Fabrication effects on Niobium oxidation and surface contamination in Niobium-metal bilayers using X-ray photoelectron spectroscopy

This study utilizes X-ray photoelectron spectroscopy to non-destructively evaluate 17 metal capping layers for their ability to prevent niobium oxidation and surface contamination during fabrication processes, ultimately identifying resilient candidates that improve the performance of superconducting resonators.

The Gist

Imagine you are trying to build a super-fast, super-precise computer that runs on the laws of quantum mechanics. To make this work, you need a special kind of "wire" made of a metal called Niobium. This wire is part of a tiny circuit called a qubit (the basic unit of a quantum computer).

However, there's a problem. Niobium is like a very hungry sponge. As soon as it touches the air, it instantly "eats" oxygen and forms a layer of rust (oxide) on its surface. In the world of quantum computers, this rust is a disaster. It acts like a noisy, sticky floor that steals energy from the computer, causing it to make mistakes.

The scientists in this paper asked a simple question: "How can we protect this hungry Niobium sponge from getting rusty while we build the computer?"

Here is their story, broken down into simple steps:

1. The "Hat" Strategy (Capping Layers)

To stop the Niobium from rusting, the researchers tried putting a "hat" on it. They took the Niobium and covered it with a very thin layer (5 nanometers thick—imagine a sheet of paper folded 10,000 times) of different materials.

They tested 17 different types of hats:

• Noble Metals: Like Gold, Platinum, and Palladium (think of these as fancy, expensive hats).
• Hard Metals: Like Titanium, Tungsten, and Tantalum (think of these as tough work boots).
• Ceramics/Nitrides: Like Titanium Nitride (think of these as ceramic tiles).
• Alloys: Mixtures of metals.

2. The "Stress Test" (Fabrication)

Building a quantum computer isn't just about putting the hat on; it's about what happens after. The researchers had to put these samples through the same rough-and-tumble processes used in real factories:

• The Sauna (Annealing): Heating the samples to 200°C to clean them.
• The Chemical Shower (Resist Stripping): Soaking them in hot, harsh chemicals to remove sticky glue used during manufacturing.
• The Acid Bath (Acid Cleaning): Dipping them in strong acids (like HF or Nanostrip) to scrub off any remaining dirt.

3. The "Magic X-Ray Glasses" (XPS)

How did they know if the Niobium underneath was safe? They used a tool called X-ray Photoelectron Spectroscopy (XPS).

Think of XPS as a pair of magic X-ray glasses.

• The glasses can see through the "hat" (the cap layer).
• They can tell if the Niobium underneath is still shiny metal or if it has turned into dull, rusty oxide.
• They can even tell how deep the rust has gone.

4. The Results: Who Passed and Who Failed?

The researchers found some surprising winners and losers:

• The Fancy Hats Failed: The Gold, Platinum, and Palladium hats looked great at first, but when the samples went into the "Sauna" (heating), oxygen sneaked right through the gaps in the hat and rusted the Niobium underneath. Verdict: Too porous for heat.
• The "Eroding" Hats: Some hats, like Aluminum and Zirconium, were so reactive that the "Chemical Shower" (resist stripping) ate the hat away entirely, exposing the Niobium to rust. Verdict: Too soft for chemicals.
• The Nitride Trouble: Hats made of Nitrides (like Titanium Nitride) were tricky. The X-ray glasses got confused because the hat itself looked a bit like rust, making it hard to tell if the Niobium was safe.
• The Superheroes:
  • Tantalum (Ta) and Tantalum Nitride (TaN): These were the champions. They didn't let oxygen through during heating, and they survived the acid baths without getting damaged. They kept the Niobium shiny and rust-free.
  • Molybdenum (Mo) and Tungsten (W): These were also good, but only if you didn't use the specific "Chemical Shower" that eats them.

5. The Final Test: The Microwave Race

To prove their findings, they built actual microwave resonators (tiny circuits that vibrate like a guitar string) using the best hats (Tantalum and Tantalum Nitride).

They measured how much energy these circuits lost.

• The Control Group (Niobium with no hat) lost a lot of energy (high noise).
• The Tantalum Group lost significantly less energy. The "hat" worked! It kept the surface clean, and the quantum computer part ran much more efficiently.

The Big Takeaway

This paper is like a consumer report for quantum computer builders. It tells us:

1. Don't use Gold or Platinum hats if you plan to heat your chips; they let oxygen in.
2. Don't use Aluminum or Zirconium hats if you plan to use strong cleaning chemicals; they dissolve.
3. Use Tantalum or Tantalum Nitride. They are the tough, reliable hats that keep the Niobium safe from rust, no matter how rough the factory process gets.

By choosing the right "hat," we can build better, faster, and more reliable quantum computers.

Technical Summary

1. Problem Statement

Superconducting qubits and resonators, particularly those based on niobium (Nb), suffer from dielectric losses primarily attributed to surface oxides and two-level systems (TLS). While Nb is chemically resilient and has a high critical temperature, it forms a non-self-limiting, lossy oxide ($Nb_2O_5$) that degrades device performance.
Current mitigation strategies include capping layers, surface passivation, and acid treatments. However, standard fabrication processes—specifically annealing, resist stripping, and acid cleaning—can compromise the integrity of these capping layers, leading to oxygen diffusion, interface oxidation, and surface contamination. There is a lack of rapid, non-destructive methods to screen capping layers for their ability to withstand these specific fabrication steps while preserving the underlying Nb interface.

2. Methodology

The authors employed X-ray Photoelectron Spectroscopy (XPS) as a rapid characterization tool to evaluate 17 different capping layers on 60 nm Nb films. The study utilized the depth sensitivity of XPS (probing the top 5–10 nm) to non-destructively assess the oxidation state of the Nb-metal interface.

Experimental Workflow:

• Sample Preparation: 60 nm Nb films were deposited on high-resistivity silicon wafers, followed by in-situ sputtering of a 5 nm capping layer.
• Capping Layers Tested:
  • Metals: Al, Au, Hf, Mo, Pd, Pt, Ta, Ti, W, Zr.
  • Nitrides: NbN, ZrN, HfN, TiN, TaN.
  • Alloys: Ti-W, Zr-Y.
• Fabrication Stress Tests: Samples underwent three distinct processing steps on the day of measurement:
  1. Annealing: 200°C for 1 hour in both air and vacuum (< $5 \times 10^{-6}$ Torr) to simulate thermal processing and hydrocarbon removal.
  2. Resist Stripping: Soaking in an AZ 300T heated chemical strip bath (NMP-based with TMAH) at 80–90°C, followed by sonication and rinsing.
  3. Acid Cleaning: Dipping in 2% HF, 10:1 Buffered Oxide Etch (BOE), and Nanostrip (sulfuric/peroxymonosulfuric acid mixture).
• Device Validation: A subset of promising candidates (Ta, Zr, TaN, TiN) was used to fabricate superconducting resonators. These were measured in a dilution refrigerator (500 mK) to extract internal quality factors ($Q_i$) and loss tangents ($\delta$) at medium and high photon powers.

3. Key Contributions

• High-Throughput Screening: Demonstrated XPS as an effective, non-destructive method to rapidly screen capping layers for oxygen diffusion barriers and chemical resilience without fabricating full devices.
• Process-Specific Failure Analysis: Decoupled the effects of thermal annealing, chemical stripping, and acid cleaning on specific material classes, identifying which layers fail under which conditions.
• Correlation of Surface Chemistry to Device Performance: Linked specific surface contamination (C, N, Na, Ca, Si) and oxide states directly to resonator loss metrics.
• Identification of Optimal Materials: Narrowed down the vast material space to a few viable candidates for Nb capping in quantum devices.

4. Key Results

A. Oxidation and Annealing

• Noble Metals (Au, Pd, Pt): Performed poorly. Despite not oxidizing themselves, they failed to prevent oxygen diffusion through grain boundaries, leading to significant Nb oxidation after air annealing.
• Refractory Metals (Al, Hf, Mo, Ta, Ti, W, Zr): Generally effective at preventing Nb oxidation during annealing.
  • Zr: Formed a fully oxidized $ZrO_x$ layer but successfully prevented Nb oxidation underneath.
• Nitrides: Generally promising due to nitrogen filling vacancies. However, XPS fitting was complicated by the formation of Nb-Nitride peaks ($NbN_x$) overlapping with suboxide peaks.
  • HfN and TaN: Showed no new suboxide formation after annealing.
  • TiN and ZrN: Showed minor increases in suboxides.
  • NbN: Contained significant oxide, making it a poor candidate.

B. Chemical Strip Bath (AZ 300T)

• Erosion: Mo, W, and Ti-W films were significantly eroded by the heated strip bath, exposing and oxidizing the underlying Nb. These are incompatible with this process.
• Contamination: The primary contaminants were Carbon (C), Nitrogen (N), Sodium (Na), Calcium (Ca), and Silicon (Si).
  • Best Performers: Au, Pd, Zr, and Zr-Y showed the least surface contamination, suggesting they could improve resonator performance without requiring acid cleaning.
  • Unresolved F-contamination: Fluorine was detected on Hf and HfN samples despite no F-based processing, likely due to chamber background.

C. Acid Cleaning

• Damaged Layers: Al, Hf, Ti, Zr, HfN, and Zr-Y were etched/damaged by all three acids (HF, BOE, Nanostrip), exposing the Nb.
• Partial Survival: Mo, TiN, and W survived HF and BOE but were etched by Nanostrip.
• Resilient Layers: Tantalum (Ta) and Tantalum Nitride (TaN) showed no damage and no Nb oxidation after exposure to any of the three acid treatments.

D. Resonator Performance

Resonators were fabricated using Ta, TaN, TiN, and Zr caps:

• Ta and TaN: Exhibited lower medium-power loss ($\delta_{MP}$) than the uncapped control, confirming their effectiveness as diffusion barriers.
• TiN: Had comparable $\delta_{MP}$ to the control but showed higher high-power loss ($\delta_{HP}$), suggesting a higher density of GHz-frequency TLS.
• Zr: Showed higher loss than the control. Although it prevented Nb oxidation, the thick $ZrO_x$ layer itself likely introduced significant dielectric loss.

5. Significance and Conclusion

This study provides a critical roadmap for the fabrication of high-coherence superconducting quantum devices using Niobium.

• Material Selection: Tantalum (Ta) and Tantalum Nitride (TaN) are identified as the superior capping layers. They effectively block oxygen diffusion during annealing, withstand aggressive chemical stripping and acid cleaning, and result in resonators with lower dielectric loss.
• Process Integration: The findings indicate that noble metals (Au, Pd, Pt) are unsuitable as thin (5 nm) caps due to grain-boundary diffusion, and that Zr, while chemically stable, introduces too much dielectric loss via its own oxide.
• Methodological Impact: The work validates XPS as a vital, rapid feedback loop for materials science in quantum computing, allowing researchers to optimize fabrication protocols before committing to time-consuming device fabrication cycles.

The data suggests that for Nb-based qubits, the use of Ta or TaN capping layers, combined with standard acid cleaning and resist stripping, offers the most robust path toward minimizing surface-induced losses.