Structural effects on the performance of epoxide-based negative-tone molecular resists

As feature sizes of individual transistors continue to decrease on integrated circuits (ICs), issues such as pattern collapse and photoacid blur become increasingly problematic during the lithography step of IC fabrication. Typically, polymers are utilized as the photoresist, the material responsible for defining a pattern to be transferred to the silicon substrate beneath it. Chemically amplified resists (CARs) have traditionally been used due to their high sensitivity compared to non-CAR resists and ease of processing. They come with several drawbacks though, which can potentially be solved by using molecular resists, which are much smaller and have a more well-defined composition than polymeric resists.¹ Several years ago, it was demonstrated that a negative-tone, epoxy-based, crosslinked molecular resist exhibited increased resistance to pattern collapse compared to a noncrosslinked resist, without evidence of swelling as is often seen in polymeric crosslinked resists.² Since then, both positive- and negative-tone crosslinked molecular resists have been developed to exploit these advantages.³ 

There are many chemistries that can be used to design a negative-tone crosslinked resist, but one that utilizes epoxide homopolymerization has been chosen in previous papers due to several benefits. First, epoxide homopolymerization is a mass-persistent process, so outgassing, which can potentially contaminate exposure optics, is less-likely to occur compared to deprotection-based resists, which outgas due to their imaging chemistry. Second, in the absence of chain transfer events, the photoacid is active only for a short time and is quickly consumed by epoxides, which then become the reactive centers as oxonium ions. This reactive center will eventually become diffusion-limited as the crosslinking proceeds, which can help limit crosslinking outside of exposed regions and potentially control blur.⁴ Much of the work on epoxy-based, negative-tone molecular glass resists has concentrated upon resists that were not soluble in 0.26N tetramethylammonium hydroxide (TMAH), since these molecules do not possess an ionizable functional group such as a phenol or carboxylic acid.^5–7 Thus, TPOE-3Ep was designed to contain a single phenol to provide an epoxide-containing negative-tone molecular resist that is soluble in standard 0.26N TMAH.⁸ 

TPOE-3Ep exhibited good 248 nm deep ultraviolet (DUV) and extreme ultraviolet (EUV) sensitivity and contrast in TMAH development, but when patterns were resolved, several problems emerged. The first issue observed was delamination of films during TMAH development which was not observed in methyl isobutyl ketone (MIBK) development, and so the material required an underlayer to eliminate this delamination. At low doses, network formation was likely incomplete, which might have allowed the TMAH to wet and etch the resist/substrate interface.⁹ Any network at low doses would likely have a very low crosslink density which would result in a material with lower mechanical properties than a fully crosslinked feature. Additionally, bridging occurred in the low-dose regions of the resist because the solubility transition for TMAH development occurs at lower doses than in MIBK.

Due to these issues, there is a need to understand how structural changes can alter the patterning behavior of epoxide-based materials so that resists can be designed to successfully pattern under a variety of processing conditions. TMAH solubility is primarily driven by epoxide-phenol reactions that consume the phenol, while development in organic solvent is primarily driven by molecular weight increases as the resist crosslinks. For TMAH development of aqueous base-soluble epoxide resists, ideally the solubility transition for TMAH development would occur at a dose where network formation is nearly complete, thereby reducing the amount of microbridging observed and potentially eliminating delamination and the need for an underlayer. Additionally, since highly crosslinked features are desired, if lines are resolved in TMAH at doses well below where complete network formation occurs, the patterned features will likely not be as mechanically robust as those at higher doses. Thus, it is important for us to identify a ratio of epoxides to phenols that results in similar solubility transitions in both organic solvent and in 0.26N TMAH developers.

Using the tetraphenolic TPOE core, this ratio can be varied, and it is expected that increasing the number of phenols relative to the number of epoxides will increase the shift in dose-to-clear (E₀) away from zero dose and reduce the amount of insoluble material in the low-dose regions of the resist. This is because increasing the number of phenols will likely increase the number of crosslinking events required to insolubilize the material in TMAH. Ideally, the insolubility transition in these materials will occur when network formation is almost complete, which would manifest itself with the TMAH and MIBK DUV contrast curves being similar. The proposed molecules to study are shown in Fig. 1.

The molecules were chosen because data for TPOE-3Ep have already been gathered, and this core allows us to study three different phenol:epoxide ratios: an excess number of phenols (TPOE-1Ep), a balanced ratio of phenols and epoxides (TPOE-2Ep), and a case where there are more epoxides than phenols (TPOE-3Ep). Calixarenes are also a potential candidate to use as cores for this process, but it is extremely difficult to isolate individual functionalities due to the large number of hydroxyl groups on the core, as well as the numerous conformations these molecules can adopt, which further complicates isolation of individual functionalities.

The goal is to design a molecule that has increased TMAH solubility at low extents of conversion, to avoid bridging in between patterned lines. To quickly screen molecules for their TMAH solubility, a calculation of the value of a parameter called log D was performed. Briefly, log D is calculated using the pK_a of the ionizable portions of the resist (phenols in the present compounds) and the pH of the developer solution (approximately 13.3 for 0.26N aqueous TMAH). A full explanation is beyond the scope of this paper, and the full details of this calculation are described elsewhere.¹⁰ A compound with a functional group that can be deprotonated by TMAH (such as a phenol or carboxylic acid) has a log D value of less than 2, and it is considered soluble in 0.26N TMAH. Figure 2 shows the log D values for the product of a single epoxide-phenol crosslink for TPOE-3Ep and TPOE-2Ep. After a single epoxide-phenol crosslinking reaction for TPOE-3Ep, the material is expected to become insoluble based upon this predictive calculation. For TPOE-2Ep, however, after a single epoxide-phenol crosslink, the product of that reaction is still quite soluble in 0.26N TMAH, and it would likely take several crosslinks to render insoluble. Thus, from a design standpoint, TPOE-2Ep and TPOE-1Ep are expected to have much higher TMAH solubility than TPOE-3Ep, which could help reduce bridging between patterned lines.

Beyond the initial motivation for improving the performance of a material analogous to TPOE-3Ep, several additional molecules were investigated to better understand how the number of epoxides and glass transition temperature (T_g) affect the patterning performance of this class of materials. Presented here is the 248 nm DUV performance, T_g values, and 100 keV e-beam patterning of over 10 new epoxide-containing negative-tone molecular resists.

A general synthetic procedure for the resists is represented by the synthesis of TPOE-2Ep in Fig. 3. Unless otherwise noted, all reagents were ordered from either Sigma Aldrich or TCI America and used as received without further purification. Full synthetic details of all compounds are included in the supplementary material.²² Once synthesized, compounds were analyzed via a Bruker 300 MHz ¹H NMR spectrometer, and mass spectra were obtained at the Georgia Tech Mass Spectroscopy Facility.

TPOE: 1,1,2,2-Tetrakis(p-hydroxyphenyl)ethane (TPOE) was synthesized by the acid-catalyzed condensation between phenol (Sigma Aldrich) and glyoxal. Phenol (4M equivalents) was dissolved in acetonitrile, and then glyoxal (40 wt. % in water, 1 eq.) and 2-mercaptopropionic acid (0.05 eq.) were added. 10 ml of concentrated sulfuric acid was then added to the solution with stirring. The reaction vessel was then placed into an oil bath set to 70 °C and stirred for 48 h and then allowed to cool to room temperature. The solution was poured into excess acetone and a precipitate was formed, which was filtered using vacuum filtration. The solid was then washed three times with deionized water and once with acetone until a white solid was obtained, which was then dried in a vacuum oven overnight at 50 °C. ¹H NMR (300 MHz, methanol-d4, δ): 6.99 ppm (d, 8H), 6.65 ppm (d, 8H), 4.55 ppm (s, 2H). Mass spectrometry (electron ionization) m/z: [M]+: 199.

TPOE-2Ep: A multistep synthesis is necessary to obtain partially functionalized resists, as shown in Fig. 3. The first step is introduction of alkenes onto the TPOE core by the reaction of 1 equivalent of TPOE and 2.2 equivalents of allyl bromide in the presence of 4 equivalents of potassium carbonate and a catalytic amount of 18-crown-6 ether in methanol. Then, the remaining phenols are protected by tert-butyl dimethylsilyl groups. The alkenes are then transformed into epoxides using oxone in the presence of sodium bicarbonate and a catalytic amount of acetone. After epoxides were synthesized, the protecting groups were removed using tetrabutylammonium fluoride (1.0M in tetrahydrofuran) to afford TPOE-2Ep. ¹H NMR (300 MHz, CDCl₃) δ (ppm): 6.98 (d, 4H), 6.90 (d, 4H), 6.60 (d, 4H), 6.55 (d, 4H), 4.50 (s, 2H), 3.80 (m, 2H), 3.30 (m, 2H), 2.80 (dd, 2H), 2.60 (dd, 2H).

Resists were dissolved in propylene glycol monomethyl ether acetate (PGMEA) with 5 mol. % (triphenylsulfonium hexafluoroantimonate) TPS-SbF₆ (Midori-Kagaku) to create 3 wt. % solutions. Resist solutions were filtered using 0.2 μM Teflon membrane filters. These solutions were used to spin cast resists onto freshly O₂-plasma cleaned silicon wafers (University Wafer) to produce approximately 45 nm-thick films. The films were subjected to a postapply bake (PAB) of 90 °C for 2 min in ambient atmosphere and then exposed to 248 nm DUV light using an Oriel Instruments 500 W Hg-Xe arc lamp with 248 nm band-pass filter. The films were then subjected to a postexposure bake (PEB) for 1 min and then developed. For solvent development, the exposed resists were developed with MIBK for 30 s, followed by a rinse with isopropyl alcohol. For aqueous base development, the resists were developed with 0.26N TMAH (AZ300) for 30 s, followed by a rinse with deionized water. Resist film thicknesses were measured using an M-2000 Woolam Ellipsometer. A normalized remaining thickness (NRT) curve was generated by plotting the ratio of the film thickness after development/drying step to the thickness obtained after the PAB. Error bars represent one standard deviation of triplicate measurements.

For e-beam patterning, films were patterned using a JEOL JBX-9300FS electron-beam lithography system with a 100 keV acceleration voltage and a 100 pA current. E-beam contrast curves were obtained by measuring film thicknesses using a Tencor P15 profilometer and then plotting ratio of the thickness remaining after development/drying to the post-PAB thickness. The patterned wafers were imaged using a Carl Zeiss Ultra 60 scanning electron microscope (SEM) with a 2 keV accelerating voltage.

Thin films of approximately 45 nm thickness (as measured by ellipsometry) were spin cast onto freshly O₂-plasma cleaned silicon wafers ordered from University Wafer. A M-2000 Woolam Ellipsometer equipped with a heating stage was used to measure the change in film thickness as temperature increased. The film thickness increase was plotted against the temperature, and the inflection point of the curve was determined to be the glass transition temperature.

It has previously been demonstrated that a resist that contains as few as two epoxides is able to resolve high-resolution patterns using e-beam lithography, based upon the 9,9-bis(4-hydroxyphenyl)fluorene (BHPF) core.⁵ In order to test the lower-limit of crosslinking in TMAH-developable, epoxide-based materials, the patterning performance of BHPF-1Ep and trihydroxyphenyl ethane (THPE)-1Ep was investigated. In the contrast curves of BHPF-1Ep in PGMEA development (Fig. 4), very little conversion is seen, based on the low NRT values, and even at a PEB temperature of 150 °C, only 10% of the original film thickness remains. Thus, a compound with only one epoxide and phenol appears to be unable to form an insoluble network.

Surprisingly, even though the log D value of this material was calculated to be 1.83, a film of the material was not soluble in 0.26N TMAH developer (a log D value lower than 2 generally indicates that the resist will be soluble in 0.26N TMAH developer). The material was placed in 0.26N TMAH developer with no PAB to discount the possibility of thermal reactions between epoxides and phenols leading to insoluble films, and the film remained insoluble. The ¹H NMR spectrum of BHPF-1Ep in Fig. 5 confirms that there are both epoxides and phenols present on the molecule. Since BHPF-1Ep did not yield high NRT values in PGMEA and was insoluble in 0.26N TMAH, it was deemed unlikely to successfully pattern, and no patterning data were obtained.

A similar situation is seen in THPE-1Ep contrast curves (Fig. 6) in both PGMEA and 0.26N TMAH development. Although this molecule has three crosslinkable functional groups, it only achieves a maximum of approximately 10% remaining film thickness in each development solvent. Various temperatures, PEB times, and development solvents were explored, but none of them produced enough insoluble material to warrant further investigation of this resist.

Unlike the other 1Ep compounds, TPOE-1Ep showed an increase in NRT with dose as shown in Fig. 7 and appears to have a sufficient number of functional groups to achieve an insoluble network. However, due to the high T_g of TPOE-1Ep (86 °C), an extraordinarily high PEB of 170 °C is required to reach an NRT of 1 in PGMEA development. Unfortunately, the material swells to an NRT value of 2 when developed in 0.26N TMAH. This enormous swelling is likely due to the large number of phenols present on the core, which are likely not being completely consumed through epoxide-phenol reactions. To increase consumption of phenols, the PEB was raised 10° to 180 °C. While the swelling was eliminated, crosslinked material remained on the wafer at doses as low as 0 mJ/cm². At this temperature, there is apparently sufficient thermal energy to promote epoxide-phenol crosslinking in the absence of a photoacid.

In the contrast curves for TPOE-2Ep in Fig. 8, a similar situation is seen where increasing the PEB from 120 to 150 °C causes thermal crosslinking in unexposed regions of the resist film. At a PEB of 120 °C, there is insufficient insoluble material to reach an NRT of 1 in PGMEA development. While increasing the PEB to 150 °C does increase the NRT to 1 in PGMEA development, and thermal crosslinking appears in 0.26N TMAH development at a dose of 0 mJ/cm². This behavior is also likely due to the high T_g of TPOE-2Ep (77 °C).

To obtain high NRT values while also avoiding thermal crosslinking in unexposed regions, there appears to be a need to balance base solubility with glass transition temperature. Literature reports show that for small phenolic molecules, inclusion of a methyl or methoxy group adjacent to the hydroxyl group of a phenol can decrease the melting point and glass transition temperature, possibly due to disruption of hydrogen bonds.^11–13 Inclusion of a methoxy group results in the phenols adding at the ortho position to the ethyl bridge during synthesis, and so only cores with methyl groups were used in this study to limit changes to the structural features of the resists. Incorporating methyl groups at the ortho position in TPOE resists may lead to a reduction in T_g so that the resists can be crosslinked at temperatures much lower than where epoxide-phenol thermal crosslinking occurs. To that end, several new molecules were synthesized based on the tetramethylphenyloxy etane (TMPOE) core and are shown in Fig. 9.

In Fig. 10, the DUV contrast curves of the new methylated TMPOE resists are compared to their TPOE predecessors, which lack a methyl group. As a whole, the new TMPOE materials achieve much lower ultimate NRT values than their TPOE counterparts in both PGMEA and 0.26N TMAH development. The methyl groups can cause a reduction in NRT in one of two possible ways. First, placement of the methyl group adjacent to the phenol likely provides steric hindrance to the phenol such that it is prevented from crosslinking with protonated epoxides, reducing the overall amount of crosslinking in the exposed films.¹⁴ Second, if the phenols cannot crosslink into the network, they will be present in the final crosslinked film, which will likely increase the T_g of the growing network due to hydrogen bonding. Such an increase in T_g can cause the film to become vitrified more quickly at a similar PEB, which limits the diffusion of the growing network, reducing the final NRT value.⁴ 

These results suggest that, to successfully design a TMAH-developable epoxide resist, there are two resist parameters that must be carefully balanced: its base solubility and its T_g. The first is influenced by the number of ionizable phenols versus nonionizable portions of the resist (such as epoxides and other structural features which lack an ionizable phenol). Glass transition temperature is influenced by a variety of factors, but in epoxide materials, it has been seen that many smaller resists tend to have a lower T_g than those based on the TPOE core.¹⁵ 

Figure 11 contains the structures of the molecules synthesized in this paper, along with their T_gs, as measured via spectroscopic ellipsometry. We chose to measure the resists via ellipsometry instead of a traditional differential scanning calorimetry (DSC) for two reasons: (1) the T_g often produces a vanishingly weak signal for small molecules in a DSC and (2) thin-film confinement effects can drastically alter the T_g of a compound, due to substrate–material interactions and the presence of a free surface at the top of a film.¹⁶ For example, it has been shown that for thin films of PMMA, the T_g increases as the film thickness decreases due to the hydrogen bonding interactions between the silicon substrate and the methacrylate-containing polymer.¹⁷ In contrast, poly(α-methylstyrene) was shown to have a lower T_g as its film thickness decreased due to the absence of preferential substrate interactions and the presence of a free surface.¹⁶ 

A general increase in T_g is seen in these resists across all cores as the number of phenols is increased, as shown in Fig. 11. The only molecule to defy this trend is BHPF-1Ep, which has a near-identical T_g (58 °C) to BHPF-2Ep (57 °C).¹⁵ There is only a very weak correlation between molecular weight and T_g, likely because in these molecules, a decrease in molecular weight is often the result of removal of epoxide groups, which introduces phenols, which raise the T_g by increasing the amount of hydrogen bonding in the resist films.^18,19 Inclusion of methyl groups adjacent to the phenol in the TMPOE molecules appears to lead to a very slight reduction of the T_g of the resists, possibly through disruption of hydrogen bonding.¹³ However, in TMPOE-3Ep, an identical T_g was observed compared to TPOE-3Ep, and this is potentially due to competing effects between disruption of hydrogen bonding and restriction of the rotational freedom of the molecules.²⁰ As more phenols are introduced into the molecule, the degree of hydrogen bonding would also increase, so the effect of T_g reduction due to hydrogen-bond disruption is more pronounced in TMPOE-2Ep and TMPOE-1Ep.

TMPOE-3Ep has an identical T_g to TPOE-3Ep, which could possibly be due to opposing effects of restriction of rotational freedom and disruption of hydrogen bonding.¹³ TMPOE-2Ep and TMPOE-1Ep both show a slight decrease in glass transition temperature due to the inclusion of the methyl groups.

By designing a material with both increased base solubility and a lower T_g, it might be possible to design a resist with increased base solubility while increasing network conversion at low doses. Such a molecule is shown in Fig. 12. The T_g of THPE-2Ep is 41.1 °C, which is much lower than the T_g of any of the base-soluble materials reported here. The exception is THP-2Ep, which is structurally identical to THPE-2Ep, but THP-2Ep lacks a methyl group in the center of the resist. THP-2Ep dewet from both unprimed silicon wafers and wafers treated with an underlayer, and such a low T_g has been problematic with resists studied in the past.¹⁵ The 248 nm DUV contrast curve in Fig. 12 shows that the resist achieves a high network conversion (as evidenced by the high NRT in the MIBK curve) at a PEB of 90 °C. The resist achieves an NRT of 1 in 0.26N TMAH at a dose of 7 mJ/cm², while in PGMEA development, it achieves an NRT of 0.8 at the same PEB.

The e-beam contrast curve of THPE-2Ep with MIBK development is shown in Fig. 13(a) for MIBK development and shows that the material reaches a maximum NRT of approximately 0.7 at a dose of 100 μC/cm². Figure 13(b) shows the e-beam patterns of THPE-2Ep when developed in MIBK, and the material appears capable of resolving 30 nm lines. Similar to TPOE-3Ep, THPE-2Ep dewets from bare silicon during TMAH development, and no patterns were initially resolved due to severe delamination on unprimed silicon wafers. However, using an underlayer based upon poly(4-hydroxy styrene) and TPOE-4Ep corrects the adhesion issue, as shown in the SEM images in Fig. 14. Blur in this material appears to be an issue since the patterned lines are larger than their nominally patterned sizes. This is likely due to the low T_g of this material, making it easy for the growing chain ends to diffuse more easily throughout the film in nominally unexposed regions.

In Fig. 15, the effect of the PEB temperature on this material is shown. In Fig. 15(a), the lines were nominally patterened at 30 nm 1:1 (line:space), but the lines resolved larger than this. However, by lowering the PEB temperature 10° to 50 °C, THPE-2Ep resolves correctly sized lines, as shown in Fig. 15(b), but the material requires slightly higher doses in order to resolve patterns. At a PEB of 50 °C, the minimum resolution appears to be 30 nm 1:1 (line:space) patterns, and the main mode of failure below this feature size appears to be bridging. The 30 nm resolution of THPE-2Ep is near the resolution of the previously reported TPOE-3Ep, which achieved a resolution of 26 nm lines using EUV lithography. Additives such as photodecomposable nucleophiles have been shown to reduce bridging in epoxide compounds, so future work will likely focus on formulating THPE-2Ep with these additives to reduce bridging.²¹ 

This study has shed light on the various structural features that affect the crosslinking behavior of 0.26N TMAH-soluble epoxide-based molecular resists. The results have shown that increasing the number of phenols results in an increase in the T_g of the resists to the point that very high PEB temperatures are necessary, which causes crosslinking in both exposed and unexposed regions. Methyl groups were introduced adjacent to the hydroxyl groups of TPOE resists in an attempt to reduce the T_g of these compounds, but only a slight reduction in T_g was observed. In addition, the resulting TMPOE resists achieved a much lower NRT than the previous TPOE resists, likely due to steric hindrance preventing the phenols from being incorporated into the crosslinked network. THPE-2Ep was designed to have a low T_g so that it could achieve high NRT values at modest PEB temperatures. The 100 keV e-beam results show that THPE-2Ep is presently capable of resolving roughly 30 nm lines when developed in both MIBK and 0.26N TMAH. Blurring appears to be minimized when using the slightly lower PEB of 50 °C, but a higher dose is necessary to maintain the same resolution. Future work will involve optimizing the processing and formulation conditions of this new resist to identify its resolution limit and also develop other low-T_g epoxide molecular resists.

The authors would like to gratefully acknowledge Devin Brown for his assistance with obtaining e-beam patterns.