Maskless Photolithography May Offer Cost Advantage

Hank Hogan

Conventional lithography achieves sub-100-nm features by using short-wavelength light, immersion technology and optical enhancements such as phase-shift masks. But the dramatically increasing cost of the latter makes alternative, maskless lithographic techniques attractive. Maskless lithography also could lead to cost-effective small-scale manufacturing.

Figure 1. This photoresist can be removed by optical exposure and chemical processing (left) or by direct ablation (right).

Recently, scientists at Laser Zentrum Hannover eV in Hannover, Germany, and at Universität Hannover in Garbsen, Germany, demonstrated an approach to the technology that can achieve 100-nm features with commercial femtosecond lasers.

Diffraction is the limiting factor in conventional lithography, and even with clever enhancement techniques, features can be no smaller than perhaps half the optical wavelength. The researchers in Germany, however, achieved 100-nm features with femtosecond pulses of infrared (~800 nm) radiation.

Figure 2. Holes in a 50-nm gold layer lying atop a 25-nm chromium layer deposited on silicon dioxide are revealed in this scanning electron microscope image. The holes were created by ablating the photoresist with femtosecond laser pulses and then ion-beam etching the exposed metal. The variation in hole size results from laser energy fluctuations.

Because a photoresist has a well-defined threshold for optical exposure, a pulse whose peak fluence is slightly more than that threshold will expose the photoresist only at the central part of the beam. Moreover, femtosecond pulses can trigger a nonlinear response in the photoresist, increasing the discrimination between above- and below-threshold regions. Theoretically, these two mechanisms should allow arbitrarily small feature sizes, but in reality, beam-pointing instabilities and laser-intensity fluctuations impose a practical limit. Nonetheless, the scientists have already demonstrated features down to 100 nm and expect the future improvements in this technology will enable even smaller features.

Of course, lasers can be and often are used to process materials directly, so why bother with the photoresist? Why not use the laser to write the pattern directly on the material?

Figure 3. Grooves in a photoresist deposited on lithium niobate were created by exposing the resist to multiple pulses and subsequent chemical removal. Periodically poled lithium niobate could be created by filling the grooves with aliquid electrolyte and applying a sufficiently strong electric field.

The answer lies in the fact that the melt dynamics of direct processing always result in the formation of a burr. Chemically or ion-beam etching the material under the removed photoresist is a much cleaner process than directly ablating the material with a laser.

There are two approaches to processing a photoresist with a laser: directly ablating it or exposing it and then removing it chemically (Figure 1). The same logic applies here; ablation is a messier process than chemical removal. Unfortunately, the scientists found that the threshold for single-pulse photoexposure was negligibly below that for ablation in both photoresists they investigated.

Figure 4. Using maskless photolithography, the researchers created a microscale reproduction of Pablo Picasso's 1955 drawing of Don Quixote.

Thus, they could ablate the photoresist with a single pulse or expose it to light for subsequent chemical removal with multiple laser pulses. The drawback to using multiple pulses is that it increases processing time. After the photoresist has been patterned, the same pattern-transfer processes that are used in conventional photolithography can be applied.

Using commercial femtosecond Ti:sapphire laser systems from Femtolasers Produktions GmbH of Vienna, Austria, the scientists produced ~100-nm features in metal films coated onto various substrates, including lithium niobate, silicon dioxide and silicon (Figures 2 and 3).

From a practical perspective, the processing speed is a crucial parameter. To maximize speed, the researchers designed an on-the-fly processing system in which the workpiece is in constant motion under the laser, rather than stopped for each pixel. They calculate that fabricating a 500 × 500-pixel array would require ~70 hours with a stop-and-go system. With an on-the-fly system, the same job could be completed in ~25 minutes, and even faster with a higher-repetition-rate laser system.