High coherence fluxonium manufactured with a wafer-scale uniformity process

Fei Wang, Kannan Lu, Huijuan Zhan, Lu Ma, Feng Wu, Hantao Sun, Hao Deng, Y. Bai, Feng Bao, Xu Chang, Ran Gao, Xun Gao, Guicheng Gong, Lijuan Hu, Ruizi Hu, Honghong Ji, Xizheng Ma, Liyong Mao, Zhijun Song, Chengchun Tang, Hongcheng Wang, Tenghui Wang, Ziang Wang, Tian Xia, Hongxin Xu, Ze Zhan, Gengyan Zhang, Tao Zhou, Mengyu Zhu, Qingbin Zhu, Shasha Zhu, Xing Zhu, Yaoyun Shi, Hui‐Hai Zhao, Chunqing DengarXiv 2024 · DOI · arXiv

TL;DR

Researchers have developed a novel "overlap process" for fabricating fluxonium qubits that achieves nearly 100% yield and remarkable uniformity across entire wafers. By moving away from traditional angle-evaporation techniques, they produced fluxonium qubits with energy relaxation times ($T_1$) exceeding 1 millisecond, proving that high-performance fluxonium is ready for large-scale, CMOS-compatible manufacturing.

What Problem Does This Paper Address?

While the transmon is currently the "industry standard" superconducting qubit, the fluxonium qubit is a formidable challenger. It offers much longer coherence times and higher anharmonicity (which makes it easier to distinguish between the 0 and 1 states).

However, fluxonium has a "fabrication tax." To work, it requires a superinductor—typically an array of about 100 Josephson junctions (JJs) connected in series. Traditional fabrication uses "double-angle evaporation," where metal is evaporated at different angles through a suspended mask. This method is notoriously difficult to scale: it’s hard to keep junction sizes uniform across a large wafer, and the resulting "shadow" structures are fragile. To build a quantum computer with thousands of qubits, we need a fabrication process that is robust, uniform, and compatible with standard semiconductor (CMOS) manufacturing.

Methods

The authors introduced an overlap junction process. Unlike angle evaporation, this method involves:

  1. Two-Step Patterning: The bottom and top electrodes are created in two entirely separate lithography and deposition steps.
  2. Vacuum Break & Cleaning: Crucially, the vacuum is broken between steps. To ensure a high-quality interface, they use a calibrated Argon (Ar) ion milling process to strip native oxides off the bottom aluminum electrode before forming the aluminum oxide barrier and depositing the top electrode.
  3. Compact JJA Design: They utilized a compact Josephson Junction Array (JJA) geometry. By shrinking the loop size of the array, they reduced the qubit's susceptibility to magnetic flux noise from surface defects.
  4. Material Choice: They used tantalum as the base metal on sapphire substrates, a combination known for low dielectric loss.

Datasets

As an experimental physics paper, the "datasets" consist of fabrication statistics and cryogenic measurements:

  • Uniformity Data: 1,408 resistance measurements across 88 test dies on a 4-inch wafer to verify yield and resistance spread.
  • Qubit Metrics: Detailed coherence and parameter characterization for 8 specific fluxonium qubits (labeled A through H).
  • Comparison Data: Results from a "Manhattan-style" JJA layout were compared against the "compact" layout to isolate the effects of geometry on flux noise.

Key Insights and Ideas

  • Breaking Vacuum is Fine: A common fear in qubit fabrication is that breaking vacuum between layers introduces contaminants. This paper shows that with precise Ar ion milling, you can create interfaces that support millisecond coherence times.
  • Geometry Matters for Noise: The move to a "compact" JJA design was vital. The researchers found that Manhattan-style layouts had significantly higher flux noise ($A_\Phi > 3.7 \mu\Phi_0/\sqrt{\text{Hz}}$) compared to the compact design ($\sim 2.4 \mu\Phi_0/\sqrt{\text{Hz}}$). They attribute this to the compact design's smaller loop perimeter, which picks up less noise from surface spins.
  • Refining the Loss Model: By measuring $T_1$ across a wide range of frequencies, the authors showed that the Two-Level System (TLS) model (where loss is temperature-independent) fits the data better than traditional frequency-dependent dielectric loss models. This provides a clearer roadmap for further improving coherence.

Key Results

  • Yield & Uniformity: The process achieved ~100% yield. The relative standard deviation (RSD) in junction resistance was < 5% for the small phase-slip junction and < 2% for the entire 100-junction array across a 2-inch wafer.
  • Record Coherence: The "best" device (Qubit G) reached an energy relaxation time $T_1 = 1.168$ ms and a dephasing time $T_{2,echo} = 0.943$ ms at the flux frustration point (197 MHz).
  • Low Noise: They achieved state-of-the-art dielectric loss tangents ($\tan \delta_C \approx 10^{-6}$) and low flux noise levels, comparable to the best results achieved via much more difficult fabrication methods.

Limitations and Future Work

While this is a major step toward scalability, a few hurdles remain:

  • Lithography: This study used e-beam lithography and "lift-off" techniques, which are standard in labs but not ideal for high-volume industrial CMOS fabs. The authors suggest that future work should move toward photolithography and subtractive etching.
  • TLS Fluctuations: They observed occasional "dips" in $T_1$ caused by Two-Level Systems (defects) drifting into resonance with the qubit. While they could mitigate this by shifting the qubit's "idle" frequency, a more permanent solution involves further material science improvements to eliminate these defects entirely.
  • Scaling to 300mm: The current work was demonstrated on 2-inch and 4-inch wafers; the next logical step is transitioning the overlap process to industrial 300mm (12-inch) silicon or sapphire wafers.

Summary updated

Abstract

Fluxonium qubits are recognized for their high coherence times and high operation fidelities, attributed to their unique design incorporating a superinductor, which is typically implemented using an array of over 100 Josephson junctions. However, this complexity poses significant fabrication challenges, particularly in achieving high yield and junction uniformity with traditional methods. Here, we introduce an overlap process for Josephson junction fabrication that achieves nearly 100% yield and maintains uniformity across a 2-inch wafer with less than 5% variation for the phase slip junction and less than 2% for the entire junction array. We use a compact junction array design that achieves state-of-the-art dielectric loss tangents and flux noise levels, as confirmed by multiple devices. This enables fluxonium qubits to reach energy relaxation times exceeding 1 millisecond at the flux frustration point. This work paves the way for scalable high coherence fluxonium processors using CMOS-compatible processes, marking a significant step towards practical quantum computing.
BibTeX
@misc{wang2024high,
  author = {Fei Wang and Kannan Lu and Huijuan Zhan and Lu Ma and Feng Wu and Hantao Sun and Hao Deng and Y. Bai and Feng Bao and Xu Chang and Ran Gao and Xun Gao and Guicheng Gong and Lijuan Hu and Ruizi Hu and Honghong Ji and Xizheng Ma and Liyong Mao and Zhijun Song and Chengchun Tang and Hongcheng Wang and Tenghui Wang and Ziang Wang and Tian Xia and Hongxin Xu and Ze Zhan and Gengyan Zhang and Tao Zhou and Mengyu Zhu and Qingbin Zhu and Shasha Zhu and Xing Zhu and Yaoyun Shi and Hui‐Hai Zhao and Chunqing Deng},
  title = {{High coherence fluxonium manufactured with a wafer-scale uniformity process}},
  year = {2024},
  eprint = {2405.05481},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2405.05481},
}

Loading related content...

Figure