A summary of the structural, chemical, and physical property differences observed between qubits on a single chip in this study. The technique used to measure each feature or property is provided in parentheses. The only three materials-level variations found to correlate with device-to-device coherence spread are indicated in red and techniques identified as potential proxies for predicting qubit performance are indicated in blue.
Image credit: Fermilab
The Science
Continued improvements in scaling of superconducting quantum platforms requires reducing the wide performance variations frequently observed between neighboring qubits on a single chip. In a coordinated blind study of 22 transmon qubits across 7 chips, researchers tracked structural, chemical, and physical variations using specialized characterization tools after measuring qubit lifetimes (T1). The study successfully mapped a hierarchy of microwave losses.
The primary performance driver is the substrate trench depth, which represents the vertical distance of the recesses etched into the underlying substrate immediately adjacent to the edges of the superconducting metal electrodes. Specifically, at depths below ~20nm, small variations dramatically modulate the electric field stored in lossy niobium (Nb) oxides (Nb2O5). At larger trench depths where this effect saturates (>70nm), secondary structural features such as the sidewall footer angle (α) dominate. Electromagnetic simulations confirm that a broader, more tapered decline (α ~ 30 degrees) increases the electric field stored in lossy oxides by 20%–30% compared to a sharp angle (α ~ 10 – 15 degrees). For bare Nb devices, non-uniform growth of the lossy native surface oxide acts as an additional independent source of performance spread.
The Impact
This work provides actionable engineering solutions for the fabrication of multi-qubit processors with predictable coherence times across chips. To desensitize qubits from these localized material variations, we suggest that future fabrication processes incorporate deep substrate trenches (> 70 nm), highly directional/anisotropic etching to maintain low α, and immediate surface encapsulation to prevent non-uniform niobium surface oxidation. Furthermore, the study identifies terahertz nanoimaging and nanospectroscopy and low-temperature magneto/optical imaging as powerful, non-destructive proxy tools capable of predicting qubit performance levels at cryogenic temperatures without requiring full microwave measurements at mK temperatures.
Summary
Precise measurements of the qubit quality factor across 22 superconducting transmon qubits on 7 distinct chips revealed significant device-to-device performance variation among neighboring, geometrically identical devices. A comprehensive blind study incorporating nanoscale structural, chemical, and physical characterization was developed to build a clear hierarchy of microwave loss. It is shown that variations in the substrate trench depth are the dominant source of performance spread below a 20nm threshold, where minute depth fluctuations directly modulate the electric field participation at the metal sidewalls. When trench depths exceed 70nm and this effect saturates, variations in the sidewall footer angle (α) take over as the primary performance driver, with broader, more tapered edge profiles (α ~ 30 degrees) increasing surface energy participation by 20%–30% relative to sharp angles (α ~ 10-15 degrees). Moreover, it is possible for spontaneous, non-uniform native oxide growth (Nb2O5) to act as an independent source of decoherence on unencapsulated niobium devices, a mechanism that can be suppressed via structural surface encapsulation. Importantly, the study identifies that terahertz nanoimaging and nanospectroscopy and low-temperature magneto/optical imaging can serve as predictive proxy tools to capture localized electric field concentrations at sharp boundaries. This approach provides clear, actionable engineering targets for fabrication optimization, which is crucially important for eliminating coherence variations and achieving reproducible performance across multiqubit chips. In addition, the entire transmon device geometry can be screened non-destructively at low temperatures using these proxy techniques, thus enabling an efficient quality control protocol prior to mK microwave testing.
Institutions
Fermi National Accelerator Laboratory, Rigetti Computing, Ames National Laboratory, NIST Boulder, Northwestern University, National Physics Laboratory
Citation
Akshay A. Murthy, Mustafa Bal, Michael J. Bedzyk, Hilal Cansizoglu, Randall K. Chan, Venkat Chandrasekhar, Francesco Crisa, Amlan Datta, Yanpei Deng, Celeo D. Matute Diaz, Vinayak P. Dravid, David A. Garcia-Wetten, Sabrina Garattoni, Sunil Ghimire, Dominic P. Goronzy, Sebastian E. de Graaf, Samuel Haeuser, Mark C. Hersam, Peter F. Hopkins, Dieter Isheim, Kamal Joshi, Richard Kim, Saagar Kolachina, Cameron J. Kopas, Matthew J. Kramer, Ella O. Lachman, Jaeyel Lee, Peter G. Lim, Andrei Lunin, William Mah, Jayss Marshall, Josh Y. Mutus, Jin-Su Oh, David I. Olaya, David P. Pappas, Joong-mok Park, Ruslan Prozorov, Roberto dos Reis, David N. Seidman, Zuhawn Sung, Makariy Tanatar, Mitchell J. Walker, Jigang Wang, Maxwell Wisne, Haotian Wu, Lin Zhou, Shaojiang Zhu, Anna Grassellino, and Alexander Romanenko Identifying materials-level sources of performance variation in superconducting transmon qubits. Appl. Phys. Rev. 13, 021416 (2026). https://doi.org/10.1063/5.0299931
Funding Acknowledgement
This work was supported by the U.S. Department of Energy, Office of Science, National Quantum Information Science Research Centers, Superconducting Quantum Materials and Systems Center (SQMS), under Contract No. 89243024CSC000002. Fermilab is operated by Fermi Forward Discovery Group, LLC under Contract No. 89243024CSC000002 with the U.S. Department of Energy, Office of Science, Office of High Energy Physics.