Mark P. Davidson

High-resolution optical overlay metrology

Optical methods are often thought to lose their effectiveness as a metrology tool beyond the Rayleigh criterion. However, using advanced modeling methods, the conventional resolution limitations encountered in well-defined edge-to-edge measurements using edge thresholds do not apply. In fact, in this paper we present evidence that optics can be used to image and measure features as small as 10 nm in dimension, well below the imaging wavelength. To understand the limits of optical methods we have extensively studied both linewidth and overlay metrology applications. Although overlay applications are usually thought to involve pitch or centerline measurements of features from different process levels, some target designs present optical proximity effects which pose a significant challenge. Likewise, line width measurements require determination of the physical edges and geometry which created that profile. Both types of measurements require model-based analysis to accurately evaluate the data and images. In this paper we explore methods to optimize target geometry, optical configurations, structured illumination, and analysis algorithms with applications in both critical dimension and overlay metrology.

Investigation of the effects of charging in SEM-based CD metrology

Scanning electron microscopes are considered the most likely tool for future CD metrology down to 0.1 micron linewidths and below. Charging effects on insulating materials are a long standing problem for electron microscopes. The shrinking design rules are making the measurement errors caused by charging more significant. In this paper a model is proposed which incorporates charging effects into a Monte Carlo simulation model. The model stems from the notion of beam induced conductivity, an established phenomenon whereby an insulator becomes conducting for a brief period of time after being hit by a primary electron. The insulator becomes conducting only within the interaction volume of the primary electron. So after multiple scans of the primary beam has occurred, it can be expected that because of the transient beam induced conductivity that the resulting charge distribution will be such a to create an equipotential surface where significant primary beam dose has occurred. This concept is applied to resist by treating the top region of the resist as a negatively charged potentials. The substrate is given a different potential In general different materials can be expected to have different potentials. One important consequence is that the corners of the resist line, if they are sharp, have strong electric fields and they repel the beam electron. We calculate the electrostatic fields given the resist geometry,then we calculate the beam deflection caused by this field, we remap Monte Carlo simulation data to fold in this effect, and finally we compare with some experimental data to see if this charging effect can account for the apparent resolution degradation that occurs at the edges of resist lines with scanning electron microscopes.

High-resolution optical metrology

Recent advances in optical imaging techniques have unveiled new possibilities for optical metrology and optical-based process control measurements of features in the 65 nm node and beyond. In this paper we discuss methods and applications that combine illumination engineering and structured targets which enable sensitivity to nanometer scale changes using optical imaging methods. These methods have been investigated using simulation tools and experimental laboratory apparatus. The simulation results have demonstrated substantial sensitivity to nanometer changes in feature geometry. Similar results have now been observed in the laboratory. In this paper we will show simulation data to motivate the use of low numerical aperture and structured illumination optical configurations. We will also present the basic elements and methods which we are now using in the design of an optical tool specifically designed for these types of measurements. Target configurations which enhance the scattered electromagnetic fields will be shown along with experimental verification of the methodology. The simulation and experimental apparatus is used to explore and optimize target geometry, optical configurations, and illumination structure for applications in both critical dimension and overlay metrology.

Evaluation of New In-Chip and Arrayed Line Overlay Target Designs

Two types of overlay targets have been designed and evaluated for the study of optical overlay metrology. They are in-chip and arrayed overlay targets. In-chip targets are three-bar two-level targets designed to be placed in or near the active device area of a chip. They occupy a small area in the range of 5 μm2 to 15 μm2 and have line widths, which are nominally device dimensions. The close proximity of the line features result in strong proximity effects. We have used two well-established theoretical models to simulate and study the effects of proximity on overlay measurements. In this paper, we also present a comparison of optical overlay results with scanning electron microscope measurements. Arrayed targets have also been designed to improve and enhance the optical signal for small critical dimension features. We have also compared theoretical simulations of arrayed targets to experimental results. In these comparisons we observe a significant variation in the location of the best focus image with respect to the features. The through-focus focus-metric we have implemented in the current work to determine the best focus image shows interesting properties with potential applications for line width metrology and process control. Based on simulation results, the focus-metric is sensitive to changes in line width dimensions on the nanometer scale.

Comparison of measured optical image profiles of silicon lines with two different theoretical models

In this paper, we describe a new method for the separation of tool-induced measurement errors and sample-induced measurement errors. We apply the method to standard overlay target configurations. This method is used to separate the effects of the tool and sample errors in the measured optical intensity profiles and to obtain the best estimate of the correct intensity profile for a given sample geometry. This most accurate profile is then compared to calculated profiles from two different theoretical models. We explain the modeling in some detail when it has not been previously published.

New method to enhance overlay tool performance

New methods to enhance and improve algorithm performance and data analysis are being developed at NIST for overlay measurement applications. Both experimental data and improved theoretical optical scattering models have been used for the study. We have identified error sources that arise from (i) the optical cross talk between neighboring lines on an overlay target (ii) the selection of the window size used in the auto-correlation and (iii) the portion of the intensity profile that is used in the overlay calculation (defined as a truncated profile). Further, we suggest methods to optimally minimize these error sources. We also present a relationship between tool-induced shift (TIS) and the asymmetry in the intensity profile.

Parametric uncertainty in optical image modeling

Optical photomask feature metrology and wafer exposure process simulation both rely on optical image modeling for accurate results. While it is fair to question the accuracies of the available models, model results also depend on several input parameters describing the object and imaging system. Errors in these parameter values can lead to significant errors in the modeled image. These parameters include wavelength, illumination and objective NAs, magnification, focus, etc. for the optical system, and topography, complex index of refraction n and k, etc. for the object. In this paper each input parameter is varied over a range about its nominal value and the corresponding images simulated. Second order parameter interactions are not explored. Using the scenario of the optical measurement of photomask features, these parametric sensitivities are quantified by calculating the apparent change of the measured linewidth for a small change in the relevant parameter. Then, using reasonable values for the estimated uncertainties of these parameters, the parametric linewidth uncertainties can be calculated and combined to give a lower limit to the linewidth measurement uncertainty for those parameter uncertainties.

Accuracy in optical image modeling

Wafer exposure process simulation and optical photomask feature metrology both rely on optical image modeling for accurate results. The best way to gauge the accuracy of an imaging model is to compare the model results with an actual image. Modeling results, however, depend on several input parameters describing the object and imaging system, such as wavelength, illumination and objective NAs, magnification, focus, etc. for the optical system, and topography, complex index of refraction n and k, etc. for the object. Errors in these parameter values can lead to significant differences between the actual image and the modeled image. Because of these parametric uncertainties, one would hope and expect the models to be far more accurate than such a comparison might indicate. An alternative used here is to compare different imaging models with each other. While the parameter nominal values should be chosen to represent real objects and instruments, they will be identical for both models and contribute no uncertainty to the conclusions. Admittedly not a complete or satisfactory answer to the question of image modeling accuracy, such a differential comparison at least places a meaningful number on modeling differences and sets a limit on modeling accuracy.

Calibration strategies for overlay and registration metrology

Critical dimensions in current and next generation devices are driving the need for tighter overlay registration tolerances and improved overlay metrology tool accuracy and repeatability. Tool matching, performance evaluation, and a move towards closed-loop image placement control all place an increase on the importance of improved accuracy and calibration methodology. In response to these industry needs, the National Institute of Standards and Technology (NIST) is introducing a calibrated overlay wafer standard. There are, however, a number of calibration requirements, which must be addressed when using these standards. These include identification of the best methods for evaluating uncertainties when using traceable, calibration artifacts, proper data acquisition and analysis, and the best calibration strategy.

Optical linewidth models: then and now

In the late 1970s Dr. Diane Nyyssonen demonstrated that the NIST could optically calibrate photomask linewidth standards that were narrower than the classical resolution limit of a conventional bright-field microscope. She equated the unknown edge position on the observed image profile to the known edge position on a theoretically calculated image profile of that line. Since at that time, there was no other way to accurately identify the position of the geometrical edge of micrometer-sized lines on their observed optical images, NIST would not have been able to issue accurate photomask linewidth standards without her theoretical model. NIST has initiated a program to re-examine Nyyssonens model to see how well it meets todays requirements for linewidth standards. Fortunately, one of the authors conferred with Dr. Diana Nyyssonen about her model before her untimely death, and he was able to improve the utility and accuracy of her model. He removed some of her assumptions and improved the efficiency of computations to the point where they could be done on a desktop computer. This paper will detail the result of the initial comparison of the Nyyssonen and Davidson models as applied to photomasks and will identify any significance of the differences as applied to the calibration of NIST photomask standards.