If you’ve been following the evolution of advanced packaging, you know that the industry is pushing boundaries like never before. From high-performance computing to industry-upending AI devices, the demand for smaller, faster, and more powerful chips is driving innovation at every level.
If you’ve been following the evolution of advanced packaging, you know that the industry is pushing boundaries like never before. From high-performance computing to industry-upending AI devices, the demand for smaller, faster, and more powerful chips is driving innovation at every level. One of the unsung heroes in this transformation: Glass carriers.
These carriers are becoming essential for applications involving high-bandwidth memory (HBM), 2.5D/3D integration, and chiplet architectures. During the manufacturing process, glass carriers serve as mechanical support for thin wafers and panel-level packages. Why? Glass carriers are noted for their warpage resistance, superior rigidity, and thermal stability. This combination of glass’ exceptional flatness and rigidity enables the precise placement of dies and interposers. Additionally, glass is optically transparent, which allows through-glass alignment during bonding and stacking, a critical capability for 3D integration where multiple layers must be accurately registered.
The benefits of glass carriers, however, come with several challenges, none of which should come as a surprise to anyone who has ever handled glass, whether in the fab or at home. Glass is fragile and, as such, is prone to surface defects, subsurface inclusions, and residual stress. Each of these can negatively impact die attachment quality, interconnect reliability, and die yield.
Let’s take a look at three major yield-killing culprits.
Surface defects such as particles, pits, and scratches are among the most common issues and may occur during glass carrier handling and processing, compromising the structural integrity and performance of advanced packaging assemblies (Figure 1). Particles can interfere with the bonding process, leading to poor adhesion or electrical discontinuities, while pits and scratches can propagate stress points that weaken the carrier during thermal cycling or molding.
However, subsurface inclusions and organic contamination, which are often introduced during reclaim or cleaning, pose more critical challenges. Inclusions within the glass can create localized stress concentrations, while organic residues can reduce UV transmission and cause bonding failures. These contaminants are particularly problematic in high-density interconnect environments where optical clarity and surface purity are critical.
Figure 1: Common glass carrier defects
In addition to surface and subsurface defects, residual stress represents a concern. Over time, these stress points, manifesting during thermal processing or mechanical handling, can lead to cracks or delamination, undermining the thermo-mechanical integrity of the entire package.
These potential challenges are compounded each time a glass carrier is reused in an effort to reduce overall packaging costs. Fortunately, technologies have been developed to address this obstacle. These technologies integrate AI-driven defect classification, real-time analytics, and adaptive scanning modes to maintain throughput without sacrificing accuracy, enabling manufacturers to detect surface anomalies, subsurface inclusions, and stress-induced defects with unprecedented precision.
Today’s wafer-based inspection platforms utilize laser scatterometry and imaging techniques to inspect for nanometer sized defects on a variety of opaque and transparent/semi-transparent substrates. These substrates may be suitable for either R&D or high-volume advanced IC substrate (AICS) and fan-out panel level processing (FOPLP) environments. Proprietary inspection technology with multiple detection channels and advanced signal processing algorithms is applied to achieve accuracy and reliability in glass carrier inspection.
Figure 2: Results of top (blue) and bottom (red) defect mapping.
With each channel optimized to capture unique scattering and reflection signatures, the technology differentiates between surface and subsurface defects, as well as stress-related anomalies, with remarkable accuracy. Surface particles, scratches, pits, bumps, surface contamination, film or bulk wafer stress, voids/inclusions can be detected, measured, characterized, and imaged. One of the most significant capabilities of this technology is the ability to conduct simultaneous top, bottom, and internal defect mapping, a critical need for transparent and semi-transparent substrates where defects can occur across multiple planes (Figure 2).
Beyond defect detection, Angstrom-level film thickness measurement provides precise control over surface coatings and residual layers. This capability is particularly valuable in the glass reclaim process where even minor variations in film thickness can impact UV transmission and bonding performance. By enabling accurate defect detection and grading, only glass carriers meeting stringent quality standards are returned to production.
By introducing technologies that mitigate risks by providing comprehensive defect mapping and stress analysis, manufacturers are able to maintain the mechanical and thermal integrity required for next-generation devices. This capability is especially valuable in markets such as AI devices, high-performance computing, and automotive electronics where reliability is non-negotiable. With this combination of advanced optical technology and robust algorithmic analysis, manufacturers can successfully achieve higher yields, lower costs, and greater confidence in their packaging processes.
As packaging complexity grows and the use of glass carriers increases, inspection systems that combine multi-depth defect mapping and stress analysis will become indispensable for ensuring yield and reliability in AI and HPC devices. With the explosive growth in AI-driven data centers and advanced packaging architectures, manufacturers need solutions that combine accuracy, speed, and cost efficiency. The laser-based wafer inspection technology discussed in this blog meets several glass carrier challenges head-on while enabling advanced packaging houses to maintain defect-free glass carriers in support of next-generation advanced packaging.
The future of glass carriers is clear: with the right technologies at the ready, manufacturers have the tools and the means to meet the growing needs of the AI and HPC markets.
Jason Lin is Director of Product Marketing at Onto Innovation.
We’d love to connect with you.
Looking to learn more about our innovative solutions and capabilities? Our team of experts is ready to assist you. Reach out today and let’s starts a conversation about how we can help you achieve your goals.
"*" indicates required fields
AI workloads are pushing the boundaries of compute, memory, and interconnect architectures, and to meet these goals, manufacturers are rapidly accelerating advanced logic and DRAM development. Chief among these innovations: gate-all-around (GAA) logic transistor and vertical gate (VG) DRAM, two device architectures that promise higher performance, improved power efficiency, and greater scalability.
However, the arrival of these device architectures introduces new levels of manufacturing complexity brought on by increasing high aspect ratios (HAR) and the continued shrinking of device dimensions by roughly 20–30% per generation. Fortunately, new capabilities in optical critical dimension (OCD) metrology are up to the challenge of measuring and monitoring features, such as gate profiles, film thickness, and structural uniformity, at the nanometer scale. With the right tools on hand, manufacturers can maintain high volume process control for GAA logic and VG DRAM in the AI era (Figure 1).
Figure 1: GAA (left) and vertical gate DRAM (right) present new challenges for process control (pictured above)
These new capabilities are enabling the delivery of accurate, high throughput measurements across complex 3D geometries like GAA and VG DRAM by offering smaller spot size, higher signal sensitivity, and enhanced precision.
But before we discuss these new capabilities, let’s dive deeper into the challenges of both device architectures. We’ll start with GAA.
GAA transistors feature HAR channels and stacks of multi-layered nanosheets, in some cases as thin as 10nm. The manufacturing complexity lies in controlling individual nanowires and accurately characterizing buried nanowires within three-dimensional architectures. Precise control over nanowire dimensions is critical for achieving the desired electrical performance, including higher transistor speeds and lower power consumption. As such, GAA manufacturers need metrology solutions capable of extracting unique signals with high sensitivity and precision from each layer, even those deeply buried under semi-opaque stacks.
While GAA logic presents unique challenges in transistor scaling, VG DRAM introduces equally complex demands in memory architecture. VG DRAM involves vertical nanowire etch, multi-layer ultra-thin film deposition, sidewall trim, and buried bitlines, electrical connections that are embedded deep within the silicon substrate rather than being placed on the surface, helping reduce the footprint of each memory cell. Additionally, VG DRAM for HBM4 features smaller cell blocks. Addressing these challenges requires an OCD system with a significantly smaller spot size in order to conduct direct on-device measurements. This, unfortunately, reduces the signal strength of optical systems.
For both GAA logic and VG DRAM devices, incorporating real-time stabilization technologies and advanced optical modeling can improve measurement precision and enable tighter process windows. In addition, AI-guided analysis allows fabs to create robust measurement recipes faster and with greater accuracy for highly complex device structures like nanosheets and VG DRAM cells. Analysis software also enables more accurate measurements that are scalable across multiple tools, while supporting smart manufacturing and predictive analytics for advanced AI applications.
Many of today’s OCD metrology tools, however, have limited SNR to measure these increasingly smaller changes in the complex device structure, while other systems do not have sufficient information to separate signals from multiple parameters (e.g., dimensions from individual nanowires) in complex device structures. Furthermore, most OCD metrology systems feature large spot sizes that do not fit into the reduced DRAM cell block area for HBM4.
Without the proper advanced metrology solutions addressing these challenges, manufacturers risk variability in critical dimensions and compromising device performance and yield. As such, more data channels and enhanced SNR at higher speed are needed to overcome the reduced signal strength due to smaller geometries and increased signal complexity from higher nanowire stacks.
Onto Innovation has recently added a number of optical and algorithm innovations to its OCD arsenal, such as new multi-polarization angle data channels with real-time stabilization technologies to enhance information content and SNR. These new capabilities offer advanced optical modeling, improving measurement precision, and robustness. Furthermore, the addition of AI-guided analysis software allows fabs to create robust measurement recipes faster and with greater accuracy for highly complex device structures. This includes nanosheets and VG DRAM cells. The use of analytical software also enables more accurate measurements and is scalable across multiple tools, all in support of smart manufacturing and predictive analytics for advanced AI applications.
As the semiconductor industry advances toward AI-optimized architectures, the need for more precise and adaptive process control has become critical. The new OCD capabilities discussed in this blog enable tighter control over individual nanowires in GAA logic and allow for more accurate measurements within increasingly compact VG DRAM cell blocks. Combined, these innovations represent a foundational shift in how the industry measures, monitors, and optimizes the most advanced semiconductor technologies.
Jiangtao Hu is a product marketing senior director at Onto Innovation.
We’d love to connect with you.
Looking to learn more about our innovative solutions and capabilities? Our team of experts is ready to assist you. Reach out today and let’s starts a conversation about how we can help you achieve your goals.
"*" indicates required fields
The FAaST system is a versatile, non-contact electrical metrology platform, with an option to combine micro and macro corona-Kelvin technologies together with digital surface photovoltage (SPV). It enables high-resolution dielectric and interface measurements across a wide range of dielectric materials, supporting both R&D and high volume manufacturing.
The primary application of non-contact CV metrology is monitoring dielectric properties during IC manufacturing. Unlike conventional electrical measurements, it requires no sample preparation, eliminating the need for MOS capacitor structures. This reduces metrology cost and enables fast data feedback in both R&D and manufacturing environments.
The corona-Kelvin method uses a corona discharge in air to deposit an electric charge (DQC) on the wafer surface. A vibrating Kelvin-probe then measures the resulting surface voltage (V), enabling determination of the differential capacitance (C= DQC/DV). By monitoring surface voltage in both dark and illuminated conditions, the system separates two key components: dielectric voltage (VD) and semiconductor surface potential (VSB), enabling determination of flat band voltage (VFB).
Analysis of the resulting charge-voltage data yields electrical parameters, including trap density (Dit), flat band voltage (Vfb), dielectric charge (Qtot), dielectric capacitance (CD), Equivalent Oxide Thickness (EOT), leakage current, and tunneling characteristics.
As your partner for innovative solutions, we’re always here for you.
Discover how our cutting-edge semiconductor solutions are engineered to meet your most complex challenges: delivering performance, reliability and innovation where it matters most.
"*" indicates required fields
The MBIR system is a revolutionary in-line, non-destructive infrared reflectometry system that enables critical process control of high aspect ratio structures, films and epitaxial structures. It meets the needs of leading-edge customers with its high speed and process coverage.
As more high aspect ratio processes are used in multiple industry segments, there are metrology needs for monitoring of related processes, including dimensions and properties of carbon film hard masks and etched 3D structures.
The MBIR system delivers high-throughput, low COO, non-contact, non-destructive measurements of dimensions and uniformity of layers and etched structures used in integrated circuit manufacturing. The small spot size makes the tool suitable for measurements of scribe line test structures as in-line process control. The unique technology and analysis capability simplifies system calibration requirements and removes the effect of substrate variations for key layer measurements.
While the software contains advanced features for measurement recipe and analysis model creation, it has a user-friendly interface and implementation that allows the fab customers to create and manage the recipe system for MBIR tool fleets.
Thickness map from amorphous carbon film
As your partner for innovative solutions, we’re always here for you.
Discover how our cutting-edge semiconductor solutions are engineered to meet your most complex challenges: delivering performance, reliability and innovation where it matters most.
"*" indicates required fields
The FAaST Digital SPV system provides a fast, non-contact, and preparation-free method for full wafer imaging of contamination in silicon. High resolution maps of diffusion length and iron (Fe) concentration are generated in minutes, setting the industry standard for precision and sensitivity in Fe contamination control, reaching the E7 cm-3 range.
There is no disputing the detrimental effect of metallic contamination on the integrity of the critical gate oxide used in integrated circuits. During high temperature processing, contamination in the silicon wafer often precipitates at the Si/dielectric interface or segregates into the dielectric—both scenarios can cause premature device failure and reduced yield. As device dimensions shrink, the tolerance for contamination decreases, requiring ever-lower background levels of metals like iron (Fe). Over the past 25 years, the IC industry has reduced typical Fe concentrations by more than three orders of magnitude, yet further reduction is essential, especially for applications like CMOS image sensors.
The FAaST Digital SPV system addresses this challenge with industry-leading sensitivity and speed. It provides a fast, non-contact, and preparation-free method for full-wafer imaging of contamination. High-resolution maps of minority carrier diffusion length and Fe concentration are generated in minutes, enabling fabs to detect and control contamination at levels as low as the E7 cm⁻³ range.
Figure 1. Typical background Fe concentration in new IC Fablines (blue) and the state-of-the-art SPV detection limit (red)
As your partner for innovative solutions, we’re always here for you.
Discover how our cutting-edge semiconductor solutions are engineered to meet your most complex challenges: delivering performance, reliability and innovation where it matters most.
"*" indicates required fields
The CnCV system enables wafer-level characterization of WBG materials without test device fabrication, reducing time and cost. As a mercury-free alternative to MCV, it eliminates contamination concerns. The enhanced Kinetic CV mode with UV-assisted corona charge neutralization achieves high throughput and precision for fast, reliable process control.
The CnCV system utilizes a novel constant surface potential corona charging, which enables the precision required over a large voltage range. The patented technology includes charge- and photo-assisted modes, especially suited for speed and precision on WBG materials and structures, including SiC, Ga2O3, GaN, and AlGaN/GaN HEMT. Additionally, Corona-Kelvin characterization includes electrical properties of dielectrics and interfaces of films on SiC and GaN epi layers. An automated top-side edge contact (TSEC) is also available enabling characterization of WBG on insulating/semi-insulating substrates. Automated bias-temperature stress (BTS) measurements are also available with the CnCV system, providing a fast, noncontact way to quantify the reliability of passivated SiC and GaN.
Beyond typical CV type parameters, the full wafer corona approach allows for QUAD (quality, uniformity, and defect) mapping. The electrical defect imaging, QUAD-EDI, mode is especially designed for SiC. It provides a unique means for quick screening of epi electrical defectivity enabling improvement in device yield prediction.
Figure 1. QUAD-EDI Map on final metallized device wafer after Merged Schottky PiN diode fabrication identifying failed dies.
As your partner for innovative solutions, we’re always here for you.
Discover how our cutting-edge semiconductor solutions are engineered to meet your most complex challenges: delivering performance, reliability and innovation where it matters most.
"*" indicates required fields