In our recent paper, “Diffraction-limit-breaking digital projection lithography via multi-exposure strategies for high-density nanopatterning”, we report a digital projection lithography strategy based on alignment-free multiple exposures and spatially uniform layout decomposition. The work demonstrates that high-density nanostructures—previously considered impractical for DMD-based projection systems—can be fabricated with high fidelity and a wide process window.

While the final results appear straightforward, the path to this work was shaped by a long-standing frustration with a problem many of us had quietly accepted as unavoidable.

Figure 1 | Schematic diagram of multiple exposure digital projection lithography system and layout breakdown

The Starting Point: When Density Becomes the Enemy

Digital micromirror device (DMD)–based projection lithography has long been valued for its flexibility, speed, and maskless operation. Yet in practice, we repeatedly encountered the same bottleneck:

As pattern density increases, image fidelity collapses.

When feature spacing approaches the diffraction limit, high spatial-frequency information is inevitably lost during projection. What follows is familiar to anyone working with dense layouts—pattern merging, extreme depth-of-focus sensitivity, and vanishing process margins. Incremental optical optimization did little to help. At some point, it became clear that this was not an engineering oversight, but a fundamental limitation of single-exposure thinking.

Instead of asking how to further sharpen a single exposure, we began asking a different question:

Do we really need to project all spatial frequencies at once?

The Key Insight: Separating Information in Time, Not Space

The conceptual turning point came from a simple observation.
While the optical system filters spatial frequencies, the photoresist integrates exposure dose over time.

This led us to a counterintuitive idea: if a dense pattern is decomposed into multiple sparse sub-layouts—each individually compatible with the optical transfer function—could the resist effectively “reconstruct” the full pattern through sequential exposure?

Crucially, DMD systems offer a unique advantage here. Because the micromirror array is digitally addressed and mechanically static, multiple exposures can be performed without any physical alignment. This meant that, in principle, temporal multiplexing could replace spatial overload—without introducing alignment errors.

What initially sounded like a conceptual workaround gradually revealed itself as a fundamentally different way of thinking about digital lithography.

Beyond Exposure Count: Why Uniformity Matters

Early experiments with simple multi-exposure schemes produced mixed results. While resolution improved, fabrication consistency remained unstable. Some regions printed cleanly; others collapsed with minor defocus.

This inconsistency forced us to look deeper—not at optics, but at layout statistics.

We realized that even after decomposition, non-uniform spacing within sub-layouts leads to spatially varying depth-of-focus and dose sensitivity. In other words, not all sparse patterns are equally printable.

This motivated the development of the gradient-descent-based layout decomposition with spatial uniformity (GD-LDSU) algorithm. By explicitly optimizing spacing uniformity while resolving minimum-distance violations, we could systematically stabilize the process window. Implementing the algorithm on GPU hardware allowed us to scale it to realistic, large-area layouts.

At this stage, the work transitioned from an exposure trick into a coupled algorithm–process framework.

Figure 2 | Results of multiple exposure experiments on two-dimensional chip metal layer

From Periodic Lines to Real Chip Layouts

Validation was critical. Periodic line arrays provided a clean benchmark: multi-exposure reduced the minimum resolvable pitch from ~0.5λ/NA to ~0.3λ/NA—well beyond conventional expectations for projection lithography.

But the real test came with two-dimensional, high-density chip metal layouts. Achieving a minimum gap of a single DMD pixel (75.6 nm) without pattern collapse was a moment we distinctly remember—not because it broke a record, but because it confirmed that the strategy worked under realistic, irregular geometries.

Equally important was what did not happen: the process window did not collapse. Dose tolerance remained surprisingly robust, even for the densest regions.

Reflections: Limits Are Often Architectural, Not Physical

Looking back, this work reinforced a lesson we have encountered repeatedly in micro- and nanofabrication:

Many perceived “limits” are not imposed by physics alone, but by the structure of how we pose the problem.

By shifting from single-exposure optimization to multi-exposure information management—and by treating layout decomposition as a process variable rather than a preprocessing step—we found room to maneuver within constraints that once felt rigid.

Looking Ahead

We believe this strategy opens several directions worth exploring:
from adaptive exposure sequencing and layout-aware process control, to integration with advanced photoresists and hybrid projection systems. More broadly, it suggests that temporal degrees of freedom in digital lithography remain largely underexploited.

High-density patterning will only become more demanding.
Sometimes, progress begins not by pushing harder against a limit—but by stepping around it.

Paper Information
Zi-Xin Liang, Jing-Tao Chen, Yuan-Yuan Zhao*, Wen-Hui Li, Jing Zhou, Xuan-Ming Duan*
Diffraction-limit-breaking digital projection lithography via multi-exposure strategies for high-density nanopatterning
Microsystems & Nanoengineering, 12, 18 (2026)
DOI: 10.1038/s41378-025-01131-x