Qudit Quantum Error Correction

In the paper's words

Extends the Ultrametric Foundation thesis into quantum computing. Formalizes geometric error confinement on tree-topology quantum processors.

The abstract, verbatim, and its opening sentence is the claim. Every line below is quoted from the text.

Numbers it reports

1. This is because: With 343 qubits at $p = 10^{-3}$, the expected number of errors per shot is ~0.34, and when 2+ errors occur, they land in different major tree branches purely by random chance — the tree cannot "confine" errors that are independently distributed.

   from 2.5 Simulation Results

2. Thermal management: 400 qubits in a 3D stack generate heat; cryogenic cooling at millikelvin temperatures must handle this load [established constraint] .

   from 6.2 3D Integration Challenges

3. We formalize three theses: (I) a tree-topology quantum processor confines errors to subtrees via the strong triangle inequality, elevating the error-correction threshold; (II) a 3D-integrated superconducting architecture realizing a 7-ary tree of depth 3 (343 physical qubits) is manufacturable with current fabrication technology; (III) an on-chip bank of prime-frequency resonators enables spectral engineering of the tree's error-confinement properties.

   from Abstract

4. The numerical simulation estimates: $$p_{\text{th}} \approx 2.0 \times 10^{-4}$$ This is significantly lower than the surface code threshold ($p_{\text{th}} \approx 1.1 \times 10^{-2}$ on a 2D grid).

   from 2.3 Threshold Scaling

5. The tree topology does NOT improve the threshold for independent errors — the geometric confinement advantage requires correlated error models [CODE-EXECUTED — threshold estimated via Monte Carlo simulation at 2500 shots/point, 50 error rates in $[10^{-5}, 10^{-1}]$] .

   from 2.3 Threshold Scaling

6. Parameter Value Source $T_1$ (relaxation) $\gt 300\,\mu\text{s}$ Coaxial 3D cavities, Al 6N purity [established] $T_2^{\text{echo}}$ $\gt 200\,\mu\text{s}$ Demonstrated at MIT-LL [established] Gate fidelity (single-qubit) $\gt 99.9\%$ State of art transmons [established] Gate fidelity (two-qubit tree edge) $\gt 99.5\%$ Coaxial cavity-mediated; lower than planar due to cavity coupling [speculative] Coherence per TSV hop $\gt 99.9\%$ Superconducting TSVs demonstrated at MIT-LL [established]

   from 3.2 Transmon Parameters

How it is done

1. Rather than laying qubits in a 2D grid (the standard approach for surface codes), connect them in a tree topology that mirrors their tensor-product structure.

   from 1.3 From Mathematical Structure to Physical Topology

2. Consider an $n$-qubit tree where physical connectivity follows the tensor-product hierarchy.

   from 2.1 Error Model

3. Theorem 1 (Subtree Error Confinement).

   from 2.2 Confinement Theorem

4. However, color codes map the hierarchy onto a 2D lattice, whereas the tree-topology approach makes the hierarchy physical.

   from 5.2 Color Codes

What would falsify it

1. The thesis would be disconfirmed if: A depth-2 tree of 8 transmon qubits connected via coaxial cavities, with tree-only gates (no grid interactions), shows no statistically significant improvement in logical error rate over an 8-qubit 2D nearest-neighbor layout with identical qubit parameters ($T_1, T_2$, gate fidelity).

   from 2.6 Disconfirmation Condition

2. The thesis predicts that coaxial cavity coupling preserves coherence better than planar coupling due to reduced cross-talk and the tree's natural error confinement [my conjecture] .

   from 3.3 Fabrication Feasibility

Where the claim stops

1. An error affecting one subtree cannot propagate asymmetrically to different branches — it is geometrically confined.

   from 1.2 Ultrametric Distance on Qubit Registers

2. The tree topology does NOT improve the threshold for independent errors — the geometric confinement advantage requires correlated error models [CODE-EXECUTED — threshold estimated via Monte Carlo simulation at 2500 shots/point, 50 error rates in $[10^{-5}, 10^{-1}]$] .

   from 2.3 Threshold Scaling

3. This is because: With 343 qubits at $p = 10^{-3}$, the expected number of errors per shot is ~0.34, and when 2+ errors occur, they land in different major tree branches purely by random chance — the tree cannot "confine" errors that are independently distributed.

   from 2.5 Simulation Results

Where to read

1. Abstract119 words · 1 min
2. 1. The Ultrametric Structure of Quantum State Space324 words · 1 min
3. 2. Thesis I: Geometric Error Confinement996 words · 4 min
4. 3. Thesis II: Physical Realizability377 words · 2 min
5. 4. Thesis III: Prime-Frequency Spectral Engineering291 words · 1 min
6. 5. Comparison with Existing Quantum Error Correction195 words · 1 min
7. 6. Limitations and Open Problems216 words · 1 min
8. 7. Connection to the Ultrametric Foundation Thesis107 words · 1 min
9. 8. Prototype Implementation69 words · 1 min
10. Changelog9 words · 1 min