Storing Quantum Data Without Constant Fixing

In the paper's words

Scalable quantum computing is limited by the thermodynamic overhead of active error‑correction cycles, which require repeated syndrome extraction, low‑latency decoding, and large ancilla resources. A quantitative gap exists: no fully derived account shows how non‑Archimedean (ultrametric) geometry can passively confine errors, compute exact code parameters, logical‑failure probabilities, and the associated cooling budget. We formalize logical qudits on finite balls of the Bruhat–Tits tree T_p+1, derive the exact ball‑size formula n_t=1+frac(p+1)(p^t-1)p-1, the conditional distance d=2t+1, and the barrier height B(d)=lceil p/2rceil^d (which equals 2^d for p=3) for odd p.

The abstract, verbatim; every line below is quoted from the text.

What it claims

that ultrametric geometry yields exponential passive suppression—would be falsified if an experiment measuring logical error rates for a depth‑3 tree with p=3 and physical error rate p_err=0.40 observed a logical error probability significantly above the bound p_L(3)=6.55×10^-4 (e.g., exceeding three standard deviations of the binomial expectation for N=10⁴ trials). Observation of logical errors at a rate comparable to the physical error rate would also falsify the barrier model.

Numbers it reports

1. Thermal‑activation analysis gives $p_L(d)\le e^{-B(d)}$ at $k_{\mathrm{B}}T=1$, and a cooling‑budget requirement $\Delta\ge1.727\,k_{\mathrm{B}}T$ per site for a $10^{-12}$ logical‑error target at depth 4.

   from Abstract

2. Thermal‑activation bounds $p_L(d)\le e^{-B(d)}$ and cooling‑budget projection $\Delta\ge1.727\,k_{\mathrm{B}}T$ for a $10^{-12}$ target (Section 4.5).

   from 1. Introduction

3. The standard active surface‑code paradigm and its thermodynamic overhead are detailed in [10], motivating the search for passive alternatives.

   from 2. Background and Related Work

4. The number of vertices follows from the tree’s branching factor $p+1$: $$ n_t = 1 + \frac{(p+1)(p^{t}-1)}{p-1}. $$ The distance is $d=2t+1$, because any error affecting $\le t$ vertices cannot change the majority vote of the root’s neighbourhood [12].

   from 3. Methods

5. Adding higher‑order terms changes the value only in the third significant digit, yielding $$ P_{\mathrm{pass}}^{(2)}\approx6.73\times10^{-7}. $$ The corresponding suppression factor is $\varepsilon/P_{\mathrm{pass}}^{(2)}\approx1.49\times10^{3}$.

   from 4.3 Passive logical‑failure probability $P_{\mathrm{pass}}$

6. Assuming independent physical error probability $p_{\mathrm{err}}$, the worst‑case logical error bound is $$ p_L(d)=p_{\mathrm{err}}^{\,B(d)}. $$ With $p_{\mathrm{err}}=0.40$: $d=3$: $p_L(3)=0.40^{8}=6.5536\times10^{-4}$. $d=4$: $p_L(4)=0.40^{16}=4.295\times10^{-7}$.

   from 4.4 Barrier height and logical‑error bounds

How it is done

1. Formal definition of a $p$‑adic ball code on $T_{p+1}$ and proof of the error‑confinement theorem (Section 4.1).

   from 1. Introduction

2. Linear‑optics error averaging [3] proposes a protocol that reduces noise in linear‑optics networks by averaging over multiple paths.

   from 2. Background and Related Work

3. Our analytical pipeline proceeds in three stages: Geometric model.

   from 3. Methods

4. The Bruhat–Tits tree $T_{p+1}$ is a regular graph where each vertex has $p+1$ neighbours.

   from 4.1 Code‑size derivation

What would falsify it

1. The main claim—that ultrametric geometry yields exponential passive suppression—would be falsified if an experiment measuring logical error rates for a depth‑3 tree with $p=3$ and physical error rate $p_{\mathrm{err}}=0.40$ observed a logical error probability significantly above the bound $p_L(3)=6.55\times10^{-4}$ (e.g., exceeding three standard deviations of the binomial expectation for $N=10^{4}$ trials).

   from 7.2 Falsification criteria

2. Observation of logical errors at a rate comparable to the physical error rate would also falsify the barrier model.

   from 7.2 Falsification criteria

3. These results convert the ultrametric passive‑QEC conjecture into explicit, falsifiable numbers and identify symmetric‑encoder construction as the remaining open task.

   from Abstract

Where the claim stops

1. The distance $d=2t+1$ derived in Section 4.2 assumes a symmetric encoder that respects branch‑permutation symmetry; no explicit construction is provided [12].

   from 7.1 Limitations

2. Independent depolarising noise per vertex is assumed.

   from 7.1 Limitations

3. Correlated errors across neighboring vertices or burst errors would violate the independence assumption and could increase $P_{\mathrm{pass}}$ [12].

   from 7.1 Limitations

Where to read

1. Abstract158 words · 1 min
2. 1. Introduction254 words · 1 min
3. 2. Background and Related Work272 words · 1 min
4. 3. Methods230 words · 1 min
5. 4. Analysis405 words · 2 min
6. 5. Results1,104 words · 5 min
7. 7. Discussion245 words · 1 min
8. 7. Practical Implications and Engineering Pathways330 words · 1 min
9. 9. Conclusion113 words · 1 min
10. Provenance and AI Involvement225 words · 1 min