1,1,1,5,5,5-Hexafluoroacetylacetonate Copper(I) Poly(vinylsiloxane)s

Department of Chemistry, Imperial College London, South Kensington Campus, London SW7 2AZ, U.K., University of Surrey Ion Beam Centre, Guildford, Surr...
1 downloads 0 Views 465KB Size
Chem. Mater. 2006, 18, 2489-2498

2489

1,1,1,5,5,5-Hexafluoroacetylacetonate Copper(I) Poly(vinylsiloxane)s as Precursors for Copper Direct-write Jenny Rickerby,† Aliz Simon,§ Chris Jeynes,‡ Trevor J. Morgan,| and Joachim H. G. Steinke*,† Department of Chemistry, Imperial College London, South Kensington Campus, London SW7 2AZ, U.K., UniVersity of Surrey Ion Beam Centre, Guildford, Surrey, GU2 7XH, U.K., Institute of Nuclear Research of the Hungarian Academy of Sciences (ATOMKI), H4001 P.O. Box 51, Debrecen, Hungary, and Department of Chemical Engineering, Imperial College London, South Kensington Campus, London SW7 2AY, U.K. ReceiVed September 19, 2005. ReVised Manuscript ReceiVed March 18, 2006

Polymeric Cu(I) precursors which deposit metallic Cu films via thermally induced disproportionation have been investigated as inks for copper direct-write to address deposit haloing, a problem when monomeric CVD precursor are ink-jetted. Cu(hfac) was complexed to poly(methylvinylsiloxane) and poly(divinylsiloxane) and these novel precursor compounds were fully characterized by spectroscopic and thermal techniques and their ability to deposit Cu films was assessed. The polysiloxanes precursors did deposit metallic Cu films, which partitioned from the polysiloxane residues. Detailed particle induced X-ray emission (PIXE) and Rutherford backscattering spectrometry (RBS) data analysis provided valuable insight into the various stages of the deposition process. The mechanistic insight will be useful in the development of higher performing metal precursor inks.

Introduction Photolithographic etching processes are currently producing metallic tracks, such as those required for the manufacture of printed circuit boards.1,2 These processes have proven to be extremely successful despite the large capital investment required to implement the necessary multistep procedures.3 A faster and cheaper alternative to the established technology could be provided by direct patterning approaches, which would result in a single production step required for the deposition of metallic tracks.4-7 A variety of printing techniques such as screen, gravure and microcontact printing, reel-to-reel, and ink-jet among others have been investigated as novel patterning techniques in this regard.8-14 The use of * To whom correspondence should be addressed. E-mail: j.steinke@ imperial.ac.uk. † Department of Chemistry, Imperial College London. § Institute of Nuclear Research of the Hungarian Academy of Sciences. ‡ University of Surrey Ion Beam Centre. | Department of Chemical Engineering, Imperial College London.

(1) Andricacos, P. C. Interface 1999, 32. (2) Percin, G.; Khuri-Yakub, B. T. IEEE Trans. Semicond. Manuf. 2003, 16, 452. (3) Moreau, W. M. Semiconductor Lithography, 3rd ed.; Plenum: New York, 1991. (4) Rickerby, J.; Steinke, J. H. G. Chem. ReV. 2002, 102, 1525 (5) Cuk, T.; Troian, S. M.; Hong, C. M.; Wagner, S. Appl. Phys. Lett. 2000, 77, 2063. (6) Heule, M.; Vuillemin, S.; Gauckler, L. J. AdV. Mater. 2003, 15, 1237. (7) Luisier, A.; Utke, I.; Bret, T.; Cicoira, F.; Hauert, R.; Rhee, S. W.; Doppelt, P.; Hoffmann, P. J. Electrochem. Soc. 2004, 151, C535. Chabinyc, M. L.; Wong, W. S.; Arias, A. C.; Ready, S.; Lujan, R. A.; Daniel, J. H.; Krusor, B.; Apte, R. B.; Salleo, A.; Street, R. A. Proc. IEEE 2005, 93, 1491. Noh, C. H.; Son, H. J.; Kim, J. Y.; Hwang, O. C.; Song, K. Y.; Byk, T. V.; Sokolov, V. G.; Kim, J. B. Chem. Lett. 2005, 34, 82. (8) Teng, K. F.; Vest, R. W. IEEE Trans. Compon. Hybrids Manuf. Technol. 1988, 11, 291. (9) Yamaguchi, K.; Sakai, K.; Yamanaka, T.; Hirayama, T. Precis. Eng.J. Am. Soc. Precis. Eng. 2000, 24, 2.

ink-jet is particularly attractive as it allows the generation and modification of any desired layout through simple manipulation of a computer graphic.8-12,15,16 These printing methods require liquid or solution-based ink formulations, which after deposition are transformed on the substrate surface through chemical activation. Colloidal metal formulations, as well as metal particulate suspensions, have been shown to create metal tracks through sintering at temperatures in excess of 300-400 °C.17-20 A combination of a metal powder and a reactive metal organic precursor has led to a significant reduction of the required sintering temperature reported to be as low as 275 °C.19 Printing metal tracks with an ink-jet printer has been demonstrated, requiring subse(10) Nur, H. M.; Song, J. H.; Evans, J. R. G.; Edirisinghe, M. J. J. Mater. Sci.-Mater. Electron. 2002, 13, 213. (11) Szczech, J. B.; Megaridis, C. M.; Gamota, D. R.; Zhang, J. IEEE Trans. Electron. Packag. Manuf. 2002, 25, 26. (12) Magdassi, S.; Bassa, A.; Vinetsky, Y.; Kamyshny, A. Chem. Mater. 2003, 15, 2208. (13) Delamarche, E.; Vichiconti, J.; Hall, S. A.; Geissler, M.; Graham, W.; Michel, B.; Nunes, R. Langmuir 2003, 19, 6567. Felmet, K.; Loo, Y. L.; Sun, Y. M. Appl. Phys. Lett. 2004, 85, 3316. (14) Schulz, D. L.; Curtis, C. J.; Ginley, D. S. Electrochem. Solid State Lett. 2001, 4, C58. (15) Rozenberg, G. G.; Steinke, J. H. G.; Gelbrich, T.; Hursthouse, M. B. Organometallics 2001, 20, 4001. (16) Calvert, P. Chem. Mater. 2001, 13, 3299. de Gans, B. J.; Duineveld, P. C.; Schubert, U. S. AdV. Mater. 2004, 16, 203. Geissler, M.; Xia, Y. N. AdV. Mater. 2004, 16, 1249. (17) Kamyshny, A.; Ben-Moshe, M.; Aviezer, S.; Magdassi, S. Macromol. Rapid Commun. 2005, 26, 281. Drummond, T.; Arbor, A.; Ginley, D. U.S. Patent 5,132,248, 1992. (18) Kydd, P. H.; Wagner, S.; Gleskova, H. Material and method for printing high conductivity electrical conductors and other components on thin film transistor arrays. U.S. Patent 6,274,412, 1999. (19) Kydd, P. H.; Jablonski, G. A.; Richard, D. L. Low-temperature method and compositions for producing electrical conductors. WO 03/003381, 2001. (20) Gonzalez-Blanco, J.; Hoheisel, W.; Sicking, J. Preparations containing fine-particulate inorganic oxides. WO 00/20519, 2000.

10.1021/cm052103s CCC: $33.50 © 2006 American Chemical Society Published on Web 04/22/2006

2490 Chem. Mater., Vol. 18, No. 10, 2006

quent annealing to 150-200 °C.21,22 We have developed a single-step process of ink-jet printing, a reactive organometallic liquid operating in real-time conversion mode using a substrate temperature of 150 °C without the need for a post-treatment step.15 Detailed analysis of the drop deposits revealed the presence of a metal halo surrounding the major Cu-rich deposit. Haloing is clearly an undesirable feature as it increases the “footprint” of the deposited precursor ink droplet, thereby limiting the minimum feature size of metal tracks available through this process. We speculated that polymeric metal organic precursors may provide the means of suppressing the haloing effect through their reduced volatility. Poly(vinylsiloxane)s are structurally similar to highly reactive Cu(I) CVD precursors (hfac)Cu(VTMS) and (hfac)Cu(VTMOS) and poly(siloxane)s have been known to be thermally degradable, making them ideal polymers for use as ligand polymers.23,24 Employing polymeric precursors as potential inks may also provide a novel opportunity to modify the viscosity behavior of ink formulations, replacing the usual “passive” viscosity modifiers with a “reactive” alternative. Our first results on the synthesis and deposition characteristics of polymeric polynuclear precursor and the implications on ink design optimization are reported here. Experimental Section General Information. All commercially bought chemicals were used as received unless stated otherwise. 1,1,1,5,5,5-Hexafluoroacetylacetone was dried over 4 Å molecular sieves (activated by heating to 400 °C/0.1 mmHg). Copper(I) oxide was dried by heating to 400 °C (0.1 mmHg) immediately prior to use. Dichlorodivinylsilane (92%) was purchased from ABCR. Methylvinylsiloxanediol (1) and divinylsiloxanediol (2) were prepared according to literature procedures.25,26 NMR spectra were acquired using Bruker AC250 and Bruker DRX400 spectrometers and deuteriochloroform (99.9%D, Cambridge Isotope Labs) and deuteriobenzene (D6, 99.6%-D, Aldrich) as solvents. Infrared spectra were obtained using a Mattson Satellite 3000 FTIR spectrometer. Solids were either prepared as evaporated films or, in the case of more sensitive compounds, crushed between two plates. Chemical ionization (CI) and electron impact (EI) mass spectra were obtained using a VG Autospec. GPC samples were prepared in chloroform (2-5 mg/mL) and chromatograms were obtained using a Viscotek Tri SEC model 302 system with triple detection (refractive index, light scattering, and viscosity detector model). The equipment was operated isothermally at 30 °C with a constant flow rate of chloroform (AnalaR, Merck) of 1.0 mL/min. A set of guard columns (Polymer Laboratories, PLgel, 5 µm Guard, 50 × 7.5 mm) and two analytical columns (Polymer Laboratories, PLgel, 5 µm MIXED-C, 300 × 7.5 mm) were used and 100 µL of sample solution were injected by an autosampler. Narrow MW polystyrene standards (Polymer Laboratories) were used for conventional calibration. Mn, Mw, and PDIs are quoted as polystyrene equivalents. (21) Rozenberg, G. G.; Bresler, E.; Speakman, S. P.; Jeynes, C.; Steinke, J. H. G. Appl. Phys. Lett. 2002, 81, 5249. (22) Jeynes, C.; Rozenberg, G. G.; Speakman, S. P.; Steinke, J. H. G. Nucl. Instrum. Methods Phys. Res. Sect. B-Beam Interact. Mater. Atoms 2002, 188, 141. (23) Camino, G.; Lomakin, S. M.; Lazzari, M. Polymer 2001, 42, 2395. (24) Camino, G.; Lomakin, S. M.; Lageard, M. Polymer 2002, 43, 2011. (25) Cella, J. A.; Carpenter, J. C. J. Organomet. Chem. 1994, 480, 23. (26) Okawa, T. Method for preparation of diphenylsiloxane-dimethylsiloxane copolymers. EP 0 693 521 A1, 1996.

Rickerby et al. Synthesis of Poly(methylvinylsiloxane) (3). To vigorously stirred 1 (0.40 g, 3.7 mmol) was added one drop of concentrated H2SO4. The mixture was stirred until it thickened. The reaction was then quenched via the addition of saturated aqueous NaHCO3 and the reaction mixture was extracted with ether (2 × 10 mL). The combined organic extracts were washed with water (2 × 10 mL), dried over MgSO4, and filtered and the solvent was removed in vacuo. The resulting viscous clear liquid was dissolved in THF (20 mL) and stirred for 18 h with TMSCl (0.1 mL). Volatiles were removed in vacuo. This yielded 0.21 g (55%) of a clear colorless viscous liquid. 1H NMR (250 MHz, CDCl3): δ 0.05-0.21 ppm (3H, m, CH3), 5.68-6.10 ppm (3H, m, CHdCH2). 13C NMR (100 MHz, CDCl3): δ -0.1 to 0.6 ppm (terminal CH3), -0.2 ppm (CH3), 133.1-133.50 ppm (CHdCH2) 136.0-136.7 ppm (CH). IR spectrum υmax (cm-1): 3055, 3017, 2964, 1598 (CdC), 1408, 1260 (Si-Me), 1079 (Si-O), 1028 (Si-O), 1008, 959, 792. GPC data: Mn ) 27700, Mw ) 70500, PD ) 2.54. Synthesis of Poly(divinylsiloxane) (4). This reaction was carried out using the procedure described for compound 3. 2 (0.39 g, 3.4 mmol) was polymerized using a drop of concentrated H2SO4 as the catalyst. The resulting viscous liquid was dissolved in THF (20 mL) and stirred for 18 h with TMSCl (0.1 mL) and volatiles were removed in vacuo. Then 0.26 g (78%) of product was isolated as a pale brown oil. 1H NMR (400 MHz, CDCl3): δ 0.13 ppm (3H, m, CH3), 5.88-6.11 ppm (6H, m, CHdCH2). 13C NMR (100 MHz, CDCl3): 1.9 ppm (CH3), 133.3-135.8 ppm (CHdCH2). IR spectrum υmax (cm-1): 3056, 3016, 2976, 2952, 1927, 1744, 1597 (CdC), 1407, 1274 (Si-Me), 1089 (Si-O), 1007, 962, 750. GPC data: Mn ) 6500, Mw ) 31200, PD ) 4.8. Synthesis of (hfac)Cu[poly(methylvinylsiloxane)] (5). Product 5b as an example. Copper(I) oxide (0.6 g, 4 mmol, 0.7 equiv) was slurried with 3 (0.50 g, 5.8 mmol) and pentane (30 mL) in a flamedried Schlenk tube. The suspension was degassed by bubbling argon through for 0.5 h. 1,1,1,5,5,5-Hexafluroracetylacetone (0.80 mL, 5.8 mmol) was added in one portion, and the reaction mixture was stirred at 0 °C for 15 h. The mixture was then canulla filtered, and the solvent was rapidly removed in vacuo. This yielded 1.3 g (68%) of a yellow viscous oil, which froze when cooled in a CO2/acetone bath. 1H NMR: 250 MHz (C6D6); δ ) 0.17-0.21 ppm (3H, m, CH3), δ ) 0.29 ppm (0.03H, s, end group CH3), δ ) 4.20-4.70 ppm (3H, m, CHdCH2), δ ) 6.19 ppm (0.85H, s, CH). 13C NMR: 62.5 MHz (CDCl3); δ ) -0.91 ppm (end group CH3), δ ) 0.91 ppm (CH3), δ ) 90.11 ppm (CH), δ ) 92.00-105.00 ppm (CHd CH2), δ ) 117.97 ppm (q, CF3, 1JC-F ) 283 Hz), δ ) 178.05 ppm (q, CO, 1JC-F ) 35 Hz). IR spectrum: υmax/cm-1; 3446, 3275, 3149, 3060, 2962, 2934, 2876, 1639 (CdO), 1609 (CdC), 1557, 1531, 1468, 1348, 1258 (SisMe), 1205, 1149, 1101 (SisO), 1025. Synthesis of (hfac)Cu[poly(divinylsiloxane)] (6). Product 6b as an example. Using the procedure detailed for 5, 1,1,1,5,5,5hexafluroracetylacetone (1.15 mL, 8.20 mmol, 2.00 equiv) was added to a slurry of copper(I) oxide (0.82 mL, 5.7 mmol, 1.4 equiv) and 4 (0.40 g, 4.1 mmol) in pentane (20 mL). The reaction mixture was stirred at 0 °C for 19 h and then filtered, and the solvent was removed in vacuo. This yielded 0.8 g (59%) of a pale yellow/green powdery solid. 1H NMR: 250 MHz (C6D6); δ ) 0.28 ppm (0.01H, s, CH3), δ ) 4.31-4.94 ppm (6H, m, CHdCH2), δ ) 6.19 ppm (0.54H, s, CH). 13C NMR: 62.5 MHz (CDCl3); δ ) 1.07 ppm (CH3), δ ) 90.63 ppm (CH), δ ) 91.00-104.00 ppm (CHdCH2), δ ) 117.97 ppm (q, CF3, 1JC-F ) 282 Hz), δ ) 178.63 ppm (q, CO, 1JC-F ) 33 Hz). IR spectrum: υmax/cm-1; 3062, 3019, 2956, 1937, 1639 (CdO), 1607 (CdC), 1557, 1531, 1468, 1408, 1346, 1258 (SisMe), 1207, 1150 (SisO), 1100, 1008, 969, 802, 745, 673.

Polymeric Cu(I) Precursors for Copper Direct-Write

Chem. Mater., Vol. 18, No. 10, 2006 2491

Table 1. Viscometry Measurements Used To Calculate µsp for Solutions of 3 and 5b polymer 3 run sample weight/g conc/[M] run/s average t µr µsp m

hexane

1 2 3

347 348 347 347.3

polymer precursor 5b

run 1

run 2

run 3

run 4

run 1

run 2

run 3

run 4

0.0637 0.030 352 352 351 351.7 1.013 0.013

0.0867 0.040 359 359 362 360.0 1.037 0.037

0.902 0.042 362 361 362 361.7 1.041 0.041

0.1258 0.058 377 378 376 377.0 1.086 0.086

0.0216 0.007 349 349 352 352.0 1.011 0.011

0.0336 0.011 353 353 352 352.6 1.015 0.015

0.0567 0.018 357 354 355 355.3 1.023 0.023

0.081 0.025 360 361 359 360.0 1.036 0.036

0.294

Viscometry. Capillary viscometry was used to measure solution viscosities for hexane solutions containing precursor 5b and its parent polymer 3. Solutions containing different concentrations of the precursor as its parent polymer were passed through an Ubbelohde suspended level dilution viscometer. The average of three runs recorded for the liquid to pass between the two marked levels was used as t. The measured values for t and subsequent calculations to determine µsp are included in Table 1. The graph of specific solution viscosity (µsp) plotted against solution concentration from these data is shown in Figure 3. The following equations were used to determine [µ] (limiting viscosity number) for each sample: c is concentration, k′ is the Huggins constant, a shape-dependent factor, and k′′ is another shape-dependent factor.27 µr ) (t/t0) ) (µ/µ0)

(i)

µsp ) µr - 1 ) (t - t0)/t0

(ii)

(µsp/c) ) [µ] + k′[µ]2c/µ0)

(iii)

Using eq (i), the run time values t for each solution were used to calculate µr with t0 (t0 is the time taken for neat solvent to run through the capillary). These values were converted into µsp using eq (ii), which in turn enables [µ] to be determined from eq (iii) by plotting c against µsp/c and using a linear graph equation y ) mx + c as shown in Figure 3. Thermal Analysis. TG analysis was carried out on a PerkinElmer TGA 7 thermogravimetric analyzer and DSC thermograms were obtained using a Perkin-Elmer Pyrus 1 instrument. Samples were heated under a N2 atmosphere using an identical heating program for both analyses; initially the samples were maintained isothermally at 50 °C for 5 min and then were heated from 50 to 500 °C at 5 °C/min (TGA). Tg and Tm of the polymeric samples was determined by heating at 5 °C/min (DSC). Cu Deposit Preparation and Characterization. The Cu deposits were prepared by applying neat precursor to a glass microscope slide and heating to 300 °C in an N2 atmosphere. All polymers were applied as pure compound, oil, or solid, and the solid melted before conversion to Cu. RBS and PIXE characterization methods were used simultaneously. 2.07 MeV 4He2+ ion beam focused to 6 × 8 µm2 (scan range max 2000 × 2000 µm2) using an Oxford triplet system at the Surrey ion beam analysis facility was used.28-30 The chamber pressure was kept at about 4 × 10-6 mbar. An Ortec implanted Si detector (ULTRA) had a 155° scattering angle, and 16.5 msr solid angle (50 mm2 active are at a distance 55 mm from sample) was applied for RBS. The total system resolution (including kinematical broadening) was 25 keV. A standard Si(Li) detector with a Be window of 72 mm2 active area map set at 135° to the beam, 25 mm far from the sample (115 msr solid angle) for the PIXE measurements. (27) Cowie, J. M. G. Polymers: Chemistry and Physics of Modern Materials, 2nd ed.; Blackie: London, 1997.

The samples were mounted in the beam line without any preparation. The data were collected using the OM_DAQ hardware and software (Dan32 for PIXE). Maps were collected with fixed time per pixel.28 The RBS data were analyzed using the DataFurnace software.31 Some chemical assumptions were imposed on the data (see discussion by Jeynes et al.32). These assumptions are represented in the extracted profiles. The fits assume that the molecules specified were used. Thus, if the fit was done with CuO, this implies only that the data are consistent with CuO; it makes no specific claims about the actual oxidation state of the Cu in the sample. However, the fits are otherwise generally model-free.

Results and Discussion Precursor Synthesis and Characterization. Poly(methylvinylsiloxane) (3) and poly(divinyl siloxane) (4) were prepared in two steps, via hydrolysis of their corresponding dichlorosilanes followed by acid-catalyzed condensation polymerization as shown in Scheme 1.33 The polymers were then end-capped with an excess of TMSCl. Both polymers were oily liquids and their molecular weights (Mn) were determined by GPC (conventional calibration with polystyrene standards) to be 27700 and 6500 Da, respectively. Since these polymers were prepared by condensation polymerization, their PDIs (2.5 and 4.8) were not narrow and include oligomers. Polymeric precursors 5 and 6 were prepared using an adapted version of Doyle et al.’s copper(I) oxide synthesis for (hfac)Cu(L) complexes, which requires strictly anaerobic conditions to prevent oxidation of Cu(I) to Cu(II).34 Precursor 5 was a viscous yellow oil at room temperature whereas 6 was a pale yellow/green powdery crystalline solid. NMR spectroscopic analysis of these precursors showed the characteristic upfield shift of the vinylic signals which is associated with complexation of Cu(hfac) to an alkene.35 1H NMR spectra for three batches, with different levels of (28) Grime, G. W.; Dawson, M. Nucl. Instrum. Methods Phys. Res. Sect. B-Beam Interact. Mater. Atoms 1995, 104, 107. (29) Jeynes, C.; Puttick, K. E.; Whitmore, L. C.; Gartner, K.; Gee, A. E.; Millen, D. K.; Webb, R. P.; Peel, R. M. A.; Sealy, B. J. Nucl. Instrum. Methods Phys. Res. Sect. B-Beam Interact. Mater. Atoms 1996, 118, 431. (30) Simon, A.; Jeynes, C.; Webb, R. P.; Finnis, R.; Tabatabian, Z.; Sellin, P. J.; Breese, M. B. H.; Fellows, D. F.; van den Broek, R.; Gwilliam, R. M. Nucl. Instrum. Methods Phys. Res. Sect. B-Beam Interact. Mater. Atoms 2004, 219, 405. (31) Barradas, N. P.; Jeynes, C.; Webb, R. P. Appl. Phys. Lett. 1997, 71, 291. (32) Jeynes, C.; Barradas, N. P.; Marriott, P. K.; Boudreault, G.; Jenkin, M.; Wendler, E.; Webb, R. P. J. Phys. D 2003, 36, R97. (33) Xu, C. H.; Feng, S. Y. React. Funct. Polym. 2001, 47, 141. (34) Doyle, G.; Eriksen, K. A.; van Engen, D. Organometallics 1985, 4, 830. (35) Shin, H.-K.; Suwon, K.-D. Solutions of precursor copper compounds for copper film deposition by chemical vapor deposition and their preparation. Euro. Pat. 989,133, 2000; p 17 pp.

2492 Chem. Mater., Vol. 18, No. 10, 2006

Rickerby et al.

Cu(I) complexation, of precursor 5 and two of 6 along with the uncomplexed polysiloxanes are shown in Figure 1; Table 2 summarizes the reaction stoichiometries used in their preparation and associated 13C NMR shifts of the vinylic carbons. It became apparent that the upfield shift in the vinyl groups was occurring progressively as the degree of complexation increased. Typically the range of chemical shifts was from 6 ppm for the uncomplexed vinyl group to the region of 4.5 ppm for higher degrees of complexation and this trend was mirrored in the 13C NMR spectra with shifts of the alkene carbons of around 135 ppm upfield to 100 ppm. Our observation is consistent with previous reports including those by Shin and Suwon who used 1H NMR alkene shifts to determine solution stabilities of CVD precursor (hfac)Cu(1-pentene) in the presence of excess 1-pentene.35 Shin and Suwon assessed increased stability as a function of NMR shift of the terminal alkene signals from δ ) 4.48 ppm for (hfac)Cu(1-pentene) to δ ) 5.48 ppm with a 50% excess of 1-pentene stabilizing the complex.35 Rozenberg et al. noted an upfield shift in the tetravinylsilane proton signals between multinuclear complexes [(hfac)Cu]2(TVS) (δ ) 4.7-5.1 ppm) and [(hfac)Cu]4(TVS) (δ ) 4.3 ppm).15 Precursor 6b was vastly more air-sensitive than 5b despite having a lower degree of complexation. Controlling the degree of complexation proved only moderately successful, as Kodas and Hampden-Smith observed reaction stoichiometry is difficult to control when preparing (hfac)Cu(L) precursors using Cu2O, with reaction time being the most influential factor.36 Complexing Cu(hfac) to poly(siloxane)s has allowed us to directly observe the effect of the degree of complexation

of Cu(hfac) on the corresponding 1H and 13C upfield shift. These trends are shown in Figure 2. The degree of complexation is quoted as a percentage of Cu(hfac) complexed vs the number of polymer vinyl groups present and was calculated as a ratio of the integral of the hfac signal to the integral of the vinyl group signal in the 1 H NMR spectra. Precursor 5 displays a linear relationship between the change in chemical shift of the vinyl signal and the increase in the degree of complexation whereas 6 deviates from a linear trend. This deviation at around 45% complexation is unlikely to be a result of steric crowding as Cu(hfac) attempts to complex to the second vinyl group on the repeat unit because the polymer is far too flexible. A more plausible reason for this trend is that at degrees of complexation below 50% Cu(hfac) is complexing to two vinyl groups in a coordination mode similar to that of Cu in (hfac)Cu(COD).37 Above 50% complexation Cu(Hfac) coordination to a single vinyl group has to be invoked or alternatively a dynamic equilibrium between intra- and possibly interchain coordination (cross-linking) is occurring. Some possible evidence to support the case for intra- or interchain Cu(hfac) coordination came from DSC measurement of Tg; values determined for both the parent polymers and precursors are summarized in Table 3. The parent polysiloxanes possess Tgs below -130 °C but complexation of Cu(hfac) to 3 increases the Tg of 5b to -18.9 °C with two small melt transitions and 6b possesses two melt transitions in close succession at 75.8 and 77.4 °C. The occurrence of crystallinity in a sample of polysiloxane is interpreted to indicate strong interactions are occurring between polymer chains, perhaps through enhanced chain packing and/or intra- and interchain cross-linking modes, decreasing rotational and translational mobilities. The addition of bulky Cu(hfac) side groups to the polymer seems unlikely to be sufficient to account for such a vast difference in polymer chain mobility. These polymeric Cu precursors proved to be a good starting point for assessing the suitability of polymeric precursors for direct-write applications. Their preparation compared to that of other highly sensitive Cu(I) precursors was facile. In terms of loading, 5b for example contains more Cu(I) per weight of precursor than (hfac)Cu(VTMOS).35 Though higher degrees of complexation inevitably render the precursor more air-sensitive, the increase in Tg exhibited as a result of Cu complexation also means that these precursors can easily be stored at low temperature in the solid state which vastly improves their shelf life over liquid precursors. Furthermore, the polymeric nature of the Lewis base ligands in these precursors has the added benefit that it is possible to isolate the precursor with excess stabilizing ligand still present, unlike conventional Cu(I) CVD precursors as the Lewis base ligand is removed by evaporation. Viscometry. The effect complexation of Cu(hfac) has on the solution properties of the polymer was investigated for two reasons; first, since these precursors are intended for

(36) Kodas, T. T.; Hampden-Smith, M. J. The Chemistry of Metal CVD; VCH: Weinheim, 1994.

(37) Chi, K. M.; Shin, H. K.; Hampden-Smith, M. J.; Duesler, E. N.; Kodas, T. T. Polyhedron 1991, 10, 2293.

Scheme 1. Silane Diols Were Prepared through the Controlled Hydrolysis of Dichlorosilanesa

a (i) Ether, water, 0 °C, yield 1 ) 52%, yield 2 ) 71%. polysiloxanes 3 and 4 were prepared via acid-catalyzed condensation polymerization. (ii) Concentrated H2SO4, 100 °C. The polymers were then end-capped. (iii) THF TMSCl, yield 3 ) 55% (n ) 322), yield 4 ) 78% (n ) 66). Complexation of Cu(hfac) to 3 and 4. (iv) Cu2O, Hhfac, pentane, 0 °C.

Polymeric Cu(I) Precursors for Copper Direct-Write

Chem. Mater., Vol. 18, No. 10, 2006 2493

Figure 1. 1H NMR spectra of Cu(hfac) complexing to the vinyl groups of polymer 3 and 4. (a) Uncomplexed polymer 3 m ) 0. (b) Complexed polymer 5a m ) 0.46. (c) Complexed polymer 5b m ) 0.85. (d) Complexed polymer 5c m ) 0.97. (e) Uncomplexed polymer 4 m ) 0. (f) Complexed polymer 6a m ) 0.39. (g) Complexed polymer 6b m ) 0.54. Table 2. Reaction and Characterization Data for Polymer Precursors 5 and 6a polymeric precursor

equiv. Hhfac

equiv. Cu2O

reaction time (h)

precursor yield (%)

Cu(hfac) complexed (m)

δ 13C NMR of vinyl group (ppm)

δ 13C NMR of vinyl group in parent polymer (ppm)

5a 5b 5c 6a 6b

0.7 1.0 1.1 1.5 2.0

0.35 0.70 0.70 0.75 1.40

6 15 23 6 19

57 68 71 48 59

0.46 0.85 0.97 0.39 0.54

107-118 94-104

133-137 (3)

104-124 92-102

133-135 (4)

a Molar equivalents of Hhfac and Cu O reacted with 3 and 4 to give 5 and 6; the degree of complexation and 13C NMR shifts of the vinyl groups are also 2 quoted.

linear polymer; deviation from a linear relationship generally implies that branching or chain aggregation (in the limit cross-linking) is present in the polymer. Solutions of polymer 3 and precursor 5b were analyzed; the data are shown in Table 1 and the graph of µsp against concentration is shown in Figure 3.

Figure 2. Correlation between the amount of Cu(hfac) complexed to the chemical shift observed in the 1H NMR spectra. (a) Cu(hfac) complexing to 6 and (b) Cu(hfac) complexing to 5.

direct-write applications, it is important to have some knowledge of the likely effects complexation has on ink viscosity. Second, viscometry can be a useful method for determining whether a polymer is linear or branched and could possibly have provided information as to the presence of concentration- and loading-dependent intermoleculer chain interactions since the molar mass of the polymer precursor is known and can be used as a calibration point. Dilute solutions nornally display a linear increase in specific viscosity (µsp) with increasing concentration of a

As expected, polymer 3 exhibits a linear increase in µsp with increasing concentration. The solutions of precursor 5b, despite being more dilute than those of 3, displayed a higher µsp and clearly do not follow a linear relationship. This is an unexpected concentration-dependent behavior due to effects arising from the Cu-complex coordination behavior. Whether this behavior arises from inter- or intramolecular aggregation could not be determined from the 1H NMR data of 5. As the time scales for dynamic processess to be observable by NMR and viscometry are quite different, further studies are needed to correlate the data sets to each other. However, the unexpected viscosity behavior obviously has important implications for controlling and optimizing ink viscosity for direct-write and merits further studies. Thermal Analysis of Precursor Polymers. TG and DSC analysis were used to quantify the disproportionation process for precursors 5 and 6. These analyses were conducted under identical conditions (heating from 50 °C at 10 °C/min) under an atmosphere of N2. The combined thermograms are shown

2494 Chem. Mater., Vol. 18, No. 10, 2006

Rickerby et al.

Table 3. Tg Data for Polysiloxanes 5 and 6 with and without Cu(hfac) Complexed

a

polymer precursor

uncomplexed Tg(°C)

∆Cp (J g °C)

complexed Tg (°C)

∆Cp (J g °C)

complexed Tm (°C)

∆H (J/g)

5b 6b

-134.9 -139.9

6.31 0.50

-18.9 no Tg

0.29 no Tg

-38.5 and -46.9 75.8 and 77.4

0.10 and 0.13 1.57a

Total ∆H for both Tm’s.

Figure 3. Specific solution viscosities (µsp) plotted against solution concentration. Measurements were carried out in hexane at 25 °C. (a) Precursor 5b and (b) parent polysiloxane 3.

Figure 4. Thermal analysis of precursor 5 conducted at a heating rate of 10 °C/min. (a) DSC of 5b, (b) TG analysis of parent polymer 3, (c) TG analysis of 46% complexed precursor 5a, and (d) TG analysis of 85% complexed precursor 5b.

350 °C, indicating incomplete thermal decomposition. Under these heating conditions it is likely that volatile oligomers are evaporating and a certain amount of desirable depolymerization is occurring but this process is incomplete. As for the precursor polymers, exotherms were detected by DSC for sample disproportionation at 210 °C for 5b and 215 °C for 6b, which correspond well to the main weight loss observed in the TG thermograms (Figure 4 and Figure 5). The precursors show extensive weight loss, leaving behind only around 10-20% of the sample as residue, a lower value than expected if one assumes polymer decomposition to be independent from Cu loading levels and a high level of Cu(I) disproportionation. The weight loss percentages obtained from the TG data analysis of 5 and 6 are summarized in Table 5 along with calculated values for the residue percentages. Two methods were used to obtain these values; the first assumes that the same percentage of oligomer and polymer degradation products evaporates from the precursor samples as evaporates from the parent polymers. The second assumes that the evaporating oligo- and polymers carry away complexed Cu(hfac) (Table 5). These processes are summarized in Figure 6 where precursor disproportionation liberates volatile siloxane oligomers and generates Cu(hfac)2 and a contaminated Cu film. The measured residue percentages are still about 10% lower than expected and two possible explanations were considered to account for this discrepancy. Cu(hfac) could be desorbing from the precursor or the polymer may be depolymerizing more extensively in its precursor form than on its own. The desorption theory was originally considered because when these precursors were first used to prepare Cu films, haloed Cu mirror was observed to form extensively around the main deposits. We initially thought that this could only be happening if Cu(hfac) was desorbing from the precursor since the precursor being a polymer should not be volatile. This possibility came to light when Han et al.’s study on the kinetics of laser-assisted CVD of (hfac)Cu(VTMS) was concidered.38 They summarized the surface chemistry involved to be consistent with the following three reaction stages: (hfac)Cu(VTMS)(g) + σ f Cu(hfac)(s) + VTMS(g) (1)

Figure 5. Thermal analysis of precursor 6 conducted at a heating rate of 10 °C/min. (a) DSC of 6b, (b) TG analysis of parent polymer 4, (c) TG analysis of 39% complexed precursor 6a, and (d) TG analysis of 54% complexed precursor 6b.

in Figure 4 and Figure 5 and a summary of the findings are given in Table 4. A TG analysis thermogram for Cu(I) CVD precursor (hfac)Cu(COD) is also shown in Figure 4 for reference as are TG analyses of parent polymers 3 and 4. Parent polymers show losses of about 40% of their total weight (3 (60%); 4 (58%)) in the temperature range 90-

Cu(hfac)(s) f Cu(hfac)(g) + σ

(2)

2Cu(hfac)(s) + σ f Cu(hfac)2(g) + 2σ + Cu38

(3)

Cu(hfac) binds to the vacant surface site (σ) by releasing the VTMS ligand, and the bound Cu(hfac) species can either desorb or disproportionate. (38) Han, J.; Jensen, K. F. J. Appl. Phys. 1994, 75, 2240.

Polymeric Cu(I) Precursors for Copper Direct-Write

Chem. Mater., Vol. 18, No. 10, 2006 2495

Table 4. Summary of Decomposition Data Acquired via Thermal Analyses of (hfac)Cu[polysiloxane] Precursors 5 and 6 TG analysis

DSC

precursor

% Cu(hfac)

onset Tdec (°C)

75% weight loss (°C)

residue %

onset Tdec (°C)

finish Tdec (°C)

∆Hdec (J/g)

3 5a 5b 4 6a 6b

0 46 85 0 39 54

90 180 180 100 180 170

280a 313 246 300b 303 243

60.7 20.4 9.7 58.0 19.2 15.7

330 c 210 320 and 390 c 215

370 c 380 390 and 440 c 340

-56 c -287 -30 and -44 c -148

a

At 39% weight loss. b At 41% weight loss. c No data acquired.

Table 5. Precursor Weight Losses Observed by TG Analysis for Precurors 5 and 6 and Calculations Estimating the Cu Content of the Residue calculation/measurement

5a

5b

6a

6b

degree of complexation from 1H NMR weight % loss of parent polymera measured total weight loss % weight % of residue weight % of polymer in residue weight % of Cu in residue

46 40 79.6 20.4 25 7

85 40 90.3 9.7 16 9

39 42 80.8 19.2 28 8

54 42 85.3 15.7 23 8

total % residue weight % of polymer in residue weight % of Cu in residue

32 24 4

25 16 5

36 27 3

31 23 4

total % residue

28

21

30

27

calculated amount of residue method 1 calculated amount of residue method 2 a

Weight loss recorded from TG analysis of parent polymers 3 and 4.

Figure 7. Formation of the smallest cyclic product via the rearrangement of siloxane bonds during thermal depolymerization of poly(dimethylsiloxane) in the temperature range 460-600 °C in a nitrogen atmosphere.23,24

Figure 6. Proposed pathways for the thermally induced decomposition of (hfac)Cu[poly(siloxane)] precursors.

Relative to laser-assisted CVD, the conditions used to convert the precursor samples during TG and DSC analyses are far too mild to effect desorption so depolymerization being more extensive than expected is a much more reasonable explanation for this excessive weight loss. Figure 7 shows the mechanism by which poly(dimethylsiloxane) is thought to thermally depolymerize. Thermally induced disproportionation is, overall, an exothermic process for these precursors despite a large endothermic contribution caused by evaporation of volatile byproducts. In addition to byproduct evaporation there is

evidence that evaporation of some Cu(I)(hfac) complexed to volatile oligomers is occurring. As complexation of Cu to volatile species cannot on its own explain the excessive weight losses observed by TG analysis, we also conclude that the polymers are depolymerizing more extensively than expected. It would be interesting to see whether using a more monodisperse, oligomer free polysiloxane would still produce haloed Cu deposits. The onset temperature for disproportionation is higher than desired as a rule since sub 200 °C was our target. Only more reactive 6b shows that it is possible to come close to this temperature range since TG analysis for this sample was comparable with a monomeric precursor like (hfac)Cu(COD) as shown in Figure 4. Cu Film Characterization. Despite the thermal analysis data indicating that the Cu films contain, at best, only 23% metallic Cu, the films prepared using both precursors exhibited metallic luster and indicated unambiguously the formation of copper metal. The Cu films used for ion beam analysis were prepared by applying the neat precursor to a glass microscope slide and heating the slide on a hot plate, in a nitrogen atmosphere, to around 300 °C. Both precursors readily converted to shiny Cu films. These deposits were visibly haloed by a shiney translucent ring of a Cu mirror. The reverse side of the glass slide revealed clear, brown oily residues. This rather intriguing observation led us to conclude that the Cu films were forming as a crust over the precursor droplet, leaving polymer residues encased below this layer.

2496 Chem. Mater., Vol. 18, No. 10, 2006

Rickerby et al.

Figure 8. When heated on a glass slide, the precursors appear to form a crust of shiny metallic Cu at the precursor/atmosphere interface. Trapped volatile byproducts then erupt through the crust, damaging the Cu film.

Figure 10. Selected area scans for Cu deposit prepared from precursor 5c on a C substrate. (a) and (b) RBS spectrum of halo region with composition plot. (c) and (d) RBS spectrum of center of deposit with composition plot. (e) and (f) RBS spectrum of outer ring of deposit with composition plot.

Figure 9. Cu deposit prepared from precursor 5c on a C substrate (scan area ) 1500 × 1500 µm2). (a) RBS surface Cu atom map (channel 242-261, top 50 nm). (b) RBS medium energy Cu atom map (channel 214-241, next 150 nm). (c) RBS low energy Cu atom map (channel 186-213, next 150 nm). (d) PIXE Cu atom map. (e) PIXE Si atom map. Areas used to produce selected area scans are marked: the triangle for the halo, the star in the central region, and the cross in the annular region (Figure 10).

Volatiles evaporated up through the Cu crust, causing visible craters to form on the surface. The salient features of this process are illustrated in Figure 8. RBS and PIXE spectra were obtained for Cu deposits prepared from precursors 5c and 6b. As thermal analysis of these Cu precursors determined that there would be residual polysiloxane, glass substrates proved to be problematic for establishing a suitable RBS data fit for atom distributions in the Cu deposits. Therefore, a Cu deposit prepared from precursor 5c on a C substrate was used to eliminate this complication and allow more accurate and reliable data fitting. RBS and PIXE atom maps for a deposit prepared using precursor 5c on a C substrate are shown in Figure 9 and were used to establish data-fitting parameters. This deposit was not prepared with control of deposit size and topology; a Cu halo surrounds the deposit (the surrounding Cu in the PIXE maps) and there is a thicker annular region around the indented center of the deposit. Selected area scans of the three differing regions, the halo, the center of the deposit, and the outer ring were used to determine chemical composition and layer thickness within these regions. The Cu deposit on C allowed the relative percentages of Cu, O, and Si to be determined as the substrate was effectively invisible. The spectra acquired from the Cu deposit on C are shown in Figure 10 along with the depthatomic percentage plots used to generate the spectral fits. These were subsequently used to determine stoichiometry differences of various species as a function of depth.

Fitting simulated atomic compositions to these spectra enables the distribution of Cu and polysiloxane residues to be established as a function of depth. The halo, for example, is Cu-rich (∼Cu2O, with Cu areal density (thickness) of 247 × 1015 Cu/cm2) and Si is completely absent, which is consistent with a vapor phase deposition process with subsequent Cu oxidation in the time between film preparation and analysis. The deposit contained two very distinct regions; the central region is surface Cu (oxide) rich with ∼25% of (surface) siloxane contamination whereas the outer ring is less Cu rich (CuO is predominant) with increasing quantities of siloxane residues. The abundance of siloxane residues in the ring region increases with depth, from 40% on the deposit surface to 100% at the substrate surface. By contrast, in the central region the siloxane residues are most abundant at the deposit surface and are scarce at depth, only reaching 250 nm (sample depth in this region was 500 nm). The large amount of silicon on the surface is depleted at about 500 nm into the sample spot. There are two possible explanations for these data: the surface contains a significant contribution of silicon or it could be due to an artifact resulting from a porous sample spot. In this case thickness at the pore in energy terms is zero, and anything underneath the pore will appear at high energy (the silicon surface). When this deposit was prepared analyzing composition and Cu content were of greater priority than topology and dimensions and therefore the origin of the two regions in the deposit has not been established. It could be speculated that as the precursor droplet converts to Cu on the heated substrate the center of the droplet collapses washing outward, hence the difference in substrate depth between these regions. A similar deposit profile was reported by Jeynes et al. for ink-jetted Cu(I) precursors and this was attributed to an impact wave washing outward as the jetted droplet hit a heated surface.22 The RBS spectra themselves are also very similar in profile and Jeynes et al. report the central region being rich in Cu and the outer ring containing lesser amounts of Cu and greater amounts of impurities.

Polymeric Cu(I) Precursors for Copper Direct-Write

Figure 11. Cu deposits prepared from precursors 5c (scan area ) 1500 × 1500 µm2) and 6b (scan area ) 250 × 250 µm2) on glass substrates. (a) RBS medium energy Cu atom map (channel 214-241, upper 150 nm) for a deposit prepared from precursor 5c. (b) PIXE Cu atom map for a deposit prepared from precursor 5c. (c) RBS medium energy Cu map of a deposit prepared using precursor 6b (channel 214-241, upper 150 nm). (d) PIXE Cu map of a deposit prepared using precursor 6b. (e) PIXE line scan across the deposit prepared from precursor 6b. Areas used to produce selected area scans are marked.

Figure 12. RBS selected area scans of Cu deposits prepared from precursors 5c and 6b on glass substrates and their respective composition plots. (a) Halo region of a Cu deposit prepared from precursor 5c (205e15 Cu atoms/ cm2). (b) Halo region of a Cu deposit prepared from precursor 6b (40e15 Cu/cm2 or 4.7 nm Cu). (c) and (d) The central region of the deposit prepared from 5c. (e) and (f) The outer region of the deposit prepared from 5c. (g) and (h) The central region of the deposit prepared using 6b.

Using C as the substrate enabled us to fit the data unambiguously and also proved helpful for fitting the data from Cu deposits on glass more reliably. A suitable data fit was established for the different regions of the deposit prepared using 5c. The halo could be fitted using only Cu and O (Si was below the detecton limit); and the deposit composition was consistent with Cu oxides and SiOC3H6 residues. These data-fitting parameters were then used to treat the RBS data acquired for the deposit prepared on glass using precursors 5c and 6b, for which the atom maps, the spectra, and fitted spectra are shown in Figures 11 and 12. The atom

Chem. Mater., Vol. 18, No. 10, 2006 2497

maps for the deposit prepared using precursor 6b are included in Figure 11; this deposit was quite thin and a neighboring deposit is visible on the left of the scanned area. On glass, the siloxane residues in these Cu deposits are masked by the substrate signal; however, using the data fit parameters established on the C substrate, it is possible to disentangle the substrate signals from the siloxane residues. By including the substrate as Si25O60Ca3MgAlNa10, it is possible to achieve a good spectral fit using the same assumptions as were used for the C substrate deposit. The unambiguous data fit parameters established with the C substrate were subsequently applied to Cu deposits produced using precursors 5c and 6b on glass substrates. Not only was the deposit topography similar, a thin central region and a thicker ring around the outside of the deposit, but also the atomic compositions were consistent with the deposit produced using the same precursor on C. The central region is richer in Cu (Cu2O) than the outer ring. Siloxane residues are underlying in the central region but appear to lie close to the substrate surface in the annular region. Obtaining physical thickness information of the annular region inevitably indicates porosity since the depth scale assumes full density. The most likely explanation for the “intermixing” of the glass substrate signal with the deposit is due to variable thickness over the beam spot. The deposit produced using precursor 6b did not display the outer ring topography and proved to be less Cu rich. This could be due to this precursor being significantly more air-sensitive than 5c, hence, more impurities being incorporated into the film. The difference in topography might result from the deposit being far smaller and therefore effects such as waves traversing the liquid precursor droplet become less important. Cu deposits were also prepared on substrates with a range of electronic properties using precursor 5c. In CVD, deposition on nonconducting substrates is generally far more difficult than on conducting substrates; therefore, it was necessary to determine whether the substrate had an effect on these Cu precursors and method of deposition. RBS Cu atom maps of similarly prepared deposits from precursor 5c are shown in Figure 13 for comparison. The general topography in all cases involved a thin Cu rich central region and a thicker, more Cu poor perimeter region. From this it was concluded that the substrates’ electronic properties have no effect on this deposition process with respect to deposit formation. Additionally, no obvious variations in deposit composition were observed for the different substrates. What is not known is to what extent deposition conditions have an effect. Overall, these deposits were found to be rich in Cu, which possibly oxidized during the time between deposit preparation and analysis. Oxidation prevented an accurate determination of the quantity of impurities, incorporated during precursor conversion, from being made. Attempts were not made to establish deposit conductivities as the oxide abundance and porosity is likely to be too great. Considerable optimization of parameters in the deposition process such as substrate temperature could possibly yield conducting Cu deposits.

2498 Chem. Mater., Vol. 18, No. 10, 2006

Rickerby et al.

Figure 13. Low energy RBS Cu atom maps (channel 186-213, 0-150 nm above Si surface) of deposits prepared using precursor 5c on different substrates. Al scan area ) 2000 × 2000 µm2, Si (n-type) scan area ) 1000 × 1000 µm2, Si (p-type) scan area ) 1500 × 1500 µm2, C scan area ) 1500 × 1500 µm2, glass scan area ) 1500 × 1500 µm2.

Conclusions [(hfac)Cu]poly(siloxane) precursors were synthesized as a new class of Cu(I) precursor compounds targeted specifically for direct-write applications. Precursor polymers were fully characterized by spectroscopic and thermal techniques and their ability to deposit Cu films was assessed. The Cu ink viscosity increased with the level of Cu centers loaded onto the polymer. Though these precursors did not prevent deposit haloing when converted in their liquid precursor form, they did provide valuable information on how liquid precursors convert thermally to copper on a heated substrate. TGA and DSC data allowed us to formulate a deposition mechanism which was fully supported by RBS and PIXE data analyses. Haloing of the deposits prepared from these precursors most likely results from evaporation of

Cu(hfac) complexed to small (cyclic) oligomers. These oligomers could readily be prevented from forming by polymer fractionation or the use of higher molecular weight polysiloxanes. The available data suggest that film formation in these precursors proceeds at the atmosphere/droplet interface and is presumably driven by evaporation of Cu(hfac)2. Acknowledgment. We would like to express our gratitude to the EPSRC and Patterning Technology Limited (PTL) for funding this research project. A.S. is grateful for funding from the 5FP project NAS-MICRO-XRF (Contract No. G4RDCT2000-00402) as well as from the Hungarian Research Foundation (OTKA) under Contract No. T034381. CM052103S