Oxide-nitride heteroepitaxy for low-loss dielectrics in superconducting quantum circuits

This paper demonstrates that heteroepitaxial $\gamma$-Al$_2$O$_3$ grown on TiN via pulsed laser deposition forms a high-quality, single-crystal dielectric with an intrinsically low two-level system loss of $(2.8 \pm 0.1) \times 10^{-5}$, establishing it as a promising materials platform for reducing dielectric losses in superconducting quantum circuits.

The Gist

The Big Picture: Building a Better Quantum Computer

Imagine you are trying to build a super-fast, super-smart computer that uses the laws of quantum physics to solve problems no regular computer ever could. This is called a superconducting quantum computer.

The "brain" of this computer is made of tiny circuits called qubits. Think of these qubits as incredibly sensitive dancers. To dance perfectly, they need a perfectly smooth, silent floor. If the floor is bumpy or noisy, the dancers trip, lose their rhythm, and the dance fails.

In the current technology, the "floor" these dancers stand on is made of a material called amorphous aluminum oxide. The problem is that this material is like a cracked, dusty sidewalk. It's disordered and full of tiny, invisible "gremlins" (scientists call them Two-Level Systems or TLS) that jump around and steal energy from the dancers, causing the computer to make mistakes.

The Solution: A Perfect Crystal Floor

The researchers in this paper asked a simple question: What if we built the floor out of a perfect, smooth crystal instead of a cracked sidewalk?

They decided to build a "sandwich" of materials using a technique called Pulsed Laser Deposition (PLD). Think of PLD like using a high-powered laser to zap a target, turning it into a fine mist that rains down onto a surface, building up layer by layer.

The Sandwich Recipe:

1. Bottom Bun: A layer of Titanium Nitride (TiN), which is a super-conducting metal (it conducts electricity with zero resistance).
2. The Filling: A layer of Aluminum Oxide (Al₂O₃), but specifically a crystalline version (γ-Al₂O₃). This is the "perfect floor."
3. Top Bun: Another layer of Titanium Nitride.

They grew this sandwich on a sapphire crystal base, ensuring every layer lined up perfectly with the one below it, like stacking perfectly aligned bricks.

The Challenge: Keeping the Layers Clean

Growing these layers is tricky. When you heat up metals and ceramics to grow them, they often start to "melt" into each other. Oxygen atoms from the ceramic filling might leak into the metal buns, or nitrogen might leak out.

If this happens, the "perfect floor" gets dirty, and the gremlins (TLS) come back.

The Team's Trick:
They used a special "cleaning" process before they started. They baked the machine and used a titanium target to act like a magnet for oxygen, sucking up any stray air molecules in the chamber. This ensured the metal layers stayed pure and didn't get oxidized.

The Results: A Super-Smooth Dance Floor

To check their work, they used powerful microscopes and X-rays (like a super-advanced medical CT scan) to look at the layers.

• The Good News: The layers were perfectly aligned (epitaxial). The metal and the ceramic met at sharp, clean boundaries with almost no mixing.
• The Chemical Check: They looked at the atoms and confirmed the materials were exactly what they were supposed to be, with very few "impurities."

The Big Test: Measuring the Loss

Now, they needed to see if this new "crystal floor" actually worked better. They built tiny microwave resonators (think of them as tiny tuning forks) using their new sandwich material.

They measured how much energy the "gremlins" stole from the system.

• Old Material (Amorphous AlOx): The gremlins were very active. The loss was high.
• New Material (Crystalline Al₂O₃): The gremlins were almost gone!

The Result: The new material had 100 times less loss than the old material.

• Analogy: If the old material was a noisy room where you had to shout to be heard, the new material is a library where you can whisper and be heard perfectly.

Why This Matters

This discovery is a game-changer for two reasons:

1. Better Qubits: Because the "floor" is so smooth, the quantum dancers (qubits) can dance for much longer without tripping. This means the computer can do more complex calculations before making a mistake.
2. Smaller Devices: Currently, these quantum computers are huge because the circuits need to be large to avoid interference. Because this new material is so clean, the researchers can make the circuits much smaller (like shrinking a house down to the size of a shoebox) without losing performance. This is essential for building computers with thousands of qubits, which is the goal for the future.

Summary

The team successfully built a "perfect crystal sandwich" for quantum computers. By using a laser to grow these layers and keeping them chemically pure, they created a material that is 100 times better at holding onto quantum information than what we use today. It's a major step toward building a practical, powerful quantum computer that fits in a room rather than a warehouse.

Technical Summary

1. Problem Statement

Superconducting (SC) qubits are a leading platform for fault-tolerant quantum computing, but their coherence times are currently limited by dielectric losses. The incumbent material for Josephson junction (JJ) tunnel barriers is amorphous aluminum oxide (AlOx).

• The Issue: Amorphous AlOx lacks long-range order and stoichiometry, hosting a high density of parasitic Two-Level Systems (TLS). These TLSs act as energy sinks, causing decoherence in qubits.
• The Challenge: While single-crystal dielectrics theoretically offer lower TLS densities, integrating them into SC circuits is difficult. Many candidates (e.g., GaAs, AlN) suffer from piezoelectricity or interdiffusion issues with superconducting electrodes (e.g., Niobium oxidizes easily). Furthermore, existing high-coherence designs often have large footprints, hindering the scaling required for multi-qubit quantum processors.

2. Methodology

The authors propose a heteroepitaxial trilayer stack: TiN / $\gamma$-Al$_2$O$_3$ / TiN, grown on an $\alpha$-Al$_2$O$_3$ (sapphire) substrate.

• Fabrication:
  • Deposition: All layers were grown via Pulsed Laser Deposition (PLD) in a single vacuum cycle to prevent contamination.
  • Substrate Prep: Sapphire substrates were annealed at 1100°C.
  • TiN Growth: A critical "baking" step was employed where the chamber was heated to 900°C while ablating a Ti target to getter residual oxygen and water, ensuring near-stoichiometric TiN growth.
  • Structure: Two samples were created with varying dielectric thicknesses:
    1. Thin: 62.2 nm TiN / 13.5 nm Al$_2$O$_3$ / 53.8 nm TiN.
    2. Thick: 63.0 nm TiN / 58.3 nm Al$_2$O$_3$ / 68.1 nm TiN.
• Characterization Techniques:
  • Structural: X-ray Reflectivity (XRR), High-Resolution X-ray Diffraction (XRD), Reflection High-Energy Electron Diffraction (RHEED), and Scanning Transmission Electron Microscopy (STEM) with 4D-STEM diffraction.
  • Chemical: X-ray Photoelectron Spectroscopy (XPS) with depth profiling, Time-of-Flight Secondary Ion Mass Spectrometry (ToF-SIMS), and Electron Energy-Loss Spectroscopy (EELS).
• Device Design:
  • LEPPC: Lumped-Element Resonators with Parallel-Plate Capacitors were fabricated.
  • Geometry: The design uses a parallel-plate capacitor geometry with a filling factor near unity, ensuring that measured losses are dominated by the bulk dielectric ($\gamma$-Al$_2$O$_3$) rather than the inductor or interfaces.
  • Air Bridge: An aluminum air bridge connects the top capacitor plate to the inductor, minimizing dielectric participation from other materials.

3. Key Contributions

• First Direct Measurement: This work reports the first direct measurement of the intrinsic dielectric loss of epitaxial $\gamma$-Al$_2$O$_3$ in a superconducting circuit.
• Heteroepitaxial Integration: Demonstrated the successful growth of a high-quality, single-crystal $\gamma$-Al$_2$O$_3$ dielectric sandwiched between TiN superconducting electrodes, overcoming previous interdiffusion issues seen with other superconductors (like Nb).
• Device Miniaturization: The LEPPC devices have a footprint of $2.4 \times 10^{-2}$ mm$^2$, representing a two-order-of-magnitude reduction compared to standard coplanar waveguide (CPW) resonators, addressing the scalability bottleneck.

4. Key Results

A. Material Quality and Structure

• Crystallinity: The stack exhibits high-quality heteroepitaxy. The bottom TiN grows epitaxially on sapphire, followed by cubic $\gamma$-Al$_2$O$_3$ (spinel-like structure, space group $Fd\bar{3}m$), and a top TiN layer.
• Interfaces: STEM and XRD confirm sharp interfaces with minimal roughness.
• Chemical Integrity:
  • TiN: Highly stoichiometric in the bulk (measured $T_c$ = 4.8 K).
  • Interdiffusion: Minimal anion interdiffusion. However, a thin 1–2 nm interlayer of mixed $Ti_xO_yN_z$ was observed at both TiN/$\gamma$-Al$_2$O$_3$ interfaces, attributed to oxygen migration during growth.
  • Oxidation: The TiN bulk remains resistant to oxidation, unlike other SC electrodes.

B. Dielectric Loss Performance

Using microwave lumped-element resonators at 10 mK, the team extracted the intrinsic TLS loss ($\delta_{TLS}^0$):

• Thin Sample (13.5 nm): $\delta_{TLS}^0 = (3.5 \pm 0.2) \times 10^{-5}$
• Thick Sample (58.3 nm): $\delta_{TLS}^0 = (2.8 \pm 0.1) \times 10^{-5}$
• Comparison: This represents an improvement of nearly two orders of magnitude over the amorphous AlOx used in standard JJs (which typically exhibit $\delta_{TLS}^0 \approx 1.4 \times 10^{-3}$).
• High-Power Performance: The devices showed exceptionally high quality factors at high power ($Q_{max} \approx 6.4 \times 10^5$ for the thick sample), though a power-independent loss mechanism (likely the interfacial $Ti_xO_yN_z$ layers or twinned domains) was observed.

5. Significance

• Coherence Improvement: By replacing amorphous AlOx with epitaxial $\gamma$-Al$_2$O$_3$, this platform offers a pathway to significantly extend qubit coherence times, a critical requirement for fault-tolerant quantum computing.
• Scalability: The compact LEPPC geometry enables high-density integration of qubits, essential for building large-scale quantum processing units (QPUs).
• Material Platform: The study validates Transition Metal Nitrides (TiN) combined with epitaxial oxides as a robust, wafer-scale compatible materials platform. This is particularly promising for emerging architectures like merged-element transmons and Microwave Kinetic Inductance Detectors (MKIDs).
• Industrial Relevance: The use of PLD and standard lithography suggests compatibility with industrial fabrication processes, unlike some alternative approaches (e.g., van der Waals transfer).

In conclusion, this work establishes that heteroepitaxial oxide-nitride stacks are a superior alternative to amorphous dielectrics, offering a combination of ultra-low microwave loss and compact device footprints necessary for the next generation of quantum hardware.