🔬 physics

Enhanced electron-beam lithography to reduce the frequency scatter of 200-400 GHz superconducting microstrip resonators for on-chip filterbank spectrometers

This paper demonstrates that optimizing electron-beam lithography by reducing the beam step size and avoiding main-field stitching significantly reduces both random and systematic frequency scatter in 200–400 GHz superconducting microstrip resonators, thereby enabling the high spectral resolution required for next-generation integrated superconducting spectrometers.

Original authors: L. G. G. Olde Scholtenhuis, D. J. Thoen, J. J. A. Baselmans, K. Karatsu, L. H. Martin, S. Vollebregt, A. Endo

Published 2026-06-23

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: L. G. G. Olde Scholtenhuis, D. J. Thoen, J. J. A. Baselmans, K. Karatsu, L. H. Martin, S. Vollebregt, A. Endo

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to build a massive, high-tech library of sound. In this library, every book represents a specific slice of the universe's "soundtrack" (light waves coming from distant galaxies). To read these books, you need a row of tiny, perfectly tuned radio receivers (filters) sitting side-by-side. Each receiver is designed to listen to one specific note, and they need to be spaced out so precisely that there are no gaps between them. If there's even a tiny gap, you miss a whole chapter of the story.

The scientists in this paper were trying to build this library for the "DESHIMA" project, which aims to listen to the universe using superconducting chips. However, they hit a wall: their receivers were slightly out of tune. Sometimes they were too close together, and sometimes too far apart. This "frequency scatter" meant they were missing parts of the cosmic story, limiting how clear their picture of the universe could be.

The Problem: The "Pixel" Grid

To build these tiny receivers, the team uses a machine called an electron-beam lithography system. Think of this machine as a super-precise paintbrush that draws the circuits onto a silicon chip.

The paper identifies two main ways this "paintbrush" was messing up the tune:

The "Seam" Problem (Mainfield Stitching):
The machine can't draw a huge picture in one go. It has to draw it in small squares (called "mainfields") and then stitch them together, like a quilt.
- The Analogy: Imagine drawing a long line across a quilt. If you stop at the edge of one square and start again on the next, your hand might shake slightly, or the squares might not line up perfectly. The line ends up crooked.
- The Result: In their first device, some filters crossed these "seams." The filters on one side of the seam were shifted to a lower pitch than the ones on the other side. It was like having two rows of instruments where one row was slightly out of tune with the other.
The "Step" Problem (Beam Step Size):
The machine doesn't draw a smooth, continuous line; it places tiny dots very close together to create the illusion of a line. The distance between these dots is the "Beam Step Size" (BSS).
- The Analogy: Imagine trying to draw a smooth circle using only a ruler and a pencil. If you take big steps (large BSS), your circle looks jagged and blocky. If you take tiny, microscopic steps (small BSS), the circle looks perfectly smooth.
- The Result: In their first device, the "steps" were too big (6.25 nanometers). This made the shape of the filters slightly "jagged" and inconsistent, causing random variations in their pitch.

The Solution: Smoother Steps and No Seams

The team built three new devices to test their theories.

Device 1: Had the "seam" problem (filters crossed the stitching lines) and the "big step" problem.
Device 2: They fixed both issues. They rearranged the design so no filters crossed the seams, and they told the machine to take tiny, tiny steps (reducing the step size from 6.25nm to just 1.00nm).
Device 3: A copy of Device 2 to prove the results weren't a fluke.

The Results

The improvement was dramatic. By making the "steps" smaller and avoiding the "seams":

The Systematic Shift Disappeared: The difference in pitch between the two sides of the chip vanished. The "seam" was indeed the culprit for the big, predictable error.
The Random Jitter Vanished: The random, messy variations in pitch dropped by about 70%. The "jagged" lines became smooth, and the filters were much more consistent.

Why This Matters

The paper concludes that to build the next generation of these super-sensitive cosmic listeners, you need nanometer-scale control. You can't just "guess" the dimensions; you have to control the drawing process with extreme precision.

By fixing the drawing tool (the lithography), they improved the "tuning" of the filters by a factor of four. This means future versions of these spectrometers can pack the filters much tighter together without leaving gaps. This allows astronomers to capture a continuous, unbroken view of the universe's history, rather than a fragmented one with missing pieces.

In short: They stopped the "paintbrush" from shaking and stopped the "quilt" from having crooked seams, resulting in a perfectly tuned library of the cosmos.

Technical Summary: Enhanced Electron-Beam Lithography for Superconducting Microstrip Resonators

Problem Statement
Integrated superconducting spectrometers (ISSs) are critical for submillimeter-wave astronomy, offering scalable architectures for large-scale spectroscopic surveys of dusty star-forming galaxies. However, the spectral resolution and completeness of these devices, particularly those utilizing on-chip filterbanks, are currently limited by the precision of resonant frequency spacing. In practice, frequency scatter often results in a resolution of $F/\Delta F < 500$ , creating gaps between adjacent filter passbands that lead to spectral incompleteness. While numerous factors (layer thickness, kinetic inductance, etching profiles) can shift resonance frequencies, the specific physical origins of the frequency scatter in superconducting microstrip filters remain poorly understood. This uncertainty hinders the realization of next-generation spectrometers capable of matching the emission-line widths of massive galaxies ( $F/\Delta F \approx 300–3000$ ).

Methodology
To isolate the causes of frequency scatter, the authors fabricated three 200–400 GHz filterbank spectrometers (Devices D1, D2, and D3) with identical filterbank designs but systematically varied electron-beam lithography (EBL) parameters. The devices utilize NbTiN/Al hybrid kinetic inductance detectors (KIDs) coupled to I-shaped microstrip bandpass filters.

The study focused on two specific EBL parameters:

Mainfield Layout: Large patterns are divided into "mainfields" stitched together by moving the substrate stage. Misalignment at these boundaries can distort geometries.
- Device D1: One side of the filterbank (Side B) had mainfield boundaries crossing the filter patterns, while the other side (Side A) contained filters fully within single mainfields.
- Device D2 & D3: All filters were positioned entirely within single mainfields, eliminating mainfield-edge crossings.
Beam Step Size (BSS): The grid spacing on which the beam places exposure points.
- Device D1: BSS set to 6.25 nm.
- Device D2: BSS reduced to 1.00 nm (increasing write time by a factor of ~40).
- Device D3: A copy of D2 fabricated to verify reproducibility.

Post-fabrication, the devices were measured at 120 mK. The frequency response was characterized by measuring the KID phase response to a single mm-wave signal generated by a photomixing continuous-wave terahertz source. The transmission of individual filters was normalized against a wide-band KID response to calculate the Normalized Spacing Ratio (NSR), defined as the ratio of the measured frequency gap to the designed gap. An NSR of 1 indicates a perfect match to the design.

Key Results
The analysis of the NSR distributions revealed two distinct modes of frequency deviation:

Systematic Frequency Shift (Mainfield Stitching):
In Device D1, a significant systematic offset was observed between Side A and Side B. Filters on Side B (crossing mainfield boundaries) were shifted to lower frequencies relative to Side A. The absolute difference in mean frequency between the sides was $|\mu_A - \mu_B| = 1.831$ . This indicates that mainfield stitching introduces a deterministic geometric distortion.
Random Frequency Scatter (Beam Step Size):
The standard deviation of the frequency scatter ( $\sigma$ ) was significantly reduced by optimizing the BSS.
- In D1 (6.25 nm BSS), the average standard deviation was $\sigma \approx 0.53$ .
- In D2 (1.00 nm BSS), the standard deviation decreased by approximately 70% to $\sigma \approx 0.16$ .
- Device D3 confirmed these results with $\sigma \approx 0.14$ , demonstrating that the improvement is reproducible and not due to uncontrolled fabrication variations.

By eliminating mainfield crossings and reducing the BSS, Device D2 achieved a negligible systematic offset ( $|\mu_A - \mu_B| = 0.003$ ) and a four-fold improvement in overall frequency spacing precision compared to the initial configuration.

Significance and Claims
The paper demonstrates that lithography-induced frequency scatter is the dominant factor limiting spectral completeness in current ISS implementations. The authors conclude that:

Mainfield stitching is the root cause of systematic frequency shifts between groups of resonators.
Beam step size (BSS) on the nanometer scale is the primary driver of random frequency scatter.
Nanometer-scale control of lithographic processes is imperative for realizing next-generation integrated superconducting spectrometers with higher spectral resolution.

The authors note that with the reduced scatter observed in D2 ( $\sigma \approx 0.16$ ), the devices approach the tolerable limit ( $\sigma_{max} \approx 0.2$ ) required to ensure that the minimum absorbed power across the entire band exceeds two-thirds of the peak absorption for 95% of the filters. This improvement is essential for enabling continuous spectral coverage without gaps, thereby matching the spectral resolution requirements for high-redshift galaxy evolution studies.

The Problem: The "Pixel" Grid

The Solution: Smoother Steps and No Seams

The Results

Why This Matters

Technical Summary: Enhanced Electron-Beam Lithography for Superconducting Microstrip Resonators

More like this