As the semiconductor industry confronts the growing complexity of manufacturing ever-smaller transistors, traditional lithographic scaling faces increasing resistance from the laws of physics. The era of simply shrinking feature sizes to gain performance is fading. In its place, engineers are exploring inventive alternatives to continue delivering better chips without relying solely on dimensional scaling. One such alternative gaining traction is spectral engineering, the use of different light wavelengths in lithography and metrology to unlock new levels of precision and control. In this layered conversation about what defines technological progress, Erik Hosler, an innovation strategist focused on consumer-centric technology development, reminds us that the true test isn’t just in feature size, but in experience.
This subtle yet powerful reframing lies at the core of spectral engineering’s promise. The technique isn’t about creating smaller transistors outright, but about improving visibility, control, and error correction during the fabrication process. As feature sizes push atomic limits, it becomes increasingly difficult to detect faults, measure features accurately or etch patterns consistently using conventional wavelengths. By leveraging broader parts of the electromagnetic spectrum, engineers can now see and shape at scales previously obscured, enabling continued advancement in chip design even without shrinking every layer.
A Shift from Shrinking to Seeing
Moore’s Law has long been interpreted as a race to the bottom, a relentless march toward smaller transistors. But in the last decade, this framing has developed. Today, progress is often about how well we can understand and manipulate what we already have, not just how small we can make it.
Spectral engineering exemplifies this shift. Instead of focusing only on the size of features, it focuses on how features are detected and interpreted during fabrication. By using different light wavelengths such as ultraviolet, infrared, and others, engineers can enhance metrology, improve defect detection, and refine alignment techniques during patterning.
Philip Leray of IMEC recently noted the untapped value in this approach, pointing out that “there are a lot of wavelengths to be exploited and to bring value.” These spectral ranges, previously considered fringe or specialized, are now central to the conversation about how to extend the scaling roadmap.
Seeing Through New Lenses
At its core, spectral engineering offers a simple but transformative concept: different wavelengths reveal different things. Just as switching to night vision or infrared can reveal what is invisible to the naked eye, applying various light spectra to chip manufacturing unveils hidden defects, inconsistencies, or alignment problems that standard optical tools might miss.
For example, using shorter wavelengths can help resolve finer details during photolithography. Meanwhile, longer wavelengths might be better suited for detecting subsurface irregularities or interactions between materials. Multi-wavelength or hyperspectral imaging goes a step further by combining multiple observations into a rich, multidimensional view of the chip.
These new methods do not just help monitor the process. They enhance it in real time. Engineers can dynamically adjust exposure, compensate for variations, and fine-tune materials interactions based on insights derived from different wavelengths.
Reinventing Resolution Without Shrinking
Traditional lithographic scaling was all about physical resolution, pushing the limits of optics to pattern ever-smaller shapes. But resolution can also be functional, not just physical. Spectral techniques allow engineers to refine the accuracy, fidelity, and predictability of existing lithographic processes.
It matters because, as we enter the sub-5nm world, physical scaling alone can no longer guarantee improvement. Random variation, or stochastics, begins to dominate. Errors are becoming more frequent. Process windows shrink. That is where better sensing and monitoring, powered by spectral diversity, comes into play.
Instead of trying to solve the problem with higher numerical aperture optics or new lithography equipment, engineers can reshape how precision is defined. It expands the limits of current technologies by understanding them better.
Scaling by Understanding, Not Just Building
What makes spectral engineering particularly compelling is that it aligns perfectly with the industry’s broader shift toward system-level thinking. Scaling is no longer about transistor size. It is about the entire ecosystem, including design tools, materials, software, and measurement strategies.
That is where perception plays a crucial role. If a device performs better, lasts longer, or supports new capabilities, even if its transistors have not shrunk, then most would agree that innovation has occurred. Consumers rarely question the method. They care about outcomes.
Erik Hosler emphasizes, “So long as that expectation is met, then Moore’s Law, in a sense, is still alive.”
This insight reframes the engineering conversation. Success is not about strictly following a scaling roadmap. It is about delivering a consistent and compelling user experience, however that may be achieved.
Integrating Light and Learning
Spectral strategies do not exist in isolation. They often work best when combined with other technologies like machine learning, which can analyze the vast amounts of data generated from multi-wavelength inspection. AI-assisted metrology, for instance, can spot subtle correlations across different spectral bands and flag process drifts before they become defects.
This integration of optics, data science, and materials engineering is where the real value emerges. Together, these elements form a feedback loop that improves every stage of the chip production cycle, from design to final inspection.
Multicolor lithography is another example. It allows engineers to use multiple exposure wavelengths in the same process step, introducing another degree of freedom. While still in initial stages, this approach could eventually enhance feature contrast and minimize edge placement errors, enabling more intricate designs without relying on more aggressive physical scaling.
A Broader Definition of Progress
As with many post-Moore’s Law strategies, the value of spectral engineering lies in its ability to redefine progress. Instead of chasing ever-smaller nodes at all costs, the industry is now exploring more intelligent ways to innovate by measuring better, understanding more, and adapting quickly.
This new philosophy embraces the idea that better tools and smarter data can take us further than raw scaling. With spectral engineering, the chip fabrication process becomes not just more precise, but more adaptive. It learns, responds, and develops, much like the devices it produces.
In this context, the pursuit of Moore’s Law becomes less about physics and more about creative problem-solving. How do we continue to deliver value? How do we squeeze more from the same space? Spectral engineering helps answer those questions in practical, measurable ways.
When Light Leads the Way
The death of Moore’s Law has been declared many times. However, what we are really witnessing is a transformation. The industry is moving from linear, physics-driven scaling to multidimensional, intelligence-driven optimization, and spectral engineering plays a vital role in that transformation.
By leveraging a broader range of light, engineers can better detect, understand, and control the microcosmic world of semiconductors. It does not just preserve the pace of innovation. It enhances it, offering new routes forward that were previously inaccessible. And most importantly, it keeps expectations alive. As long as the chips keep getting smarter, faster, and more efficient, Moore’s Law, in spirit if not in strict definition, remains relevant.
