Research

PERSPECTIVE

Our research perspective

Complex problems in mathematics and theoretical physics can contain aspects that do not require detailed examination, while even a slight oversight in others can substantially alter the conclusion. In these fields, the existence of a proof and an understanding of why it works—its underlying structure—are distinct matters. UCL uses mathematical approaches to investigate these differences.

We explore when a correct answer can be reached without examining every element. When the desired conclusion still cannot be drawn, we carefully investigate which clues are missing and the fundamental reasons why.

VISUAL EXAMPLE

Recognizing without seeing everything.

A visual metaphor
  1. Only a fragment is visible.

    Fine visual noise obscures the image, leaving just a red fragment visible.

  2. Understand the noise.

    Consider how the brightness and distribution of dots affect visibility.

  3. Choose where to look.

    Look at features that help distinguish the object: spots and the wing seam.

  4. Enough to recognize.

    Much of the noise remains, but the visible features are enough to recognize a ladybird.

A red fragment alone is not enough.

Investigate which clues are missing, and why identification remains uncertain.

QUESTIONS

Questions at the heart of our research

  • Under what conditions can we reach a correct answer without examining every element?
  • When a conclusion cannot be reached, which clues are missing, and why?
ABSame responseNot yet distinguishableABDifferent responsesNew sensor: A onlyAdd 1

This focus on where to look to distinguish candidates connects to research on observation and test placement, and conditions for identification.

Graph theory

  • Identifying codes & dominating sets

    How can fewer observations still distinguish locations?

    Sensor placement, missing reports, and the cost of observing while moving.

    Published identification results
  • Triangle packing & covering

    Can we prove Tuza's conjecture for a wider range of graphs?

    Relations between packing triangles and covering them with edges.

    Preprint · arXiv

Quantum mechanics

  • Quantum state identification & certification

    Can partial measurements certify the whole system's state?

    We study the conditions under which partial measurements can certify closeness to a target state.

    Preprint · Zenodo
  • Quantum measurement & incompatibility

    Which processes preserve measurement incompatibility?

    How processing a quantum system affects measurements that cannot be performed jointly.

Information & computation

  • Group testing & fault identification

    How can additional tests distinguish faults at minimum cost?

    Where to add individual tests when pooled tests on pairs cannot distinguish faults.

RESEARCH ARCHIVE

Publications

All publications11papers

Newest first

Quantum mechanicsExamining & extending prior research

Band-sum corrections and finite-mesh evaluation

Can the electronic-band conjecture be proved, and under what conditions can transition signs be evaluated on a finite mesh?

Summary & scopeClose details

Contribution to prior research
Zenodo · Preprint

Overview

We prove Conjecture 1 in the original electronic-band construction of Ono and Shiozaki (Theorem 2). Theorem 6 evaluates the source’s one-cell transition signs in symmetry class AI, retaining the original conventions and admissibility throughout each mesh interval. AI here denotes a symmetry class, not artificial intelligence.

Scope and verification

The general proofs are analytic. The accompanying HTML verification guide and three exact-arithmetic checks illustrate identities and assumptions in the displayed examples. The result does not certify topology from arbitrary discrete samples alone or provide a complete classification of phases or higher-cell invariants.

Read the paper and verification materials (Zenodo)

Band-sum corrections and finite-mesh evaluation of the Ono–Shiozaki invariants

Quantum mechanicsOriginal research

Quantum marginals and sharp error guarantees

When local observations are close to a target, how close must the full quantum state be?

Summary & scopeClose details

Zenodo · Preprint

Submission to arXiv is planned.

Overview

For specified quantum states, we prove sharp limits on the global error guarantees obtainable from local observations. With the same twelve-qubit target and observation map, the strength of the guarantee depends on whether competing states may be entangled.

Scope of the result

For a fixed twelve-qubit product target and all 66 two-qubit marginals, the sharp trace-norm error exponent is 1/2 for competitors separable across the six:six cut and 1/4 for unrestricted competitors. Adding the three-qubit marginals of one six-qubit block restores exponent 1/2 for unrestricted competitors. Sharpness concerns the worst-case asymptotic exponent as the observation error tends to zero.

Sharp robustness of quantum marginal reconstruction under composition

Graph theoryExamining & extending prior research

Identifying-code density bounds on two infinite grids

How sparsely can vertices be marked if nearby marks must distinguish every vertex?

Summary & scopeClose details

Zenodo · Preprint

An arXiv version is planned.

Overview

We improve density bounds for identifying codes in the hexagonal and truncated trihexagonal grids. The lower bounds also cover nonperiodic configurations; explicit constructions provide upper bounds.

Scope of the result

The hexagonal-grid lower bound is 18/43 + 1/79310. For the truncated trihexagonal grid, the unrestricted minimum lies between 1829/4263 and 45/104. The exact unrestricted minimum remains open in both grids.

Read the paper and verification materials (Zenodo)

Density bounds for identifying codes in the hexagonal and truncated trihexagonal grids

Graph theoryExamining & extending prior research

Exactly which hexagonal strip widths attain density 1/3?

How does strip width change the density needed for locating-domination?

Summary & scopeClose details

Zenodo · Preprint

Overview

For an infinite hexagonal strip, the minimum locating-dominating density is 1/3 exactly when the width is divisible by 3. At the other widths, the strip boundary forces a positive density excess.

Scope of the result

For width k not divisible by 3, the excess is of order 1/k². Exact densities are proved for widths 7, 8 and 10, with lower bounds covering nonperiodic configurations. The general exact-density conjecture remains open.

Read the paper and verification materials (Zenodo)

Locating-dominating sets in hexagonal strips: minimum density 1/3 exactly at widths divisible by 3

Graph theoryExamining & extending prior research

Exact repair costs for structured observation failures

When some observations fail, which combinations of repair costs are feasible?

Summary & scopeClose details

Zenodo · Preprint

Overview

For a three-dimensional majority rule, we determine the joint trade-offs between net observations, edits and acquisitions. The paper gives all feasible joint budgets, their Pareto frontier, and constructive repairs.

Scope of the result

The result applies to the specified family of structured five-site failures, reference observation mask, majority rule and repair region. It does not claim the same formulas for arbitrary failure sets.

Read the paper and verification materials (Zenodo)

Exact repair trade-offs for structured observation failures in a three-dimensional majority rule

Graph theoryOriginal research

Tuza's conjecture and triangle overlap

How can triangle-overlap conditions extend the class of graphs satisfying Tuza's inequality?

Summary & scopeClose details

arXiv:2611.11128v1 · Preprint

Overview

We prove Tuza's inequality for finite simple graphs under specified triangle-overlap and density conditions. We also localize the density condition to the subgraph induced by the vertices of the Hifumi core, as defined in the paper.

Scope of the result

The inequality holds when each edge belongs to at most four triangles, the edges belonging to four triangles form components with at most two edges, and the maximum average degree is at most 25/2. The paper gives a stronger strict bound and its conditions. This is not a proof of the general Tuza conjecture.

Triangle-overlap cores and density bounds for Tuza's conjecture

Graph theoryExamining & extending prior research

Width-eleven strip

What sensor density is sufficient for a hexagonal strip eleven rows wide?

Summary & scopeClose details

Overview

Sensors must distinguish every unselected vertex in an infinite strip eleven rows wide. This paper determines the exact minimum locating-dominating density: 48/143.

Details

The lower bound 48/143 holds for all locating-dominating sets, including nonperiodic ones. A 26-column periodic construction selects 96 of 286 vertices and attains equality. This settles width eleven in Gomes and Wakabayashi's Conjecture 9, not the conjecture at all widths.

View the paper and verification materials on Zenodo

The exact locating-dominating density of the hexagonal strip of width eleven

Graph theoryExamining & extending prior research

Walks on the king grid

How many visits are needed to identify locations while moving, when revisits also count?

Summary & scopeClose details

Overview

A walk on the king grid may move horizontally, vertically, or diagonally. Its visited vertices must form an identifying code. This paper bounds the number of visits, counting repeated visits to the same vertex.

Details

For an n×n board with n≥6, the minimum number of visits M_n satisfies 28n²/113 − 7n ≤ M_n ≤ n²/4 + 36n. The limit of M_n/n² exists and lies between 28/113 and 1/4. Whether the limit equals 1/4 remains open.

View the paper and verification materials on Zenodo

Bounds for dynamic identifying codes in the king grid

Graph theoryExamining & extending prior research

Error detection: king grid

How many detectors are needed when one detector's positive report may be lost?

Summary & scopeClose details

Overview

Each detector senses the eight neighboring vertices of the king grid, excluding its own vertex. This study improves the density lower bound for DET:OLD sets under the specified model of a false-negative report at an unknown detector.

Details

Every DET:OLD set has lower limiting graph-ball density at least 79/192 at every center. Together with the known upper bound, the minimum lies between 79/192 and 13/30. No periodicity is assumed, and the exact minimum remains open.

View the paper and verification materials on Zenodo

An improved lower bound for error-detecting open-locating-dominating sets in the king grid

Graph theoryExamining & extending prior research

Error detection: square grid

How many detectors are needed when reports distinguish a detector's own location from its neighbors?

Summary & scopeClose details

Overview

On the square grid with four neighbors per vertex, a detector reports different values for its own location, an adjacent location, and all other locations. This paper improves the density lower bound when a nonzero report at one unknown detector may be lost.

Details

Every DET:LD set has lower limiting graph-ball density at least 168/295 at every center. With the known upper bound, the minimum lies between 168/295 and 7/12. The result includes nonperiodic sets; the exact minimum remains open.

View the paper and verification materials on Zenodo

An improved lower bound for error-detecting locating-dominating sets in the square grid

Graph theoryExamining & extending prior research

Kagome grid

Can sensors distinguish every location in the kagome grid without sensing their own vertices?

Summary & scopeClose details

Overview

In the kagome grid of triangles and hexagons, every vertex, including sensor locations, must have a distinct nonempty set of neighboring sensors. This paper establishes the exact minimum open-locating-dominating density of 2/5.

Details

The known universal lower bound 2/5 is explicitly transferred to the lower limiting graph-ball density of every OLD set at every center. A periodic construction selecting 12 of 30 vertices attains equality. The lower bound does not assume periodicity.

View the paper and verification materials on Zenodo

The exact open-locating-dominating density of the kagome grid