Reducing TLS loss in tantalum CPW resonators using titanium sacrificial layers

The authors demonstrate that depositing an ultrathin titanium sacrificial layer on tantalum films acts as a solid-state oxygen getter to chemically reduce the native oxide interface, subsequently removed to yield tantalum coplanar waveguide resonators with internal quality factors exceeding 1.5 million, representing a threefold improvement over untreated devices.

The Gist

Imagine you are trying to build a super-precise musical instrument, like a violin, but instead of wood and strings, you are using a tiny piece of metal on a silicon chip to store energy. In the world of quantum computing, this "violin" is called a resonator, and its job is to hold onto a single particle of energy (a photon) without losing it. The better it holds onto that energy, the longer the quantum computer can think before it makes a mistake.

For a long time, scientists have used a metal called Tantalum (Ta) because it's very good at this job. However, even Tantalum has a flaw: when it touches the air, it instantly grows a thin, invisible layer of "rust" (oxide). Think of this rust not as a solid shield, but as a fuzzy, messy carpet full of tiny, sticky traps. These traps are called Two-Level Systems (TLS). Every time the energy tries to vibrate, it gets snagged by these sticky traps, causing the signal to fade away. This is called "loss."

The Problem: The Sticky Rust

The paper explains that while Tantalum's natural rust is better than other metals, it's still too messy. It creates too many of these sticky traps, limiting how long the quantum computer can stay "coherent" (focused). Scientists have tried to clean this rust off or cover it with a protective blanket (a "capping layer"), but these methods often leave behind a messy interface or add new problems.

The Solution: The "Sacrificial" Bodyguard

The researchers came up with a clever, temporary trick using a different metal: Titanium (Ti).

Think of the Titanium layer as a sacrificial bodyguard or a temporary shield.

1. The Setup: They take the Tantalum metal and lay down a tiny layer of Titanium on top of it. This layer is incredibly thin—only 2 atoms thick (about 2 Angstroms).
2. The Action: Titanium is like a hungry sponge for oxygen. As soon as the metal is exposed to air, the Titanium "eats" the oxygen before it can reach the Tantalum. Instead of the Tantalum growing its own messy, sticky rust, the Titanium reacts with the oxygen to change the chemistry of the surface. It essentially forces the Tantalum to grow a much smoother, cleaner, and less "sticky" surface layer.
3. The Removal: Once the device is built and the surface chemistry has been fixed, the scientists wash away the Titanium bodyguard using a special chemical bath (Buffered Oxide Etchant). The Titanium is gone, but the improvement it made to the Tantalum's surface remains.

The Result: A Clearer Signal

The paper reports that by using this "sacrificial" Titanium trick, they were able to clean up the surface significantly.

• Before: The standard Tantalum devices had an internal quality factor (a score for how well they hold energy) of about 0.4 to 0.5 million.
• After: The devices treated with the Titanium trick scored an average of 1.5 million, with some reaching over 2 million.

This means the new devices hold onto their energy three to four times longer than the old ones. It's like upgrading a violin string that was fraying and losing sound to a pristine, high-quality string that rings out clearly for much longer.

Why This Matters

The researchers found that this method works because it specifically targets the "rust" that forms where the metal meets the air. They also found that if you leave the Titanium on too long or don't wash it off completely, the device actually gets worse (because the Titanium itself can become a source of mess). But when done right—using a tiny layer, washing it off, and then heating the device gently—it creates a much cleaner surface.

In short, the paper demonstrates a simple, practical way to make quantum circuits "sing" longer and clearer by using a temporary, hungry metal layer to clean up the surface before the final product is finished. This doesn't require changing the whole design of the computer; it just tweaks the surface chemistry to reduce the "sticky traps" that cause errors.

Technical Summary

Technical Summary: Reducing TLS Loss in Tantalum Resonators Using Titanium Sacrificial Layers

Problem Statement
Superconducting quantum circuits, particularly those utilizing tantalum (Ta), have achieved significant coherence times, yet they remain limited by dielectric loss from two-level systems (TLS). While Ta offers a chemically stable native oxide superior to niobium, the amorphous nature of the native Ta pentoxide (a-Ta2O5) at the metal–air (MA) interface continues to generate TLS dipoles that degrade the internal quality factor ($Q_i$) of resonators and qubits. Previous strategies, such as capping layers (e.g., gold or magnesium oxide), have shown modest success but often introduce additional interfaces or fail to prevent oxide regrowth during fabrication. Furthermore, standard etching techniques can remove surface oxides but cannot prevent their spontaneous re-oxidation, leaving the MA interface as a primary source of decoherence.

Methodology
The authors propose a surface engineering approach utilizing an ultrathin (2 Å) titanium (Ti) sacrificial layer to chemically modify the Ta oxide at the MA interface. The fabrication workflow involves:

1. Substrate Preparation: High-resistivity silicon substrates are prepared with 200 nm sputtered $\alpha$-Ta films.
2. Sacrificial Layer Deposition: Native Ta oxide is removed via Ar$^+$ ion beam milling, followed by the in situ deposition of a 2 Å Ti film atop the pre-sputtered Ta.
3. Fabrication: Standard positive-tone optical lithography and CF$_4$-based reactive ion etching are used to pattern coplanar waveguide (CPW) resonators. The Ti layer acts as a kinetic barrier and oxygen getter during these steps.
4. Post-Processing:
   • Buffered Oxide Etchant (BOE): A long BOE treatment (20 minutes) is applied to completely remove the Ti layer, leaving behind a chemically reduced Ta oxide surface.
   • Annealing: Samples undergo high-temperature annealing (700°C for 10 hours) to further suppress TLS loss.
5. Characterization: The team fabricated a matrix of devices varying in Ti inclusion (2 Å), BOE duration (1 min vs. 20 min), and annealing (yes/no). Resonators were characterized via power-dependent and temperature-dependent $S_{21}$ measurements to extract $Q_i$ and TLS loss tangents. X-ray Photoelectron Spectroscopy (XPS) was used to verify the chemical state of the surface at various stages.

Key Contributions

• Novel Surface Modification: The paper introduces a method where a removable Ti layer acts as a solid-state oxygen getter, preferentially reacting with ambient oxygen to alter the formation dynamics of the native Ta oxide, rather than acting as a permanent capping layer.
• Process Integration: The method is fully compatible with existing thin-film deposition and microfabrication workflows, requiring no changes to qubit architecture or circuit layout.
• Chemical Control: The study demonstrates that the Ti layer must be fully removed (via long BOE) to be effective; residual Ti leads to increased loss, confirming the layer's role as a temporary modifier rather than a permanent interface.

Results

• Quality Factor Enhancement: Resonators treated with the 2 Å Ti layer, long BOE, and annealing (Sample T-LA2) achieved an average single-photon internal quality factor ($Q_i$) exceeding 1.5 million, with a maximum of 2.14 million.
• Comparative Performance: These results represent a 2–4 fold increase in $Q_i$ compared to identical devices lacking the Ti layer (e.g., Sample R-LA, which had $Q_i \approx 0.57$ million).
• TLS Loss Reduction: The filling-factor-adjusted TLS loss tangent ($F\delta_{TLS}$) was reduced from $1.11 \times 10^{-6}$ (bare Ta, annealed) to $0.58 \times 10^{-6}$ (Ti-treated, annealed).
• High-Power Performance: At high photon numbers ($\langle n \rangle \ge 10^6$), the Ti-treated devices showed a $Q_i$ exceeding 13 million, a 3–9 fold improvement over reference devices.
• Robustness: The process mitigated potential damage from the initial argon ion milling, with un-annealed Ti/Ta samples outperforming bare Ta controls, suggesting the Ti layer protects the surface during fabrication.
• Temperature Dependence: Measurements confirmed that at operating temperatures (~25 mK), losses are dominated by TLS rather than thermal quasiparticles, which only become significant above 400 mK.

Significance and Claims
The paper claims that this method highlights the critical role of interfacial oxide chemistry in superconducting loss. By directly addressing the MA interface through atomic-scale surface engineering, the authors provide a practical route to extending $T_1$ lifetimes in superconducting qubits without altering device architectures. The results reinforce that reducing TLS loss is achievable through materials-based approaches that are scalable and compatible with standard fabrication. The authors note that while the current process involves a high thermal budget (annealing), future work could explore thicker Ti layers for better process tolerance and reduced thermal steps. The method is presented as a direct, scalable alternative to circuit-level optimization strategies for enhancing qubit coherence.