fragments or (ii) to couple a ligand fragment and a whole ligand. Production of ketones from primary alkoxy species has not been observed over copper or copper(1)oxide,21-%titanium oxide (Ti02),25 or zinc oxide (ZnO).% Primary alkoxy specieson strontium titanate, glass, and copper(I1)oxides have not been studied, so it is possible that migration or coupling reactions to form ketones do occur on these surfaces and/or in the presence of a copper(I1) metal center. However, this scenario seems unlikely. A residual spectrum containing the 1750-cm-1product is shown in Figure 4. Comparison with infrared spectra obtained in this laboratory and with published spectra's show that the 1750-cm-l carbonyl stretch is not acetone, formaldehyde, acetaldehyde, dimethylethylamine, dimethylamine, trimethylamine, propionaldehyde, methyl ethyl ketone, diethyl ketone, or ammonia. The peaks in the residual spectrum are consistent with those expected for (dimethylamino)ethanal, (CH&NCH2CH=O. Absorption due to a CEO stretch could occur at 1750cm-1. There is only one peak between lo00 and 1100 cm-1 (at 1039cm-I), as expected for a compound with C-N bonds but no C-0 bond. The peak at 1271 cm-l can be reasonably assigned to a C-N stretch of a tertiary N,Ndimeth~1amine.l~ Peaks between 1150 and 1300 cm-l are Wachs, I. E.; Madix, R. J. J. Cntal. 1978,53, 208. Bowker,M.; Madix, R. J. Surf. Sci. 1982, 116, 549. Cox, D. F.; Schulz, K. H. J. Vac. Sci. Technol. A , 1990,8,2599. Schulz, K. H.; Cox, D. F. J. Phys. Chem. 1993,97,647. Kim, K. S.; Barteau, M. A. J. Mol. Catal. 1990,63, 103. (26) Vohs, J. M.; Barteau, M. A. Surf. Sci. 1989, 221, 590.
(21) (22) (23) (24) (25)
CVD of Cu from Copper(Il) Dimethylamino Ethoxide
absent from the infrared spectra of non-amine aldehydes (e.g., acetaldehyde). However, in this region the spectrum is quite similar to that of dimethylethylamine. In addition, the two-pronged shape of the C-H vibration absorptions (just below 3000 cm-1) is common to amines and aldehydes but not to ketones. Thus the difference spectrum is most likely of an aldehyde containing a tertiary dimethylamine group. The major thermal decomposition product of copper(I1) dimethylamino ethoxide is thus indirectly identified from the IR spectrum as (dimethylamino)ethanal. Note that the spectrum cannot be reproduced by adding the spectra of dimethylethylamine and any of the other aldehydes or ketones considered as possible products. Unfortunately, we could not obtain either (dimethy1amino)ethanal or a published infrared spectrum of it. By the end of the decomposition experiment, the precursor tube was coated with a deposit ranging in color from yellow to rose-red, suggestive of copper metal and/or copper(1) oxide. The red regions turned black, a color characteristic of copper(I1) oxide, shortly after exposure to ambient air. The deposits dissolved quickly in aqua regia but appeared untouched by acetone. This suggests the deposits were largely inorganic not organic. Thus, there is qualitative evidence for deposition of mixed oxidation state copper on the tube walls. In summary, copper dimethylamino ethoxide begins to decompose at 150 "C to (dimethy1amino)ethanal. (The same product is observed when (dimethy1amino)ethanol contacts 150 "C strontium titanate.) As the tube temperature approaches 200 "C, significant (dimethylamino)ethanol is detected as well as (dimethy1amino)ethanal. Above 210 "C, a new product appears, which may contain a carbon-nitrogen or carbon-carbon double bond. This is probably the same product formed by the alcohol at 350 "C and may result from fragmentation of the alkoxy group. Deposition on Strontium Titanate. On the basis of the results of decomposition in the precursor tube, substrate temperatures of 150, 200, and 260 "C were selected for study. No deposition was expected below 150 "C. Deposition was expectedat 260 "C, but contamination due to ligand fragmentation was a concern. A t 200 "C, deposition was expected, but decomposition was not expected to involve significant ligand fragmentation. Preliminary experiments showed that the major products at 200 OC are (dimethy1amino)ethanol and (dimethylamino)ethanal, formation of which involvesno carbon-carbon bond cleavage. Thus, high film purity was expected at 200 "C. As expected, no deposition occurred a t substrate temperatures below 150 "C. Deposition at 150 "C resulted in translucent, nonconductive yellow films. Auger analysis shows the films to contain copper and to be very thin, perhaps 1-5 nm thick, based on the sampling depth of Auger spectroscopy. The titanium signal from the substrate is clearly visible after short ion bombardment. Carbon is present and decreasing in concentration throughout the films. This suggests that postdeposition contamination persists throughout the films, and no region in the depth profile is identifiable as characteristic of bulk film. Thus, it is impossible to determine the oxidation state of the as-deposited films, Infrared analysis during deposition a t 150 "C shows a tiny peak near 3545 cm-1, which indicates (dimethylamino)ethanol is formed in smallamounts,but most peaks
Chem. Mater., Vol. 5, No. 12, 1993 1705
I!
I
Figure 5. IR spectrum during deposition at (a) 150 "C,(b)200 "C,and ( c ) 260 OC.
in the spectrum are characteristic of the alkoxide. The 1750-cm-' carbonyl stretch is present only in early spectra (approximately first 12 h of experiment); see Figure 5a. The substrate has a yellow cast after about 12 h under deposition conditions and does not change further in color. Thus, the 1750-cm-1peak, earlier identified as (dimethylamino)ethanal, is associated with deposition. Mass spectral analysis provided no additional information. The reason for the apparent halt in deposition after 12 h is unknown. Deposition occurs on strontium titanate at substrate temperatures of 200-210 "C. The substrate appears to be evenly coated with a pink, opaque, conductive film. Resistivities on the order of 200 p a cm were measured by four-point probe. (Resistivity of bulk copper is 1.678 p a cm.) A film stored in laboratory air darkened to a rose color over 3 weeks. After 10 weeks, the film was nearly black. Surface analysis was performed on this film after one week of exposure to air. The results of surface analysis after ion bombardment (toremove surface contamination) were in agreement for films stored in air and for films stored in a vacuum desiccator for 1 week. Auger analysis of the films was performed after a short ion bombardment to remove postdeposition contaminants. Figure 6 is a typical Auger electron spectrum. Carbon and nitrogen are below detectable levels (less than 1atom ?6 1. The small amount of oxygen present may be due to recontamination of the surface by oxygen in the vacuum system between ion bombardment and analysis. This phenomenon is common on nickel and aluminum surfac88.27 Figure 7 shows the copper 2p and copper LVV peaks collected during XPS. The absence of shakeup satellites in the copper 2p region and the shape and position of the (27) Frank Cromer, Department of Chemistry,V.P.I. & S.U.,personal communication.
Young et al.
1706 Chem. Mater., Vol. 5, No. 12, 1993
Figure 6. Auger electron spectrum of film deposited at 200 "C. I
A
Red, opaque films deposited on strontium titanate at substrate temperatures of 250-270 "C. There was no observable difference in the IR spectra when the stage temperature varied between 250 and 270 "C. Although XPS showed copper present only in the metallic state,the films were only 60-75 atom % copper. Oxygen and carbon contamination persisted throughout the films. The minimum carbon content in individual films ranged from 10 to 30 atom %. Nitrogen was absent. The films were conductive, however, the resistivity as measured by fourpoint-probe was 2 orders of magnitude greater than for films grown at 200 "C, which is attributed both to poor connectivity between grains and to contamination. Films grown at 250 "C showed better adhesion to strontium titanate than those grown at 200 "C but were still easily scratched off. A typical IR spectrum taken during deposition at 250270 "C (Figure 5c) shows mostly peaks characteristic of copper dimethylamino ethoxide. (Dimethylamin0)ethanol and (dimethy1amino)ethanal are present, as is the C-N or C==Cstretch observed in the preliminary decomposition experiments. IR spectroscopy also shows carbon monoxide as a product, while mass spectral analysis indicates carbon monoxide and carbon dioxide. The formation of CO and CO2 implies ligand fragmentation is occurring.
Discussion 350
'
3 b
I
3k2 I 3b Binding Energy, eV
3b4
I
l
0
r
932.8
A 952.7
(b)
~
1, I
I
1
1
Figure 7. XPS spectrum of film deposited at 200 O C : (a) Cu LW region; (b) Cu 2p region.
copper L W Auger transition peak confirm that copper metal is the only oxidation state present.28 Scanning electron microscopy (SEM)reveals a film of uniform, spheroidal grains, between 0.25 and 0.5 pm in diameter, densely packed on the smooth, strontium titanate surface. The film appears to be about 1pm thick. Adhesion of the film to the substrate is poor. The film is easily scratched. The high resistivity of the film is attributed to poor connectivity between the grains. The IR spectrum collected during deposition (Figure 5b) shows primarily copper dimethylamino ethoxide, along with some (diemthy1amino)ethanol. At 200-210 "C, the 1750-cm-1peak ((dimethy1amino)ethanal)is more intense relative to experiments at 150 "C. The 1750-cm-' peak also persists throughout the 2 days of an experiment. During this time, the film becomes increasingly rich in color. As at 150 "C, (dimethy1amino)ethanal production is associated with deposition. (28) (a) Tobin, J. P.;Hirschwald,W.; Cunningham,J. Appl. Surf.Sci. 1983,16,441. (b) Fleisch, T. H.; Zajac, G. W.; Schreiner, J. 0.;Mains, G.J. Appl. Surf. Sci. 1986, 26, 488.
Deposition Chemistry: Major Deposition Products. The major products included in the proposed reaction pathway are copper metal, (dimethylamino)ethanol, and (dimethy1amino)ethanal. These products are associated with copper deposition from copper dimethylamino ethoxide at temperatures rangingfrom 150to 270 "C. Ultrahighvacuum studies of primary alcohols adsorbed on meta121i22 and oxide23-26surfaces provide additional evidence for our proposed MOCVD reaction pathway. Alcohols adsorbed on surfaces21-2seither may desorb intact or may lose the alcoholic hydrogen to form a surface alkoxy species. There are two common reaction pathways available to surface alkoxy species: 8-hydride elimination to form the aldehyde and hydrogen recombination to form the alcohol. Hydrogens on the carbon 8 to the surface (a to the oxygen) are activated by the presence of the oxygen heteroatom.22 One of these @-hydrogensmay be lost to the surface, which triggers formation of a double bond between the oxygen and carbon. The resulting aldehyde desorbs. The hydrogen released onto the surface may react with another surface alkoxy group to form the alcohol, which desorbs. Similar reactions of the alkoxy ligands are expected when copper dimethylamino ethoxide adsorbs on the surface of strontium titanate or the growing film: @-hydrideelimination to form (dimethy1amino)ethanaland reaction with surface hydrogen to form (dimethylaminoh ethanol. A schematic of the reaction pathway proposed to occur during MOCVD with copper dimethylamino ethoxide appears in Figure 8. Note that the reacting alkoxy ligands, normally drawn as bidentate, have been drawn as unidentate for convenience. The available analytical techniques do not allow the orientation and structure of the adsorbed alkoxide to be determined. However, it is reasonable to expect an equilibrium between the unidentate and bidentate forms, especially where the copper center can achieve higher coordination by interaction with the surface. A similar equilibrium between intramolec-
CVD of Cu from Copper(II) Dimethylamino Ethoxide
Chem. Mater., Vol.5, No. 12, 1993 1707
ularly hydrogen bonded and nonintramolecularly hydrogen bonded (dimethy1amino)ethanol exists.14 The short-term fate of the eliminated @-hydrogenis uncertain. It may be bound to the surface of the substrate or the growing film. This is certainly reasonable, based on surface chemistry of alkoxy species derived from adsorbed alcohols. However, it is also possible that the hydrogen remains attached to its own metal center to form a copper hydride. Reductive elimination of the alcohol, leaving the copper metal atom behind, would follow. With the available data, it is impossible to tell whether abstracted hydrogens are mobile on the surface and react with neighboring adsorbed species or whether they remain attached to their metal center and undergo further reaction there. One might question how a copper species with a formal oxidation state of 2+ can deposit clean copper metal without a co-reductant, as is required by copper(I1) @-diketones.loFormation of (dimethy1amino)ethanol involves reductive elimination of an alkoxy ligand and a hydrogen, analogous to reductive elimination of ligands from metal hydrides.29 Elimination of (dimethylaminolethanol resulh formally in reduction of the surface, just as reductive elimination from a molecular metal hydride results formally in reduction of the metal center. Production of copper metal, (dimethylamino)ethanol, and (dimethy1amino)ethanal is consistent with known surface chemistry and metalorganic chemistry. Reductive elimination of the alcohol and j3-hydride elimination of the aldehyde are interdependent. The fact that both pathways are available to a primary alkoxy species allowsthe copper(11) alkoxide to decompose cleanly without a reactive cofeed. In contrast, a similar hydride elimination pathway is not available to P-diketonate ligands. The @-diketoneis the only stable decomposition product formed without ligand fragmentation. An external hydrogen source is required to form the ketone from the ketonate. Without a reactive cofeed, @-diketonatefragments remain bound to the metal center. Thus, clean films of copper metal
cannot be deposited from copper(I1)@-diketonateswithout a reactive cofeed to provide hydrogen. Hydrogen atoms eliminated in the formation of (dimethy1amino)ethanal may recombine on the surface and be released as H2 (undetectable in our deposition system), or may combine with alkoxy ligands to form (dimethylamino)ethanol. In the absence of hydrogen sources other than @-hydrideelimination (such as adsorbed hydroxyl groups), the ratio of (dimethy1amino)ethanalto (dimethylamino)ethanol produced by thermal decomposition of copper dimethylamino ethoxide should be at least one-to-one. Realistically, some excessof (dimethy1amino)ethanol may be produced in our experimental system, because all traces of water cannot be removed from the deposition apparatus. Standards of (dimethy1amino)ethanal are not available, so IR spectra cannot be quantified. It was hoped that NMR analysis of products condensed from the reactor outlet gas in a liquid nitrogen trap would reveal the ratio of aldehyde to alcohol. Unfortunately, (dimethylaminolethanal was not detected by NMR, although (dimethy1amino)ethanol and some small peaks due to other unidentified species were detected. The aldehyde may be consumed by aldol condensation en route to the trap, producing a hydroxyaldehyde product such as (CH& NCH2CH(OH)C(N(CH3)2)CH0.30 Such products would condense in the reactor outlet line, never reaching the trap. The aldehyde is also vulnerable to reactions in the trap, especiallywhen the collected solids (at liquid nitrogen temperature) are melted for NMR analysis. In addition to aldol condensation, there is the possibility of hemiacetal formation, since the aldehyde will be in alcohol solution.30 However, the aldehyde product is clearly observed by in situ IR spectrometry. Thermal Deposition Chemistry: High-Temperature Deposition Products. At a substrate temperature of 260 "C, a new product is detected by a peak at 1640cm-l in the IR spectrum. This peak suggests the presence of a C=C or C=N bond. Formation of the new species is related to ligand fragmentation, since its appearance is accompanied by the production of CO and COz and by inclusion of carbon and oxygen in the film. Without a more specific identification, the proposed reaction pathway is necessarily speculative. The peak is too small to allow further identification by IR, and mass spectral analysis revealed no additional information. Recall that the (dimethy1amino)ethoxy ligand has a heteroatom at each end that may interact with the surface (substrate or growingfilm) or the metal center. Discussion thus far has emphasized bonding via the oxygen and the resulting activation of @-hydrogensfor elimination. The nitrogen may also interact with the surface. A hydrogen from an amine methyl group, or possibly an entire amine methyl group, may be lost to the surface. This can trigger formation of C=N bonds and fragmentation of the (dimethy1amino)ethoxygroup. Another possibility is that multiple hydrogens may be lost to the surface, rather than just one @-hydrogen. Formation of C=C bonds may compete with formation of C=O bonds. Olefin formation might lead to ligand fragmentation and film contamination. Alternatively, the alkoxy species may deoxygenate, and the resulting hydrocarbon dehydrogenate, leaving hydrocarbon fragments and carbon to be incorporated in the film.
(29) Cotton, F.A.;Wilkineon, G. Aduanceddnorganic Chemistry, 3rd ed.; Wiley: New York, 1972.
(30) Morrison, R.T.;Boyd, R. N. Organic Chemistry, 4th 4. Allyn ; and Bacon: Boston, 1983;pp 867-869.
Me Me
Me Me copper dimethylaminoethoxlde
J
beta-hydride elimination
H
aldehyde
c u elimination r
L b
&Me alcohol
Figure 8. Reaction pathway for copper dimethylaminoethoxide.
Young et al.
1708 Chem. Mater., Vol. 5, No. 12, 1993 Studies of carboxylate^,^^ a l k o x i d e ~and , ~ ~aldehydes32 on CuzO have revealed some of the processes by which oxygenated organics decompose to leave surface carbon. Some or all of these processes may occur on strontium titanate or on the growing film and contribute to film contamination a t higher temperatures. Reduction (deoxygenation) of surface oxygenate species to surface hydrocarbons on CuzO has been observed from 0 to 325 "C. The nature of the hydrocarbons formed is not known, but their dehydrogenation above 225 "C is clearly observed over C u ~ 0 . Thus, ~ ~ surface * ~ ~ oxygen ~ ~ ~released by reduction of surface alkoxy species would be available for production of CO and COz from surface hydrocarbons and for inclusion in the film. In addition, above 325 "C, surface carbon extracts lattice oxygen from CuzO to form CO and C02.31932 Extraction of lattice oxygen might be important in early stages of film growth on strontium titanate above 250 "C, while the oxide surface is exposed but is not expected to be important once the oxide is covered by copper metal. Extraction of lattice oxygen allows complete burnoff of surface carbon from CUZOin ultrahigh vacuum at 525 0C.24*31332However, at deposition temperatures of 250270 "C, on a growing metal film, carbonaceous species are expected to remain. Although film contamination above 250 "C is described as "severe", the ligand fragmentation reaction is actually a minor one. There is one carbon atom for every two or three copper atoms in the films, while the precursor contains eight carbon atoms for each copper atom. Most of the ligands are probably eliminated virtually intact, gaining or losing only hydrogen. This highlights the sensitivity of film quality to deposition chemistry. Even a minor reaction can have a large effect on film purity. Thermal Deposition Chemistry: Alternative Reaction Pathways. The @-hydridelreductiveelimination pathway explains the available data in a satisfactory manner. It is reasonable based on the chemistry of alkoxy species on surfaces in ultrahigh vacuum. It explains how clean metal films deposit from a copper(I1) precursor. However, we must consider the possibility of another mechanism, because no published IR spectra exist to confirm our in situ identification of (dimethylamino)ethanal as a product. One alternative explanation is that the aldehyde is actually a minor deposition product, and the alcohol is the major organic product of deposition. If this is the case, some source of hydrogen other than @-hydride elimination from the ligand must exist. Two potential hydrogen sources were considered and ruled out. A leak in the system is one possible hydrogen source; a source ruled out by thorough testing of the system and by reasoning. If the system leaks, water will be admitted. Water is a ready source of hydrogen and will lead to production of alcohol from an alkoxide. However, alkoxides are water-sensitive at any temperature. If hydrolysis were the major reaction producing alcohol, alcohol would be obvious in the infrared spectra below 150 "C. (Dimethy1amino)ethanol would be produced all the time, not just during film deposition. A system leak is not the source of hydrogen for alcohol production. Another possible hydrogen source is complete ligand fragmentation. An occasional ligand may completely decompose on the surface of the growing film. This would
release 10 hydrogen atoms, allowing formation of 10 alcohols from the decomposition of five alkoxide molecules. In exchange for five copper atoms deposited, four atoms of surface carbon would be deposited. (Surface carbons produced by dehydrogenation are expected to remain in the film unless the temperature exceeds 500 "Cand a source of oxygen is available.) In fact, less than one surface carbon atom deposits for every 100 alkoxide molecules decomposed at 200 "C (based on film analysis by Auger spectroscopy). I t is difficult to visualize a reaction that could release large amounts of hydrogen (to produce alcohol) and produce no other species detectable by vaporphase infrared analysis or by surface analysis of the film. In this experimental system, the reaction pathway cannot be established beyond doubt. However, production of (dimethy1amino)ethanalby @-hydrideelimination, and the related production of (dimethy1amino)ethanol by reductive elimination, is the pathway most consistent with the data collected and with the known chemistry of surface alkoxy species. Substrate Effects on Deposition. Films grown from copper dimethylamino ethoxide on strontium titanate are composed of spheroidal grains with diameters of 1pm or less. They adhere poorly to the substrate and appear poorly connected to one another. Because these small grains have a high surface area, the films are susceptible to oxidation in air. Because the grains are poorly connected, the sheet resistivities are quite high. The resistivity of bulk copper is 1.678pfl cm.33 The resistivity of pure (as measured by Auger electron spectroscopy and X-ray photoelectron spectroscopy) copper films deposited a t 200 "C is about 200 pfl cm. Perhaps another substrate would interact favorably with the growing film and yield lower resistivity.
(31) Schulz, K. H.; Cox,D.F. J. Phys. Chem. 1992,96,7394. (32) Schulz, K. H.; Cox, D. F. J . Phys. Chem. 1993, 97, 3555.
(33) Weast, R. C.,Ed. CRC Handbook of Chemistry and Physics, 67th e d . ; C R C Press: Boca Raton, F L , 1986; p F-120.
Conclusions Thermal deposition from copper dimethylamino ethoxide in inert atmosphere does not occur on strontium titanate below 150 "C. At 150 "C, the resulting thin films are nonconductive. At substrate temperatures of 200210 "C, conductive films of pure copper metal deposit on strontium titanate while higher substrate temperatures (250-270 "C) cause significant oxygen and carbon contamination. The sheet resistivity of pure (as measured by Auger spectroscopy) copper films deposited on strontium titanate at 200 "C exceeds that of bulk copper by nearly 2 orders of magnitude due to poor connectivity among the copper grains. The major organic products of thermal deposition of copper from copper dimethylamino ethoxide are identified by in situ infrared spectroscopyas (dimethy1amino)ethanal and (dimethy1amino)ethanol. The reactions that produce the aldehyde and alcohol are interdependent. The aldehyde is formed by @-hydrideelimination, which releases a hydridic hydrogen from the ligand. This hydrogen reacts with another alkoxy ligand and the resulting reductive elimination produces alcohol and copper metal. The interdependence of the two reactions leads to deposition of clean metal films without a reactive cofeed. The @-hydridelreductiveelimination pathway remains important at 260 "C. I n situ infrared spectroscopy shows that (dimethylamino)ethanol and (dimethy1amino)ethanal
CVD of Cu from Copper(I0 Dimethylamino Ethoxide are produced. However, additional reactions which produce C = C or C-N species, carbon monoxide, and carbon dioxide also occur, as indicated by in situ infrared and post-reactor mass spectroscopy. Formation of double bonds and production of CO and COZ indicates ligand fragmentation that leads to carbon and oxygen contamination of the film. The nature of the carbon contamination, and the intermediates which lead to its formation, could not be determined. Films deposited near 260 "C are less conductive than those deposited at 200 "C by 2 orders of magnitude, because of carbon and oxygen contamination. The reactions that lead to deposition of copper from copper(I1)dimethylamino ethoxide are the same reactions observed during ultrahigh-vacuum studies of alkoxy specieson surfaces. Arecent study of copper tert-butoxide deposition" also shows chemistry previously observed for alkoxy species on surfaces. Work done in our laboratorylZ supports the identification of isobutylene as the major
Chem. Mater., Vol. 5, No. 12, 1993 1709 decomposition product of copper tert-butoxide.l' (8hydride elimination is not an available reaction pathway for tertiary alkoxy species.) Ultrahigh-vacuum studies of ligand species adsorbed on surfaces may be valuable in general for predicting decomposition chemistry of metalorganic precursors but cannot provide the temperature vs mechanism data that are accessible with in situ studies of deposition chemistry. Acknowledgment. The authors acknowledgeDr. Kirk H. Schulz, currently at the University of North Dakota, for collection of LEED data; Dr. S. B. Desu at Virginia Polytechnic Institute and State University for use of his four-point probe apparatus; Tina A. Handlos and Brian Risch at Virginia Polytechnic Institute and State University for assistance in obtaining SEM photographs; and Kelly E. Matthews and Christopher P. Roy a t Virginia Polytechnic Institute and State University for collection of NMR spectra.
