SECURITY THROUGH DECOHERENCE

Quantum
Interference
Resolved

A machine-learning approach to quantum interference: characterising a device’s noise process and recovering coherent signal from its readout.

Get in touch See how it works
SCROLL

Every quantum system is an open system. What it records is a signal entangled with its environment. The task is to tell them apart.

§ 1 — THE PROBLEM

Decoherence as the limiting factor in quantum computation.

No physical qubit is perfectly isolated. Coupling to the environment causes the off-diagonal terms of the density matrix to decay, converting superposition into classical mixture on timescales characterised by T₁ and T₂. Gate imperfection, crosstalk and measurement error compound the effect, so that the observed signal is the intended computation composed with the device’s noise channel.

STATE · DECOHERENT
FIG. 1 — TWO-SOURCE INTERFERENCE. RANDOM PHASE NOISE WASHES OUT THE FRINGES; REMOVING IT RESTORES VISIBILITY.
1.1

Decoherence

Energy relaxation (T₁) and dephasing (T₂) progressively destroy phase information, bounding the depth of any circuit that can be executed reliably.

1.2

Crosstalk

Unintended coupling between neighbouring qubits and control lines introduces correlated errors that independent-noise models fail to capture.

1.3

Readout error

State discrimination at measurement is imperfect; assignment errors bias estimated expectation values even when the preceding circuit is exact.

§ 2 — METHOD

Learning the noise channel, then inverting its effect.

Rather than assuming a fixed parametric noise model, Telecount0 learns the error process of each device from its own data and uses that estimate to recover the underlying signal.

RAW READOUT RESOLVED SIGNAL
FIG. 2 — RAW READOUT AND MITIGATED ESTIMATE OF THE SAME OBSERVABLE (SCHEMATIC).
  1. 2.1

    Characterise

    Collect readout and calibration data continuously, forming an empirical record of how the device departs from ideal behaviour.

  2. 2.2

    Model

    Fit a learned representation of the noise process — its sources, correlations and temporal drift — specific to the device under study.

  3. 2.3

    Mitigate

    Apply the model to estimate the noise-free signal, returning coherent observables suitable for downstream computation.

§ 3 — RESEARCH

A research programme, documented for scrutiny.

Qubits drift as their environment changes through the day. Telecount0 pairs pulse-level correction with autonomous AI agents that watch for that drift and cancel it as it happens. Validated over 72 hours on a 127-qubit IBM superconducting processor, against standard and static-mitigation baselines.

0.89
Kernel purity, up from 0.61 unmitigated
3.8×
Lower average gate error than unmitigated circuits
92.8%
Prediction accuracy (F1), up from 68.2% unmitigated
Telecounto Research Group, IBM Quantum Research Fellow, Center of Excellence in Quantum Technology. Dynamic Quantum Drift Mitigation via Autonomous Agentic Pulse-Level Hamiltonian Inversion for High-Fidelity Educational Data Mining. FULL PAPER ON REQUEST
§ 4 — SECURITY

Security through decoherence.

Oracle and IBM systems for quantum states.

Quantum-resistant security and cryptography, built on a proprietary Kyber stack. A quantum state cannot be observed without being disturbed, so any interception shows up in the signal.

  • MEETS
  • NIST
  • SOTA2
PURITY
1.000
θ
0.0°
φ
0.0°
A Bloch sphere: the qubit's state vector precesses and shrinks toward the centre as it decoheres, then is restored.
EARLY ACCESS

Correspond with the research team.

For collaboration, technical briefings or access to preprints. Tell us about your device and we will follow up.

Telecount0 is a division of Petri Research.