Chip Singulation Laser Dicing: Intelligent Process Monitoring in Real-Time

By Kristin Rinortner | Translated by AI 5 min Reading Time

Related Vendor

In wafer dicing, not only are more precise laser processes important but also intelligent process monitoring. Researchers at Chiba University have now presented an interesting approach: they measure the recoil force generated during material removal. This could make laser dicing a self-monitoring manufacturing process.

Laser Dicing: From the recoil force of the laser pulse upon impact on the semiconductor, a reproducible process signal can be derived in real time, enabling automated control of wafer dicing.(Image: Ciba University / CC BY-SA 4.0)
Laser Dicing: From the recoil force of the laser pulse upon impact on the semiconductor, a reproducible process signal can be derived in real time, enabling automated control of wafer dicing.
(Image: Ciba University / CC BY-SA 4.0)

The separation of semiconductor chips from a processed wafer is a crucial step in the backend of semiconductor manufacturing. In this process, the wafer is cut into so-called dies using a diamond saw, which form the actual semiconductor components. Wafer dicing today significantly determines the yield, reliability, and cost-effectiveness of a process. At the same time, wafers are becoming thinner, and mechanical reserves are decreasing. As a result, traditional diamond sawing methods are reaching their limits.

The industry is currently developing several methods in parallel. In addition to traditional laser ablation cutting, stealth dicing, waterjet-guided laser dicing, and hybrid processes are gaining importance. The goal of all developments is to further reduce material removal, avoid microcracks, and simultaneously significantly increase throughput.

Gallery

Laser cutting, where individual chips are separated using focused laser pulses on the semiconductor wafer, is increasingly preferred as sensitive, more ductile materials can be processed with minimal mechanical stress. However, achieving an adequate yield requires precise process control.

With increasing process speed, however, the processing state, i.e., the penetration depth of the laser pulse, can only be monitored to a limited extent during laser operation. To this day, many systems rely on optical inspection systems or downstream quality controls. This is precisely where the research of Chiba University comes into play.

Evaluate Process Signals in Real Time

The approach of the Japanese scientists led by Professor Hirofumi Hidai from the Graduate School of Engineering at Chiba University focuses on recoil force. This is a reaction force that is generated on the surface of a material when it is rapidly heated and vaporized by a laser pulse. According to the law of conservation of momentum, an equal and opposite force acts on the material in this process.

The Japanese use the recoil force as a direct process control signal for laser dicing. Instead of cameras or confocal sensors, a piezoelectric force sensor exclusively registers the counterforce generated during material removal. This allows conclusions to be drawn about the actual machining process.

The results of the study will be published on November 1, 2026, in Volume 203 of the journal "Optics and Laser Technology."

Recoil force as an in-situ process signal

Chiba University demonstrates for the first time that the recoil force generated during laser dicing can be used as an in-situ process signal. It correlates with the processing depth, detects wafer breakthrough, and thus enables continuous monitoring of laser dicing without additional optical sensors. The authors view this as a promising approach for future closed-loop controls in semiconductor manufacturing.

That's why this is important: The boom in generative AI is increasing the demand for high-performance processors and HBM memory. At the same time, nearly all leading semiconductor manufacturers are adopting chiplet architectures and advanced packaging. As a result, the number of high-value dies per wafer, and thus the economic value of each individual dicing step, is rising. Process monitoring is becoming a productivity factor.

Recoil Force Reveals the Machining Progress

For the experiments, the researchers used a nanosecond-pulse Nd-YAG laser with a wavelength of 1064 nm and a pulse duration of 25 ns. To measure the recoil force, the scientists utilized a piezoelectric force sensor. Simultaneously, the actual processing depth was measured confocally. This allowed the measured recoil force to be directly compared with the actual material removal.

The results show an almost linear relationship between pulse energy and recoil force. Above a pulse energy of approximately 3 mJ, plasma effects limit further material removal. The recoil force remains a reproducible process signal and provides real-time information about the process.

Determine Machining Depth in Real Time

During multi-pulse machining, the recoil force continuously decreases. The cause is the increasing defocusing of the laser beam with growing drilling depth. The authors derived a mathematical model from this, which calculates the machining depth directly from the measured force.

With an average relative deviation of around 24%, the method does not yet replace high-precision measurement technology. However, the depth information is generated directly during the machining process. And this is achieved without a camera, without additional measurement optics, and independently of the contour (deep or narrow) of the cutting grooves.

Clearly Detect Wafer Cutting

Particularly interesting is the behavior just before the wafer is completely cut through. The course of the recoil force changes significantly at this point. This allows the exact moment when a wafer is fully cut to be determined without the need for visual inspection of the sample.

Subscribe to the newsletter now

Don't Miss out on Our Best Content

By clicking on „Subscribe to Newsletter“ I agree to the processing and use of my data according to the consent form (please expand for details) and accept the Terms of Use. For more information, please see our Privacy Policy. The consent declaration relates, among other things, to the sending of editorial newsletters by email and to data matching for marketing purposes with selected advertising partners (e.g., LinkedIn, Google, Meta)

Unfold for details of your consent

Especially in highly automated production lines, this could reduce the number of unnecessary laser pulses. At the same time, the risk of damaging dicing films or subsequent substrates decreases.

A Step Towards the Semiconductor Manufacturing of the Future

The true significance of the work lies in its potential for real-time process monitoring. In semiconductor manufacturing, processes are not only monitored but are also intended to be automatically adjusted. Closed-loop systems adapt laser power, focal position, or feed rate during machining to match the actual process state. This requires robust sensor signals that are directly obtained from the manufacturing process.

The recoil force arises directly from the interaction between the laser and the material, making it significantly closer to the actual process than downstream image analysis.

In the future, force measurement could be combined with AI-supported control algorithms. These would detect deviations early and automatically optimize laser parameters. This could not only reduce waste but also lower energy consumption and increase process stability.

Industrial Series Production Not Yet Foreseeable

Several challenges remain before industrial series application is feasible. The researchers particularly highlight the dynamics of the measurement system, higher pulse frequencies, and the influence of scanner movements as open fields of development. Additionally, the transfer to high-speed systems with repetition rates of several kilohertz still needs to be investigated.

Nevertheless, the research project already points to a direction in which modern laser processes could develop: away from open machining workflows and toward self-monitoring manufacturing systems that continuously assess their own process state.

Conclusion: Recoil Force as a Means of intelligent Process monitoring

The presented recoil force measurement is far more than an academic idea. It addresses one of the central topics of semiconductor manufacturing: intelligent process monitoring in the era of advanced packaging, chiplets, and AI accelerators. While traditional sensor systems observe the process externally, the new approach utilizes a signal that arises directly from the laser-material interaction.

If the process can be further developed for industrial series production, it could lead to a new sensor type for smart fabs—robust, cost-effective, and directly integrated into the manufacturing process. This would make laser dicing a decisive step towards autonomous semiconductor manufacturing. 

Article in Optics & Laser Technology