Design and Operation of Wafer-Scale Packages Containing >500 Superconducting Qubits

This paper presents a wafer-scale packaging architecture capable of supporting over 500 superconducting qubits on a single 3-inch die, demonstrating that large-scale integration can be achieved without compromising device performance while enabling high-throughput metrology for optimizing quantum processor fabrication.

The Gist

Imagine you are trying to build a super-computer that uses the laws of quantum physics to solve problems impossible for today's machines. The "brain" of this computer is made of tiny, fragile units called qubits. Think of these qubits as incredibly sensitive butterflies; they can do amazing things, but if you touch them too hard, or if the room is too warm, or if there's too much noise, they stop working.

The problem is that to build a useful quantum computer, you need millions of these butterflies. But right now, scientists can only fit a few dozen on a single chip without them crashing into each other or getting too hot.

This paper is about a team at Oxford Quantum Circuits who built a giant, high-tech "house" (a package) that can hold over 500 of these quantum butterflies on a single 3-inch silicon wafer (about the size of a dinner plate). They didn't just build a box; they engineered a sanctuary that keeps the butterflies safe, cool, and able to talk to each other.

Here is how they did it, broken down into simple concepts:

1. The "Anti-Noise" Shield (Stopping the Echo)

Imagine you are in a large, empty gymnasium. If you whisper, the sound bounces off the walls and comes back to you as a confusing echo. In a quantum computer, these "echoes" are called parasitic modes. They are stray radio waves that bounce around the metal box, confusing the qubits and causing errors.

• The Solution: The team filled their "gymnasium" with hundreds of tiny, metal pillars (like a dense forest of trees).
• The Analogy: Imagine trying to shout across a room. If the room is empty, your voice echoes. But if you fill the room with thousands of soft pillows or trees, the sound gets absorbed or scattered before it can bounce back. These pillars act like a "sound-dampening forest" for radio waves. They push the annoying echoes up to a frequency so high that the qubits can't hear them, leaving the qubits in a quiet, peaceful environment.

2. The "Thermal Contraction" Puzzle (The Shrinking Jigsaw)

Quantum computers must be kept at temperatures colder than outer space (near absolute zero). However, the package is built at room temperature.

• The Problem: When you cool things down, they shrink. Different materials shrink at different rates. Imagine a metal lid and a glass plate glued together. If the metal shrinks faster than the glass when cooled, the glass might crack, or the lid might crush the delicate chips inside.
• The Solution: The team used computer simulations to predict exactly how much every piece would shrink. They designed the "jigsaw puzzle" with extra space (gaps) so that when the pieces shrink, they don't smash into each other. They also used a special soft metal (Indium) that acts like a flexible gasket, allowing the parts to slide slightly without breaking the electrical connection.

3. The "Multiplexing" Highway (One Road for Many Cars)

Usually, to control one qubit, you need one wire. If you have 500 qubits, you need 500 wires. But a quantum fridge is so small and cold that you can't fit 500 wires going in; it would bring in too much heat and melt the computer.

• The Solution: They created a multiplexing system.
• The Analogy: Think of a busy highway. Instead of building 500 separate roads to 500 different houses, they built one super-highway with 56 "exits." At each exit, there is a special filter (a traffic cop) that directs the signal to 9 different houses (qubits) at once.
• The Result: They can control and read 500 qubits using only a handful of cables. This is like sending one letter that contains instructions for 500 different people, rather than writing 500 separate letters.

4. The Results: A Super-Performing Team

They tested this new "house" with over 100 qubits and found:

• Longevity: The qubits stayed alive (coherent) for about 100 microseconds. In the quantum world, this is like a marathon runner finishing a race without tripping.
• Accuracy: They could read the state of the qubits with 97.5% accuracy.
• Temperature: The qubits were kept at a frigid 36 millikelvin (colder than deep space).

Why Does This Matter?

This isn't just about building one big computer yet. It's about testing and learning.

Imagine a car manufacturer trying to build a million perfect cars. If they can only test one car at a time, it takes forever. But if they can test 500 cars at once on a single track, they can quickly see which parts are failing, which materials are best, and how to fix the manufacturing process.

This package allows scientists to test hundreds of qubits simultaneously. It acts as a high-speed feedback loop. If a specific part of the chip is failing, they can spot it immediately. This helps them figure out how to manufacture the millions of perfect qubits needed for a true, fault-tolerant quantum computer.

In short: They built a massive, ultra-cold, noise-canceling, shrink-proof apartment complex for 500 quantum butterflies, allowing them to live happily and work together. This is a crucial step toward building the quantum computers of the future.

Technical Summary

1. Problem Statement

The realization of fault-tolerant quantum computers requires scaling superconducting qubit counts from hundreds to millions. However, current packaging technologies struggle to support large-scale integration due to several critical challenges:

• Parasitic Microwave Modes: Large sample spaces (e.g., 3-inch wafers) host low-frequency cavity "box modes" that overlap with qubit frequencies, causing Purcell decay, crosstalk, and dephasing.
• Material Loss: Packaging materials (dielectrics, conductors, seams) introduce loss channels that degrade qubit coherence times ($T_1, T_2$).
• Thermal Contraction: Differential thermal contraction between packaging components (aluminum, sapphire, PCB) during cooling from room temperature to millikelvin temperatures can cause mechanical failure or misalignment of qubit-resonator coupling.
• Thermal Load: High wiring density required for large qubit arrays increases the heat load on dilution refrigerators, potentially preventing the system from reaching necessary operating temperatures (<20 mK).
• Manufacturing Feedback: Current testing methods often lack the statistical power to identify process outliers or optimize fabrication across large sample sizes.

2. Methodology and Design

The authors designed and operated a wafer-scale package capable of housing over 500 qubits on a single 3-inch die.

A. Core Architecture

• Qubit Type: The package supports "Coaxmon" qubits (transmons formed from coaxial capacitive pads) on a low-loss sapphire substrate.
• Multiplexing: Instead of individual control lines for every qubit, the design uses a 9-to-1 multiplexing architecture. The package contains 56 triangular multiplexing cells on a PCB. Each cell couples 9 qubit-resonator pairs to a single readout line via capacitive pins.
• Mechanical Structure:
  • A monolithic aluminum lid and base form a superconducting cavity.
  • Pillars: Machined pillars in the base pass through apertures in the wafer and spacer, creating a galvanic short between the lid and base.
  • Indium Joints: Indium is used at the pillar interfaces to ensure superconducting thermal and electrical contact while accommodating thermal contraction.
  • PCB Integration: The PCB (Rogers dielectric with copper traces) sits between the spacer and the base, acting as a Purcell filter and signal routing layer.

B. Mitigation Strategies

• Box Mode Suppression: The array of pillars acts as a microwave metamaterial, shifting the fundamental cavity mode from 2.4 GHz (in a bare cavity) to ~12 GHz. This moves the cavity modes out of the qubit (4–6 GHz) and readout (10 GHz) frequency bands, preventing Purcell decay and crosstalk.
• Loss Budgeting: The authors used Finite Element (FE) simulations to calculate Energy Participation Ratios (EPR) for various loss channels (dielectric, conductor, seam, and Purcell loss). They optimized the distance between the qubit and the lossy PCB to minimize dielectric and conductor losses.
• Thermal Contraction Management: FE thermal modeling was used to simulate the differential contraction of aluminum and sapphire from 300 K to 4 K. The design incorporates specific tolerances (gaps) around the pillars to prevent mechanical collision or breakage during cooling.
• Thermal Load Modeling: Simulations estimated the heat load for a fully wired system (including per-qubit drive lines and parametric amplifiers). The results indicated that even with a full wiring payload, the heat load (~3 µW at the mixing chamber) is well within the cooling power of commercial dilution refrigerators.

3. Key Contributions

1. Wafer-Scale Packaging: Demonstrated a functional package supporting >500 qubits on a single 3-inch die, a significant step toward industrial-scale quantum processor manufacturing.
2. Multiplexed Readout without Per-Qubit Control: Validated a high-throughput architecture where qubits are read out and characterized via multiplexed lines without individual drive lines, significantly reducing wiring complexity and thermal load.
3. Statistical Characterization Framework: Introduced a methodology for high-throughput feedback, using large-N datasets to identify manufacturing outliers and optimize fabrication processes, rather than just reporting the "best" qubit.
4. Material Interface Validation: Provided experimental lower bounds on the conductivity of Aluminum/Indium joints, confirming they are low-loss and suitable for superconducting packaging.

4. Experimental Results

The package was tested with a 3-inch die containing >500 qubit-resonator pairs. Measurements were performed on subsets of the qubits (105 qubits for coherence, 54 for readout fidelity).

• Coherence Times:
  • Median $T_1$: ~97 µs (measured over 105 qubits).
  • Median $T_{2e}$: ~129 µs (measured over 104 qubits).
  • Spatial Analysis: No significant spatial trends were found in coherence times across the wafer, suggesting uniform manufacturing quality. The coherence limits were found to be dominated by the material stack (aluminum pads and sapphire substrate) rather than packaging-induced losses.
• Readout Performance:
  • Fidelity: Median readout fidelity of 97.5% (error ~2.5%) for 54 qubits.
  • Error Budget: The dominant error source was qubit decay ($T_1$) during the readout pulse. The authors note that introducing parametric amplifiers and faster readout could further reduce this error.
• Effective Temperature:
  • Median effective qubit temperature: 36 mK (measured over 54 qubits).
  • This represents state-of-the-art performance for passive reset, indicating excellent thermal isolation and low heating from the package.
• Bootstrapping Analysis: The authors performed a statistical bootstrapping analysis showing that while median coherence times can be determined with small sample sizes, accurately determining the minimum coherence times (critical for identifying "error hotspots") requires large sample sizes (approaching half the total population).

5. Significance and Impact

• Scalability: This work proves that large-scale integration of superconducting qubits is feasible without compromising device performance. The packaging induces negligible additional loss compared to on-chip materials.
• Manufacturing Optimization: The ability to measure hundreds of qubits simultaneously provides a rich statistical dataset. This allows manufacturers to identify process outliers (the "tails" of the coherence distribution) and refine fabrication techniques to improve yield and uniformity.
• Roadmap to Fault Tolerance: By demonstrating that a package can support >500 qubits while maintaining millikelvin temperatures and high coherence, this architecture provides a viable pathway for the intermediate steps toward fault-tolerant quantum computers, which require millions of physical qubits.
• Community Utility: The design serves as a reference for high-throughput testing tools, enabling the community to move beyond single-qubit characterization to process-level optimization.

In conclusion, the paper presents a robust, thermally managed, and electromagnetically optimized packaging solution that successfully bridges the gap between small-scale quantum experiments and the massive scale required for practical quantum computing.