An advanced technology that uses 13.5nm light, extreme UV (EUV) lithography uses mirrors and masks to pattern features 32nm wide or even smaller onto microchips, allowing the creation of denser and thus faster, more powerful computers. Before this technology can be widely applied, however, the industry must find a way to reduce the defect density of the masks.1 The sources of defects must be mitigated or eliminated at each step of the mask manufacturing process, which requires methods of identifying defects. Our research, with experimental results at visible wavelengths and simulations in the deep UV, suggests that 3D data sets from through-focus scanning optical microscopy (TSOM) could inexpensively provide information about buried defects that are either difficult or impossible to detect using current methods.
Mask defects can be categorized by whether they change the phase or amplitude of reflected EUV light. Amplitude defects are located on top or near the top of the multilayer mask structure. Phase defects, however, are generally formed either by the deposition of multiple layers over substrate defects (particles or pits) or by particles added during the multilayer deposition. Buried defects are harder to detect than surface defects.
Repair and mitigation techniques differ by defect type. Most commercial inspection tools, however, provide no information about amplitude or phase change. Characterization techniques such as electron microscopy, ion microscopy, or atomic force microscopy can provide information on surface topography but are unable to report the size and depth location of a defect underneath multiple layers. Also, electron and ion microscopes provide limited or no information about the core of the buried defects. Methods with unique capabilities, such as actinic inspection tools (AITs), exist but are limited in scope, purely experimental, resource intensive, and time consuming. Therefore, EUV mask development has a critical need for a low-cost, high-throughput technique that can detect a defect’s size, depth, and type. This data is integral to the feedback cycle of defect repair mechanisms.
One such recently developed technique is TSOM (see Figure 1).2–8 Conventionally, a single best-focus image is used for any type of optical analysis; blurry out-of-focus images are not considered optimal for high-precision applications. The TSOM method, however, demonstrates the usefulness of out-of-focus images for dimensional analysis. The concept of this method was first introduced in 2006.2 The set of through-focus optical images forms a 3D optical data set, which the TSOM method uses for 3D shape analysis of nanometer-scale targets.4, 5
Through-focus optical interactions commonly have distinct sensitivity to different variations in the physical dimensions of nanometer-scale targets. A TSOM image facilitates the analysis of these optical interactions. As a result, we used simulations to demonstrate for the first time that TSOM’s through-focus 3D optical data offers excellent performance for detecting patterned defects as small as 5nm.6
TSOM demonstrated the potential to detect defects on EUV masks as small as 15nm and provide information about how they change the amplitude and phase of light. The technique may also be able to determine depth location. Figure 2(a) is a simulation, using optical imaging modeling software, of applying the TSOM method to detect a bump-type phase defect with a 21nm spherical-equivalent volume diameter (SEVD). Simulations indicate that the TSOM method should have sufficient signal strength over experimental noise to detect phase defects as small as 15nm SEVD using 193nm wavelength light. A pit defect exhibits a distinct color reversal in comparison to the bump defect, as shown in Figure 2(b), clearly distinguishing the two.
We substantiated these simulated TSOM results experimentally using a low-cost optical microscope operating at the visible wavelength of 520nm to identify defects on an EUV mask as bump-type. The same defects could not be detected using a scanning electron microscope due to the lack of depth contrast. Shifting to shorter wavelengths such as 193nm will improve the sensitivity of TSOM.
Nondestructively detecting the depth of buried defects (pits or particles) in a multilayer EUV stack is a challenging task, but the simulated TSOM images for such buried particle defects do show sensitivity to the depth of particles. This suggests the possibility of inferring the depth of defects within multilayered masks, which is essential information for defect repair schemes. We are in the process of validating experimentally the simulation results by building a controlled blank EUV mask test structure having particles of known size placed at different known depths in the multilayer structure. This validation will be extremely beneficial for the defect repair process, resulting in major savings in time and cost.
In summary, the TSOM method has the potential to detect phase defects on EUV masks as small as 15nm SEVD, distinguish pits from bumps, and infer the depth of buried particles. The TSOM method is economical because it requires only a 193nm optical microscope (which is inexpensive compared to an AIT), and has high throughput. As a result, TSOM appears to be a good candidate method for detecting defects on EUV lithography masks.