This PDF file contains the front matter associated with SPIE Proceedings Volume 12953, including the Title Page, Copyright information, Table of Contents, and Conference Committee information.

The lithographic imaging performance of contact holes is limited by the efficient use of light and contrast fading caused by 3D mask effects. ”Split pupil” exposures have been proposed to mitigate contrast fading for line-space-patterns. We present a simulation study investigating the extendibility of split pupil exposures to dense arrays of contacts on dark field and light field masks using different mask absorber options. Our simulations indicate that the combination of split pupil exposures and low-n/low-k absorbers can offer comfortable imaging performance for arrays of 10 nm square contacts with a pitch of 20 nm on a dark field mask. These results indicate the potential of combining low-n absorbers and split pupil exposure strategies to enable high-NA EUV lithography to reach its ultimate optical resolution limits.

In a previous work, contributors to Critical Dimension (CD) variability intra-field were shown to come from reticle, contrast fading coming from reticle M3D fading and scanner optics fading. In addition, intra-field best focus shifts (BF) were reported. It called for a holistic assessment and control of total CD uniformity (CDU). In this work we expand the experimental validation total CDU in two ways (1) we study the local reticle variations and its translation to wafer variability by adding Aerial Image Measurement System (AIMS) into the metrology loop (2) we investigate if the observed best focus differences have impact on the 3D aspects of the resist: resist profile and local CDU at bottom and top of contact hole were measured by Atomic Force Microscopy (AFM).

High-NA EUV lithography is being prepared for the next stage of volume production of state-of-the-art integrated devices. First wafer exposures on ASML’s EXE5000 are expected early in 2024. Beyond assessment of the benefit of high-NA by simulation, ZEISS AIMS EUV offers the potential to compare the imaging benefit of 0.55NA to the established 0.33NA, through aerial image analysis of dedicated mask patterns. The recently available capability of high-NA imaging on AIMS® EUV was applied to compare options for imec’s logic patterning roadmap, specifically for tip-to-tip structures (T2T). Beyond direct comparison of 0.33 and 0.55 NA, low-n absorber was compared to conventional Ta-based absorber. Moreover, in view of anamorphic imaging at high-NA, T2T pattern orientation was compared, i.e., either along the 4X exposure slit direction or along the 8X scan direction. Lastly pattern tonality, i.e., darkfield versus lightfield, were evaluated side by side. The comparisons were made for selected, yet not optimized, dipole-like sources. Beyond normalized intensity log-slope (NILS) for the line-space part, the through-focus analysis comprised ILS and required bias for shrinking T2T size. The results show that 0.55NA provides clear advantages, but their variation among absorber type, T2T pattern orientation and tonality highlight the potential of preferred combinations. Such are suggested as starting points for further optimization.

Transistor architectures below the 7 nm node are significantly enabled by EUV lithography, with single exposure EUV processes simplifying small pitch patterning processes. However, EUV stochastics are a significant hurdle in meeting aggressive process assumption targets and achieving high yields. In particular, line edge roughness and line width roughness (LER and LWR) at EUV patterned gate have been identified as key limiters of device performance and yield within these nodes. Here, we study the impact of different illumination schemes on gate LER and LWR. We specifically utilize NILS to target LER and LWR reduction, with high NILS observed to primarily reduce high frequency roughness. Post etch, a largely illumination independent reduction in the mid and high frequency regimes is observed. Finally, impact of illumination on long channel gate patterning is assessed and a NILS independent LWR response is observed both post development and etch.

The stochastic effect in contact single patterning is one of the primary challenges in extending into sub-40nm pitch with 0.33NA EUV. EUV stochastic defects induced by EUV photon shot noise are known to strongly correlate to image contrast. Mitigation of Mask3D induced contrast fading is one of the key solutions to enable further shrink, while maintaining sufficient defect-free process latitude. Wavefront and pupil co-optimization is designed to compensate the Mask 3D phase error that leads to contrast fading. For application in HVM, the newly developed Pupil/Mask/Wavefront co-optimization gives the best imaging performance while maintaining the illumination efficiency and decreasing the rms wavefront for the final optimal wavefront to ensure there is no negative impact on the rest of the patterns that are not included in the optimization. In this paper, we investigate how to apply Pupil/Mask/Wavefront co-optimization to improve the image contrast of a sub-40nm pitch contact hole array, including in-resist verification. We will first explain the fundamentals of Mask 3D fading mitigation via phase injection for a 1D feature and how to extend this concept to 2D features. We will compare the effectiveness of new Pupil/Mask/Wavefront co-optimization versus Zernike Z5 or Z6 only phase injection method. Finally, we will show the potential benefit in combination with using a low-n phase shifting mask for which the optimum image contrast is achieved with the co-optimized wavefront, pupil and mask.

The small depth of focus of high-NA EUV systems asks for robust focus metrology and possibly even focus control. Fast optical focus metrology is possible with dedicated focus-sensitive targets that make use of mask-3D effects. It is beneficial to connect this optical focus measurement to the focus behavior of actual device structures. Focus errors of device structures can be determined by measuring Pattern Placement Errors (PPE) with e-beam since a focus error usually lead to a layout-dependent PPE. By using a large field of view SEM we can capture a large variety of pattern layouts in 1 image acquisition. This large pattern variety creates a lot of diversity resulting in a robust “on-device” focus measurement.

Extreme ultraviolet lithography (EUVL) system at 0.33 numerical aperture (NA) has been used for high volume manufacturing of advanced technology node devices. As pattern pitch shrinks, low-k1 imaging becomes necessary but it typically suffers from reduced imaging contrast. In the case of low-k1 imaging, critical dimension (CD) variation on EUV mask enhances on wafer, which is called Mask Error Enhancement Factor (MEEF). CD variation for tighter line and space (L/S) features on the photomask has taken up a large fraction of wafer CD variation. Therefore, it is important to balance improvement of Normalized Image Log Slope (NILS) and MEEF. Understanding the mask contribution to EUVL is one of the important initiatives to further optimize 0.33-NA EUVL employability towards pushing down pitch scaling. The optical properties of the EUV mask absorber, which are the refraction index n and extinction coefficient k, have potential to improve NILS and MEEF. Low-n attenuated phase-shift mask (attPSM) is expected to mitigate mask three dimensional (M3D) effects and to enhance the image contrast compared to traditional Ta-based absorber mask. In this study we experimentally evaluated the MEEF on the drawn feature size, illumination, and mask absorber properties to characterize the impact of these optical factors on MEEF for the appropriate condition using the NXE:3400B EUV scanner. This paper covers two types of mask absorber materials which are Ta-based and low-n attPSM. We experimentally demonstrated that low-n absorber EUVL mask is expected to improve the balance of MEEF at vertical L/S through pitch due to suppression of best focus variation. We also tried to identify the suitable absorber optical properties in terms of MEEF and NILS on both 0.33-NA and 0.55-NA EUVL through rigorous lithographic simulations.

Improving local critical dimension uniformity (LCDU) as well as productivity in the extreme pitch of EUV layers is the most important challenge in DRAM device. In general, automated source optimization (SO) process is limited to investigate all possible candidates for the best LCDU performance, like fixing the amount of mirror in specific sigma and generating uncommon pole shapes. In this paper, we present LCDU improvement method in hexagonal contact array patterns based on controlling the center sigma position in the EUV pupil. We demonstrated that the LCDU was improved when the source points were located on the outer edge side of pupils regardless of normalized image log slope (NILS). This indicated the mirror location in the source is significant for LCDU in the similar NILS conditions. Moreover, it was confirmed that the CD along the critical direction increased as source points moved toward the center of the 3-beam imaging area, thus the changed CD ellipticity could further improve LCDU or cut down ACI defect source. The results showed a similar tendency in both the binary mask (BIN) and phase shift mask (PSM) even to differ type of resist. Overall, we verified the correlation of LCDU and ADI CD ellipticity through the sigma center position control, and we believe this approach will contribute to future research on improving LCDU.

The line edge roughness (LER) of the mask absorber increases as the critical dimension (CD) approaches the mask’s imaging limit. Mask LER has an impact on the corresponding wafer LER, especially in the case of EUV lithography because of the inverse dependence on exposure wavelength. Low-frequency mask LER is transferred directly to the wafer LER while high-frequency mask LER impacts the wafer image by reducing the image log slope (ILS). In this study, we designed a programmed random LER module and fabricated it on a state-of-the-art EUV mask to introduce controlled variations in LER amplitude while maintaining a similar spatial frequency to the reference mask LER. The unbiased 3-σ mask LER was extracted from the 36nm pitch line/space design, showing that programmed LER is transferred to mask LER. The mask was exposed with the ASML NXE:3400B EUV lithography scanner under focus exposure matrix (FEM) conditions. The transfer of mask LER to wafer LER exhibited a similar trend in programmed jog amplitude and step. Wafer LER increases when the exposure conditions deviate from the best dose or best focus, as the mask error enhancement factor (MEEF) increases and is proportional to the LER transfer function. It was observed that programmed LER is filtered through the illumination system, by comparing the power spectral density (PSD) of mask and wafer LER. The programmed LER led to increased wafer defects, specifically, wafer 3-σ unbiased LER above 2.4nm starts to increase wafer defects.

Background: Patterning of very tight pitches suffers from stochastic variations that can impact yield. Different patterning processes with lower stochastic variations are preferred when those lower variations have a quantifiable benefit in terms of device yield or performance. Aim: Here two different process flows, a traditional EUV patterning flow and one involving directed self-assembly (DSA) rectification, will be compared to determine the differences expected in device failure rates, with the failure mechanism being the shorting of a via hole to the wrong feature. Approach: These device failure rates will be based on a rigorous edge placement error (EPE) model taking stochastic variations into account, leading to predictions of device failure and the definition of an overlay process window (OPW): the range of overlay errors that keeps the device failure rate above a minimum specified value. Results: For the patterning of 18 nm pitch line/space patterns contacted with 12 nm wide vias, the EUV process flow produces a 2.5 nm OPW, while the DSA rectification process expands that OPW significantly to 4.0 nm. Conclusions: Using a rigorous EPE modeling approach fed by accurate stochastics measurements, the significant benefits of the DSA rectification process have been quantified.

As technology nodes further shrink, high yield becomes more and more challenging to achieve. Photoresist resolution (see Figure 2) and CD uniformity are two core yield limiters. This can be improved with the use of Lam’s EUV dry photoresist system[1], where the dry develop technique is used to replace wet develop. This process is also less prone to pattern collapse[1]. Another advantage of this new photoresist system consists in its higher dose sensitivity compared to conventional photoresists which leads to higher throughput[1][2]. High resolution process needs to be combined with optical proximity correction (OPC) to enable cutting edge EUV patterning. OPC relies on the capability of a model to predict accurately the behavior of such an innovative process. The purpose of this study is to provide a comprehensive quantification and characterization of the accuracy of an OPC model fitted for this process ADI and AEI (after carbon open). We generated wafers exposed with an OPC calibration reticle, processed with Lam’s dry deposited and dry developed photoresist. The anchor is a horizontal line-space grating at pitch 32nm (Figure 2). We acquired CDSEM images of more than 1800 features ADI (after dry develop) and AEI (after carbon open). A subset of the features were collected through FEM ADI. To better understand the characteristics of the dry photoresist system, we used metrics such as proximity, MEEF, DoF and EL. ADI OPC models were calibrated using ASML Tachyon software. Finally, we evaluated the OPC model accuracy of both Tachyon FEM+ (physically motivated) and Newron (machine-learning) engines, ADI for pitch 32nm BEOL metal layer use case.

It is crucial to predict hard failure in photolithography process to determine design rules and process condition in the product development stage. Accurate prediction of hard failures through simulation have powerful effects such as shortening the product development period and improving mass production yield. Previously, parameters used to determine whether a pattern is expected to fail include NILS (Normalized Image Log-Slope), image contrast, or chemical distribution in the photoresist. However, these methods are almost infeasible because the accuracy becomes low as process condition changes and calibration process of chemical distribution is too complicated. In this paper, a novel method using optical parameters and machine learning is proposed to predict hard failures of ADI (After Development Inspection) patterns, and this methodology was evaluated in the process of applying inorganic photoresist.

With state-of-the-art EUV lithography moving to the 3 and 2 nm nodes, yield control and connected to that particle contamination control are crucial aspect of High-Volume Manufacturing (HVM). While much progress has been made in recent years, the continuously tightening node requirements translate into ever more stringent requirements on particle contamination control. Besides for lithographic scanners, operating in a low background pressure hydrogen gas environment, particle contamination control is also important for space exploration (which also operates in low pressure environments), where particles may lead to malfunctioning moving parts, loss of solar power generation, and human health hazards. A key factor in release of particles in these low-pressure environments is the ionization of the low-pressure background gas by energetic photons (for instance EUV in lithographic scanners, and broadband energetic radiation in space), and resulting plasma with fast electrons of 25 eV and above. Experiments show that these electrons can mobilize and remove particles on most materials, and that the governing effects strongly depend on the substrate material, coating and surface finishing. This paper will discuss work on understanding and modeling these effects and describe possible solution paths to improve particle contamination control, both for lithographic scanners and for space exploration.

EUV lithography is rapidly being pushed to its resolution limit, where tradeoffs are heightened between resolution, throughput, and stochastics. Mitigation strategies include attenuated phase shift masks (aPSMs) and thinner high-k absorbers. Furthermore, multilayer bandwidth and phase shift may need to be reassessed. All these improvements relate to mask 3D effects (M3D), arising from several causes: First, phase shift vs pitch, which sets the aPSM target phase shift around 1.2pi instead of pi. Second, absorber thickness effects which directly relate to the promise of high-k absorbers. And third, multilayer effects like bandwidth and phase vs angle. We propose to quantify these effects in simulation for different EUV scanner generations (0.33 NA, high-NA, or hyper-NA). Finally, we will propose measuring these effects with the EUV Tech ENK (EUV n/k tool) using actinic scatterometry. The complexity of M3D suggests that new metrics of multilayer and absorber performance beyond reflectivity will need to be considered. Actinic scatterometry provides a promising route to measuring M3D due to its sensitivity to wavelength, angle, and feature size.

Lasertec released an actinic patterned mask inspection (APMI) system named ACTIS in 2019 and has since been providing it as an EUV mask inspection solution for use at mask shops and semiconductor fabs in high-volume manufacturing. Actinic inspection is a type of inspection that utilizes 13.5nm EUV, the wavelength of light used in EUV lithography. ACTIS performs high-resolution, high-throughput inspection of EUV masks and detects mask defects that print on wafers in the lithography process (printable defects). One of the key advantages offered by APMI is its ability to perform highresolution through-pellicle inspection. EUV pellicles are expected to be used in semiconductor fabs, and ACTIS can detect printable defects on EUV masks even when a pellicle is attached to them. High-NA EUV lithography will be used at the technology nodes of N2 and beyond. The projection optics of high-NA EUV lithography are anamorphic optics, which have different scales of magnification for the horizontal and vertical axes. Lasertec has developed a next-generation model of ACTIS for high-NA mask inspection. It uses higher NA objective optics to meet the requirements of high-NA EUV mask inspection. This paper discusses the next-generation ACTIS for high-NA EUV lithography and its inspection performance on programmed defect masks that are similar to actual production masks. The paper also discusses the characteristics of a light source and illumination optics required for APMI.

EUV blank is a complicated system including substrate, multilayer, cap layer, absorber, and so on. All those structures contribute to the lithographic performances. In this paper, lithographic impacts of cap layer were simulated. By evaluating lithographic performances for both a binary type mask and a phase shift type mask, impact of cap layer thickness was clarified especially for best focus range of PSM. Second, impacts of cap layer recess induced by mask and lithographic processes were evaluated. Bayesian optimization method was applied to optimize lithographic performances efficiently. The results show thinner cap thickness enables to improve dose aware resolution (NILS × Threshold^0.5) and its stability. On the other hand, thinner cap thickness may cause worse durability for the processes such as etching and repairing. So balance between lithographic performances and film durability should be considered.

EUV stochastic pattern anomalies are a major obstacle in future EUV lithography, and they depend on processes such as development, underlayers, extrinsic anomalies, and resist materials. This paper explores the impact of molecular clustering on EUV stochasticity using the discrete dev/etch model since the above factors can impact stochasticity through mechanical interactions among nearby molecules. Pattern formations with larger/denser molecular sub-clusters (SCs) suppress stochastic pattern anomalies because forming such sub-clusters requires more photons. Stochastic pattern anomalies increase with thinning resist films, however, since the dispersion in the sub-cluster ratio within a small volume of exposed resist film increases as the volume size approaches the cluster size. LWRs and defect probabilities in thin resist films decrease with increasing the sub-cluster density requirements. Underlayers (ULs) and extrinsic anomalies in resist materials influence stochasticity through their interactions with molecular SCs. Interfacial defects are highly dependent on conditions of UL surface and sub-cluster formation. Extrinsic defects and resist fluctuations with a scale longer than the SC size generate defect floors or magnify stochastic defect probabilities, while those shorter than the SC size have little impact. Controlling sub-cluster requirements may help keep LWR and defect probability sufficiently low, but this generally is at the expense of resist sensitivity.

The combination of High NA EUV anamorphic projection optics and unchanged mask-blank size result in a “High NA field” with a maximum size of 26x16.5 mm² at wafer level. Therefore, to create a die larger than the High NA full field, two images are stitched together. So-called in-die stitching is enabled by a combination of design, mask, OPC, process, and scanner solutions. We present an overview of our learnings about at-resolution stitching based on a representative experimental study at NA=0.33, in preparation for tomorrow’s NA=0.55. For a pitch 28nm vertical line-space, optimum conditions are confirmed experimentally to create a robust stitch. A P28 LS is measured post-stitching utilizing either a Ta absorber or a low-n absorber. For the latter, the higher reflectivity is experimentally mitigated by using sub-resolution-gratings. We also quantify the imaging impact of the transition between the absorber and the black border in the stitching region.

The new high numerical aperture (NA) Extreme Ultraviolet Lithography (EUVL) with a NA of 0.55 is being developed at ASML, which is using an anamorphic projection system with the demagnification of 4× in xdirection and 8× in y-direction. Compared to the traditional 0.33NA EUV scanner with full-field image size of 26 × 33mm², 0.55NA EUVL reduces the exposure field size to half-field (26 × 16.5mm²), due to this 8× demagnification in y-direction and the reticle size remaining unchanged (six-inch square). Therefore, in-die stitching between two exposures is needed for the applications requiring larger than half-field size. To achieve in-die stitching in practical applications at advanced node, performing model based optical proximity correct (OPC) is an essential step. Therefore, a complete process modeling and OPC flow is required. To build an accurate OPC model, the interaction effects between two stitching fields require some special considerations, such as aerial image interaction, optical proximity effect among the stitching patterns, mask absorber reflection, black border proximity effect, as well as the stray light from the neighboring fields effect. All these effects must be captured by specific models. In this paper, we will investigate the in-die stitching effects and solutions through simulation and wafer data. Thus, to collect the wafer proof data, various stitching test patterns have been designed and placed on imec test masks, and the wafer data will be obtained on imec 0.33NA EUV scanner.

Extreme ultraviolet (EUV) lithography systems have become one of the representative system nodes in the semiconductor industry. As EUV systems have been widely implemented to shrink integrated circuits, the importance of overlay control is increasing as much as patterning control. In terms of overlay control in lithography systems, a projection optics module is a key factor in determining the distortion of the overlay. In this paper, we present characteristics of EUV projection optics affecting overlay and propose a methodology predicting distortion of the overlay by calculating pattern shift using aberrations of EUV projection optics as an input. Also, the non-telecentricity of EUV systems is taken into account while calculating pattern shift as understanding the optical feature of EUV is required to explain the results of calculated distortion. Experimental results of dynamic random-access memory (DRAM) devices are presented to understand different behaviors between the projection optics structure of DUV and EUV scanners.

EUV lithography has been introduced in semiconductor fabrication and maximizing yield and throughput is extremely important. One key enabler is the use of a high-transmission pellicle to hold particles out of the focal plane and thereby minimize their impact on imaging. Imec initiated the development of a promising pellicle based on a network of carbon nanotubes (CNT). This CNT membrane offers the advantage of very high EUV transmission (> 95 %) and durability compatible with the EUV scanner power roadmap. Moreover, wafer printing with a CNT pelliclized mask on ASML’s EUV scanner at imec has been successfully demonstrated with good printing performance. Since the CNT pellicle is only a few tens of nanometers thick and suspended over an area of tens of centimeters, a major challenge of the pellicle is to control and optimize its mechanical stability and robustness when used in the EUV scanner. The pellicle rupture probability depends on a multitude of parameters, including pressure changes during mask loading and unloading, thermal expansion during exposure, initial stress/strain variations over the large pellicle, membrane degradation in the hydrogen plasma environment, and thickness of the pellicle. In this paper, the mechanical pellicle characterization as a function of the pressure changes for different CNT membranes is presented. The characterization is based on small-size sample evaluation using a bulge test method. By applying controlled plasma to such samples, it was possible to characterize the membranes not only as freshly fabricated but also after exposure to EUV scanner-like conditions. Additionally, the parameters obtained from small samples could be correlated to the actual movement during scanner manipulation. These measurements enable a fundamental understanding of CNT membranes and how they will behave in an industrial environment.

Background: The increasing demands on computational lithography and computational imaging in the design and optimization of lithography processes necessitate rigorous modeling of EUV light diffracted from the mask. Traditional electromagnetic field (EMF) solvers are inefficient for large-scale technology problems, while deep neural networks rely on a huge amount of expensive rigorously simulated or measured data. Aim: In order to overcome these constraints, we explore the potential of physics-informed neural networks (PINN) as a promising solution for addressing complex optical problems in the field of EUV lithography. Approach: We extend the existing MaxwellNet to simulate the light diffraction from typical reflective EUV masks. The coupling of the predicted diffraction spectrum with image simulations enables the evaluation of PINN performance in predicting relevant lithographic metrics and typical mask 3D effects. Results: The results of modeling near- and far-field diffraction using PINN showcase a good performance in terms of convergence behavior, stability, accuracy, and a significant speed-up (up to ×10000) compared to the rigorous 3D mask simulation using an established numerical EMF solver. In contrast to other machine learning approaches, PINN is able to accurately simulate the near field, learns the involved physics, and captures the optical and mask-induced 3D effects. PINNs can predict lithographic process windows with sufficient accuracy. Conclusions: Differently from numerical solvers, once trained, generalized PINN can simulate light scattering in several milliseconds without re-training and independently of problem complexity. This opens up the capabilities for partially coherent imaging simulations without the Hopkins approach, source optimization, and fast investigation of mask 3D effects.

To print ever smaller features at high contrast, projection lithography technology has been developed to allow use of shorter wavelength light and to increase numerical aperture (NA) from 0.33 to 0.55. After enabling EUV wavelengths, to keep up with the scaling trends the industry would now again like to increase the NA. Since the depth of focus (DoF) is inversely proportional to the square of NA, in hyper NA EUV lithography (EUVL), we anticipate that the total available DoF in the lithography process would be further limited. Therefore, within the constrained DoF budget, it is necessary for lithographers to minimize the unwanted best focus (BF) variations generated from different pitches on a photomask. In this paper we identify how the mask 3D (M3D) effect induced BF variation through pitch behaves according to changes in the pattern orientation and mask tonality for hyper NA EUVL. We study how various focus shift mitigation strategies can be combined to align best focus and enhance the image contrast for hyper NA EUVL.

To enable cost-effective scaling of technology nodes and extend Moore’s law for at least another decade, ASML has developed the High NA EUV platform. With an increase of the numerical aperture (NA) from 0.33NA to 0.55NA, High NA EUV brings multiple benefits to the semiconductor market, such as reduction of process complexity, yield improvement and higher resolution. This paper presents the High NA EUV ASML roadmap, providing a comprehensive overview of the systems being developed to support our customers’ nodes, and showing how we maximize 0.33NA (NXE) and 0.55 NA (EXE) platforms commonality, making High NA an evolutionary step on EUV technology. We will also give an overview of the progress and status of the first High NA EUV platform, the EXE:5000. Several systems have been now fully built in the ASML factory, which deliver the first performance data and integration learnings to support shipment to our customers. In parallel, first common learnings from the imec - ASML joint High NA Lab will be reported out, enabling early process development and accelerating the industry eco-system (mask, resist), as it is essential to the successful introduction of High NA EUV.

With the first delivery from ASML to Intel end of 2023, the next generation of EUV systems with ZEISS optics has found its way to the first high-end chip manufacturer. The increased numerical aperture (NA) of 0.55 compared to stateof-the-art EUV optics with a NA of 0.33, the new generation of NA=0.55 EUV scanners allow to support Moore’s law for at least another decade. Besides the new NA=0.55 EUV systems, ZEISS and ASML also continuously improve the current EUV scanners by enabling higher throughputs combined with improved imaging performance. We will present an overview over the status and key facts of the new NA=0.55 EUV optics as well as the latest performance improvements achieved for EUV systems with NA of 0.33. Furthermore, we will provide a glance at the EUV optics roadmap and provide give an outlook on what can be expected for EUV optics in future.

Digital Scanner (DS) is an optical maskless exposure tool with a SLM (Spatial Light Modulator) and a DUV solid-state laser with wavelength of 193 or 248 nm. There are two configurations of SLM: a tilt SLM, in which each micro-mirror pixel tilts to change the amplitude of reflected light; and a piston SLM in which each micro-mirror pixel moves along optical path to change the phase of reflected light. Both types are applicable for DS, but piston SLM has a better image contrast due to strong phase shift effect. A DS proof-of-concept tool (DS-POC) with piston SLM and exposure wavelength of 193 nm was developed, which has a similar imaging resolution with the DS248, a tool planned as the first DS product for lithography of 180 nm node or below. Exposure results of 180 nm node logic patterns by DS-POC are presented. Process window analysis of the logic patterns by simulation shows better performance of piston SLM than tilt SLM on exposure latitude. CD accuracy of less than 10% was experimentally demonstrated for resolution chart of L/S with CD from 150 nm through 300 nm using piston SLM at DS-POC.

In this paper we will report on the most recent immersion scanner innovations to improve scanner matching overlay. These are realized by improvements in e.g. optical column distortion, wafer alignment and system-metrology. We will elaborate on scanner solutions for wafer handling/chucking of warped wafers. Furthermore, to enable cost-of-ownership reduction, system design implementations driving larger scanner productivity (wafer per hour) will be presented.

The compatibility of Free-Electron Lasers (FELs) with Extreme Ultraviolet (EUV) scanners is a pivotal consideration as EUV lithography advances. Unlike plasma sources that emit incoherent light, FELs offer relatively coherent light with distinct characteristics including narrower bandwidth, higher repetition rates, and shorter pulse durations. This study elaborates on the integration process of FEL sources with existing scanner systems without necessitating modifications to the scanner’s illumination or projection optics. We delve into the coherence and speckle considerations crucial for maintaining image quality and minimizing defects in photolithography. Our findings reveal that the spatio-temporal splitting of FEL pulses, facilitated by the scanner’s illumination system, effectively mitigates coherence and speckle-related challenges. Moreover, the analysis of potential pulse damage reveals that despite the high peak power of FELs, the design of the scanner ensures that the risk of damage to optics is minimal. This research shows that it is practical and possible to use FEL sources in EUV scanners, enhancing their performance and efficiency in semiconductor manufacturing.

An accelerator-based free electron laser (FEL) source of 13.5nm light can, among other benefits, provide more power, polarized light and no out-of-band radiation. This paper explores how the extra power, polarized light and lack of out-of-band radiation could be used in future scanner designs. The differences between light generated from a laser-produced plasma (LPP) source and an FEL source are discussed. Rigorous mask simulation, OPC and source optimization are used to assess the impacts of polarization on contrast and process window for lines/spaces and multiple 2D patterns. Polarization is shown to provide contrast and process window benefits at half-pitches below 8nm. A full-field, isomorphic, 0.55NA scanner design is considered as a case study showcasing the benefits of FEL-based EUV light sources.

EUV lithography systems are now fully deployed in the high-volume manufacturing of leading edge semiconductor devices. In this paper, we review the performance of ASML’s current generation light sources in the field and preview the next step in EUV source performance for the NXE:3800E system. The NXE:3800E system marks a substantial step forward in scanner productivity, delivering a remarkable increase of (+60 WPH) in throughput, made possible by an increase in EUV source power (to >500W). This significant increase in power was achieved through improvements in the droplet generator, higher power CO2 drive laser, improved collector design, and enhancements in our plasma controls required for higher plasma power. Details of these developments and their impact on system design and performance will be discussed, along with recent high-power performance demonstrations of the overall integrated EUV light source system.

Advanced packaging was applied during the early stages of CSWLP (Chip-Scale-Wafer-Level-Package) development for mainly package form-factor reduction. However, advanced packaging is used not only for package size reduction but also for many remarkable features including fan-out wafer level packages that are used for mobile application processor to increase interconnect counts and reduce chip height. 2.5D silicon interposer technology is also used for Graphics Processing Units (GPU) and Artificial Intelligence (AI) chips to interconnect System-on-Chips (SoC) and cache memory to provide wide bandwidth computing power. Advanced packaging will also play a key role in the upcoming Chiplet era. Canon developed our first i-line stepper for advanced packaging in 2011. Since then, we have expanded our tool lineup to support customer demands with developments supporting large size packages and panel-level packaging. In this paper, we compare advantages between wafer-level packaging and panel-level packaging. In addition, we study bonding error budgets for fine pitch bump package in the upcoming Chiplet era. We will compare bonding errors among Silicon interposers, Organic interposers, and Glass interposers and point out the importance of lithography tool distortion reduction to realize less than 10 µm bump pitch packages. Furthermore, we will report on our low distortion patterning solution, the FPA-5520iV LF2 advanced packaging stepper.

As the semiconductor industry progresses towards the 2nm logic technology node in pursuit of improved chip performance and density, the demand for minimum pitch scaling in the back-end-of-line (BEOL) interconnect becomes crucial. Imec N3 logic design rules defined a minimum Metal 2 (M2) layer pitch of 30 nm, representing 2nm technology nodes. To further enhance semiconductor integrated circuit performance, attention is shifting towards advanced mask materials for current 0.33 NA EUV scanners. Low-n masks have been shown to improve extreme ultraviolet (EUV) imaging performance in terms of Local-CDU (LCDU), reduced mask 3D effects and improved optical contrast compared to a Tabased mask. In our study, we observed notable enhancements in optical contrast for real logic designs using a low-n mask. Our findings demonstrate an impressive LCDU of 5.5 nm and CGDU of 5.5 nm for Place’n’Route (PnR) structures at a pitch of 32. Furthermore, we successfully printed tip-to-tip (T2T) features as small as 20 nm on the wafer for regular tip-to-tip structures that didn’t get any Optical proximity Correction (OPC). These advancements mark significant progress towards manufacturability and developing a holistic patterning approach for random logic metal with EUV.

To keep up with the pace set by Moore's law, an innovative standard cell architecture called CFET has been proposed recently. Its technical challenge is to stack transistors on top of each other to achieve higher density. Nevertheless, the targeted nodes still require very small dimensions in terms of pitches, critical dimensions (CD) and tip-to-tip, but also in terms of geometries. In this paper we explore the patterning of a 2D local interconnect, Middle of the Line (MOL) layer with aggressive pitches and spaces that has been foreseen as a possible option for this CFET architecture. Multiple patterning solutions are proposed including 1- EUV print with multiple colors, 2- Spacer assisted solutions with multiple cut patterns. Finally, we evaluate the benefit of using 3- High NA EUV lithography as a potential candidate for this type of layer.

In state-of-the-art lithography equipment, isolating vibrations and maintaining vacuum integrity at the highest level is crucial while minimizing outgassing and contamination. This is essential for optimizing tool performance. Due to their viscoelastic properties, elastomers, primarily perfluoroelastomers or FFKMs, are commonly employed in sealing and damping applications in high-end semiconductor equipment. This work introduces two advanced simulation modeling techniques. The first technique predicts the lifetime of elastomer seals based on the elastomers' compression set, including temperature effects. The second technique simulates the damping properties of elastomers to optimize damper design. Predicting the lifetime of sealing systems can help determine the optimum maintenance periods and increase profitability in costly wafer processing. Predicting the damping behavior of elastomers can help accelerate the equipment design and development process at semiconductor OEMs.

Inverse lithography technology (ILT) has been the focus of research for several years due to its ability to produce theoretically optimal mask shapes. However, its widespread adoption has been hindered by the complex computational techniques involved and the challenges associated with writing curvilinear ILT mask shapes. To enhance manufacturability, one approach involves streamlining the curvilinear mask shapes by converting them into simplified, ‘Manhattanized’ rectangular shapes using a mask rule constraint (MRC) compliant ILT method. The aim of this study is to examine the influence of mask fracture sizes on manufacturability and to assess the impact of traditional optical proximity correction (OPC), curvilinear ILT, and Manhattanized ILT on the quality of photographic images. The study focuses on challenging cell structures, and generating pattern shapes using traditional OPC, curvilinear ILT, and Manhattanized ILT with MRC compliance. Fracture sizes were varied from 5nm to 25nm in increments of 5nm. To compare mask manufacturability across different patterns, several factors were evaluated, including data fracturing, writing time, and metrology. Photographic image quality metrics, such as NILS (Normalized Image Log-Slope), MEEF (Mask Error Enhancement Factor), EPE (Edge Placement Error), PV-band (Process Variation Band), and CDOF (Common Depth of Focus) were also assessed. The comprehensive analysis aims to provide a better understanding of the trade-offs between different approaches and guide future improvements in mask manufacturability and image quality.

Recently, as technology nodes continue to shrink in the manufacturing of integrated circuits (ICs), performance demands for lithographic scanners have consistently risen. As critical specifications for scanners, both overlay and focus must be limited to the nanometer range. The stage-positioning accuracy along x and y axes directly affects overlay, while the accuracy along z-axis directly influences focus. The moving average (MA) and moving standard deviation (MSD) of position error during exposure can be used to evaluate the stage-positioning accuracy. The reticle and wafer stages of scanners are typically multi-input multi-output(MIMO) motion systems. The interaxis coupling of MIMO motion systems being a significant challenge to achieving high positioning accuracy. The state-of-the-art MIMO control strategy employed for the reticle and wafer stages involves the initial implementation of a decoupling controller to eliminate inter-axis coupling, followed by independent single-input-single-output(SISO) control design for each individual axis. The existing decoupling approaches neglect the influence of force ripple of actuator, which contributes to unsatisfactory decoupling performance. On the other hand, the efficiency of existing decoupling approaches relies on the acceleration of reference trajectory. The reference trajectories for x and y axes during exposure have high speed and acceleration, while z-axis reference trajectory has relatively low acceleration, this results in unsatisfactory decoupling performance for z-axis. In this study, a dynamic decoupling controller is proposed to further reduce the inter-axis coupling. Specifically, the controller’s input signal is independent of the reference trajectory. Furthermore, considering the force ripple of actuator, a data-based iterative update algorithm is used to approximate the optimal controller coefficients. The reduction in computational complexity and the convergence properties of the algorithm are demonstrated from a model-based perspective. Numerical simulation and Hardware-in-loop(HIL) testing with established approaches confirm that, within a limited number of iterations, this approach improves z-axis stage-positioning accuracy by over 20%, while maintaining the stage-positioning accuracy of x and y axes. The proposed dynamic decoupling controller benefits on focus performance improvement.

In the manufacturing of advanced integrated circuits, ArF immersion lithography remains the primary tool for the patterning process. With the increasing requirements for critical dimension uniformity (CDU) in advanced technology nodes, every nanometer of CD variation becomes crucial. Over the past decades, scanner manufacturers have continuously adopted many technologies to improve the performance of DUV light sources, aiming to reduce CD variation caused by bandwidth, central wavelength, and other factors as much as possible. Previous investigations into the impacts of source bandwidth on imaging performance have provided the foundation for this work. In previous simulations of limited light bandwidth, Gaussian and Lorentzian symmetric line shapes were often used to represent the laser spectrum. However, the emitted beam from excimer lasers always exhibits some level of asymmetry in reality. This study focuses on the impact of spectral asymmetry on imaging performance. We conducted researches on the effects of various laser bandwidth shapes on the imaging of the critical patterns such as line space and other 2D patterns. We introduced an asymmetry parameter Δn to characterize the spectrum change. In the modeling part of this study, backward propagation (BP) neural network was established to predict the effects of a certain range of asymmetric spectra. We conclude that although the laser bandwidth in modern scanners is relatively low, the asymmetry of the laser spectrum does have certain impacts on CD variations depending on the mask pattern.

Gigaphoton Inc. presents a Sn-LPP EUV light source for mask inspection tools. This light source uses a minimum-mass Sn droplet generator with an in-line Sn fuel supply system, a double-pulse laser irradiation scheme with precise shooting control and a debris mitigation technology with H2 buffer-gas flow. During 1500 hours of continuous operation, a brightness of 120W/mm2 sr at the plasma point, a stable EUV energy 3σ-value below 5% and a high availability of 99% have been obtained without reflectivity degradation of the EUV collector mirror. We are currently developing key components for higher repetition rate to further increase the brightness.

In the implementation of EUV lithography, the stochastic effects in photoresist patterning are significant. The stochastic characteristic of Chemically Amplified Resist (CAR) for EUV demands novel modeling methods instead of the continuous model used in DUV. The previous model is directly derived from Gillespie algorithm, which sorts all cells into logarithmic classes based on the magnitude of the propensity functions. It takes a lot of time to update the classes in each iteration. Moreover, it splits the whole system only once, and for large systems a logarithmic class can include still larger number of cells, which can also take up a large amount of computing time. In this study, a new stochastic EUV resist model is proposed to improve the speed of PEB simulation by using a modified minimal process algorithm, which employs a splitting method called cascade classification. It divides evenly all cells into small groups in some arbitrary manner and each group is further divided evenly into smaller ones, and so on, until the smallest groups include a sufficiently small number of cells. The research shows that the new algorithm is more efficient than the previous one while maintaining accuracy.

The defect in multilayer of mask blank can degrade the image quality in EUV lithography. It is caused by particles deposited under or inside the multilayer, which leads to deformation of the multilayer structure. This deformation can impact the reflectivity of the multilayer and cause amplitude changes and phase shifts, resulting in pattern shift or critical dimension error in EUV lithography imaging. In order to improve imaging performance, medium size defects can be mitigated by pattern shifting techniques, while small size multilayer defects can only be compensated by mask repair to meet the imaging quality requirements. Therefore, it is necessary to investigate the effect of defect size on lithography imaging performance. In this work, we evaluate the impact of conformal defect with different sizes on the imaging of three classical test patterns: Line Space, Tip to Tip, and Tip to Line. By representing the defect by using ℎtop, 𝑤top, ℎbot, and 𝑤bot, the defects on different central and edge positions in the arc slit of exposure field are investigated. Then, the data gathered from lithography simulation is used to model the parameters of defect and the imaging result indicators such as CD. The corresponding relationship between the defect size and the imaging CD is derived through a mathematical model. From the constrains of the lithography imaging performance indicators, the tolerance of the parameter of the multilayer defect is further predicted inversely.

Plasmonic lithography can amplify the evanescent wave resonance at the mask and participate in imaging by exciting surface plasmon polaritons (SPPs), breaking the diffraction limit in traditional lithography. Source Optimization (SO) technology is widely used to compensate for imaging distortion in traditional lithography. This paper proposes an effective SO model for plasmonic lithography under the compressed sensing (CS) framework. To accelerate the algorithm, the SO is formulated as an underdetermined linear problem, where the number of equations is much smaller than the source variables. We selected lines, contacts, and complex test patterns to verify the imaging improvements and superiority of the model. The results indicate that compared to the annular sources, optimized sources can achieve better imaging results and higher imaging contrast. This provides favorable conditions for the large-scale application of plasmonic lithography.

Roughness cannot be ignored as feature sizes shrink with Moore's Law, since it has potential to influence the device's performance. The roughness is usually described by line edge roughness (LER) and line width roughness (LWR). LER is the deviation of a feature edge from its ideal shape and is defined as three times the standard deviation, the deviation from the average line width is defined as LWR. For a certain critical dimension (CD) and pitch, there are many factors that can contribute to the roughness in the lithography process, such as source, mask optimizations, photoresist types and its processing, etc. An in-depth insight of the roughness formation mechanisms is essential to improve LER. This study employs photoresist process simulation to analyze LER, offering an efficient alternative to silicon data collection. Simulation analysis is carried out to examine the key factors influencing LER, including quencher concentration, photoacid diffusion length, PEB temperature. Concurrently, the study also delves into the impact of photoresist resin molecular composition and the development process on roughness. By using simulation to understand and predict roughness, the research provides insights into optimizing lithography parameters, thereby improving process stability and minimizing roughness formation.

Super-polished optics are crucial components for extremely short wavelength applications. One of the main applications are illumination optics. Short operational wavelengths of ~13,5 nm lead to high demands on substrate’s surface quality, especially for microroughness features in the high spatial frequency range (HSFR). The excellent surface quality is essential for high-performant reflective mirrors. Typically, mirror substrates for EUV applications are made of glass, glass-ceramics and silicon. This paper describes an alternative approach by using metallic substrates made from electroless Nickel-Phosphorous (NiP) plated Aluminum alloys. The X-Ray amorphous NiP alloy enables several polishing and correction techniques to reduce the surface roughness down to the addressed sub-nm range. The process chain of metal optics includes ultra-precise diamond turning (DT) and chemical mechanical polishing (CMP) steps. Microroughness (HSFR) of < 0.2 nm RMS (AFM, 1 × 1 µm²) by CMP could be generated on flat and curved surfaces as well as on freeform optics. The process chain will be verified by surface characterization techniques, e.g. atomic force microscopy and white light interferometry. The data is evaluated using the established PSD (power spectral density) analysis method. Conclusions about the surface errors concerning the different processing technologies can be verified. This paper shows that super-polished metallic freeform optics are suitable for curved EUV illumination optics.

Traditional approaches to quantifying stochastic EUVL variability for the use in stochastic-aware OPC correction and verification algorithms are tested against the stochastic EUVL via failure probability calculation by the Calibre stochastic model. The purpose is to check if these simplified approaches, based on the image metrics (e.g., NILS) or PV band parameters may provide an accurate failure probability prediction for a broad variety of layouts typical for the via layers of modern ICs. We also present the examples of verification of the Calibre stochastic model failure probability predictions against the brute force Monte Carlo trials.

To achieve higher resolution extreme ultraviolet lithography targeted toward sub10nm, reflective projection scanner image numerical apertures (NA_i) are being increased beyond the current value of 0.33 to 0.55, and upwards of 0.75 as a desirable target. Bragg reflectors using alternating silicon and molybdenum that have heretofore been coated as periodic multilayers cannot achieve desired reflected amplitudes as corresponding 0.25X mask numerical apertures (NA_m) are accordingly increased. Additionally, TM polarized image modulation decreases with NA, which becomes significant at 0.55 and above. We present here the optimization of non-regular alternating, or aperiodic, silicon-molybdenum multilayer reflective coatings that can achieve improved amplitude and polarization performance through angle as higher-NA EUVL lithography is pursued. Through the use of rigorous EM computation paired with a genetic optimization method, we show that amplitude apodization can be recovered to 60% peak reflectance for NA_m values up to 0.2 (corresponding to NA_i values of 0.8), while at the same time achieve a TE degree of polarization (DOP) exceeding 40%.

Extreme-ultraviolet (EUV) lithography is a crucial technology in semiconductor manufacturing, and the development of effective pellicles is essential to prevent mask contamination and ensure patterning accuracy. Traditional approaches to improving pellicle transmittance have faced limitations, prompting exploration into novel strategies such as CNT, graphitelike film, or structural modification. In this study, we investigate the mechanical stability and imaging impact of porous pellicles, which can overcome the limitations of conventional structures. Our findings reveal that while porous pellicles induce stress variations, overall residual stress is almost maintained. Imaging simulations demonstrate minimal impact on pattern fidelity, highlighting compatibility with existing lithographic systems. Additionally, we fabricated the silicon nitride porous pellicles and measured EUV transmittance. Experimental results confirm significant increases in EUV transmittance with porous structures, validating their potential for next-generation lithography applications.

Since high-NA extreme ultraviolet (EUV) exposure tools have a half field size, two masks are necessary for exposing a large device which is larger than the half field of 26 mm × 16.5 mm. However, it takes long time until exposing the 2nd half field since the 2nd mask would be loaded after exposing the 1st half fields on many wafers. This may cause poor characteristics at the stitching area. Therefore, an introduction of a two-mask stage is proposed. This can drastically reduce the time until exposing the 2nd half field. However, since a conventional scan with two masks makes two separate exposed areas, a new scan method is necessary in order to contact the two half fields.

In this submission, we introduce the performance and attributes of our broadband EUV source, which is generated using a copper-based laser-produced plasma. This source is currently operational in EUV Tech’s HVM reflectometer and multi-angle reflectometer / scatterometry (ENK) tools.

The problem of radiation damage on photomask becomes more serious in the semiconductor technology using the 193nm ArF excimer laser, this phenomenon of radiation damage is regarded as a serious issue for semiconductor device fabrication. Wafer yield loss could be caused by the mask CD variation and the transmittance loss due to MoSi film radiation damage on the photomask, many studies have been done about the mechanism and the solution of the radiation damage such as haze and Cr migration so far. However, we still need to clarify some abnormal types of growing defects such as scum defects induced by MoSi film radiation damage. In this paper, we introduce a study of ArF Phase Shift Masks radiation damage-induced scum defects. The impact of radiation damage induced scum defects caused mask CD variation found by AIMS-intensity and AIMS-CD. Additionally, the defect's growing location is stronger related to the distance to frame region or scribe line with the Cr layer of the photomask. Also, this defect is not generally removed through the conventional wet cleaning process but it only could be removed by a kind of dry treatment with a wet process. Base on the analysis data and previous kinds of literature we bring forward the possible defects generation mechanism.

In parallel to the development of EUV lithography equipment and processes towards High-NA there is a still growing interest to study and understand the role of 3D mask absorber driven effects on imaging. Alongside with appropriate exposure and patterning techniques, the choice of a suitable mask absorber material has large impact on the process window optimization. The development of solutions to the challenges combined with the EUV reflective mask typically is based on rigorous imaging simulations, but aerial image metrology can be a supplementary way to verify the role of mask 3D effects on lithography. For that purpose, the ZEISS AIMS EUV actinic mask measurement represents a unique and variously usable way for the qualification of the current as well as the next generation High-NA mask printing performance under scanner-equivalent optical column conditions. In this paper, we show results of AIMS® EUV measurements at Ta-based absorber masks where we identify a significant impact of mask 3D driven phase shifts on the aerial images. For a standard High-NA dipole setting, the intensity profiles of a dense 18 nm vertical lines and spaces structure show an initially unexpected line splitting through focus that can be reproduced by rigorous imaging simulations. In contrast, no splitting is found if the experiment is described within a 2D model. A follow-up experiment on a similar dense vertical 20 nm lines and spaces clip with dedicated monopole illumination demonstrates a prominent asymmetric line splitting for each of the single pole positions, again in accordance with simulation. For a center monopole setting we see a frequency doubling of the aerial image for one particular focus position. To describe the aerial image behavior, we applied an analytical model for three-beam-imaging in the presence of defocus and mask 3D effects. Consequently, the observed through-focus behavior can be explained by various phase shifts to the three-beam imaging cross-terms involved.