Revolutionizing Process Integration

Sub-100 nm design rules require the deliberate integration of components to maintain profiles and CDs.

Image process integration

The goal of image process integration is to make each component of the imaging system work synergistically with other imaging components to produce focus-exposure process windows large enough to be used cost effectively in a manufacturing environment. Previously, the device design, layout, exposure tool, mask and photoresist communities had not interacted to achieve the best possible resolutions. It has not been necessary. Preliminary work out of International SEMATECH has indicated that with 'proper integration of the entire imaging process,' optical lithography may extend to 70 nm.¹ Though different technologies may have large process windows, the trick is to ensure that the windows for all critical features overlap within an exposure level and fit from level to level, Petersen said. For example, alternating phase shift masks (PSM) are appropriate for the gate level and attenuating PSMs for contacts. The problem is that they image differently; how the lens is sampled will effect image placement and quality. The question is, will overlay of the two levels be within overlay tolerance? This is one aspect of image process integration.

Fig. 1. Simulation of contact holes using an 18% attenuated PSM show (l) the intensity of an uncorrected image and (r) the effectiveness of phase edges in suppressing side-lobes. (Source: International SEMATECH)

Sub-100 nm design rules require pulling out all stops and closely scrutinizing all technologies. A wish list may look as follows: from the steppers ­ the best illumination design to improve imaging performance and minimize complexity of the photomask; from layout ­ smart, friendly, flexible designs (like Manhattan-type structures) to be able to use quadrapole illuminators for the best process window or simplify alternating PSM layout; from resist ­ chemistries to free up mask design and detune from lens aberrations; from the photomask ­ clean, defect-free imaging and minimal additional aberrations.

'Think of everything that impacts the image as an aberration,' Petersen said. From lens aberrations and reflections to etch and implant, each can create degradation of the image transfer. At the point of exposure, it is important to know all aberrations, how to control them and minimize their impact. Given available technologies, optimizing the process requires balancing tradeoffs. Shipley, in collaboration with International SEMATECH for example, is exploring new resist designs to relax side lobe constraints for better contact formation.¹ The overall brightness and image quality of a contact is determined by the brightness of a side-lobe. If the side-lobe is too bright, it will print, but if overly suppressed, it will limit image quality of the contact. The limits of tolerable brightness are a resist issue.

Recent advances at Shipley have seen resists progress from zero side-lobe intensity tolerance to side-lobe suppression within the critical focus-exposure process window. Such resist developments can improve process controls further down the line.

Photoresists

Compounding higher resolutions and lower k factors impacting resist development is the trend toward 'customization,' said Murrae Bowden, director of technology, semiconductor photopolymers group at Olin (Norwalk, Conn.). Though resist considerations are not currently included aggressively in the front end of the design process, the unique pattern requirements associated with the various layers of a device have required a fine-tuning of resists to specific applications such as dense-line, isolated-line and contact hole features. The effect is an increase in complexity for resist designers, multiplying the dimension of development efforts.

The idea of having a universal resist applicable to multiple layers is simply not tenable. Today, resist manufacturers need to know how their customers will use the resists and must anticipate the production process a few years out. While cognizant of cost of ownership, photoresist is still very much a technology driven business, Bowden said.

Photomasks

Fig. 2. AFM of scattering bars shown within a dual trench alternating PSM (r) etched to 180 ° and (l) etched deeper to 360° (Source: Photronics)

Early in image process integration, the design engineers interact with the mask makers. Choices made early in the design process have a direct effect on photomask cycle times and cost, said Ken Rygler, executive vice president, DuPont Photomasks Inc. (Round Rock, Texas). In addition to structuring the design to optimize reticle production and capabilities, control of real estate becomes more of an issue with advanced reticles such as PSMs and optical proximity corrections (OPC). For example, complex OPC masks may contain 'outrigger' or 'scatter bars' and serifs attached to corners of poly gate lines to mitigate line end shortening.

Beyond cost and real estate, the complexity of PSM and OPC can impact their manufacturability. For example, a data modification step between the design and photomask manufacturing can result in a pattern that cannot be printed on a reticle, Douglas VanDenBroeke, director of research and development at Photronics (Milpitas, Calif.), said. However, manufacturing solutions are under investigation. One characteristic of alternating PSMs that can impact manufacturing is that, in addition to shifting phase, etching the quartz mask produces a relative difference in intensity to unetched areas. In essence, any etch will decrease the amount of transmitted light. One method of bringing intensities back into balance is with an additional etch of the entire mask (Fig. 2).

Fig. 3. AFM depicts etch depths used to form a four-phase (0, 60, 120 and 180°) alternating PSM. (Source: Photronics)

With data processing, reticle manufacturing and wafer printing being pushed to their limits, design engineers need to be aware of what happens to their data and current limitations in reticle manufacturing. A simple binary mask requires one data file to define the pattern, but with PSMs, each phase included on a mask calls for separate data files (Fig. 3).

Simulations

As technology progresses, modeling-based solutions can be used to estimate the impact of processes and design in compensation schemes, said Roger Caldwell, vice president of Silicon Technology at MicroUnity (Sunnyvale, Calif.). This is particularly applicable to processes that have predictable error, a component of total error. Under certain conditions, etch, for example, has a predictable loading effect, etching isolated features faster or more slowly than dense features. If polymer-dependent, an isolated line may etch less than dense features and if reactant-dependent, may etch more.

The model-based solution essentially reduces the variation in etch by pre-compensating the design to account for the shift, but it assumes reproduceability as part of the process. The image, however, may not change the same way every time. Ultimately, the process window must be increased. This can be accomplished with software such as MaskRigger, which adds sub-resolution features to increase the photo process window, and wafflization fill geometries to increase the etch process window, Caldwell said.

Fig. 4. Model of (l) a phase shifted space and (r) an unshifted space provides an accurate electric field intensity distribution due to scattering effects from the mask. (Source: FINLE Technologies)

Specialized software packages are needed, as more complex mask topographies are used for PSM and subwavelength features for OPC. Reflections off the etched sidewalls of PSMs and diffraction effects of OPC features can perturb the wave front and alter the aerial image and therefore must be taken into consideration in mask and lithography process design. In such cases, the true optical behavior of the photomask can no longer be predicted by scaler approximations, which are commonly used today. Instead, programs such as ProMAX from FINLE Technologies (Austin, Texas) solve Maxwell's equations to predict the intensity and phase of the electric field immediately below a photomask (Fig. 4). This data can then be used with other software tools such as PROLITH/3D to estimate the resulting aerial image and ultimately the resist profile to aid in astute design and manufacturing decisions.

The real strength of modeling may be in providing greater understanding of the process. In a joint effort by FINLE and International SEMATECH², for example, alternating phase shift mask technology was shown to be considerably more than etch depth into a quartz mask and wavelength of the light source, typically thought to determine the amount of shift in phase. Using mask topography and lithography simulations, a new model was developed that indicates additional dependence on numerical aperture, partial coherence and reduction ratio of the exposure tool, pitch and duty cycle of the line pattern and reduction ratio of the resist. Previously observed in the lab, this model considers all imaging components and targets tangible controls.

Process sequence

While image process integration looks at the interaction of all imaging elements prior to actual image transfer to the wafer, fine-tuning requires that successive processes in the fabrication sequence also be considered. So in addition to a look forward, with increasing frequency, optimization of a particular process takes into consideration a look at its impact on previous process steps.

With the trend toward increasing ion implant beam currents to address throughput and overall equipment efficiency³, one study out of Texas Instruments (Dallas) and Eaton (Beverly, Mass.) looked at the impact of high beam currents, during high energy implants, on CD control and mask edge integrity⁴. The goal was to optimize the resist pretreatment, beam current and plasma ashing sequence, verifying linewidth, three sigma deviation of linewidths and mask edge definition with SEM. Resist pretreatments were shown to preserve side-wall slope, and beam currents as high as 450 µA did not degrade edge tolerances. Characterizing the resist removal process proved most beneficial with low asher particle contamination, clean wafer surfaces and ash time savings of 62% to 75%.

The refinement to process integration is extending to process control as well. For example, KLA-Tencor (San Jose, Calif.) combines inspection, metrology and analysis tools to provide an integrated approach to optimizing reticle, lithography and etch process modules. CD, overlay and defects are tracked throughout the process, with the data correlated to yield and performance results to improve process control. With sky-rocketing costs, taking a new slant on process integration is becoming a necessity.

References