X-Ray Scatterometry for Stacked Memory-Logic Signal Separation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current X-ray scatterometry techniques face challenges in accurately measuring complex, stacked semiconductor structures due to signal contamination from underlying CMOS structures, leading to high root mean squared errors and computational burdens, which hinders the adoption of transmission-based scatterometry methods for characterizing memory structures.
Innovation Solution
The approach involves modeling the scattering response of logic structures in signal space using a small number of weighted basis functions and combining it with the electromagnetic response model of memory structures to optimize parameter values, allowing for decoupled simulations and reduced computational effort.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional X-ray scatterometry techniques are used to measure stacked semiconductor structures, then measurement capability is provided, but signal contamination from underlying CMOS structures causes high root mean squared errors and reduced measurement precision
Solution Approach 1:
The patent segments the scattering signal into distinct components: a first scattering signal from the memory structure and a second scattering signal from the CMOS structure. By separating and independently analyzing these signals, the method eliminates contamination effects and improves measurement precision of the memory structure parameters.
Solution Approach 2:
The patent extracts the second scattering signal from the CMOS structure and removes it from the total measured signal. This extraction process isolates the first scattering signal from the memory structure, enabling accurate measurement without interference from the underlying CMOS structures.
2Measurement precision
If full electromagnetic response modeling is used to account for both memory and logic structures, then measurement accuracy improves, but computational burden and time increase significantly
Solution Approach 1:
The patent segments the modeling process into two independent parts: modeling the memory structure with its electromagnetic response, and modeling the CMOS structure separately with a simplified scattering model. This segmentation allows efficient computation while maintaining accuracy by treating each structure according to its dominant characteristics.
Solution Approach 2:
The patent changes the modeling approach by switching from a full electromagnetic model for both structures to a hybrid approach where the CMOS structure is modeled using ray tracing or geometric optics approximations. This parameter change in modeling methodology reduces computational complexity while preserving measurement accuracy.
3Adaptability or versatility
If complex three-dimensional structures with diverse materials are characterized, then device functionality is achieved, but characterization difficulty increases due to parameter correlation
Solution Approach 1:
The patent segments the characterization task by assigning different modeling approaches to different structural components. The memory structure uses rigorous electromagnetic modeling while the CMOS structure uses simplified geometric modeling. This segmentation reduces parameter correlation and makes characterization of complex three-dimensional structures more manageable.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This method enables more accurate and efficient characterization of memory structures by reducing fitting errors and computational time, improving the precision and speed of scatterometry measurements for complex semiconductor structures.
Implementation Method 1
an x-ray illumination source configured to generate a beam of x-ray illumination light incident on the stacked structure
Implementation Method 2
an image of light scattered from the stacked structure under measurement is detected in response to the incident illumination beam
Data Source
AI summary
Methods and systems for performing measurements of stacked semiconductor structures, e.g., stacked memory and logic structures, based on X-Ray transmission scatterometry measurement data are described herein. In some examples, the scattering response of logic structures is modelled directly in signal space by a mathematical expression including a relatively small number of weighted basis functions. The scattering response of the logic structures and the scattering response of the memory structures determined by an electromagnetic response model are combined, e.g., by summation or convolution. The combined modelled signals are compared to the measured signals at the detector to generate an error signal. The error signal is employed to drive a regression analysis employed to optimize parameter values characterizing the memory structures, values of the weighting coefficients of the signal space model, or both. In other examples, the scattering response of the logic structures is known, and a model is not needed.