Introduction

Mask design has become increasingly complex, since it is difficult to predict the printing characteristics of a mask exposure tool combination solely from knowledge of the layout geometry. Extensive use of computer programs such as Splat or Image is commonly the primary means for understanding the through-focus behavior for mask features such as parallel lines with phase shift or vias with attenuating phase shift. After mask design and fabrication, the printing characteristics of the mask are studied by printing a matrix covering a range of exposures and focus settings. These experimental test chips are then parted, and key features are measured using a scanning electron microscope; this is a time-consuming procedure.

The Aerial Image Measurement System (AIMS*)¹ was designed to provide a means for rapidly evaluating the exposure and depth-of-focus characteristics of real masks, including chrome, proximity-corrected, phase-shifted, or attenuating phase types, prior to resist validation. This device has been found to provide a new means for examining masks capable of providing useful information in a very short time before resist experiments are undertaken [3-5].

A UV microscope was set up in which the NA of the imaging lens was adjusted so that the image produced possesses the same resolution characteristics as that of a particular stepper. In order to emulate effects introduced by phase shifts or optical proximity, the masks are imaged at the wavelength of use (in the present case, 365 or 248 nm). An aperture was also placed in the illumination system so that coherence could be matched and explored. In addition, multiple off-axis apertures may be used to emulate off-axis illumination such as the Canon or Nikon approach. Initial experiments were carried out in which images were compared with extensive computer simulation and with resist features for a wide variety of cases. Good agreement was found between computer simulations and AIMS images obtained for test masks under a variety of conditions. In most cases it is sufficient to predict resist behavior via a high-contrast resist model. If desired, the image data may be used as a starting point for more extensive resist calculations.

Tool development and description

‚‚‚‚Figure 1 ‚‚‚‚Figure 2

Software package
An extensive software package was developed to provide a user-friendly graphical interface for the AIMS setup, control, data collection, and analysis. The program was written in Microsoft Visual C++** for the Microsoft Windows® operating system. The C++ class structure provides a modular environment for easy program extendibility.

The program software communicates with the AIMS tool through a serial communication port to the Zeiss MCU26 microscope stage, as well as through a custom ISA bus adapter to the Photometrics camera. The serial link controls the stage position, filter changer, and lamp selector. The AIMS program, communicating with the Photometrics camera adapter, quickly transfers image data from the camera to the computer memory.

The program provides easy management of a variety of tool control functions. The camera type, speed, and resolution, the MCU26 controller parameters, and the objective lens calibration file are initialized on start-up of the program to their last saved values. All of these parameters can be changed through program menus. An interactive graphical display shows the current position of the lamp selector, the filter changer, the lens turret, and the dark- and bright-field sliders. Additional routines provide for aperture alignment, mask focus, and mask positioning.

After loading a mask onto the stage, it is often desirable to determine the correlation between mask and stage coordinates. By using common alignment marks on the mask and following a three-point alignment process, mask rotation and magnification correction factors are determined. This permits quick and accurate mask positioning to desired mask or stage coordinates. A KLA inspection machine is commonly used to check a mask for defects. This inspection results in a report or list of potential defect sites. The list is easily imported into the AIMS program and used to move the mask swiftly from site to site to evaluate the printability of these defects.

To assist the data acquisition process, several image-capture parameters and routines are available: a quick image preview and alignment mode, image size, and exposure settings. Since features on a photomask are usually 1-3 µm in size (for 5× masks), it is not necessary to capture and store the entire 100 × 128-µm region of the mask that is optically imaged onto the camera. A subfield is often specified defining the region of interest. Once set up, single-focus images and/or through-focus images may be acquired and stored for further analysis. For documentation purposes, a data header is attached to each image file to record the mask designation, measurement conditions, and user comments.

Image analysis
A variety of image analysis and data display routines are built into the AIMS software program. Upon capture, the image is displayed on the screen using either a linear gray-scale, pseudo-color, or threshold-highlight palette, thus providing a qualitative picture of the data (Figure 3). A contour plot may then be calculated for a series of intensity threshold values (Figure 4). The contour plot provides a quick approximation of how the image would print in a high-contrast resist at specified threshold values.

‚‚‚‚Figure 3 ‚‚‚‚Figure 4

To provide quantitative views of the data, plots showing profile, linewidth versus threshold, linewidth versus defocus, and exposure defocus are available. Profile plots (Figure 5) may be calculated for a vertical or horizontal slice of the image. The mouse may be used to select a single bright or dark feature in the profile plot for further analysis. The linewidth versus defocus plot (Figure 6) is determined by calculating the width of the selected feature for specified intensity thresholds over the range of focus. This analysis method provides a quick means of determining expected resist feature linewidth variations versus defocus. Finally, the exposure defocus plot (Figure 7) is calculated by further specifying the permitted tolerance in feature size and plotting the limits in exposure over the range of focus.

‚‚‚‚Figure 5 ‚‚‚‚Figure 6 ‚‚‚‚Figure 7

These analysis methods are extremely valuable in determining the printability of mask defects. While conventional mask inspection tools may determine defect size, other factors (such as edge wall shape and defect phase) are not easily measured. The AIMS tool measures the aerial imaging performance of the mask directly, combining effects of mask amplitude, phase, and surface topography, a result not obtained by other inspection methods. Further analysis of the image may be performed by importing the data file into an advanced workstation program, for example to perform photoresist development simulations.

Photolithography involves many factors such as the reticle, exposure system, and photoresist process which influence the final result obtained on the substrate. These factors typically have a complex relationship to the final results and require numerous experiments for optimization. The typical time required to perform an experiment capable of assessing the performance of a lithography system can range from several days to several months, depending on the complexity of the experiment. The AIMS technology of rapid measurement and image analysis offers a means of emulating the lithography system by significantly reducing the time and cost required to evaluate lithographic performance.

Applications of AIMS

While a variety of PSM techniques exist (attenuated, alternating, rim, and outrigger), the attenuated and alternating techniques are the most generally studied. In the attenuated PSM, the dark areas of the mask typically transmit the exposure energy (5-10% relative intensity) with a 180° phase relative to the clear areas [6]. The alternating PSM method of Levenson et al. [1] utilizes alternating light regions to transmit exposure energy (100% relative intensity) with a 180° phase relative to a neighboring light region [7]. Additive or removal techniques may be used to create alternating phase structure (among which etched quartz technology is the most prevalent).

Inaccurate control of fabrication parameters for a PSM can have a detrimental impact on lithographic performance. There may be deviations from theory, such as transmission reduction through a phase-shifted opening which is dependent on etch roughness as well as electromagnetic scattering phenomena from the sidewalls of the etched opening. In addition to transmission, phase errors will exist if the correct material depth relative to the refractive index is not obtained. This error may be produced by etch endpoint inaccuracy during quartz removal in a subtractive quartz process, and film property imperfections may be produced by variations in an attenuated film deposition process.

Fabrication processes have been developed to address these deviations from theory in order to optimize lithographic performance. A post-etch treatment (also referred to as an etch-back process), in which an isotropic wet etch moves the etched-quartz sidewalls beneath the bordering chrome film, compensates for the strong impact of electromagnetic scattering [8]. AIMS measurements were used to fine-tune this process on a special feature in a 0.25-µm DRAM cell design [5], as shown in Figure 8. The aerial image measurement in Figure 9 demonstrates the effect of the etch-back process on the amount of optical scattering in the phase-shifted opening. The optical scattering produces an intensity transmission error, as seen on the no-etch case, by its reduced peak intensity. The peak intensity of the phase-shifted opening clearly increased relative to the non-phase-shifted opening as the wet-etch depth increased. Optimum lithographic performance was identified with an approximate etch-back of 1200 Å [5].

‚‚‚‚Figure 8 ‚‚‚‚Figure 9

Phase optimization for alternating and attenuated PSM processes was also established by applying the AIMS. Alternating PSM processes have an inherent asymmetry when the phase is not ideal. This asymmetry provides a method of extraction which relies on a comparison of the size of adjacent features through focus. This difference in size of adjacent openings of opposite phase (CD) can be expressed in terms of a phase error,

CD = A( · defocus) + B(1 - (1 - T)),

where is the phase error and T is the intensity transmission error. The phase error defines the slope of the CD curve. The A and B parameters in Equation (1) depend on the design of the alternating pattern and exposure system NA, wavelength, and partial coherence.

AIMS measurements were used to characterize an actual alternating PSM process. With the AIMS tool configured with = 248 nm, NA = 0.5, and = 0.6, aerial image measurements were taken through focus for a 0.20-µm alternating line/space grating, as shown in Figure 10. The difference in intensity peak width for lines of opposite phase was determined as a function of focus and plotted in Figure 11 at multiple thresholds. A value of phase was then extracted using Equation (1). The phase was found to be 5° from optimum, requiring an etch depth correction.

‚‚‚‚Figure 10 ‚‚‚‚Figure 11

Effects of the phase error observed in the AIMS measurements were also confirmed using stepper exposures on the Micrascan® II from SVG Lithography Systems, Inc. (SVGL). Figures 12(a) and 12(b) respectively show the 0.20-µm pattern in 0.6 µm of APEX-E resist at 0 and 0.25-µm defocus. The defocused condition clearly shows resist openings of differing width, in contrast to the 0.0 defocus condition, in which all resist openings are of uniform width. This is consistent with the aerial image, in which neighboring intensity peaks vary in size for the defocused condition.

‚‚‚‚Figure 12

The phase of an attenuating PSM is particularly difficult to ascertain, since it depends on the real and imaginary parts of the refractive index of the phase-shifting material as well as the depth into the mask substrate. The AIMS tool, however, provides a means of measuring the phase through the comparison of aerial image focus characteristics of a contact and lines [9,10]. Small contact holes receive the greatest improvement of all feature types from attenuating phase shifting, and are therefore the most sensitive to any phase error. The action of a phase error is to shift the optimal focal plane in either a positive or a negative direction. Other structures such as the line/space grating are not as sensitive to phase, producing a significantly smaller change to the feature's focus characteristics. By quantifying the difference in focus at which a maximum size is achieved for a contact and nested lines, the AIMS tool can be used to determine attenuated PSM phase errors. Fabrication parameters can then be optimized. Figure 13 shows the difference in focus for a 0.35-µm contact compared to a 0.35-µm line/space pattern as a function of phase error.

‚‚‚‚Figure 13

The ability of the AIMS tool to identify fabrication errors rapidly makes it ideal for mask qualification, in which good reticles can be separated from bad. This is demonstrated with the DRAM cell feature in Figure 8. Two masks were fabricated with this pattern using the same manufacturing process. Figure 14 shows the CD versus defocus characteristics of the two reticles. A significantly larger phase error exists for the first reticle, as exhibited by the large slope of the CD curve versus the near-zero slope of the second reticle. This conclusion was further supported with exposure-defocus analysis, which showed a 62% improvement in DOF for a 15% variation in exposure for the second reticle compared to the first. In addition, the improved performance was confirmed using stepper exposures on the SVGL Micrascan II.

‚‚‚‚Figure 14

Lithographic performance optimization is required once fabrication parameters are established and mask qualification is complete. The AIMS tool provides a rapid means of optimization for stepper parameters such as numerical aperture and partial coherence [11]. This is demonstrated for the 0.20-µm alternating line/space grating shown in Figure 10. Aerial image measurements were taken through focus with the AIMS configured at = 248 nm, NA = 0.5, and = 0.6 and 0.3. Figure 15 shows the intensity distributions, with the 0.3 partial coherence clearly having improved image contrast and through-focus performance. Quantification of the improvement is established through an exposure-defocus analysis.

‚‚‚‚Figure 15

Summary

Acknowledgments

*AIMS is a trademark of International Business Machines Corporation.

**Visual C++ is a trademark, and Microsoft Windows is a registered trademark, of Microsoft Corporation.

Micrascan is a registered trademark of SVG Lithography Systems, Inc.

References

¹ Patent applied for in 1994.

Received February 9, 1996; accepted for publication October 15, 1996