Metrology Analysis Tool For Photolithography Process Characterization At Advanced Nodes

Continued scaling of integrated circuits to smaller dimensions is still a viable way to increase compute power, achieve higher memory cell density, or reduce power consumption. These days, chip makers are using single-digit nanometer figures or even Angstrom to label their manufacturing technology nodes, which are associated with the size of features patterned during the lithography process.

The accurate measurement or metrology of these small features, either on the photomask or wafer, becomes essential for characterizing the individual patterning steps and monitoring the manufacturing process. Moreover, metrology data is the basis for building computer models that describe the lithographic process and predict the impact of process parameter variations. Those models are crucial for process development and optimization, as well as a pre-requisite to determine mask corrections, so-called optical proximity corrections, which are applied to a given mask design to print on target.

Since the features discussed here are much smaller than the wavelength of visible light, scanning electron microscopy (SEM) is used to visualize the fine structure. Within this method, the intensity distributions of backscattered electrons and secondary electrons are used to construct a grey scale image, as shown in figure 1. Topographic steps or edges of features generate a higher intensity signal and therefore appear as a white line, while unpatterned areas appear dark.

Fig. 1: SEM image.

Typically, a line or space pattern is analyzed by drawing a virtual line (aka gauge) perpendicular to the pattern and then determining the distance between two edges, measuring the critical dimension (CD) of the feature. A similar methodology is applied to contact holes or pillars to determine their diameter.

At the most advanced nodes, non-Manhattan geometries and curvilinear designs are being used to further increase the feature density within a cell or a chip. Characterizing such structures through gauge measurements could be challenging, as indicated in figure 2. By using just a few gauges (figure 2a), important characteristics might be missed (figure 2b); to sample the exact shape of the feature, multiple gauges along the outline of the structure are needed (figure 2c), increasing equipment and measurement time as well as data volume.

Fig. 2: Gauge-based vs. contour-based sampling.

An effective way to extract all information within a SEM image’s field of view (FOV) is using contours (figure 2d). Contours enable a precise characterization of all features on mask and wafer data, including size, position, and orientation. Moreover, contours allow the fitting of corner rounding parameters, or the efficient comparison of images collected under different process conditions.

That is why Synopsys has developed S-Metro, a powerful solution for automated gauge- and image-based metrology data processing and analysis. S-Metro provides high-performance, comprehensive contour extraction capabilities for mask and wafer data, which become an essential component in mask qualification or wafer process characterization, as input for model building and validation. Figure 3 shows an example of an extracted contour (figure 3a – with the original target layout (green line) and the contour of the printed pattern (red line)) and a heatmap indicating the areas with the highest deviations from the target layout (figure 3b).

Fig. 3: SEM image with (a) extracted contours and (b) heatmap analysis.

What is S-Metro?

As a part of the Synopsys mask solutions portfolio, S-Metro processes experimental data from any e-beam metrology equipment to assess and supplement a pool of data which is used to build and validate lithographic process models. The quality of those models in terms of accuracy and predictivity, matching simulation results to experimental results, strongly depends on the quality and consistency of input data.

The process window analysis capabilities on gauge-based data provide valuable insights into the robustness of the lithographic performance, supporting data consistency checks and flyer elimination. Photomask characterization delivers parameters such as corner rounding and bias information – parameters which are used to adjust the mask setup for modeling. Contours extracted from wafer data of critical clips are ideal input for building models, as those provide an efficient sampling of important design space parameters.

Key highlights of S-Metro include:

• Assess gauge-based measurement results to characterize process performance:
  • Automated numerical data import from metrology equipment.
  • Sort, group, and filter data for quick navigation.
  • Enable data cleaning and flyer elimination.
  • Determine individual and overlapping process windows.
  • Assess data consistency through statistical analysis.
• Analyze SEM images and extract contours for building lithographic models:
  • Automated import of SEM images from any source.
  • Fast and accurate alignment of images to layout (GDS, OASIS).
  • Robust, transparent, and customizable contour extraction.
  • Extensive contour analysis capabilities.
  • Automation and parallelization for high volume data processing.

Using S-Metro with photomask metrology data

With the introduction of curvilinear mask features (for instance through Inverse Lithography Technology, ILT) or curvilinear designs, the analysis of contours becomes indispensable to validate the writing process of the mask or qualify a mask for production. From the extracted contours, S-Metro can extract mask processing characteristics such as corner rounding of the absorber, or a mask bias. Those parameters can be applied to a post-OPC mask layout to obtain a more realistic mask layout which then can be used as a more accurate starting point for litho modeling and simulation.

S-Metro comes with a comprehensive set of contour analysis capabilities. Besides edge placement error (EPE) analysis through histograms or heatmaps, S-Metro can perform more complex and automated analyses, for instance on periodic patterns. Contours of individual features within the field of view can be extracted, and the position can be determined.

Users can easily perform a statistical analysis of the results, looking at CD, area, or pitch uniformity and obtain insights into placement accuracy of the array elements.

S-Metro is also able to detect and count mask imperfections. The severity of a defect can easily be checked by rigorous simulation. For the simulation setup, the extracted contour from the mask SEM image is used to construct a topographic photomask. S-Litho rigorous simulation predicts the 3D shape of the mask pattern in the photoresist on wafer.

Using S-Metro with wafer metrology data

SEM image contours are equally suited for characterizing a wafer pattering process, for instance, lithography or etch. The image of a single printed critical device clip can reveal more information on process performance and cover a large portion of the design space than a much larger amount of simple 1D and 2D gauge-based test patterns. This statement not only applies to curvilinear designs but also to Manhattan and all-angle layouts.

Using contours of critical clips in lithography model building has many advantages: Only a few clips are needed for model calibration or validation, which reduces metrology tool time as well as the time to build an accurate model. At the same time, model quality improves since the most critical features are used as a baseline for calibration.

While a conventional, gauge-based calibration methodology requires many measurement points to sample a given layout clip, one extracted contour from a single SEM image can be a lot more efficient for building a model. Figure 4 shows the schematics for a contour-based resist model calibration flow, which is applicable to both compact and rigorous models. Within an optimization loop, resist model parameters are tuned until a good match between the simulated contour and the extracted contour (reference) is achieved. The cost function for the optimization is determined by the gap between the simulated and measured contour, which will be minimized during the optimization process.

Fig. 4: Contour-based calibration flow.

S-Metro is seamlessly integrated into this flow, supporting contour extraction on calibration as well as validation sets, and the contour-to-contour analysis to calculate the cost function for a given set of resist model parameters.

Conclusion

Accurate metrology of features now measured in Angstroms is essential and even more challenging. Curvilinear assist features are already commonly used to increase resolution and improve process robustness. At the most advanced nodes, non-Manhattan geometries and curvilinear designs are being used to further increase the feature density within a cell or a chip. Even for photonics devices, where curvilinear patterns are standard, precise shape characterization and pattern fidelity are essential for optimizing device performance. Characterizing these structures through multiple gauge measurements seems to be a feasible extension of a conventional metrology approach. However, equipment setup of measurement locations becomes significantly more complex, and the risk remains that important patterning properties still might be missed.

S-Metro provides a powerful engine to extract metrology results from SEM images that have already been collected, whether on a photomask or on a wafer, independent of the used equipment. Detailed information on feature dimensions and shapes can be obtained without re-measurement, saving valuable metrology tool and engineering time. S-Metro is using those contours not only for the precise characterization of arbitrarily shaped features, but also to enable a comprehensive comparison of contours, whether that’s between contours obtained on different wafers, different masks, different exposure tools, varying process conditions, or separate fabs. Comparing simulated contours to experimental controls builds the foundations for calibrating high-quality lithography process models. S-Metro enables users to take maximum advantage and benefit from existing metrology data.