cover story

Phase Shifting and OPC Address Subwavelength Challenges Phase shifting and optical proximity correction can help designers to ensure that current manufacturing processes will render deep- submicron and nanometer designs according to specifications. by Linard Karklin

As IC feature sizes drop below 0.25 µm into the subwavelength range, IC designers and the manufacturers that support them face a new challenge in meeting IC performance and cost objectives. IC feature sizes are now smaller than the wavelength of light used in optical lithography equipment, yet no currently available equipment solutions can shrink the processing wavelength. Without significant changes in the methods used to design and manufacture advanced ICs, the move towards even smaller IC feature sizes and higher IC performance stands in severe jeopardy.

Fortunately, two emerging software technologies will begin to enable designers to achieve improved performance in the subwavelength design realm. Phase shifting enables significantly smaller geometries; optical proximity correction (OPC) fixes subwavelength distortions. Combined, these technologies offer enhanced performance levels, operating on the design layout to create photo masks with nonprinting features that extend the optical resolution and correct for subwavelength effects. Phase shifting and OPC also make it possible to utilize and extend the life of existing process technologies--an attractive alternative to relying on higher-risk and very costly process technology advances that may never come to fruition.

Designers can achieve the industry's smallest feature sizes and avoid the perils of subwavelength phenomena by working with manufacturers that support phase-shifting solutions, and by using tools that make phase- shifting conversions and optical proximity corrections to the design, feeding accurate final design images back to the simulation tools.

The subwavelength design gap

Since the early 1980s, technologists have predicted that conventional optical lithography techniques would become obsolete when IC feature dimensions became finer than the wavelength of light used to expose them during processing. Over the past 20 years, stepper, resist, and mask-making technologies have made incremental advances that have reduced exposure wavelength, but this progression hasn't kept pace with the rapid descent of IC feature size (see Figure 1). Consequently, for the first time in chip manufacturing history, we have reached the critical crossover point where IC feature sizes are at, or below, the wavelength of light of the best-in-class optical lithography equipment. In the future, the gap between the two measures will worsen before it begins to improve.

The subwavelength gap poses significant challenges to designers and fabs alike. Optical lithography equipment is at present unable to reliably resolve geometries smaller than wavelengths in the 0.25-µm range. Under these conditions, optical distortions as well as the diffusion and loading effects of subsequent resist and etch processes cause printed line edges to vary--depending upon the density, size, and location of nearby features. These so-called "proximity effects" manifest themselves in the form of unprinted patterns and distorted geometries. For designers, this phenomenon translates into line-width variations and other distortions that can significantly decrease performance--or even worse, cause missing, incomplete, or shorted structures that result in hard failure (see Figure 2). Even if manufacturers absorb these failures, they pass along yield problems in the form of higher product costs and unreliable production delivery schedules. Below 0.25 µm, IC design teams can no longer expect that the manufactured IC will reflect their design intentions.

Semiconductor manufacturers and equipment makers are working diligently on process-based remedies to the subwavelength problem, but such solutions are impractical. Today, the wavelength of a 248-nm stepper is equal in size to the 0.25-µm features it produces. Many herald the evolution of optical lithography from 248 nm to the next generation of optical wavelength, 193 nm, as the last great hope for extending the life of conventional lithographic technology. While this advancement--even if available today--could narrow the subwavelength gap, the progression to feature sizes below 193 nm is imminent, so benefits would be short-lived at best. In reality, 193-nm technology remains in development and requires not just highly advanced new steppers but new photoresist as well. New resist technology alone takes about three to four years to optimize, so 193-nm lithography will not be production-worthy for some time. Even after its introduction, 193-nm equipment will be extremely expensive, costing manufacturers over $1 billion per fab. Other equipment-based alternatives under development--extreme-ultraviolet lithography, direct-write e-beam, x-ray, and ion projection--will most likely remain unavailable for the foreseeable future.

Phase shifting

Fortunately for IC manufacturers and the designers they support, design technology has evolved. One of the two new techniques--phase shifting--improves the resolution that optical lithography can attain, producing smaller, higher-performance IC features. Phase shifting refers to the modulation of projected light at the mask level to improve the depth of focus of the stepper. Shifting the phase of the light utilizes more of the optical spectrum, thus enabling significantly finer geometries. Depending upon the wavelength of the stepper to be used, software tools shift the phase of the geometries within the IC layout for a given critical layer. The tools determine how and what areas of the layer to shift by using complex algorithms that, when designed correctly, can double the resolution of printed geometries, cutting gate length in half.

Phase shifting a design layer involves manipulating the design data to create specialized masks. After layout, the tools manipulate the layout and create a unique layer to be used only during mask generation. This special layer identifies areas of the actual mask quartz that require partial etching, shifting the phase of light by the requisite amount during exposure (see Figure 3). The tools may also generate a second mask to block out the unwanted dark lines that are a by-product of the exposure process for a phase-shifted mask.

Figure 1	The subwavelength design gap
As feature sizes descend below 0.25 µm, they fall well below the wavelength of light that conventional optical lithography equipment can handle, requiring new mask making and processing techniques.

Phase shifting can involve either one or two exposure steps during processing. Single-exposure phase shifting, as the name implies, requires only a single mask, and hence one exposure during processing. This process phase-shifts geometries in intervals of either 180-120-60 degrees or 90-270 degrees. Though single-exposure phase shifting simplifies processing, the mask making for a multiphase-shifted layer costs a great deal. The problem is that the high cost just doesn't justify the results: The single-exposure method fails to improve resolution appreciably.

Double-exposure phase shifting achieves a superior result at significantly lower cost. The approach uses two masks: first a 180-degree phase-shifted mask, then a standard mask that removes unwanted dark lines from the first exposure. The two-exposure mask set is much easier--and thus much cheaper--to manufacture. The approach enables manufacturers to use 248-nm equipment to create 0.10-µm geometries.

Figure 2	Subwavelength distortions
In the subwavelength design realm, designers can no longer be sure that their physical layout (a) will match the final silicon (b). Common subwavelength distortions include: features that don't print (1), distorted line edges (2), rounded corners (3), and silicon "scum" (4) that may cause interference or chip failure.

Today, phase shifting is practical and necessary only for the most critical layers. The complex algorithms that accomplish phase shifting apply more readily to layers with fine, isolatable structures such as transistor gate regions. Moreover, since phase shifting works best under a double-exposure process, designers need to minimize the number of phase-shifted layers, to keep processing complexity and cost within reason. So designers typically perform phase shifting only on the "gate" layer, the regions in which the poly and diffusion layers intersect. By phase shifting the gate layer, designers can effectively halve the gate length, doubling the performance they can normally attain from a standard 0.25-µm process.

Therefore, although phase shifting is a viable alternative for solving small feature size processing limitations, the use of this strategy is limited by cost and complexity. In addition, not all layers are suitable candidates for phase shifting.

The silicon eye chart

For all of its strengths, phase shifting works effectively only in conjunction with another more common class of corrective techniques, OPC. Despite the fact that it stands for "optical proximity correction," OPC today actually resolves distortions that result not only from optical proximity, but also from the diffusion and loading effects of resist and etch processing.

OPC works by adding features to layouts at the mask level to minimize the effects of optical and process distortions (see Figure 4). Such corrections range from enhancing the outside corners of the design by adding tiny squares, to trimming the inside corners of geometries to prevent excessive rounding or line shortening. OPC corrections make the final wafer image match the designer's layout and prevent fatal shorts and opens, as well as ensuring the accuracy of extracted data used in simulations.

OPC comes in two flavors--one based on models, the other on layout rules. Model-based OPC uses models of the lithography, development, and etch processes to compute the required corrections for each layout layer. The model-based approach offers meticulously thorough correction, but is extremely computation-intensive and is only as good as the model it uses. An inaccurate model creates incorrect OPC.

Rule-based OPC applies a predetermined correction based upon the local layout geometry. Complex algorithms, combined with parameters reflecting process phenomena, define the compensation of geometries needed on the mask. Once obtained, the parameters--derived from models, experimental results, or a combination of the two--can apply to all features on all layers. Rule-based OPC is more productive from a computational standpoint, but may prove insufficient for aggressive designs such as DRAMs.

An optimum approach to OPC thus uses a hybrid of the two methods--applying model-based correction to the densest areas of the IC such as datapaths or memory, and applying rule-based correction to sections less constrained by area. As SOC design proliferates, an increasing number of ICs will contain areas that require a hybrid approach.

Like phase shifting, deployment of OPC influences the design, mask making, and fabrication stages of the design-to-manufacturing flow. Manufacturers fabricate special test chips to generate OPC models of their process, models used in turn by design automation tools that manipulate design layout and create OPC-corrected mask layers.

Naturally, OPC presents a verification challenge. Current DRC tools check features that are supposed to print--but don't check the nonprinting features that OPC introduces. Today, only the mask inspection software that simulates the end silicon can find failures. Eventually, DRC tools that check OPC features will help design teams take full advantage of OPC correction.

The next move

Combining phase shifting and OPC techniques to correct and enhance a design bridges the subwavelength design gap. OPC alone can correct a design for subwavelength effects, thus assuring higher yield and better prediction of performance via simulation. For high-performance designs that require subwavelength features, phase shifting of gates reduces critical feature size by as much as 50 percent. As phase shifting technology progresses and shifting all layers becomes feasible, full die size reduction will occur. Ultimately, the combination of phase shifting and OPC makes 0.11-µm features available with today's 0.25-µm processes.

Figure 3	Phase shifting
Phase shifting eliminates blurred subwavelength images by shifting the phase of projected light on alternating features (a). The result is dramatically improved resolution, enabling designers to attain geometries half the size, printable under standard process and mask conditions (b).

When determining how best to deploy OPC and phase shifting in a design-to-manufacturing flow, a design team must consider three key factors: designer involvement, selection of manufacturer, and selection of design tools. The extent of designer involvement in the modification of the design layout bears greatly on the other two factors.

The team can certainly opt to leave the entire process of applying phase shifting and OPC to the design up to the manufacturer and mask maker. This scenario confines designer involvement to the selection of a manufacturer capable of accommodating phase-shifting and OPC adjustments to the design. The strategy can simplify the designer's role, but eliminates the possibility of post-design verification to ensure that changes made to the layout don't compromise the functional or physical integrity of the design. Segregation also prohibits the comprehension of performance enhancements, and limits the ability to take advantage of the potential design enhancements that result from the phase shifting and OPC processes.

Alternately, the design team may incorporate phase shifting and OPC into the physical design process prior to handing the design off to the manufacturer. This scenario allows designers to comprehend the implications of phase shifting and OPC adjustments in real time, and to iterate the design if warranted. Critical considerations include the selection of a manufacturer who is "subwavelength-capable" and the appropriate choice of design tools that will effectively make subwavelength adjustments and support post-layout verification.

Figure 4	Optical proximity correction
As illustrated on the layout of a 0.15-µm SRAM cell, OPC tools examine the original layout (a) and make corrections for subwavelength effects on the mask to produce a final geometry (b). Note corrective geometries added to outside and inside corners. The wafer images--uncorrected (c) and corrected (d)--show OPC's effects in silicon.

Regardless of the level of designer involvement in the subwavelength adjustment process itself, it remains critically important to select a manufacturer who supports advanced subwavelength correction techniques. A manufacturer must possess the necessary infrastructure to perform subwavelength mask adjustments and the more involved mask and wafer inspections. Almost as important are the subwavelength methods--rule versus model based OPC and single versus double-exposure phase-shifted masks--a manufacturer supports. Software tool compatibility between the designer's environment and the verification tools used by the manufacturer can ensure the smooth transfer of information and data.

Should the design team elect to perform phase shifting and OPC adjustments to the layout prior to manufacturing hand-off, the selection of design tools becomes another important consideration. Since subwavelength design adjustments are so tightly linked to manufacturing and mask making, the three disciplines should share common or highly compatible tools. The type of tools the manufacturer uses governs the phase-shifting and OPC methods they support and serves as another screening criterion for manufacturers. The highly computation-intensive algorithms require the designers to evaluate tools not only in terms of whether they accommodate advanced methods such as double-exposure phase shifting and model-based OPC, but also in terms of their ability to manage data to avoid compromising designer productivity. The tools should also contain productive links to other tools in the designer's environment as well as to the manufacturer's and the mask-maker's tools.

Today we are entering a new era in the evolution of IC manufacturing technology. The conventional lithographic processes that have enabled the phenomenal descent of IC feature sizes over the past 25 years are approaching the end of their usefulness. In the absence of an immediate replacement for conventional technology, phase shifting and OPC technologies present an exciting proposition to both designers and manufacturers. These techniques offer the powerful capability to halve the feature size, thus doubling the performance attainable from the standard 0.25-µm optical lithography process. Through these design automation techniques, optical lithography regains its vitality, and designers may continue to evolve their designs along the path predicted by Moore's law.

Linard Karklin is chief scientist at Numerical Technologies, Inc. in Santa Clara, Calif. He is an authority on advanced photolithography and simulation, and has held technical and executive positions at Sigma-C GmbH and Silvaco.

To voice an opinion on this or any Integrated System Design article, please email your message to jeff@isdmag.com.

integrated system design  May 1999