Lithography Past...

Ruth DeJule, Associate Editor -- Semiconductor International, 12/1/1998

Twenty years ago the end of optical lithography was thought to be approaching as the semiconductor industry looked ahead to 1 µm geometries. X-ray lithography appeared to be the likely successor. In 1978, 3 µm features typically patterned 75 mm wafers with 900 nm overlay and 300 nm critical dimension (CD) control specifications. Chips were a mere 18 mm² and sold at $1000/Mb. This year, as the semiconductor industry reached $147 billion, chips are 15X larger, overlay and CD control specs 12X smaller; prices have dropped 250X to $4/Mb, and optical lithography is being extended an order of magnitude to 100 nm.

1978 was a time of transition from contact aligners to mirror projection aligners, called scanners. First introduced by Perkin-Elmer, scanners made by Cobilt and Canon soon followed. The precursor to steppers used a 1X mask for exact replication of the image onto the wafer. In a single pass, the stage and reticle were scanned under a broad spectrum mercury arc lamp with distinct peaks at 365, 405 and 436 nm (g, h and i lines, respectively). Reflective optics passed all wavelengths. Exposures selectively mixed wavelengths to reduce standing waves and effectively eliminate roughness or ripples on sidewalls. Throughput was respectable at ~80 wph, and though adequate for 256 Kb technology, the limitation was resolution (~1.5 µm) and control of focus and distortion, locally.

Fig. 1. The slowdown in the technical engine may see a corresponding limit in feature size at ~25 n m. (Source: Dataquest)

Today's i-line steppers use refractive optics and requisite narrow band pass filters to ~6 nm. Lenses corrected to 365 nm control incoming light. High and low index specialty glasses are used to design lenses that have a flat band for maintaining focus. Similarly, for state-of-the-art KrF lasers used in 248 nm DUV scanners, the required lens material, fused silica, will shift focus with any change in wavelength. The laser bandwidth is therefore narrowed to a few pico meters. Today's scanners step to each image field and scan through a 5-10 mm slit at the reticle level. Full production throughput of 104 wph with 150 nm resolution on 200 mm wafers can be realized with systems such as ASML's PAS5500/700B. The first production ArF 193 nm, full-field step-and-scan system is also available (ASML's PAS5500/900). Others will follow such as Canon's FPA-5000 AS1.

In 1978, automation was limited. Most wafers were processed by placing a cassette on a loader that used rubber bands to transfer the cassette to a prealigner or a spin chuck, following a loop or track function. Wafer transfer from one process to another occurred openly in Class 1000 cleanrooms. Today, lithography is self contained with litho cells and robotics to pick-and-place among a "cluster" of lithography tools. The wafer is never touched or in contact with contaminating surfaces and is confined to a Class 1 environment.

Where will lithography be in 20 years? Non-optical technologies are very much talked about and appear imminent. However, for CDs below 30 nm, there is likely to be a continued presence of alternative approaches, each vying to reduce COO. Interferometric lithography? Holographic lensing for vertical lithography as we move beyond 2-D planar devices? As it is today, the need to reduce manufacturing costs will dominate choices, likely driving technology along the path of evolution, not revolution (see Figure). Does this mean 20 more years of optical lithography? Dataquest's principal analyst Klaus-Dieter Rinnen, gives it a better than 50:50 chance. The sheer force of human ingenuity may be the key to optical lithography's longevity.