There are three basic ingredients to all commercial DNQ-novolac photoresists: a phenolic novolac resin, a diazonaphthoquinone (DNQ) type dissolution inhibitor, and an organic casting solvent (see Figure 2).2 The novolac resin provides the physical properties required in the photoresist such as good film forming characteristics, etch resistance, and thermal stability. The DNQ makes it possible to image the photoresist by providing a photochemical route to modifying the dissolution rate of the resist in aqueous alkaline developers (see Figures 3). The organic solvent has properties that provide the ability to spin coat the resist to form uniform, glassy thin films.

 

Novolac resins are soluble in a variety of common organic solvents including cyclohexanone, acetone, ethyl lactate, NMP (1-methyl-2-pyrrolidinone), diglyme (diethyleneglycol dimethyl ether), and PGMEA (propyleneglycol methyl ether acetate). In photoresist applications, the casting solvent must be carefully selected to provide good handling, spinning, and film forming properties. The most commonly used solvent today in DNQ-novolac is PGMEA, due to its relatively benign characteristics with respect to human exposure. Resists are generally formulated with polymer loadings of 15 to 30 percent by weight with respect to the solvent content of the resist solution. The viscosity of the solution can be adjusted by varying the polymer to solvent ratio, thus allowing resists to be formulated for coating a variety of film thicknesses.2 

 

Figure 2- (above) two main components of DNQ-Novolak Resists

 

Figure 3- (Below) Development rate of unexposed and fully exposed mixtures of a DNQ type inhibitor and a novolac resin. This type of plot is commonly referred to as a "Meyerhofer plot."4

 

 

The addition of DNQ inhibitors to novolac resins results in a retardation in the dissolution rate of the mixture as shown in Figure 3, 4 and 5 As the concentration of DNQ is increased in the mixture, the dissolution rate of the mixture is further retarded. Upon exposure, the DNQ-novolac mixture typically exhibits a dissolution rate greater than the pure resin. It is this modulation in dissolution rate by means of exposure to radiation that is responsible for the function of DNQ-novolac resists. Modern DNQ-novolac resists are typically formulated with DNQ loadings of approximately 20% by weight with respect to the weight of novolac resin, and can exhibit changes in dissolution rate in excess of three orders of magnitude between their unexposed and exposed states.2,5 The nature of the interaction between DNQ dissolution inhibitors and novolac resins is complex and not well understood. Recent experimental evidence has shown that hydrogen bonding of the inhibitor to the novolac resin plays an important role in the dissolution inhibition process.6,7 In fact, although a number of models have been proposed to explain the function of DNQ-novolac resists8-12, none of these theories have been able to account for all of the phenomena associated with DNQ-novolac resists.    

 

The solvents used to develop DNQ-novolac photoresists are generally aqueous alkaline solutions. The earliest developers used for DNQ-novolac resists were metal hydroxide solutions, such as dilute KOH or NaOH. Over the years, as the semiconductor industry has become more sensitive to metal contamination, metal containing developers have been replaced by organic non-metal developers such as solutions of tetramethyl ammonium hydroxide (TMAH) in water. Today, the most common developer used in the semiconductor industry is 0.26N TMAH.2      

 

The dissolution of novolac resins in aqueous developers is not a simple polymer dissolution process like PMMA dissolving in an organic solvent such as MIBK. In order for the novolac resin to dissolve into the aqueous solution, hydroxide ions from the solution must first deprotonate some of the phenolic sites on the novolac chain.2,8 In this way, dissolution of novolac into aqueous developers is more similar to an etching process, such as copper dissolving in an acidic solution.       

 

As can be seen from Figure 3, unexposed DNQ-novolac resists have a non-zero dissolution rate. This non-zero rate results in a loss of film thickness in unexposed resist films which is termed the "dark film loss" or "dark erosion." In their completely exposed or "bleached" state, DNQ-novolac resists have dissolution rates that can vary from tens of nanometers per second to upwards of a thousand nanometers per second depending on the exact nature of the resist. The dissolution rate of DNQ-novolac resists depends on a variety of factors including the hydroxide ion concentration of the alkaline developer solution, molecular weight of the novolac resin, DNQ loading, etc.2,13-16 Between the completely unexposed and exposed states, the dissolution rate of DNQ-novolac resists generally increases in a non-linear fashion as a function of exposure dose or remaining PAC concentration in the resist film.2,17-19 There are two common methods used to describe this non-linear dissolution behavior. In the first method, the film thickness remaining after some finite development time is plotted as a function of exposure energy to generate a "characteristic curve" or "contrast curve" for the resist (see Figure 4).2,20 This description of resist performance gives rise to the common term of "resist contrast" (g). The resist contrast can be defined as shown in equation 1 with respect to the terms defined in Figure 4.  

 

Equation 1:  Lithographic Contast of Photoresist Films.  The g value for a resist is a convenient measure of the imaging resolution of a photoresist, with higher g values corresponding to higher resolution resists.  

 

Figure 4-Example of a resist contrast curve for a positive tone DNQ-novolac type resist. Normalized thickness is calculated as film thickness remaining after development divided by initial film thickness measured before immersion into developer. D0 is termed the "dose-to-clear" for the resist and g is defined as shown in equation 1.

 

With the concept of resist contrast now defined, it is now possible to discuss the importance and impact of resist contrast in the lithographic process. The most common exposure method in use today is projection printing. In projection printing, an image of the mask pattern is projected onto the resist coated wafer.21 Due to distortions introduced by the imaging system, the "aerial image" or intensity distribution at the wafer plane is not a perfect step function representation of the mask (see Figure 5).17,21 It is the job of the photoresist to translate the smoothly varying aerial image projected by the exposure tool back into a crisp binary style relief pattern of the mask features.17 It is clear from Figure 6 that a non-linear response of the dissolution rate of the resist to exposure dose is required to accomplish this imaging task. In fact, for a resist with an infinitely high contrast (i.e. a resist that possesses a step function for a contrast curve), it would be possible to resolve images of a mask as long as there was any finite difference in intensity produced at the wafer plane by the exposure tool. The contrast of a photoresist is determined by a complex set of factors, and discussions of this subject can be found in the literature.2,20 

 

Figure 5 - Comparison of real aerial image produced by projection exposure tools with ideal mask pattern.

 

Figure 6- (below) Illustration of the importance of a non-linear resist response to exposure dose. A resist that displays a linear rate versus dose behavior will reproduce the sinusoidal behavior of the aerial image, while a non-linear resist response can reproduce the desired binary image from the mask in the final resist profiles.

The second method used to describe the non-linear dissolution behavior of a DNQ-novolac photoresist is to plot the dissolution rate of the resist as a function of the fractional amount of PAC that remains in the resist after exposure (see Figure 7).18 This type of plot is often referred to as an "R(m) plot" for a photoresist. A non-linear behavior in the R(m) plot for a resist directly translates into high contrast. In order to further understand the function of DNQ-novolac photoresists, a more detailed understanding of their chemical nature is required.

Figure 7- Example of R(M) plots for two DNQ-novolac photoresists showing the difference between high and low contrast responses. A non-linear response results in a high contrast resist.

 

1. Basic DNQ Novolak Resist Function

 

2.Diazonapthoquinone Dissolution Inhibitors

 

The most common class of dissolution inhibitors used in non-chemically amplified novolac based resists belong the to the diazonaphthoquinone (DNQ) family of compounds. The basic structure of DNQ is shown in Figure 8, along with the resonance structures that are commonly written for this compound.2

Figure 8- Structure of diazonaphthoquinones, showing both the general formula used to represent these compounds and the possible resonance structures.

 

The diazonaphthoquinone chromophores used in all commercial resist formulations bear other functional groups.2 The most common of these substituted DNQs are derivatives of the 4- and 5-DNQ sulfonyl chlorides shown in Figure 9. The reactive sulfonyl chloride is generally reacted with an alcohol on an organic "ballast group" or "backbone." Figure 10 shows examples of the types of ballast groups and backbones that are used in making DNQ inhibitors. An example of a common multi-functional backbone is 1,2,3-trihydroxybenzophenone. Benzophenones, due to their absorbance in the UV wavelength ranges used for resist exposure, have recently been replaced with other more transparent backbones such as the p-cresol trimer shown in Figure 10. An example of a monofunctional phenol that is used as a ballast group is cumylphenol (see Figure 10). In some cases, the DNQ moieties may even be directly attached to the backbone of the novolac resins. These organic backbones and ballast groups serve a variety of purposes including increasing the inhibition strength of the DNQ and increasing the solubility of the DNQ in resist solutions. More detailed studies of the impact and function of the ballast groups in DNQ-type dissolution inhibitors can be found in the literature.2

 

Figure 9- Two precursors to commercial DNQ inhibitors. The chloride groups are generally reacted in a based-catalyzed esterification with an alcohol group on a ballast structure or backbone molecule to make the final inhibitor.

 

Figure 10- Examples of mono-functional ballast groups and multi-functional backbones. These compounds are typically alcohols. Multi-functional compounds may be partially or fully esterified, leading to a wide variety of possible inhibitors.

 

 

 

The DNQ family of compounds exhibit strong absorption bands in the ultraviolet wavelength range, from approximately 300 nm to 450 nm, that are utilized in the imaging of DNQ-novolac photoresists (see Figure 2.11). These absorption bands can be assigned to n-p* (S0-S1) and p-p* (S1-S2) transitions in the DNQ molecule. Due to the conjugated nature of the DNQ system, the absorption properties of DNQs are strongly affected by the nature and location of substituents.22 Figure 2.11 shows examples of the absorption spectra for DNQ-4-sulfonate and DNQ-5-sulfonate inhibitors. It is clear from this data that the absorption bands of these molecules are strongly affected by the position of the sulfonate group, with the DNQ-5-sulfonate compound being better suited to G-line exposure than the DNQ-4-sulfonate due to its strong absorption at this wavelength. An important feature of these DNQ systems is that with increasing exposure to radiation, the absorption bands of these molecules disappear as the compounds are converted to indene-carboxylic acid photoproducts. This decrease in the absorbance of photoproducts of the DNQ photolysis is termed "bleaching." This decrease in absorbance during exposure allows light to propagate to the bottom of thick and strongly absorbing resist films as the exposure proceeds, thus allowing for complete reaction of the DNQ throughput the thickness of a resist film. The choice of alcohol ballast group or backbone to which the DNQ is attached can also have a strong effect on the absorption properties of the compound. If the ballast group or backbone strongly absorbs in the UV wavelength range where the resist will be imaged, then the bleaching effect can be diminished due to the residual absorbance of the ballast or backbone groups.

 

Figure 11- Examples of absorption spectra for two types of DNQ inhibitors.2

 

In addition to the desired conversion of DNQ to an indene carboxylic acid photoproduct, there are a number of side reactions that are also possible. An example of some of the possible side reactions of DNQs that have been presented in the literature are shown in Figures 12 and 13 In general, DNQ-novolac resists are designed such that the dominant photolysis product is the indene carboxylic acid and other side reactions are suppressed. At this point, one may question the source of water that is required for the formation of the indene carboxylic acid in the photoresist. DNQ-novolac resists are in fact slightly hydroscopic, and a low concentration of water is naturally present in resist films processed in moderately humid environments. If DNQ-novolac resists are processed in very dry environments, the slower reaction of the ketene to form the phenol ester with the novolac resin can be promoted. This possibility of ester formation is an important consideration in resist processing. There are also a number of non-conventional resist processing schemes in which these side-reactions are purposely exploited to modify the behavior and imaging performance of DNQ-novolac resists.17

 

Figure 12- Examples of possible side reaction of DNQs. Notice that side reactions are possible with any of the three species (DNQ, ketene, and carboxylic acid) in the desired photoreaction sequence.

3. Novolak resins

Novolac resins are formed by the acid or metal ion catalyzed co-condensation of phenols with formaldehydes (see Figure 13).2 Acid catalysts are the preferred choice for photoresist applications due to the concern of metal contamination in semiconductor manufacturing. The aromatic phenol ring has three positions, two ortho and one para positions, that are activated toward electrophylic substitution by formaldehyde. In general, mixtures of meta- and para-cresol are used instead of phenol for the production of resist grade novolacs.2

 

Figure 13- General reaction scheme for the production of novolac resins

 

The molecular weight of novolacs used for resist production are realtively small. The number average molecular weight (Mn) of resist-grade novolacs is generally between 1000 and 3000, which corresponds to between 8 and 20 repeat units, while the weight average molecular weight (Mw) may be as large as 20,000. From this information, it is clear that the polydispersity of novolac resins used for resist applications can be quite large. The resins have such low molecular weights, that they could be referred to as oligomers or oligomeric mixtures.

 

The actual structure of novolac resins can be quite complex, even though their molecular weight is relatively low. Kamide and Miyakawa have reported that for an eleven unit phenol-formaldehyde resin, there are 35 unbranched isomers and 2842 branched structures possible.24 This large number of possible polymer structures is a result of the number of different bond-types (o,o-, o,o¢ -, p,o-, and p,o¢ -) that can be formed during the novolac synthesis. The type of linkages present in a novolac resin strongly affects the properties of the formulated resist. Hanabata and coworkers identified four factors concerning the structure of the novolac resin as having key influences in the performance of DNQ-novolac resist:25-30

 

·        Molecular weight of the resin

·        Polydispersity of the resin

·        Methylene linkage position (o,p; o¢ ,p; o,o¢ )

·        Ratio of m,p-cresol precursors

 

The influence of novolac resin architecture on the performance of DNQ-novolac resists is described in more detail in the literature.2 The influence of these factors on resist performance have been understood only recently, and fine control over the polymer structure is difficult due to the sensitivity of novolac synthesis to small changes in the experimental conditions. Therefore, control of novolac properties from batch to batch with the accuracy required for photoresist applications has not been generally possible. Most commercial resist suppliers instead resort to blending together a number of different resin batches to produce a resin with relatively constant and reproducible physical properties.2 In fact, one technique used to improve the performance of modern high-performance DNQ-novolac resists relies on the separation and blending of various novolac resins and is termed the "tandem novolac" approach.31 The first step in making a tandem novolac resin is the fractionation of a polydisperse novolac resin into fractions of different molecular weights. The blended resist resin is then made by recombining the low and high molecular weight components, while omitting the intermediate molecular weight fractions (see Figure 14). The result of this process is a novolac resin, and ultimately a resist, that generally has a higher heat resistance and exposure sensitivity than conventional novolac resins.2

 

Figure 14- Gel permeation chromotography traces of a polydisperse novolac resin and a tandem-novolac resin made from this resin by fractionation and remixing.

4. Summary

DNQ-novolac resists have been in use for over 25 years in the semiconductor industry.2,32 The novolac resin in these resists provides the desired physical properties such as etch resistance that are required in the resist, while the DNQ provides a route to image the material. Although there are only two main components to these resists, there a number of factors that can be used to modify the imaging performance of these resists. Choice and location of ballast or backbone group for the DNQ inhibitor strongly affects the absorbance behavior and dissolution inhibition performance of these compounds. The composition and structure of the novolac resin also strongly impacts the performance of a DNQ-novolac resist. The ultimate goal in designing a DNQ-novolac resist is to design a material that displays a strongly non-linear response in dissolution rate versus exposure and provides good film forming characteristics, thermal stability, and high etch resistance.

 

5. References

  1. R.R. Dammel, personal communication, 1998.
  2. R.R. Dammel, Diazonaphthoquinone-based Resists, Bellingham, WA: SPIE Optical Engineering Press, Vol. TT-11, (1993).
  3. J.J.Yu, C.C. Meister, G. Vizvary, C.B. Xu, P. Fallon, Proc. SPIE, 3333, at press, (1998).
  4. D. Meyerhofer, IEEE Trans. Electron Devices, ED-27, pp. 921, (1980).
  5. A. Reiser, H.Y. Shih, T.F. Yeh, J.P. Huang, Angew. Chem. Int. Ed. Engl., 35, 2428, (1996).
  6. C.G. Willson, W. Yeuh, M.J. leeson, T. Steinhausler, C.L. McAdams, R.R. Dammel, J.R. Sounik, M. Aslam, R. Vicari, M.T. Sheehan, Proc. SPIE, 3049, 226, (1997).
  7. C.L. McAdams, Structure Function Correlation Studies of Dissolution Inhibitors for Novolac-Based Photoresists, Masters Thesis, The University of Texas at Austin, (1996).
  8. R.A. Arcus, Proc. SPIE, 631, 124, (1986).
  9. M. Hanabata, A. Furuta, Y. Uemura, Proc. SPIE, 771, 85, (1987).
  10. M. Hanabata, Y.Uetani, A. Furuta, Proc. SPIE, 920, 349, (1988).
  11. T.F. Yeh, H.Y. Shih, A. Reiser, Proc. SPIE, 1672, 204, (1992).
  12. T.F. Yeh, H.Y. Shih, A. Reiser, Macromolecules, 25, 5345, (1992).
  13. P.C. Tsiartas, L.L. Simpson, A. Qin, C.G. Willson, R.R. Allen, V.J. Krukonis, P.M. Gallagher-Wetmore, Proc. SPIE, 2438, 261, (1995).
  14. C.L. Henderson, P.C. Tsiartas, L.L. Simpson, K.D. Clayton, S. Pancholi, A.R. Pawloski, C.G. Willson, Proc. SPIE, 2724, 481, (1996).
  15. W.D. Hinsberg, M.L. Gutierrez, Proc. SPIE, 469, 57, (1984).
  16. J.P. Huang, T.K. Kwei, A. Reiser, Proc. SPIE, 1086, 74, (1989).
  17. C.G. Willson, "Organic Resist Materials," Introduction to Microlithography, Second Edition, Ed. L.F. Thompson, C.G. Willson, and M.J. Bowden, Washington, DC: American Chemical Society, 139, (1994).
  18. F.H. Dill, W.P. Hornberger, P.S. Hauge, J.M. Shaw, IEEE Trans. Electron Devices, ED-22(7), 445, (1975).
  19. C.L. Henderson, P.C. Tsiartas, S.N. Pancholi, S.A. Chowdhury, K.D. Dombrowski, C.G. Willson, R.R. Dammel, Proc. SPIE, 3049, 805, (1997).
  20. C.A. Mack, Microelectronics Manufacturing Technology, 14(1), 36, (1991).
  21. M.J. Bowden, "The Lithographic Process: The Physics," Introduction to Microlithography, Second Edition, Ed. L.F. Thompson, C.G. Willson, and M.J. Bowden, Washington, DC: American Chemical Society, 19, (1994).
  22. C.G. Willson, R. Miller, D. McKean, N. Clecak, T. Tompkins, D. Hofer, Polymer Eng. & Sci., 23(8), 1012, (1983).
  23. A. Reiser, Photoreactive Polymers – The Science and Technology of Resists, New York: J. Wiley and Sons, (1989).
  24. K. Kamide, Y. Miyakawa, Makromol. Chem., 179, 359, (1978).
  25. A. Furuta, M. Hanabata, Y. Uemura, J. Vac. Sci. Technol., B4, 430, (1986).
  26. M. Hanabata, A. Furuta, Y. Uemura, Proc. SPIE, 631, 76, (1986).
  27. M. Hanabata, A. Furuta, Y. Uemura, Proc. SPIE, 771, 85, (1987).
  28. M. Hanabata, Y. Uetani, A. Furuta, Proc. SPIE, 920, 349, (1988).
  29. A. Furuta, M. Hanabata, J. Photopolymer Sci. Tech., 2, 383, (1989).
  30. M. Hanabata, A. Furuta, Proc. SPIE, 1262, 476, (1990).
  31. M. Hanabata, F. Oi, A. Furuta, Proc. SPIE, 1466, 132, (1991).
  32. C.G. Willson, R.A. Dammel, A. Reiser, Proc. SPIE, 3049, 28, (1997).

Contact info:
Clifford L. Henderson
cliff.henderson@chbe.gatech.edu
404 385-0525