Greg Blackman on the optics operating in the ultraviolet range and the applications they are used for
The short wavelength of UV, which makes it so suitable for certain applications, is the very thing that makes it difficult to work with as far as optics are concerned. Take semiconductor processing: being able to etch ever-smaller structures onto silicon chips requires shorter wavelengths to illuminate the lithographic mask. Advanced logic chips contain structures as small as 45nm, which is right at the boundary of what current lithographic illumination sources (typically excimer lasers operating at 193nm) can produce, because of the physical properties of light. It’s just not possible to print at resolutions much below the wavelength of light, and to fabricate 45nm structures from light at 193nm requires clever mask design that optically sharpens the profile. However, there comes a point where no amount of optical trickery can reduce structure size any further and the only way to progress is to use a shorter wavelength of light. Next-generation lithography is based around extreme ultraviolet (EUV) at 13.5nm to decrease structure sizes further and consequently fit more transistors on a chip (see panel on page 12: extreme ultraviolet).
The challenge with fabricating optics that operate in the ultraviolet range is that everything has to scale accordingly. ‘The acceptable precision for optical components for a relatively demanding application is typically a tenth of a wave, specified at 633nm,’ explains Gregory Fales, senior product line manager at optics manufacturer Edmund Optics. ‘If that same optical component is being used at 250nm instead, then its precision is only one quarter of a wave; therefore, to achieve the same precision in the UV as in the visible, everything has to be produced to a much tighter scale.’ The same applies to coatings. Simple coatings are typically designed to quarter or half wavelengths of light and therefore, in the UV, the coating thickness has to be much more accurately monitored. ‘Small fluctuations in production show up as much greater errors in the UV than they would in the visible,’ he says.
Fales identifies two common application areas for UV optics: firstly, traditional semiconductor processing, which uses deep UV or extreme UV, the optics for which are, generally speaking, very high precision and difficult to make. ‘That market has been well-served for some time and it’s the area that has driven a lot of the development in UV optics,’ he says.
The other area is in the biomedical market for spectroscopy and fluorescence microscopy. In spectroscopy, most organic materials absorb in the UV and, therefore, a lot of material identification can be made using UV light sources and UV optics. ‘In the past, prior to the development of modern UV light sources, it was a lot more difficult to carry out some of this organic material identification through spectroscopy,’ Fales comments. Likewise with fluorescence microscopy, scientists are developing fluorophores and fluorochromes that absorb UV and emit in the visible spectrum for use in various biomedical applications, from drug discovery to flow cytometry.
According to Fales, a typical UV filter, 10 to 15 years ago, would be made from a UV-transmissive coloured glass substrate, coated with a simple metal and dielectric layer to provide a narrow band. These filters would have 10 to 12 per cent transmission, and a lot of biomedical instrumentation that operated in the UV used gratings and reflective components as opposed to filters. With advances in sputtering coating technology, filters are now available with greater than 80 per cent transmission in the UV, with much better blocking and steeper slopes, among other beneficial properties. Instrument manufacturers have been able to replace some of those reflective components with filters, which simplifies the mechanics of the machine, making it smaller and cheaper to build.
One change impacting optics manufacturers has been the introduction of RoHS guidelines in the EU, which forced glass manufacturers to remove lead from their higher index glasses (flint glasses). This made them much less transmissive in the UV and, to an extent, hurt the development of UV applications, says Fales. Other suitable materials transmitting in the UV, however, have been developed, one of which is ZBLAN, a fluorine glass combined with the heavy metals zirconium, barium, lanthanum, aluminium, and sodium. The material has a low transformation temperature and can be moulded into aspheres, which is particularly beneficial for producing the high-precision optics used in biomedical applications that often have to gather light from divergent or weak fluorescent signals.
Fused silica has been the ubiquitous choice for UV optics, because it’s a relatively easy material to process and has high internal transmittance in the near-UV. It can also be formed into an asphere through diamond turning, although this increases the cost of manufacture substantially. Moving into deep UV for semiconductor processing, calcium fluoride is often the material of choice according to Fales. However, it has various thermal characteristics, which, coupled with the precision required for semiconductor applications, make it difficult to process.
There has also been a lot of development in producing low cost, low power UV LEDs and laser diodes for benchtop spectrometers and other biomedical equipment, which are much more suitable for these types of applications than more traditional xenon lamps that give off a lot of heat, or mercury arc lamps that emit many discrete semi-monochromatic wavebands.
Working with silicon
Leaving aside excimer laser semiconductor processing, there are several micromachining-type applications for UV lasers, most of which operate at 266nm or 355nm, including solar cell scribing, link processing on memory circuits and other types of very fine machining operation, typically in silicon or semiconductor materials. Silicon absorbs well at 355nm and so the penetration depth of these UV lasers is short – there is an efficient coupling of the laser energy with the surface of the semiconductor material. The short wavelength also means the beam can be focused to a small spot size allowing fine features to be processed.
‘These applications need to use high-power, either pulsed or continuous lasers, so there’s a premium on the quality of the optics and the thin film coatings,’ explains Trey Turner, vice president of technology and strategy at REO. REO, based in Boulder, Colorado, US, provides high-performance optical products.
‘Producing optics for micromachining laser systems presents a slightly different set of problems than are typically encountered with optics for excimer lasers, where spot sizes are much larger,’ says Turner. With lithography, the illumination pattern on the wafer is very small, but in the laser path, before the light reaches the objective and the imaging system, the spot sizes are relatively large. However, laser systems for micromachining typically operate at several watts of power with relatively small spot sizes and high power densities. The coatings need to have very low absorption to avoid the optics heating up – ‘the thin films need to be high quality, because of the beam power densities passing through the optic,’ he says.
Ion beam sputtering, which deposits high-quality, low-defect thin films, is typically used to coat UV laser optics for 266nm and 355nm applications. The substrate quality also has to be carefully considered. REO uses a superpolishing technique that allows for control of surface micro-roughness and sub-surface damage, which can be a factor in how much UV light the optical substrate absorbs.
REO also supplies Activated Covalent Bonding (ACB) technology, an adhesive-free process for bonding two optical surfaces together, which is particularly advantageous for UV optics, such as cube beam splitters. Typically, adhesives for joining optical components together absorb highly in the UV, so other techniques need to be employed to avoid their use. ‘There’s really no suitable adhesive that’s transmissive in the UV and absorption increases at deeper UV wavelengths,’ notes Turner.
ACB involves activating the surface of the glass in such a way that catalyses a chemical reaction across the interface. Covalent bonds are formed between the two substrates, which are chemically identical to the bonds holding the glass together.
Precision lenses for microlithography have to be qualified to a high degree of accuracy, and most optical manufacturers will use interferometry to measure the surface form. However, a lot of potential errors are introduced during the assembly of objective lenses. Objective lenses used in microlithography systems consist of multiple lenses stacked on top of each other and each lens has to be perfectly centred to the system axis. ‘High-quality lens assemblies are required for microlithography, and the tolerances for centring the lenses are even tighter working in the UV, because of the short wavelength,’ explains Dr Stefan Krey, technical director at Trioptics, a German-based manufacturer of optical test equipment. Small centring errors can result in severe deterioration of image quality, he says, such as by introducing coma errors.
Trioptics’ OptiCentric equipment, based on ultra-precision air bearings, is used to centre lenses to the system’s optical axis at sub-micron resolution. The equipment precisely rotates the lens and inspects the runout of each optical surface, making sure the lens is perfectly centred. OptiCentric also has a feature built into its software for the final inspection of complete assembled objective lenses for centring errors.
Semiconductor processing remains a large user of UV optics, but according to Fales of Edmund Optics, most of the growth in applications will be those operating above 250nm rather than in deep UV, where there’s still a lack of materials and it’s a much more expensive area. Biomedical applications operating in the 250-400nm range for instance, are, from an optics perspective, significantly easier than working below 250nm, where the choice of substrate and coating materials is limited, he says.
Extreme ultraviolet (EUV) radiation at 13.5nm is being used in next-generation lithography systems as the short wavelength has the potential to decrease structure sizes on silicon chips. EUV light is produced via plasma typically by ionising tin – one method of doing this is using a high-power CO2 laser.
The optics used in EUV lithography systems are reflective rather than transmitting, as the short wavelengths are blocked even with very thin materials. Fraunhofer Institute for Material and Beam Technology (IWS) is carrying out research on reflective coatings for EUV and X-ray optics. The institute has been working with Philips Research Laboratories, Lambda Research Optics Europe and ASML to develop an EUV multilayer mirror with low infrared reflectance at 10.6μm, the CO2 wavelength.
Conventional EUV mirrors are coated with molybdenum silicon (Mo/Si) multilayers, which reflect large amounts of infrared radiation at 10.6μm, produced by the plasma in addition to EUV. Fraunhofer IWS, however, has been using a diamond-like carbon material as a replacement for molybdenum, which suppresses IR light. ‘The coating material is transmissive to IR radiation, which passes into the substrate of the mirror, while EUV light is reflected,’ explains Dr Stefan Braun at Fraunhofer IWS.
Cutting out IR to generate a purer EUV laser beam becomes more important with higher beam intensities and in high-throughput machines. ‘Currently, EUV machines are operating at intensities in which the IR radiation reflected has little effect,’ Braun says. ‘However, increasing EUV power generates greater amounts of infrared, which has to be suppressed to avoid damaging the optics.’
Braun comments: ‘These mirrors have shown promise in suppressing IR radiation. Of course, there are different ways to achieve this; it can be done directly at the source, for instance, but DLC/Si multilayers are an effective and promising method of reducing infrared radiation in the beam.’