Variational Tensor Network Tomography Pioneers Path to Learning Elusive Topological Quantum States
Variational Tensor Network Tomography (VTNT)
A team of scientists from IBM Quantum, Princeton University, Harvard University, and Freie Universität Berlin has developed a potent new method for describing complex quantum states, especially those with exotic topological order, marking a significant advancement towards the full realization of the potential of quantum processors. Variational Tensor-Network Tomography (VTNT) is a heuristic tomographic technique that effectively learns the representations of difficult two-dimensional (2D) quantum states by utilizing machine learning and limited randomized measurements.
The advancement of quantum simulation has long been limited by the challenge of accurately expressing and validating highly entangled states, particularly those with topological order, which are characterized by fractionalized excitations and emergent gauge fields.
The effectiveness of Neural Network Quantum State Tomography (NNQST) for intrinsically long-range entangled topological states is still unknown, conventional quantum state tomography (QST) is typically not scalable, and Matrix Product State (MPS) tomography has trouble with complex entanglement patterns and 2D states.
You can also read Non-Gaussian States Improves Quantum Key Distribution
The Power of Variational Tensor-Network Tomography
The researchers used a strategy that combined randomized measurement techniques with variational optimization on tensor networks (TNs) to get beyond these challenges.
The main idea is to minimise a loss function that is determined by the difference between the model and the observations by fitting a scalable tensor network model (an MPS ansatz in the numerical demonstrations) to experimental data obtained through randomized measurements. Maximum Likelihood Estimation (MLE), which minimizes the negative log-likelihood (NLL), is used to formalize this approach.
Importantly, the procedure is very experimentally friendly and sample-efficient. All that is needed are random measurements, or snapshots of the quantum state taken in the bases.
The authors rigorously demonstrated that random measures are tomographically complete for learning any genuine pure states represented by tensor networks, which serves as a crucial theoretical foundation for this limited measurement technique. This result is important since many topological states can be selected as real pure states, including the eigenstates of real Hamiltonians that are important for quantum optimisation and condensed matter physics.
The method optionally adds a regularization term based on classical shadows to improve training robustness. This regularization penalizes discrepancies between estimations obtained directly from the measurement dataset and the model’s prediction of physical observables (such as Pauli strings).
You can also read Quantum Confinement Physics By Xinjiang Technical Institute
Successes in Simulation: Surface Codes and Spin Liquids
Numerical “experiments” mimicking two different topologically ordered systems were used to illustrate the efficacy of the Variational Tensor Network Tomography VTNT protocol:
- The Perturbed Surface Code: The Code for the Perturbed Surface A crucial example of a state utilized in fault-tolerant quantum computation, the ground state of the perturbed surface code Hamiltonian, was successfully learnt by the approach. The infidelity scaled polynomially with the number of samples for the exact surface code ground state, indicating effective learning.
- The researchers discovered that a high per-site local fidelity could be obtained with just a few thousand random measurements when examining a perturbed surface code on a strip geometry up to 45 qubits. Additionally, the infidelity was much decreased in the sample-limited regime by employing regularization based on predicted stabilizer expectation values.
- Rydberg Quantum Spin Liquid (QSL): QSL, or Rydberg Quantum Spin Liquid Characterizing the ground states of arrays of strongly interacting Rydberg atoms on a ruby lattice which can support a quantum spin liquid phase was the main goal of the second application. Since neither shows local symmetry-breaking order, it is frequently difficult to distinguish this topological phase from a simple disordered phase.
The Variational Tensor Network Tomography VTNT protocol effectively learnt the disordered phase and the spin-liquid state for a qubit system. The trained MPS model allowed the researchers to effectively predict important diagnostic features that are hard to measure directly, like the von Neumann entanglement entropy and the expectation values of nonlocal string operators. Determining the topological entanglement entropy, a crucial indicator of a gapped spin liquid, requires this entanglement information.
The outcomes validate the viability of the approach by demonstrating that it can faithfully recreate the target states with a fidelity that surpasses that of under samples.
You can also read Superconducting Josephson Junction Quantum Computing
Outlook for Near-Term Quantum Devices
The success of Variational Tensor Network Tomography VTNT offers a viable and practical route for characterizing 2D states on near-term quantum simulators, especially when using basic random measurements. The researchers offer a number of directions for future investigation. Future research could extend the methodology to true 2D tensor networks such as Projected Entangled Pair States (PEPS), even though the quasi-one-dimensional MPS ansatz was employed in the current numerical trials. Additionally, the technique might be used to investigate fermionic quantum states in optical lattice systems.
Notably, the random measurement technique is very well suited to the hardware available today, particularly Rydberg quantum simulators, where high-fidelity logical encodings make it simple to implement basis measurements, providing both error detection and state characterization. The ability to precisely estimate states via straightforward single-qubit rotations will continue to be appealing after the current NISQ era, even as quantum control advances. Additionally, Variational Tensor Network Tomography VTNT’s character presents a singular chance to get insight into experimental flaws, offering crucial input for attaining greater quality state preparation.
You can also read SpeQtral Free Space Quantum Communication Trials At NUS
