Microelectronics Technology - American Chemical Society


Microelectronics Technology - American Chemical Societypubs.acs.org/doi/pdf/10.1021/bk-1995-0614.ch019macroscopic dimens...

2 downloads 80 Views 3MB Size

Chapter 19

Silicon-Containing Block Copolymer Resist Materials

Downloaded by UNIV OF CALIFORNIA SAN DIEGO on July 15, 2016 | http://pubs.acs.org Publication Date: May 5, 1995 | doi: 10.1021/bk-1995-0614.ch019

Allen H. Gabor and Christopher K. Ober Department of Materials Science and Engineering, 214 Bard Hall, Cornell University, Ithaca, NY 14853

Polymer chains of controlled block and graft architecture have been investigated for use as microlithographic resists. For the last ten years, such research has focused on polymers having a silicon-containing block. These materials offer highly desirable properties, including excellent oxygen reactive ion etch resistance and good thermal stability due to their microphase separated structure. We review the use of silicon-containing block copolymers as resist materials and update the reader on our own research in this developing area. This will include a discussion of our most recent block copolymer resists which are developable in environmentally friendly supercritical CO2. Materials designed for use as resists-of-the-future must meet stringent requirements. In addition to resolving features less than 0.3 μπι, the resist must stand up to both dry and wet etches, develop in environmentallyfriendlysolvents, adhere well to the wafer, have high contrast and be highly sensitive to the exposing radiation. We believe that block copolymers will allow further improvements in resists so that these demands can be met. Although, block copolymer resists have been prepared and studied for the last two decades, there seems to be a general lack of awareness of this literature. Some of this is due to the limited readership of work "published" only as patents or in conference proceedings. This paper consists of three sections. In the first section we briefly review polymer phase behavior so that the reader understands what parameters need to be considered when designing block copolymer resists. In the second section we review past research on silicon-containing block copolymers designed to be microlithographic resists. We also review silicon-containing graft copolymer resists which, while not block copolymers, do have a controlled architecture. (By 'controlled architecture' we mean that the placement of the monomer units making up the polymer chain is not random but instead is regulated by the polymer chemist. The polymer chemist, for example, could prepare a graft copolymer by copolymerizing a monomer with a macromonomer or alternatively could prepare a block copolymer through a sequential addition of different monomers to a reactor.) In the final section we present a detailed account of our own research on block copolymers for 193 nm resist materials. Often, the superior properties of polymers with controlled

0097-6156/95/0614-0281$12.00/0 © 1995 American Chemical Society

Reichmanis et al.; Microelectronics Technology ACS Symposium Series; American Chemical Society: Washington, DC, 1995.

282

MICROELECTRONICS TECHNOLOGY

architecture are caused by the microphase separated structures that these types of polymers form. We therefore begin by reviewing the phase behavior of polymers.

Downloaded by UNIV OF CALIFORNIA SAN DIEGO on July 15, 2016 | http://pubs.acs.org Publication Date: May 5, 1995 | doi: 10.1021/bk-1995-0614.ch019

Section I: Phase Behavior of Polymers Different polymers are rarely miscible with each other and thus when blending two polymers, phase separation almost always occurs (1). unlike small molecules, which have a high entropie driving force for mixing, large linear chain molecules have a low entropie driving force for mixing (2). The enthalpic forces generally favor phase separation since the interaction energy is usually lower when two identical molecules are neighbors, than when two different molecules are neighbors. The morphology, which results from the phase separation of two polymers, is dependent on kinetic factors. In order to lower the free energy, the mixture decreases the area of interface between the two phases (3). This results in materials with phases of macroscopic dimensions and in many cases with disappointing properties (4). For example, the strength of the interface between the phases can be weak due to the absence of chains bridging the interface. In contrast, many of the spectacular properties of block copolymers result from the microphase separated structures that they form (5). Block copolymers consist of at least two different polymer segments, covalently connected together. A block copolymer with two blocks is termed a diblock copolymer. While the blocks often phase separate due to a low entropy of mixing, they are not able to macrophase separate due to the covalent bonds connecting them. The connections between the blocks will cause the mixture to have a larger interfacial area than the corresponding homopolymer mixture. The interfaces of the block copolymers, will be mechanically strong due to the chains of the block copolymers crossing them (6). Although the morphology of block copolymers is commonly described as microphase separated, nanophase separated is actually more descriptive since the size of aie phases is roughly equal to the radius of gyration of the blocks, typically 5-30 nm (7). A diblock copolymer, that does microphase separate, can have either an ordered or disordered microstructure depending on the magnitude of the enthalpic driving force for phase separation. In order for block copolymers to form an ordered microphase separated structure, the enthalpic driving force must be higher than the minimal enthalpic driving force needed for a mixture of corresponding homopolymers to phase separate. This is because there is an entropie penalty associated with the requirement that the joints between the two blocks be positioned at an interface (8). Also, the compositional sharpness of the interface increases with the enthalpic driving force for phase separation. The enthalpic driving force increases as XN increases where X is the Flory Huggins parameter and Ν is the number of monomelic repeat units of the chain (9). Thus, one way to increase the driving force is to increase the length of the block copolymer. The Flory Huggins parameter, X, relates the interaction energy cost associated with having the same or different nearest neighbors. For symmetric diblock copolymers (block copolymers which have two distinct blocks each occupying the same volume), when XN is much greater than 10, the interfaces between the two phases are narrow and compositionally sharp and the microstructure is ordered as shown in Figure 1 (10). As XN decreases towards 10, the width of the interface is increased but the microphase separated structure is still ordered. At approximately XN=10, there is an order-disorder transition (ODT). Block copolymers with values of XN greater than -2 but less than 10 still microphase separate, but the structures formed do not have long range order. For XN less than ~2, the two blocks are miscible with each other and the material is homogeneous. When the enthalpic driving force is large (XN greater than 10), different ordered microstructures can be obtained depending on the length ratio of die two blocks. Figure 2 shows the structures predicted theoretically for a diblock copolymer (9). The symmetric diblock copolymer forms lamellae where the thickness of each lamella is on the order of the radius of gyration of the polymer. Diblock copolymers with asymmetric blocks have different equilibrium structures. At very asymmetric block Reichmanis et al.; Microelectronics Technology ACS Symposium Series; American Chemical Society: Washington, DC, 1995.

19.

GABOR & OBER

Silicon-Containing Block Copolymer Resist Materiah

Downloaded by UNIV OF CALIFORNIA SAN DIEGO on July 15, 2016 | http://pubs.acs.org Publication Date: May 5, 1995 | doi: 10.1021/bk-1995-0614.ch019

ratios, the block copolymers form spheres of the minority phase in a continuous matrix of the majority phase. The structures shown in Figure 2 are at equilibrium because eliminating additional interfaces actually increases the free energy since such elimination would require the polymer chains to stretch further out of their random coil configuration (11).

Figure 1. A symmetric block copolymer with both a black and a white block will phase separate when the enthalpic driving force (XN) is large enough. The top pictures show an aerial view of the monomeric-unit composition. The bottom graphs show the volumefractionof black block as a function of r for the disordered structures and r perpen­ dicular to the lamellae for the ordered structures (adapted from 10).



Increasing length of black block relative to white block

Short Black Block Relative to White Block

Fairly Symmetric Block Copolymer

Short White Block Relative to Black Block

Figure 2. Microphase separated structures predicted by theory for a diblock copolymer. In this figure one block is color coded black while the other block is coded white (adapted from 9). The different microstructures formed by block copolymers act as microcomposites. For example, poly(isoprene-è-styrene) which has a microstracture of spheres of poly(isoprene) in a continuous matrix of poly(styrene) has two distinct glass transition temperatures (Tg's): a low Tg for the isoprene phase and a high Tg for the styrene phase. However, the material would not become rubbery until the Reichmanis et al.; Microelectronics Technology ACS Symposium Series; American Chemical Society: Washington, DC, 1995.

283

Downloaded by UNIV OF CALIFORNIA SAN DIEGO on July 15, 2016 | http://pubs.acs.org Publication Date: May 5, 1995 | doi: 10.1021/bk-1995-0614.ch019

284

MICROELECTRONICS TECHNOLOGY

temperature was raised above the Tg of the continuous poly(styrene) phase. Thus, a material can have the high modulus and thermal stability of poly(styrene) but take advantage of toughening by poly(isoprene). At the other extreme of monomelic composition, very different properties can be observed. For example a poly(styreneMsoprene-fc-styrene)which has a rubbery continuous phase of poly(isoprene) can have properties similar to a rubber with chemical crosslinks. The behavior occurs because the high Tg poly(styrene) phase acts as a physical crosslink for the poly(isoprene) phase. However, this block copolymer rubber is thermoplastic since when heated above the Tg of the poly(styrene) it can be reshaped (5). When block copolymers are properly designed and applied they offer the undiminished desirable properties of each phase and minimize the undesirable properties of each phase. The superior properties of block copolymers are the reason that several different groups have investigated block copolymers as resists. Section Π: Review of Silicon-Containing Copolymer Resists With Controlled Architecture Copolymers with controlled architecture have been sporadically investigated as resist materials for the last 20 years. They were and still are studied because of their promise of offering superior properties compared to corresponding random copoly­ mers, i.e. copolymers having similar molecular weight and chemical composition to that of the block copolymer, but with the monomer units in a random sequence. During the last ten years most research on block copolymer resists has focused on preparing and using them as the imageable polymer of bilevel resist schemes. Bilevel Resist Scheme. Bilevel resist schemes are effective in the intermediate to advanced stages of lithography where a single-level resist might not cover the circuitry previously deposited on the silicon wafer. The bilevel resist scheme involves planarizing the surface of a wafer with a nonimageable, organic layer and imaging the desired pattern with a thin, top-coated resist (12). There are other lithographic advantages to the bilevel resist system (13). Due to the thinness of the imageable layer, optical lithography can be performed with minimal linewidth irregularities associated with the problem of limited depth of focus. Also, when exposing an electron beam (e-beam) resist, fewer backscattered electrons "expose" the resist, since many are stopped in the planarizing layer. Finally, the bilayer system offers greater ease in obtaining high aspect ratio features (the height of a feature divided by the width) since a planarizing layer with thermal stability superior to that of the resist can be used. The lithographic steps for bilevel imaging are outlined in Figure 3 (14). The resist must offer resistance to oxygen RIE so that the imaged pattern can be transferred through the organic planarizing layer. The resistance to the oxygen RIE in most cases is achieved by using a silicon-containing polymer as the resist material. Resists with higher silicon-concentrations generally have increased oxygen RDB resistance. When the etch rate of the silicon-containing resist is at least 20 times slower than that of the organic planarizing layer, reproducible pattern transfer can be achieved (15). Thus, there is a minimal concentration of silicon necessary for the resist to have adequate oxygen RIE resistance. In an oxygen RIE, the top surface of the silicon-containing polymer is oxidized. The oxidized surface was the subject of many investigations, and it was shown that the oxidized top surface was similar in chemical bonding to silica (15,16). After a silica layer forms, the film is etched by sputtering which is a much slower process than the chemical etching of the unprotected organic planarizing layer (17). Despite all their benefits, silicon-containing resists are not used commercially. One problem associated with siloxane based polymers is their low Tg, which leads to dimensional instability of the imaged features (18). Traditionally the Tg of siliconcontaining resists was raised by randomly copolymerizing with a "high Tg" monomer. This method is a compromise in that the dimensional stability and oxygen RIE Reichmanis et al.; Microelectronics Technology ACS Symposium Series; American Chemical Society: Washington, DC, 1995.

19.

GABOR & OBER

Silicon-Containing Block Copolymer Resist Materials 285

Downloaded by UNIV OF CALIFORNIA SAN DIEGO on July 15, 2016 | http://pubs.acs.org Publication Date: May 5, 1995 | doi: 10.1021/bk-1995-0614.ch019

resistance are intermediary between those of the two parent homopolymers. In contrast properly designed block copolymers are able to be both dimensionally stable and have superb oxygen RIE resistance. Another problem with silicon-containing resists is their extreme hydrophobicity which makes aqueous base developability difficult (14). In our research, preliminary evidence points to the possibility of using aqueous base developer for block copolymer resists with high silicon-content (19).

Figure 3. Bilevel resist scheme involves imaging a pattern into a top-coated resist and then transferring the pattern through the planarizing layer. First Use of Silicon-Containing Block Copolymers as Resist Materials. Hartney et al. were the first to report using silicon-containing block copolymers as resist materials (20-22). They demonstrated that two different polymers, which can not be used as a polymer blend because of macroscopic phase separation, can be used as a resist when prepared in a block copolymer configuration. They prepared diblocktriblock mixtures of poly(p-methylstyrene-è-dimethylsiloxane) and poly(pmethylstyrene-fc-dimethylsiloxane-fr-/?-methylstyrene) which they subsequently chloromethylated. Table I shows the chemical structure for this and the other Reichmanis et al.; Microelectronics Technology ACS Symposium Series; American Chemical Society: Washington, DC, 1995.

MICROELECTRONICS TECHNOLOGY

286

polymers discussed in this review. The chloromethylated polymer had a slightly larger polydispersity than the unmodified mixture of poly(p-methylstyrene- bdimethylsiloxane) and poly(p-methylstyrene-fc-dimettiylsiïoxane-^/7-methylstyrene). Two different block ratios were synthesized. Thefirst(CPMS/DMS 1) had a relatively short poly(dimethylsiloxane) [PDMS] block and a silicon-concentration of 4.0 wt %. The second (CPMS/DMS 2) had a longer PDMS block and a siliconconcentration of 15.5 wt %. CPMS/DMS 1 and CPMS/DMS 2 were effective e-beam resists with sensitivities of 2.2 and 0.9 μΰ/αη and contrasts of 1.4 and 1.3 respectively. With CPMS/DMS 2,0.75 urn lines were imaged. The sensitivity of the CPMS/DMS 2 was actually slightly better than that of a chloromethylated poly(p-methylstyrene) homopolymer of similar molecular weight to the block copolymer, In comparison, the sensitivity of a random copolymer consisting of two different monomer units, is usually intermediate to the sensitivity of the two parent polymers. The high sensitivity was surprising since poly(dimethylsiloxane) is a low sensitivity e-beam resist Several factors could be involved in aie high sensitivity of this block copolymer resist. Because the two blocks are each in their own distinct phase, the block with a high sensitivity is not diluted by the block with a lower sensitivity. Brewer had found in 1977 a similar improvement in e-beam sensitivity for styrene-diene block copolymers (23). Hartney et al. proposed, as an explanation for the improved sensitivity, that in the microphase separated structure the PDMS microphase acted as an effective crosslink. While this is probably not true for all block copolymer resists, we believe that for crosslinking resists, a second phase can act to effectively decrease the dose needed if the developer used has a greater affinity for the block of higher sensitivity. Thus, we believe that if a developer with a greater affinity for the PDMS was used, the sensitivity improvement caused by the PDMS "crosslinks" would not occur. Hartney et al. demonstrated that the block copolymers could be mixed with chloromethylated poly(p-methylstyrene) homopolymer to modify the oxygen RIE resistance. The CPMS/DMS 2 had the slowest etch rate, etching 13 times slower than a hard baked novolak. As the percent silicon was decreased, by adding homopolymer, the etch rate increased. In their study, the weight percent silicon needed to be greater than 14 to obtain an etch selectivity ratio higher than 10.

Downloaded by UNIV OF CALIFORNIA SAN DIEGO on July 15, 2016 | http://pubs.acs.org Publication Date: May 5, 1995 | doi: 10.1021/bk-1995-0614.ch019

2

First Use of Silicon-Containing Graft Copolymers as Resist Materials. Bowden et al. in 1987 published their work on graft copolymers having a poly(methyl methacrylate) [PMMA] main-chain with PDMS grafts (24). PMMA is positive tone while PDMS is negative tone when exposed to e-beam. When exposed to e-beam the resist was negative tone. However, the resist had poor contrast and even in the best cases 40% of the resist film was removed. The authors attributed this to the slightly higher sensitivity of PDMS. The resist was negative tone, when exposed to e-beam, even when the PDMS was the minority phase. When exposed to deep UV, the resists were low sensitivity (2700 mJ/cm ) positive tone resists. The oxygen RIE resistance of the graft copolymers was intermediate to that of the parent homopolymers. 2

Block Copolymers with Deep UV Imageable Blocks. In 1987 Crivello received a patent for a novel method of preparing block copolymers containing both PDMS blocks and imageable blocks (25). He prepared several block copolymers having various imageable blocks using a novel free radical initiator which had PDMS units attached. Among the imageable blocks he prepared were poly(r-BOC styrene) which is the basis for the current generation of 248 nm resists and poly(teri-butyl methacrylate) [P(r-BMA)] which is being investigated for use as a 193 nm chemically amplified resist (see Table I). The block copolymers had approximately 10 wt % Si. Crivello was able to image 2 μπι features with these resists, using 250 nm, contact printing. The resists were developed with aqueous base even though the copolymers had hydrophobic PDMS blocks ranging from 1400 to 4500 g/mol. One problem with the PDMS radical initiators that Crivello used, and with the resulting polymers, is that they contain sSi-OR bonds (R is either an organic group or an organic polymer). The Reichmanis et al.; Microelectronics Technology ACS Symposium Series; American Chemical Society: Washington, DC, 1995.

19.

GABOR & OBER

Silicon-Containing Block Copolymer Resist Materials 287

Downloaded by UNIV OF CALIFORNIA SAN DIEGO on July 15, 2016 | http://pubs.acs.org Publication Date: May 5, 1995 | doi: 10.1021/bk-1995-0614.ch019

sSi-OR bond can be hydrolytically cleaved, as will be discussed in the =Si-OR bond stability section. Block Copolymer-Homopolymer Resist Mixtures. In 1989 Allen and MacDonald investigated the oxygen RIE resistance of block copolymer-homopolymer mixtures (26). Before their work, researchers believed it was necessary to have greater than 8 wt % silicon in the resist to obtain selectivities acceptable for reproducible pattern transfer. Allen and MacDonald proposed that the silicon-concentration only needed to be greater than 8 wt % in the top surface of the resist, and that this could be achieved using a surface active, microphase separated block copolymer. They designed and synthesized the triblock copolymer: poly(2,6-dimethyl-l,4phenylene oxide-fe-dimethylsiïoxane-6-2,6-dimethyl-l,4-phenylene oxide) [PXE-fe-PDMS-è-PXE] (see Table I) (27). Because PDMS has a lower surface energy than poly(2,6-dimethyl-1,4-phenylene oxide) [PXE], it segregates to the surface. PXE is known to be miscible with poly (styrene) (28), and the block copolymers they prepared were compatible with poly(styrene) homopolymer. One problem with PXE-è-PDMS-fr-ΡΧΕ, which is otherwise well designed, is the presence of =Si-OR bonds connecting the PXE and PDMS blocks, which results from the coupling chemistry used. As will be discussed in the ^Si-OR bond stability section, this bond is susceptible to hydrolysis, and we therefore believe that the block copolymers can have a short shelf-life. Allen and MacDonald's proposal, that uncompromised oxygen etch resistance could be obtained by using sûicon-containing block copolymer resists, did not follow smoothly from the previously performed research. As discussed in the subsections reviewing the earlier studies of Bowden et al. (24) and of Hartney et al.(21), siliconcontaining surface active block copolymers had been found to have etch resistance intermediate to that of the parent homopolymers. Indeed Allen and MacDonald measured poor etch resistance with their block copolymers, one of which had 4 wt % silicon and one of which had 5 wt % silicon. Thus, the idea turned into two questions: 1.) Why is the etch resistance of the surface active block copolymers not enhanced?' 2.) 'How thick does the top surface need to be, to obtain the desired etch resistance?' They asked the latter question since it was known that the etch rate decreases as the silica-like layer gradually builds up during the oxygen RIE. The thickness of the silicon-containing layer therefore needs to be at least thick enough to allow the protecting silica film to form. There are two ways to increase the thickness of the silicon-containing surface layer. The first is to increase the molecular weight of the PDMS block. The second is to swell the silicon-containing surface layer by adding PDMS homopolymer. Allen and MacDonald were not able to increase the molecular weight of the PDMS block due to limitations with their synthetic method. However, they were able to swell the silicon-containing surface layer by adding PDMS homopolymer of a molecular weight less than that of the PDMS block. They found that as long as r was less than 0.4, where r is the ratio of unbound to bound PDMS, homopolymer could be blended with the block copolymer without macroscopic phase separation occurring. Adding the PDMS homopolymer had a dramatic effect on the oxygen RIE rates of the polymer blends. The etch rate of the triblock which had 5 wt % silicon was only two times slower than the etch rate of poly(styrene). This same block copolymer, when blended with PDMS homopolymer (r=0.2, a total of 6 wt % silicon) was not observed to etch during the RIE treatment. A blend that consisted of a 1:1 mixture of the triblock copolymer and poly(styrene) and PDMS homopolymer (r=0.2, a total of 3 wt % silicon) had an etch rate 80 times slower than poly(styrene) homopolymer. Allen and MacDonald studied the surface composition of their blends using variable angle XPS. Based on the XPS results they estimated that the siliconenriched surface layer should be at least 5 nm thick and preferably greater than 10 nm to obtain good etch resistance. The dramatic increases in etch resistance were maximized by using oxygen RIE conditions that gave optimal selectivity. Furthermore, the block copolymer-homopolymer systems used by Allen and Reichmanis et al.; Microelectronics Technology ACS Symposium Series; American Chemical Society: Washington, DC, 1995.

288

MICROELECTRONICS TECHNOLOGY

Table I. Polymers With Controlled Architecture Studied For Use As Microlithographic Resists Authors

Block or Graft Copolymer Resists Downloaded by UNIV OF CALIFORNIA SAN DIEGO on July 15, 2016 | http://pubs.acs.org Publication Date: May 5, 1995 | doi: 10.1021/bk-1995-0614.ch019

ÇH

A

3

Hartney, Novembre and Bates

CHS

φ

Citation (20-22)

CH C1 2

ÇH

(25)

(B)

3

φ

Crivello

t-BOC ÇH

ÇH

3

CH

(25)

B