Characterization of fractionated phenolic resins used in photoresists.
Photoresists used in the manufacture of semiconductors and microchips are comprised primarily of a polymer (resin), solvent, sensitizer, and additives (1). It has been shown that phenol/formaldehyde resins are important binder resins especially useful for positive g and i-line sensitive resists (2). These resins are commonly called novolak resins. The novolak structure/property relationship is complex and provides clues for improving photoresist performance. One of the key properties of these resins is polydispersity or the distribution of molecular weight fractions and the role played in performance of the resin. Early work suggested that resins with narrow polydispersities coupled with the use of appropriate speed enhancing resin additives improved overall performance (3-6). Low polydispersity resins were made by fractionating the novolak resin by dissolving solid resin in a solvent and adding a non-solvent to separate out the undesired low molecular weight resin fractions (1, 2). Another common method of fractionating novolak resins consisted of distributing molecular weight fractions between two non-miscible organic layers (7).
Conventional fractionation methods tended to have relatively low yields; however, the photoresists made with these fractionated resins had improved performance when compared to those made with unfractionated resins. Recently, a novel method of fractionating resins using a liquid/liquid centrifuge was shown to provide higher yield of resins with similar properties to resins obtained from conventional processes (8-11). Thus the current work was undertaken to characterize the resins isolated by the different fractionation methods and to improve the efficiency and yields of the fractionation process.
Solid novolak resins were either synthesized by an acid catalyzed phenol/formaldehyde condensation reaction or were obtained commercially from vendors.
Resins were synthesized by condensation of formaldehyde with various combinations of meta-cresol, para-cresol and, in some instances, other alkyl substituted phenols such that the mole ratio of formaldehyde to total phenolic components was about 0.7/1. Reactions were run with about 0.3% of oxalic acid as the catalyst. After reacting the mixture for 6 hours at 90-95[degrees]C, the mixture was distilled at temperatures up to 200[degrees]C and finally at 200[degrees]C and 25-30 mm Hg vacuum to remove unreacted cresols. The molten resin was then poured into a crystallization dish and allowed to cool. The cooled solid mass was then broken up and powdered using a mortar and pestle.
Conventional Fractionation Method
The conventional method used in this study for fractionation was to dissolve the solid resin in an organic solvent and then add the correct ratio of a non-solvent (typically water) to precipitate the desired fractionated resin. Once the aqueous layer containing the low molecular weight resin (undesired) was removed, the fractionated resin could be used. A resist solvent was added to the resin layer to dissolve the solid precipitate and the final resin solution was obtained by distillation to remove residual water entrained in the resin during the fractionation process. Typical resist solvents were propylene glycol methyl ether acetate (PGMEA) or ethyl lactate (EL).
Liquid/liquid Centrifuge Method
An extraction technique using a specially designed liquid/liquid centrifuge was developed as the new alternate fractionation method (7-10). The centrifuge was designed by the CINC Corp. of Carson City, Nevada, but had heretofore not been used for fractionation of polymeric materials. The centrifuge was designed with sets of two inlet and two outlet ports. Resin solutions could be introduced through one inlet port while an immiscible solvent was simultaneously fed into the other inlet port. Separation by density differences led to isolation of fractionated resin from one exit port while the undesired low molecular weight materials were removed through the other port. It was shown that the technique could be used in a continuous operational mode as opposed to the common batch operation previously used in traditional fractionation schemes.
Using gear pumps, solutions of the unfractionated resin were pumped into one inlet port of the centrifuge while the extracting solvent was introduced into the other inlet port. The heavy phase outlet containing the fractionated resin and the lighter phase containing the extracting solvent along with lower molecular weight fractions of the initial resin were obtained through each of the two exit ports. Forces in the range of 100 to 1000 "G" could be obtained at the rotor wall (12).
Resin solutions useful for making advanced positive i-line photoresists could be obtained by distilling off any entrained extracting solvent and adjusting the % solids of the resultant solution. The choice of initial solvents like PGMEA or EL provided the resin in those solvents typically used in the manufacture of resists for the microelectronics industry. All that was needed for continuous operation was a continuous supply of the unfractionated resin solution and the extracting solvent.
Initial experiments showed that wide variations in the rotational speed of the centrifuge only had relatively small effect on the separations, so the experimental work centered on changes in temperature and the resin solution to extraction solvent ratio (R/H). Although the centrifuge could not be heated directly, the elevated temperature experiments were run by circulating the resin solution and extraction solvent in separate jacketed vessels until the desired temperature was obtained. Experimental conditions used for fractionating a resin prepared by condensation of m-cresol and p-cresol with formaldehyde are outlined in Table 1.
Resin solutions were prepared by dissolving 30% by weight of solid resin in PGMEA. The PGMEA solutions were diluted to 20% solids by the addition of methanol (MeOH) resulting in a 58%/42% ratio of PGMEA/MeOH for the final solvent combination (R in Table 1). Hexane was used as the immiscible extraction solvent (H). In all cases, a two-phase system could be demonstrated beforehand in simple mixing experiments. The centrifuge was fitted with two gear pumps that were calibrated and used to pump the resin solution and the extraction solvent into the inlet ports of the centrifuge.
Results show that the resin samples increased in average molecular weight and decreased in polydispersity as expected by the removal of lower Mw fractions from the broad mixture of these fractions. The Dissolution Rate (DR. A/sec) also became slower as the speed enhancing effects of the low Mw materials were removed from the mix.
In a second series of experiments, the same resin was isolated in ethyl lactate (EL) and a similar series of fractionations were run with the new fractionation method. Table 2 shows the conditions and results of those experiments. The resin solution (R) was comprised of the resin in a fixed ratio of ethyl lactate/MeOH at 58%/42% for this series of experiments. Hexane (R) was used as the extraction solvent. In general, the same trend of higher average Mw, lower polydispersity and slower photospeed was observed although the overall dispersities were somewhat higher than shown in the PGMEA fractionation.
Table 3 shows the results of a fractionation series similar to that shown in Table 2 except that the resin was synthesized using a ratio of 5 parts of m-cresol, 4 parts of p-cresol and 2 parts of 2,3,5-trimethyl phenol (TMP) condensed with formaldehyde at a ratio 0.7 mole of formaldehyde to 1 mole of total phenols and using oxalic acid as the catalyst (0.3% by weight). This resin was designated MPT. All these fractionations were done at ambient temperature (22-25[degrees]C) and overall resin solution solids of 20%. The resin solvent composition varied in this set of experiments and three similar but separate batches of starting MPT resin (control samples 1-3) were used.
In trial 1, the resin solution was mixed with the extraction solvent (water) and fed through a single port of the centrifuge and the apparatus was used to mix and separate the phases in one operation to mimic the batch process method of fractionating the resin.
Changes in Mw and DR show the wide fluctuation in the amount of lower Mw fractions that can be removed by this technique. Yields and degree of fractionation are only limited by the immiscible solvents used for the input solutions, the temperature of the fractionation and the ratio of the inlet solution feed rates.
Separation of Oligomers by Column Chromatography
In order to characterize the composition of the materials removed by the fractionation methods, the low molecular weight resin fractions were dissolved in acetone and were separated by column chromatography on a silica gel column using mixtures of hexane/acetone as the eluting solvent. Fractions were isolated and characterized by [C.sup.13] nuclear magnetic resonance, Gel Permeation Chromatography (GPC) and Gas Chromatography/Mass Spectrometry (GC/MS) on any volatile components.
The resins obtained from the heavy and light phases were characterized using GPC. Molecular weight average (Mw) and number average (Mn) molecular weights were measured using THF solvent. Four chromatography columns, 500A, 1000A, 10,000A, and 100,000A (available from Waters), were connected in series. The system was calibrated using polystyrene standards. The refractometer detector was a Waters 410 model.
Measurement of Dissolution Rates
Dissolution rates (DR) were determined with a laser interferometer in a 2.38% tetra-methyl ammonium hydroxide (TMAH) developer. A Xinix[R] Process Monitor Model 2200 Scope was used to analyze the interference scan.
[.sup.13]C NMR Analysis
Each resin fraction was dissolved in an appropriate solvent (1,4-dioxane-[d.sub.8] or DMSO-[d.sub.6]) and the spectra were run on a Varian VXR-200 spectrometer at Tilde50 MHz for [.sup.13]C nuclei. Spectra were run overnight and 15-20 K scans were obtained for analysis.
Volatile materials from the samples eluted from the column separations of the low Mw fractions isolated were analyzed by gas GC/MS.
A Hewlett-Packard Model HP5890 gas chromatograph (GC) coupled with a MSD HP 5970 mass spectrometer was used. An RXT-1 30 x 0.25 x 0.25 column was used and the temperature was held at 50[degrees]C for 2 minutes ramping up to 280[degrees]C at 5 degrees/min and then held at 280[degrees]C for 10 minutes. The injector and detector were set at 280[degrees]C and the MS scan range was 50-550. Injection volume was 1 [mu]L with a solvent delay of 2 min.
Selected resins isolated as resin solutions in either PGMEA or EL by standard batch fractionation methods or the new centrifugal fractionation technique were compared for lithographic properties.
Fractionated resins were formulated into i-line sensitive photoresists using formulations where the resins isolated by standard fractionation methods were compared to those isolated by the centrifugal method. MPT resin was chosen for direct lithographic comparison.
RESULTS AND DISCUSSION
Surprisingly, higher resin yields of comparable lithographic quality resins could be obtained with the new fractionation method and, in some instances, even improved functionality was observed. Thus it was concluded that not only the amount of low Mw materials removed but also the composition of those fractions determined the functional characteristics of the fractionated resins. Tables 1 - 3 show the Mw changes that resulted from the fractionation conditions used for the separations. Component distributions were characterized to determine the differences that might explain these results.
Analysis of structure/functional property relationships for novolak resins containing cresols has only been recently reported (13- 15). Although early [.sup.13]C NMR characterization of novolak resins focused on the aromatic portion of the spectrum (15), more recent work has centered on the regions between 11 - 23 ppm and 23 - 40 ppm where the cresol methyl group and the methylene linkages could be assigned to the possible isomers (16). We basically used the same technique for assignment of methyl and methylene groups in our study. Certainly the method is useful in judging the relative differences between fractionated resins although definitive assignments were difficult especially with the resins containing m-cresols since there were three active sites for methylene insertion.
Figure 1 shows the possible substitution positions for the methylene bridge carbon when m-cresol units are joined. The positional designations were made in relation to the phenolic hydroxy group. Under typical acid catalyzed cresol/formaldehyde condensation reactions, frontier electron density values predicted that the most active site is the C4 or para carbon followed by the C6 carbon and lastly by the sterically hindered C2 carbon (16). In addition, one might expect a fairly linear polymer since substitution at C2 is not favored. This should also limit a large amount of tertiary substitution because of steric factors (a hexa-substituted aromatic).
Douki, et al. (16) have estimated the overall ortho-ortho' (o-o') substitution by dividing the [.sup.13]C NMR peak area between the range of about 23-31 ppm by the area over the range of about 23-42 ppm assigned to the total inter-ring methylene peak absorption. The area between about 31 - 42 ppm represented the ortho-para' (o-p') and para-para' (p-p') as a percentage of the total area of the peaks between 31 - 42. Likewise, a combination of trisubstituted phenolic segments plus mono-substituted end groups were judged by dividing the extreme upfield NMR absorption area at about 14 ppm by the entire methyl absorption area between about 14 - 21 ppm. As a further, somewhat arbitrary, differentiation, methylene linkages in the range from about 31 - 35 were assigned to the o-p' substitution while the remainder were assigned to the p-p' substitution pattern.
As might be expected, the NMR spectrum for MPT resin (Table 3) containing a third phenolic component such as 2,3,5 trimethyl phenol (TMP), was more complicated due to overlap of absorptions in the ranges studied and the possible isomers increased geometrically since there were now three possible substitution sites from m-cresol and two additional sites from TMP (Fig. 2). In spite of this complexity, we used the same basic formulas to calculate the relative differences in the substitution of resins isolated by each of the two fractionation processes described earlier in this report. These relative structural differences should still be valid in judging the yield/functional advantage of one fractionation method against the other.
[FIGURE 1 OMITTED]
The peak areas for o-p' and p-p' substitution were combined in this evaluation due to the large overlap in this segment of the NMR spectrum. Steric factors should greatly favor p-p' substitution with the mcresol and TMP used in making the MPT resin; however, the para-cresol moiety can only form the o-o' methylene substitution pattern. Table 4 summarizes differences between the analyses of spectra from the fractionated resins and the low Mw fractions removed during the fractionation of MPT by both methods. The unique high field absorption at about 14 ppm was used to collectively total the end groups and trisubstituted m-cresols. Products from the trisubstituted category would also show up in the o-o', p-p' or o-p' substitution categories, thereby leading to the total percentages >100%. Overall fractionation yields of each process and the dissolution rate of the materials are also shown.
[FIGURE 2 OMITTED]
Results show that the centrifugal fractionation process gave a yield improvement of about 35% while functional properties were mostly improved as well when compared with resists made with resins from the conventional process. Since the DR of resins isolated by the new process was faster and closer to the unfractionated material, another distinct advantage was realized since less speed enhancing additive was needed to attain roughly the same processing speed for patterning micro-lithographic images in finished photoresist formulations. In essence, the resins had inherently faster apparent photospeed. Figure 3 shows lithographic comparison of a control resist (standard fractionation) vs. resist made with resin fractionated using the liquid/liquid centrifuge. Resist photospeed was improved by 20 - 25% while the resolution of dense lines was comparable. Other resist parameters (depth of focus and focus latitude) were also similar.
[FIGURE 3 OMITTED]
Thus the new fractionation protocol provides a critical raw material with improved or comparable functional properties while substantially increasing the overall yield of the material. Another potential advantage was the reduced amount of "waste" material generated (15% vs. 50%) although some recent work has suggested that low Mw material removed may be recycled for use in some resist coatings as speed enhancement additives (17).
Figure 4 shows an overlay of the GPC chromatograms for the unfractionated MPT resin (top), the fractionated resin from the centrifugal separation and the material removed during the fractionation step (bottom). The removal of low Mw material (dimer, trimer and other lower oligomers) is clearly indicated by the lower trace.
We concluded that the overall distribution of higher o-o' and lower total trisubstituted fragments of the resin isolated from the new fractionation method was undoubtedly responsible for the properties seen when used in resist formulations.
Volatile materials from the waste layer were characterized by GC/MS of the material removed during the fractionation step. Dimers of the meta or para-cresols with the 2,3,5 trimethyl phenol with a small amount of dimers of the meta or para cresols or cross products were found under these conditions (18). Figure 5 shows the mass spectrum and the rationale for the peak fragments observed.
Centrifugal fractionation removed less of the lower Mw o-o' isomers than the standard method while removing quite a bit more of the trisubstituted isomers. The net result was a much higher yield with no drop-off in functional performance of the resin. This was demonstrated by formulating the resins in resists and determining the relative properties of the resists made with these resins. Resists were formulated in an ethyl lactate (EL)/n-butyl acetate (NBA) solvent mixture using resin (fractionated by either method), a mixture of two different proprietary photoactive compounds, and a surfactant. A small amount of an oligomeric phenolic speed enhancer was added to each formulation to fine tune the desired photospeed. Since less of the lower Mw fractions were removed with the new fractionation method, slightly less amount of the speed enhancing additive was needed with this material.
[FIGURE 4 OMITTED]
Resists were prepared with the following conditions:
Coating Thickness: 0.94 [mu]m
Soft Bake: 90[degrees]C/90 seconds on a hot plate
Exposure: 0.54 numerical aperture (NA) NIKON[R] i-line stepper
Post Expose Bake: 110[degrees]C/60 seconds on hot plate
Development: AZ[R] 300 MIF TMAH developer at 23.0[degrees]C for 1 minute
In all cases, the functional properties matched or exceeded those of the resists made with resins from the standard fractionation process. Advantages such as faster photospeed and better line resolution were measured when using the resins from the new fractionation process. All of the other functional criteria (exposure latitude, linearity e.g.) were comparable.
[FIGURE 5 OMITTED]
GPC Overlay of Fractionated Resin
Novolak resins fractionated in a novel liquid/liquid centrifugal method were characterized and compared to resins obtained by traditional separation methods. Small but finite differences in the distribution of the myriad isomers isolated by each method were found by NMR and GC/MS analyses. These differences were responsible for higher overall yields of product without compromising functional properties when formulated into photoresists. Analyses of NMR spectra were used to characterize the relative differences between the isolated resins. The method utilized a method where methylene linkages between adjacent phenolic components could be characterized.
The new fractionation technique was shown to be faster, gave higher yields of functionally useful resin and may be used in a continuous process as opposed to the batch mode commonly used with existing methods.
Table 1. M/P Cresol Resin Fractionation (PGMEA). PD Trial Temp. [degrees]C rpm R/H Mw, GPC Mw/Mn Control (Unfractionated) 7500 15 1 55 1000 2 9917 5.1 2 0 5000 1 9941 5.4 3 0 1000 2 9046 5.1 4 55 1000 0.5 11710 5 5 55 5000 2 9601 5 6 25 3500 2 9536 5.1 7 25 1000 1 10069 5 8 55 3500 1 10002 5 9 0 1000 0.5 8788 5.3 10 25 5000 0.5 9846 5.1 11 25 5000 2 10274 5.3 12 0 3500 2 10115 5.8 13 0 3500 0.5 9513 5.4 14 55 5000 1 12544 6 15 0 1000 1 9803 5.4 DR Trial A/sec Control (Unfractionated) 200 1 - 2 - 3 - 4 70 5 - 6 - 7 - 8 - 9 - 10 - 11 - 12 - 13 - 14 95 15 - Fractionated Phenolic Resins Table 2. Meta/Para Cresol Fractionation (EL). Resin Feed Trial Temp. [degrees]C R/H mL/min Control (Unfractionated Resin) 1 55 2 400 2 40 2 400 3 35 0.5 75 4 25 0.5 75 5 40 2 400 6 25 2 400 7 55 2 400 8 25 2 400 9 55 0.5 75 10 55 1 200 11 25 1.25 225 12 25 1.25 225 13 55 1.5 250 14 35 1.5 250 15 45 0.5 75 16 55 0.5 75 PD Trial Mw, GPC Mw/Mn DR A/sec Control (Unfractionated Resin) 9000 15 300 1 11247 8.8 138 2 9.2 146 3 10169 8.6 146 4 11368 9.5 187 5 11728 9.3 150 6 11857 9.6 176 7 11340 4.3 140 8 11674 9.4 187 9 14155 8.1 41 10 12659 9.7 159 11 11980 9.6 190 12 11183 9.1 191 13 12032 9.3 113 14 11688 9.4 143 15 10999 9 127 16 13862 6.8 28 Table 3. MPT* Resin Fractionation (EL). Resin solution Extract R/H Trial R solvent H (mL/min) Control 1 1 MeOH/EL Water (**) 3.25 Control 2 2 EL/MeOH (***) Hexane 1 Control 3 3 EL/MeOH (***) Hexane 2 Control Sample same as Control 3 4 EL/MeOH (***) Hexane 0.5 Trial Temp, [degrees]C Mw, GPC DR, A/sec Control 1 2104 800 1 RT 5147 60 Control 2 2860 600 2 40[degrees]C 6365 270 Control 3 2383 686 3 RT 2653 584 Control Sample same as Control 3 2383 686 4 RT 2870 533 (*) Meta-Cresol\Para-Cresol\2,3,5 Trimethylphenol Resin (TMP). (**) Water added directly to MeOH/EL: 79%/21% resin solution and fed into only one inlet port for separation. (***) EL/MeOH: 58%/42% solvent used to make the 20% solids solution fed into one port while hexane was fed into the second inlet port. Table 4. Comparison of Fractionated MPT Resin. Sample % ortho-ortho' % o-p' + p-p' % Trisubst. Control Resin 77 23 14 Standard Fractionation 71 29 17 Waste Layer 81 19 Centrifugal Fractionation 74 26 10 Waste Layer 87 13 15 Sample DR, A/sec Yield, % Control Resin 690 Standard Fractionation 60 50 Waste Layer Very fast (*) Centrifugal Fractionation 500 85 Waste Layer Very fast (*) (*) DR's could not be measured in the developer used.
1. P. Van Zant, Microchip Fabrication: A Practical Guide to Semiconductor Processing. 2nd Ed., McGraw-Hill, New York (1990).
2. M. Dalil Rahman, D. P. Aubin, and D. N. Khanna, Proc. SPIE, 2724, 542 (1996).
3. M. Hanabata, F. Oi, and A. Furuta, Polym. Eng. Sci., 32, 1494 (1992).
4. M. Hanabata, Y. Uetani, and A. Furuta, Proc. SPIE, 920, 349 (1988).
5. M. Hanabata, F. Oi, and A. Furuta, Proc. SPIE. 1446, 132 (1991).
6. M. M. Cook, M. D. Rahman, and P. H. Lu, Proc. SPIE, 3333, 1180 (1998).
7. M. D. Rahman, P. Lu. M. M. Cook, W. Kim, D. N. Khanna, Proc. SPIE, 3333, p. 1189 (1998).
8. S. Wanat, M. D. Rahman, J. J. Kokoszka, and B. Narasimhan, U.S. Patent 6,121,412, Preparation of Fractionated Novolak Resins by a Novel Extraction Technique (2000).
9. S. Wanat, M. D. Rahman, B. Narasimhan, D. McKenzie, and M. Cook, Proc. SPIE, 3999, 513 (2000).
10. B. Narasimhan, S. F. Wanat, and M. D. Rahman, Polym. Eng. Sci., 40, 2251 (2000).
11. M. M. Cook, M. D. Rahman, S. F. Wanat, D. McKenzie, B. Narasimhan. R. Fea, and M. Munoz, Proc. SPIE, 3333, 1175 (1998).
12. C. Luri, Private Communication, CINC[R] Corporation (1999).
13. L. E. Bogan, Macromolecules, 24, 4807 (1991).
14. M. A. Khadim, M. D. Rahman, and D. L. Durham, Proc. SPIE. 1672, 347 (1992).
15. R. R. Dammel, Diazonaphthoquinone-based Resists, SPIE Optical Engineering Press, Bellingham, Wash. (1993).
16. K. Douki, T. Kajita, and S. Iwanaga, Proc. SPIE, 3333, 384 (1998).
17. S. F. Wanat, M. D. Rahman, D. Durham, D. McKenzie, and M. Cook, Proc. SPIE, 3678, 1185 (1999).
18. S. F. Wanat, B. Narasimhan, D. McKenzie, Z. Xiang, and M. D. Rahman, Proc. 12th Int. Conf. on Photopolym., SPE, Brookfield. Conn. (2000).
STANLEY F. WANAT * and M. DALIL RAHMAN
Clariant Corp.-AZ[R] Electronic Materials
70 Meister Avenue
Branchburg, NJ 08876
|Printer friendly Cite/link Email Feedback|
|Author:||Wanat Stanley F.; Rahman, Dalil M.|
|Publication:||Polymer Engineering and Science|
|Date:||Oct 1, 2003|
|Previous Article:||Preparation of starch-graft-polyacrylamide copolymers by reactive extrusion.|
|Next Article:||Layered silicate nanocomposites based on various high-functionality epoxy resins. Part II: the influence of an organoclay on the rheological behavior...|