Anti-Reflective Coatings: A Story of Interfaces

Peter Singer, Editor-in-Chief -- Semiconductor International, 3/1/1999

The semiconductor manufacturing industry has moved into an era where lithographic patterning can no longer be performed by simply coating the wafer with a layer of photoresist, exposing and developing that photoresist and then etching the desired feature onto the wafer's surface (the photoresist acts as a mask). Instead, it is necessary to deposit an extra layer ­ maybe two or three ­ before the photoresist is spun on. This extra layer is designed to prevent reflection of light that is transmitted through the photo-resist, reflected off the substrate and back into the photoresist, where it can interfere with incoming light and cause the resist to be unevenly exposed.

Called bottom anti-reflective coatings (BARCs), these layers are now used extensively in applications such as gate definition and for opening contacts, where it's important to keep tight control over critical dimensions (CDs). They are not used much at the interconnect level, since the layer of titanium nitride (TiN) typically used to control electromigration and prevent hillocking on aluminum lines has the extra benefit of being a good anti-reflective (AR) coating. That could quickly change, however, as tighter CD control becomes necessary even for interconnect levels.

As the industry transitions to light with shorter wavelengths ­ from 248 to 193 nm and possibly 157 nm ­ the challenges of minimizing reflections increase. 'Every time you go to a shorter wavelength, the reflectivity of the substrate becomes higher, and as a result you have more problems with interference effects that affect your ability to get consistency in your resist pattern,' said Wilbert Van den Hoek of Novellus (San Jose, Calif.). At the same time, advances in optical lithography such as phase-shift masks, off-axis illumination and optical proximity have allowed features to be printed that are smaller than the wavelength of light. This has made the elimination of CD variations caused by reflected light even more critical. 'That's another dynamic that has forced more use of the anti-reflective coatings into the semiconductor processing,' said Van den Hoek.

In the near future, another problem that's emerging is that AR coatings will be required on top of dielectric layers, which are transparent. As you'll soon see, this requires a more complex two-layer structure, where the bottom layer is reflective, and the top layer cancels out reflected light through destructive interference.

How AR coatings work

Fig. 1. Reflectivity is a function of film thickness, refractive index (n) and extinction coefficient (k). (Source: Brewer Science)

Light, of course, can be transmitted, absorbed, reflected or refracted, depending on the optical properties of the materials and the interfaces it encounters. AR coatings are typically organic materials that are spun-on in much the same way as photoresist, or they are inorganic silicon oxynitride (SiON) films deposited by plasma-enhanced chemical vapor deposition (PECVD).

Organic AR films work to prevent light reflection in a much different way than inorganic films. Organic films ­ made by companies such as Brewer Science, Clariant, Hitachi and Tokyo Ohka ­ work by matching the refractive index of the AR layer with that of the resist; if there's no difference in refractive index, then there will not be reflection at the resist-BARC interface. These organic films are also designed to absorb light, so the light that penetrates the AR coating gets absorbed before it reaches the next interface, where it could get reflected again (or, if it is reflected, absorbed before it reaches the photoresist). They are typically deposited in fairly thick layers, on the order of 600-900 Å (Fig. 1).

Inorganic BARC layers, on the other hand, work by destructive interference. 'You accept the fact that light is reflected from the resist-BARC interface, but there's also light reflected from the BARC-substrate interface. You try to make it such that those reflected beams are opposite in phase, so they cancel out,' said Van den Hoek of Novellus. This is done by adjusting three optical properties of the SiON film: refractive index n, extinction coefficient k and the film thickness. Ian Latchford of Applied Materials (Santa Clara, Calif.), said, 'We use a very stable helium-based process, which allows easy tuning of the optical properties, n, k and t, to near-zero reflectivity for any substrate and resist type. We found that helium also offers advantages in particle counts and film shelf life compared to organic films.'

The semiconductor industry began with spin-on organic AR coatings, which first came into widespread use for the 0.35 mm generation. Proponents of the organic approach point out that organic BARCs are advantageous in terms of cost of ownership, refractive index reproducibility, planarization capability (in some applications), rework capability (cost and remove), film thickness tolerance and surface control¹.

On the other hand, proponents from the inorganic camp claim that PECVD techniques enable deposition of very thin films with precise control over film thickness (±1%), allow easy adjustment of film composition and enable complex multi-layer structures to be created. SiON films also have higher etch selectivity to photoresist than organic films. Although inorganic films are generally not removed after the process, proponents say that it's not a problem.

193 nm challenges

Development work on both organic and inorganic BARC systems continues, mostly focused on developing solutions for 193 nm lithography. James Lamb of Brewer Science (Rolla, Mo.) said he doesn't expect the challenges of 193 nm to be that different from those experienced in the 'early days' of 248 nm BARC development. 'A key problem we have now is there's really no mature 193 nm resist,' Lamb said. 'It's hard to get a baseline.' Since the chemistry of the BARC material has to be closely matched with that of the resist, it's difficult to proceed until the industry standardizes on one or two resists. 'We don't see it as being a substantial material problem to produce a 193 nm BARC. We have a large portfolio of chemical platforms, and we know how to modify those for compatibility with resist.'

The challenge for inorganic films is to change the film's composition to obtain the right optical properties, since, at 193 nm, most silicon-based inorganic materials become opaque. This means that the way in which inorganic BARC films are used today, based on the principles of destructive interference, will no longer work. Van den Hoek said the industry will have to make 'dramatic changes in film composition to make anti-reflective coatings work at 193 nm. We'll move from silicon-rich SiON films where 40%+ composition of the film is silicon to films that are much more like oxide.' He said the new composition isn't difficult to perform with PECVD. 'The nice thing about PECVD of silicon oxynitride is that you can vary the composition of silicon, oxygen and nitrogen over a relatively wide range. By adjusting process conditions, you can make a film that is suitable for I-line lithography, 248 nm deep UV or for 193 nm deep UV. You can tune n, k and, of course, thickness over wide ranges to match your requirement.'

Resist interaction problems

Regardless of the type of BARC layer, be it organic or inorganic, an important criterion is that it not chemically interact or 'intermix' with the overlying photoresist. With organic BARCs, the formation of an intermixing layer is usually suppressed either by crosslinking the BARC or by using a polymer that is insoluble in common resist casting solvents.

With inorganic BARCs, there can also be adverse reactions between the film and the photoresist. The culprits are amine groups (NH₂) that are left on the surface of the SiON film after processing. Photoresists, especially deep UV photoresists, can chemically react to these amines. 'The DARC* layer on its own sometimes interacts with deep UV photoresist so that you end up with a 'foot' at the base of the resist,' Applied Materials' Latchford said. 'The deep UV resists are more sensitive to this footing than the older resists.'

Fig. 2. Most BARC organic films are deposited with ~30% to 50% conformality (left) but can also be used to achieve partial planarization, between 75% and 90% (right). (Source: Brewer Science)

One way to prevent this interaction is to deposit a thin oxide on top of the BARC before the resist is deposited. 'We can do an in-situ oxide, so we can actually mitigate that effect,' Latchford said. 'It provides an inert barrier between the resist and the DARC and has no impact on the anti-reflective properties of the DARC layer.'

Another way to avoid the intermixing problem is to use an oxygen treatment after silicon oxynitride deposition to effectively remove those amine groups only from the surface. This was the approach used by Novellus. 'Our PEARL (Plasma Enhanced Anti-Reflective Layer) process includes an in-situ nitrous oxide treatment after the final deposition before we take the wafer out of the reactor,' said Van den Hoek. He added that post-deposition treatments could play a role beyond just removing amines. 'There are some new resists that are not amine sensitive, but they interact with other features in the surface of these films. There are all kinds of plasma treatments and non-plasma treatments that can be done depending on the exact photoresist,' he explained.

Thinner resists, new etch requirements

Another common requirement for both organic and inorganic films is that they have good etch selectivity compared to the resist. If the BARC etches much faster than the resist, more resist is preserved, and the integrity of the dimension is maintained. This is particularly critical for newer resists that need to be deposited in very thin layers to accommodate the more limited depth of focus of short wavelength steppers and in situations where features need to be etched to significantly different depths. 'The BARC open or BARC breakthrough step is very critical in the sense that you must clear it completely, number one, and number two, you must not impact your resist budget too much,' said Ashish Asthana of Lam Research Corp. (Fremont, Calif.).

The advantage here goes to inorganic films. 'Because they're so thin, and also because they're conformal, they're fairly uniform all across the wafer, so you may not need a large overetch to clear it all over the wafer and all feature sizes,' Asthana said.

If etch isn't your problem, but depth of focus is, you may want to consider a planarizing organic BARC system (Fig. 2). 'If you can get good global planarization, you're getting that much depth of focus back from your system,' said Lamb. 'You've got that much margin you can give to wafer warpage and other depth of focus issues. As you go over vertical steps, you lose the advantage of conformal coatings.'

Dual damascene challenges

Fig. 3. To avoid problems associated with use of BARC layers on transparent oxides, a three-layer structure may be required. Note: DARC is a trademark of Applied Materials. (Source: IBM Microelectronics and Infineon)

One major challenge impacting how BARC will be used in the future is the shift to new lithographic techniques, first to 193 nm and shorter wavelengths and eventually to non-optical techniques. The next big challenge is to adapt to the shift to dual-damascene, low-k dielectrics and copper. Here, the biggest obstacle is patterning transparent dielectric layers that may lie over a rough topography of a reflective layer such as copper or polysilicon. 'The problem with the oxide is that it is transparent, so instead of having a reflective surface on one layer like poly or aluminum, it's more like a window. Down below, there are all kinds of reflective surfaces at multiple levels,' explained Latchford.

This makes the conventional method of controlling reflection by destructive interference inadequate. 'It's virtually impossible to design an ARC layer that can magically change phase, because it's sitting over a poly line vs. sitting over a source/drain,' said Van den Hoek. For those applications, people go to what's referred to as universal AR coatings, which are multilayers (Fig. 3). Typically, the multilayer is such that it's at least a two-layer AR coating. The bottom layer is UV opaque to mask the underlying oxide.

'Basically what we do is control the first layer with a higher n and k, so it's actually more reflective than adsorbing,' Latchford said. 'You set it up so that the reflection on the surface comes directly from the first layer you put down, the same way it is with poly and aluminum. The first layer acts as a reflective layer; the second one acts as a phase shift cancellation layer as usual.'

References

1. M. Padmanaban, et. al., 'Chemical and Lithographic Aspects of Organic Deep UV BARCs,' SPIE 23rd Annual Symposium on Microlithography, February 1998.

* Dielectric Anti-Reflective Coating (DARC) is a trademark of Applied Materials.